The Sanskrit Language (T.Burrow)

by T.Burrow

The Sanskrit Language (T.Burrow) T.Burrow १९७२

The

Sanskrit Language

T. Burrow

MOTILAL B ANARS ID ASS PUBLISHERS PRIVATE LIMITED • DELHI First Indian Edition: Delhi , 2001

First Edition: UK 1955

<D FABER & FABER LTD., LLK. All Rights Reserved.

ISBN: B 1-208-1 767-2

Also available at:

MOTILAL BANARSIDASS

41 U.A. Bungalow Road, Jawahar Nagar, Delhi 110 007 8 Mahalaxmi Chamber, Warden Road, Mumbai 400 026 120 Royapettah High Road, Mylapore, Chennai 600 004 236, 9th Main III Block, Jayariagar, Bangalore 560 011 Sanas Plaza, 1302 Baji Rao Road, Purie 411 002 8 Camac Street, Kolkata 700 017 Ashok Rajpath, Patna 800 004 Ghowk, Varanasi 221 001

FOR SAIF IN SOOTH ASIA ONLY

Printed in India

BYJAINENDRA PRAKA5HJAIN AT SHRI JAJNENDRA PRESS, A -45 NARAINA, PHASE4, NEW DELHI 110 028 AND PUBLISHED BYNARENDRA PRAKASH JAIN FOR MOTILAL BANARSIDASS PUBLISHERS PRIVATE LIMITED, BUNGALOW ROAD, DELHI 110 007 PREFACE

The discovery of the Sanskrit language by European scholars at the end of the eighteenth century was the starting point from which developed the study of the comparative philology of the Indo-European languages and eventually the whole science of modern | linguistics. In spite of this there does not exist in English any book presenting a systematic account of Sanskrit in its relation to the other Indo-European languages. One may even go further and say that there is no work in any language which adequately fulfils this purpose. Wackernagel's great work, begun sixty years ago, still remains to be completed, although, with the recent appearance of a further instalment, its completion has been brought nearer. Thumb’s Handbuch des Sanskrit which was of service to many generations of students is now very much dated, and always fell between the two stools of trying to be an elementary text-book of Sanskrit and a treatise on its comparative grammar at the same time.

On account of its antiquity and well-preserved structure Sanskrit is of unique importance for the study of Indo-European, and an up-to-date account of its comparative grammar is necessary, not only to students of Sanskrit itself, but also to those interested in any branch of Indo-European philology. Consequently when I was asked to contribute a book on Sanskrit to the series The Great Languages , it was clear that by concentrating on the study of Sanskrit from this point of view the greatest need would be met. This is particularly true since for the history of Indo-Aryan inside India, from Sanskrit down to modern times, students already have at their disposal the excellent work of Jules Bloch.

Providing a reliable account of Sanskrit in its relation to Indo-European is at the present moment not altogether a simple matter. Forty years ago there existed a generally agreed doc- trine of Indo-European theory which had been systematically presented in the early years of the century in Brugmann's Grundriss, At that time it would merely have been a question VI

PREFACE

of adopting this corpus of agreed doctrine to the needs of the student and general reader, and of the particular language described. Since then the .discovery of Hittite has revolution- ised Indo-European studies and a considerable part of the older theory has been unable to stand up to the new evidence. Consequently Indo-European studies can now be said to be in a state of flux. New theories have appeared, and are clearly necessary, but the process is not yet completed. There is no generally received body of doctrine replacing the old, and many of the fundamental points at issue remain disputed. Further- more attention has tended to be largely concentrated on phonetic questions raised by Hittite, and matters of morpho- logy, on which its evidence is also of fundamental importance, have been less exhaustively studied.

In these circumstances I have attempted to present a reason- ably consistent account of the comparative grammar of Sanskrit based on the evaluation of the new evidence, A work like this is not the .place to enter into discussion of the various conflicting theories that are in the field, if only for reasons of space, and bibliographical references have been systematically omitted. What has been written in recent years on these problems has been taken into account, and such theories as appear acceptable are incorporated in this exposition. It is hoped that it will go some way to providing an up-to-date synthesis of a subject which in its present state is hardly accessible outside the widely scattered specialist literature.

The study of Sanskrit has advanced recently in another direc- tion also. Investigation of the influence of the pre-Aryan languages of India on Sanskrit and on Indo-Aryan in its later stages, has shown that this is considerable and solid results have been achieved. As far as the structure of the language is concerned, particularly in its early stage, which is the only one relevant to the comparative study of Indo-European, this influ- ence hardly counts at all. On the other hand in the field of vocabulary it is very important that the Indo-European and non-Indo-European elements should be separated. The last chapter of the book contains a summary of the main finding^ on the part of the subject so far as established at the present stage. Future work will no doubt add more.

T. Burrow PREFACE TO THE THIRD EDITION

A number of alterations to the text of the The Sanskrit Language have been made in this edition, the principal ones being as follows. In Chapter I the latter part of Section 6 has been rewritten to conform with the now prevailing opinion that the Aryan vestiges of the ancient Near East are to be connected specifically with Indo- Aryan. Also rewritten are Section u and (in part) Section 17 of Chapter III to take account of the conclusions reached on those topics in the articles of mine which are quoted in the Appendix, Chapter VIII has been renamed Loanwords in Sanskrit, so that loanwords from Greek and Iranian (Section 2) can be dealt with in it as well as loanwords from AusLfo- Asiatic and Dravidian (Section 1). The list of loanwords from Dravidian in this chapter has been shortened by the omission of some items now' considered to. be false or dubious (

At the end an Appendix has been added containing references to the moat important contributions to the subject which have appeared since 1955, and also some supplementary notes.

September IQJ2

T. Burrow CONTENTS

PREFACE page V

I. SANSKRIT AND INDO-EUROPEAN I

II, OUTLINES OF THE HISTORY OF SANSKRIT 35

III. PHONOLOGY 67

IV. THE FORMATION OF NOUNS Il 8

V. THE DECLENSION OF NOUNS 220

VI. NUMERALS, PRONOUNS, I NDECLIN ABLES 258

VII, THE VERB 289

VIII. LOANWORDS IN SANSKRIT 374

APPENDIX TO THE THIRD EDITION 390

SELECT BIBLIOGRAPHY 399

INDEX 402

ix CHAPTER I

SANSKRIT AND INDO-EUROPEAN

§i. Indo-Aryan and Indo-Iranian

In the greater part of India today languages are spoken which are derived from a single form of speech which was introduced into India by invaders from the north-west more than three thousand years ago. The invading peoples were knowm in their own language as ivy a-, a word which is also commonly used as an adjective meaning * noble, honourable Behind them in Central Asia remained kindred peoples who eventually occupied the plateau of Iran, as well as large tracts of Central Asia. These peoples used the same name of themselves, in Avestan airy a-, and from the genitive plural of this word the modern name Iran is ultimately derived. In conformance with this usage the term Aryan is now used as the common name of these peoples and their languages ; alternatively the term Indo- Iranian is commonly used. To distinguish the Indian branch from the Iranian, the term Indo-Aryan has been coined, and as applied to language, it covers the totality of languages and dialects derived from this source from the earliest times to the present day. It is practical to distinguish three periods, Old, Middle and Modern Indo-Aryan. The classical form of Old Indo-Aryan eventually came to be designated by the term Samskrta- meaning ' polished, cultivated, correct (according to the rules of grammar) in contradistinction to Prdkrta the speech of the uneducated masses, which was the same Indo- Aryan in origin, but was subject to a process of steady change and evolution. As a term to distinguish Indo-Aryan from the non-Aryan languages the adjective ary a- was used in opposition to mlecchd- 4 barbarian In addition we may note that one of the terms for ' speech bkarati (sc. vdk) had originally an ethnic sense, meaning * language of the Bharatas ’A

1 At an early period the most prominent of the Indo-Aryan tribes, whence also the indigenous name of India bharata(-varsa). SA N S K R IT A N DI DO-EU ROPE A N Sanskrit in its narrower sense applies to standard classical Sanskrit as regulated by the grammarians but may be con veniently used more widely as equivalent to Old Indo-Aryan In this sense it covers both classical Sanskrit and the pre. classical or Vedic language. Middle Indo-Aryan, that is Prakrit in the widest sense of the term, comprises three successive stages of development : (1) The earliest stage is represented in literature by Pali, the language of the canonical writings of the Thera-yuda school of Buddhism. This is a language of the cen turies immediately preceding the Christian era. On the same level of development are the various dialects recorded in the inscriptions of Asoka (c. 250 B.C.), and also the language of other early inscriptions. (2 ) Prakrit in the narrower sense of the word or Standard Literary Prakrit, represents the stage of develop ment reached some centuries after the Christian era. It is found mainly in the Drama and in the religious writings of the Jains. The various literary forms of Prakrit were stabilised by grammarians at this period and, as a written language, it re mains essentially unchanged during the succeeding centuries. 3) Apabhramsa is known from texts of the tenth century A.D but as a literary language it was formed some centuries earlier. It represents the final stage of Middle IndoAryan, the one immediately preceding the emergence of the Modern Indo Aryan languages. The Modern languages, Bengali, Hindi Gujarati, Marathi, etc.begin to be recorded from about the end of the first millennium A.D., and from then their development can be followed as they gradually acquire their present-day form. Thus we have before us in India three thousand years of continuous linguistic history, recorded in literary documents. During the course of this period a single, and originally alien idiom has spread over the greater part of the country, and, evolving by slow degrees, has resulted in the various languages now spoken in Northern and Central India. Enormous changes have taken place during this time, and the languages we meet today are very different indeed from the ancient speech spoken by the invading Aryan tribes. Nevertheless the docu- mentation available enables us to follow in detail the various intermediate stages of development and to observe how by changes hardly noticeable from generation to generation, an original language has altered into descendant languages which superficially at any rate, are now barely recognisable as the same. SANSKRIT AND INDO-EUROPEAN 3 The earliest document of the linguistic history of Indo- ATyan is the Rgveda, which, by rough guess-work, is placed in the region of 1000 b,c. The language we find there is the source from which all later developments in India have arisen. But this language itself had evolved out of a yet earlier form of speech, by precisely the same kind of slow change and altera- tion which caused it to evolve later into something else. This earlier evolution is unrecorded by any direct documentation, but it can be reconstructed in considerable detail by means of comparison with related languages. By this method two stages in the prehistory of the language can be established : (i) By comparison of early Indo-Aryan with the very closely related Iranian, it is possible to form a fairly accurate idea of the original Indo-Iranian or Aryan language from which both have evolved. (2) By comparing Indo-Aryan and Iranian with the other Indo-European languages (enumerated below) it is pos- sible also to go beyond this, and to reconstruct in general out- line the characteristics of the original language from which all these are derived. Since Iranian in view of its very close relationship with Indo- Aryan is of the first importance for the study of Indo-Aryan philology, a short account of its distribution and documentation is desirable. The migration of the Indo-Aryans to India brought about, or perhaps was the final stage of, the separation of the primitive Aryan community into two distinct divisions which henceforth evolved separately in linguistic as in other respects. The Iranians left behind in the region of the Oxus valley 1 proceeded to expand rapidly in various directions, occupying not only the Iranian plateau which remained their centre of gravity, but also large tracts of Central Asia, extending on the one hand to the confines of China and on the other hand to the plains of South Russia. From an early period Iranian showed a much stronger tendency to differentiation into separate dialects which soon became independent languages than was the case with Indo-Aryan, which for geographical and other reasons maintained a comparative unity over most of North India for a very long period. For the old period Iranian is represented by documents in Avestan and Old Persian, and it is these texts which are of 1 A recollection of Chorasmia as their original home is preserved in the traditions of the ancient Iranians. 4 SANSKRIT AND INDO-EUROPEAN prime importance for comparison with Vedic Sanskrit. A vesta is the name given to the ancient collection of sacred writings preserved by the adherents of the Zoroastrian religion, and it is after this that the language is named. It is an eastern Iranian dialect, the exact location of which has not been precisely determined. The oldest section of the A vesta, the Gd&ds are attributed to Zoroaster himself. Concerning his date there has been much dispute, and it seems that the traditional date of the Zoroastrians themselves, which places him around 600 b.c. can hardly be correct. The language of the Gddds is no less ancient than that of the Rgveda, and for this and other reasons the composition of the two texts must belong roughly to the same period. Old Persian, a south-western dialect, and one showing tendencies to modernisation in comparison with the earliest Avestan, is preserved in inscriptions of the Achaemenian kings in a special cuneiform alphabet invented for the purpose. The relations between this ancient Iranian and the language of the Veda are so close that it is not possible satisfactorily to study one without the other. Grammatically the differences are very small ; the chief differentiation in the earliest period lies in certain characteristic and well-defined phonetic changes which have affected Iranian on the one hand and lndo-Aryan on the other. It is quite possible to find verses in the oldest portion of the Avesta l which simply by phonetic substitutions according to established laws can be turned into intelligible Sanskrit. The greater part of the vocabulary is held in common and a large list could be provided of words shared between the two which are absent from the rest of Indo-European. This resemblance is particularly striking in the field of culture and religion, and may be illustrated by a few examples : Skt. hiranya Av. zaranya- 4 gold Skt. send, Av. haend, O. Pers. haind 'army', Skt. rsU-, Av., 0 . Pers. arsti- 'spear', Skt. ksatrd Av. xsadra - ‘ sovereignty Skt. dsura-, Av. ahum -

• lord Skt. yajhi- } Av. yasna- ' sacrifice Skt. hotar Av.

zaotar- ‘ sacrificing priest Skt. soma-, Av. haoma- i the sacred drink Soma Skt. dtharvan - ' a class of priest Av. adaurvan-, dOravan- 1 fire-priest Skt. aryamdn-, Av. airyaman- ‘ member of a religious sodality In the same way we find the names of divinities and mythological personages held in common, e.g,, vSkt. Mitrd Av. mi&ra Skt. Yatna -, son of Vivdsvant Av. Yirna, son of V ivahvant- , Skt. A pirn N dp at, Av. apqm napdt SANSKRIT AND INDO-EUROPEAN 5 1 Grandson of the waters 1 (a divinity), etc. In this field, how- ever, movements of religious reform with which the name of Zarathustra is associated have tended to alter the picture from the Iranian side. For instance Av. daeva-, O. Pers. daiva corresponding to Sanskrit devd- ' god ' has acquired the meaning of f devil In the same way some Vedic divinities appear in the Avestu as evil spirits : Skt. Indra Ndsatya- : Av. Indra, Ndyhaidya-. The material for Old Iranian is somewhat restricted both as to quantity and as regards the number of dialects represented. For the Middje Iranian period, thanks mainly to discoveries of the present century, the documentation is much wider. We now have, in addition to Middle Persian proper (Pahlavi) ex- tensive documents in two important East Iranian languages which are not represented in the early period, namely Sogdian and Saka (mainly in the dialect of Khotan, but with a few texts in a neighbouring dialect). The publication and interpreta- tion of the material in these languages has progressed rapidly and successfully, but the results are not yet in the main avail- able in a form easily accessible to students of general Indo-Aryan or Indo-European philology. Eventually a considerable contri- bution should be available from this source, because, although they cannot compete in antiquity with the Avestan and Old Persian texts, they constitute independent branches of Iranian which were not previously known and therefore have preserved things which were lost elsewhere from an early period . 1 In the mediaeval period the domain of Iranian became very much restricted, mainly on account of Turkish expansion. Over large tracts of Central Asia Iranian has long since died out. It has remained principally in Iran or Persia proper, where modern Persian can look back to a continuous literary tradition of over a thousand years. On the periphery of this area, par- ticularly on the Indo-Iranian frontier, there are still many minor languages surviving in small areas, and one which is still important, namely Pasto, the official language of Afghanistan, At the other side of the territory in the Northern Caucasus Ossetic still survives from one of the numerous Iranian invas- ions of South Russia. 1 For instance the IE word lor ‘ (young) pig Lat. porous, Lith. parkas, was not previously known in Indo-Iranian, but has now turned up in Khotanese : pa'sa 6 SANSKRIT AND INDO-EUROPEAN §2. Primitive Indo-European The Indo-Iranian languages which have been briefly out- lined form together one branch of the Indo-European family. The discovery of the historical relationship of 'the members of the Indo-European family was a direct result of the discovery of the Sanskrit language and literature by European scholars towards the close of the eighteenth century. The similarity of the Sanskrit language, both in grammar and vocabulary to the classical languages of Europe is so far-reaching that scholars familiar with Latin and Greek could not fail to be struck by the resemblance. Since up to this time there had existed no clear idea of the real nature of the development of languages and of their relations with each other, the explanation of this unex- pected but quite undeniable affinity could not be provided without a completely new and scientific approach to the study of language. In his famous address to the Royal Asiatic Society of Bengal in 1786, Sir William Jones indicated in broad outlines the significance of the new discovery : ' The Sanscrit language, whatever be its antiquity, is of wonderful structure ; more perfect than the Greek, more copious than the Latin, and more exquisitely refined than either ; yet bearing to both of them a stronger affinity both in the roots of verbs and the forms of grammar, than could possibly have been produced by accident ; so strong indeed that no philologer could examine them at all without believing them to have sprung from some common source, which perhaps no longer exists. There is a similar reason though not quite so forcible, for supposing that both the Gothick and the Celtick, though blended with a different idiom, had the same origin with the Sanscrit ; and the old Persian might be added to the same family.' The truth of these remarks has been adequately demon- strated by the subsequent development of the science of Com- parative Philology, which dates from this time. During the past century and a half the languages of the Indo-European family have been the subject of intensive scientific study. The main features of the parent language have been reliably recon- structed, and the historical and prehistorical development of the several branches has been worked out in detail. The methods and principles employed have been subject to proSANSKRIT AND INDO-EUROPEAN 7 gressive improvement and refinement, and new discoveries have continued and still continue to produce a wider and deeper understanding of the subject. The methods first evolved in the study of the Indo-European languages have further been successfully employed in the study of independent linguistic families (Semitic, Finno-Ugrian, Bantu, etc.). The whole science of linguistics has come into existence as a result of the stimulus provided by the discovery of Sanskrit. The Indo-European languages are divided into ten major branches, in addition to which there are known to have been other branches which have died out without leaving adequate record. The ten major branches are as follows : L Aryan or Indo-Iranian, summarised above. II. Baltic (Lithuanian, Lettish and the extinct Old Prussian) and Slavonic (Old Church Slavonic or Old Bulgarian, Russian, Polish, Czech, Serbo-Croatian, Bulgarian, etc.). These two groups are very closely related to each crther, though not as closely as Indo-Aryan and Iranian. There are some ancient divergencies between them which make it impossible to recon- struct a primitive Balto-Slavonic language, intermediate be- tween Indo-European and the existing languages in the same way as Indo-Iranian can be reconstructed. Nevertheless in view of their many close resemblances it is convenient to group them together under a common name, Balto-Slavonic. The earliest recorded Slavonic is the Old Bulgarian of the ninth century ; Lithuanian is known only from the sixteenth century. III. Armenian, known from the fifth century a.d. IV. Albanian, known only from modern times. These four groups are collectively known as the safcm-languages for reasons which will be explained below. Opposed to them are the ctfM^m-languages, which are as follows : V. Greek, with numerous dialects. The literature begins with the Homeric poems, c. 800 b.c., but during the last twenty- five years the decipherment of documents in the linear B script and Mycenean dialect has pushed back the history of the language by 500 years. VI. Latin, which has developed into the various Romance Languages (French, Italian, Spanish, Portuguese, Rumanian, etc.). It is known in literature from c. 200 b.c., and there are scanty inscriptional remains from an earlier date. VII. Celtic, consisting of Continental Celtic or Gallic, 8 SANSKRIT AND INDO-EUROPEAN which is extinct, and Insular Celtic which is divided into Irish (Gaelic) and Brittanic (Welsh, Cornish, Breton). Literary records of Celtic begin with the Old Irish glosses of the eighth century. VIII. Germanic, which may be divided into East Germanic or Gothic (extinct), Nordic or Scandinavian, and West Germanic, to which belong the English and German languages. The earliest literary monument of Germanic is the Gothic translation of the Bible by Uifila (a.d. fourth cent.). The two major members of the family which remain to be mentioned are known from discoveries made in the present century. They are : IX. The so-called ‘ Tocharian * preserved in Buddhist manu- scripts discovered in Chinese Turkestan, dating from the sixth to the tenth centuries a.d. It is divided into two dialects which are for convenience termed A and B. X. Hittite, which is preserved in cuneiform tablets recovered from Boghaz-koi in Anatolia, the site of the capital of the ancient Hittite kingdom. The time covered by these records is the period from c. 1700 to c. 1200 B.c;., the bulk of them being dated towards the end of this period. It is the oldest recorded IE language, and at the same time in many ways aberrant from the usual type. Its discovery has raised many new and interesting problems. In addition to the major languages listed above, there existed in antiquity a considerable number of other IE languages which have become extinct and are known only from scanty remains in the form of inscriptions, proper names and occasional glosses. To put the Indo-European family into proper perspec- tive the more important of these are enumerated below. In the first place there are certain ancient languages of Asia Minor which together with Hittite form a special group. The cuneiform texts from Boghaz-koi include texts in two such languages, -Luwian and Palaean, which show close relationship with Hittite. The so-called Hittite Hieroglyphic inscriptions which have now been partially deciphered, have revealed a language which is closely related to the Luwian of the cuneiform texts. Later the Lycian language, in which there are inscrip- tions in alphabetic script, has been shown to have relationship with Luwian. Most recently the Lydian language, knowm from inscriptions from Sardis, has been shown, to belong with the above languages in the Anatolian group. SANSKRIT AND INDO-EUROPEAN 9 Fragmentary records of other Indo-European languages have been preserved from the Italian and Balkan peninsulas, and, resulting from a later movement of peoples, and separate from the above Anatolian group, in Western Asia Minor. In Italy Oscan and Umbrian are known from a fair number of inscrip- tions, and together with Latin they are classed as the Italic group. The language of the Siculi in Sicily is also considered to belong with Italic. In the south-east corner of Italy, Messapian is held to be a later introduction from across the Adriatic, possibly related to Illyrian. Venetic in North East Italy is classed as a separate Indo-European language, though showing some signs of relationship with Italic. In the Balkan peninsula the Illyrian language is known only from proper names, and there is an unsettled debate as to whether or not it is the ancestor of Albanian, The information of Thracian, the other main Balkan language in ancient times is equally scanty, and its position remains obscure. In Asia Minor fresh invasions from the Balkans brought to an end the Hittite empire (c. 1200 b.c.) and introduced new Indo-European languages into the area. Of these Phrygian is scantily preserved in inscriptions. The languages of the Indo-European family have become more widely diffused over the world than those of any other linguistic family. They also form the majority of the cultivated languages of mankind. It is not surprising therefore that the question of the original home of Indo-European has been the subject of much speculation. In the early days it was usually held that this lay in Central Asia, and that from there successive waves of emigration had carried the various members of the family to Europe. This was mainly due to the exaggerated importance attached to Sanskrit and to confusion between the primitive Aryans of whom we have spoken with the much earlier Indo-Europeans. It is as we have seen reasonably cer- tain that it was from Central Asia, more specifically the Oxus valley, that the Indians and Iranians set out to occupy their respective domains. But there is not the slightest trace of evidence or probability that the ancestors^of the Germans, Celts, Greeks and other European members of the family were ever near this area. Consequently it is now usually held that the original home lay somewhere in Europe. The main argu- ment for this is the simple but effective one that it is in Europe that the greatest number of Indo-European languages, and the 10 SANSKRIT AND INDO-EUROPEAN greatest diversity of them is to be found, and this from the earliest recorded times. At an ancient period we find enormous stretches of Asia in the occupation of Indo-Iranian, a single member of the family, and as yet little differentiated ; in Europe on the other hand a concentration of many languages occupying comparatively restricted areas, and already markedly different from each other. It follows of necessity that the pres- ence of Indo-European in the Indo-Iranian area is the result of late colonial expansion on a vast scale, while in Europe the existence of such great diversity at the earliest recorded period indicates the presence there of Indo-European from remote antiquity. It is true that the discovery of the two Tocharian dialects in Chinese Turkestan has slightly modified this picture, and it has led some to think again of an Asiatic home* But the addition of one new branch only in Asia is obviously insufficient to turn the balance. Moreover the nature of Tocharian, which has under- gone profound and far-reaching phonetic changes strongly sug- gestive of alien influence, makes it clear that this language has travelled far from its original home. Somewhat similar changes have taken place in Hittite and the allied languages of Asia Minor, and this is held to have been due to the influence of the pre-Indo-European , languages which existed in that area (Proto-Hittite, Churrian, Urartean, etc.). So we may conclude that these languages also have been brought in by invaders, and since in ancient times the distribution of languages in this area was such that the non-Indo-European languages mentioned lay to the East and the Indo-European languages to the West, it becomes clear that the direction of the invasions must have been from the West, that is to say from Europe, across the Hellespont. Further it has been pointed out that the character- istics of this Asianic branch of Indo-European are such as can only be explained by the assumption that it was separated from the main branch of Indo-European at a period very much earlier than the movements which lead to the final break-up of IE linguistic unity. This means that the earliest of all the Indo-European migrations which can be deduced from our evidence, and one that must have antedated the migration of the Indo-Iranians by a very long period of time, already points to the existence in Europe of the Indo-European tongue. Within Europe it is possible to narrow down considerably the SANSKRIT AND INDO-EUROPEAN II territorial limits within which the cradle of the Indo-European languages is to be sought. It is known with reasonable cer- tainty that the Italian and Greek peninsulas were colonised from the North. The occupation of France and the British Isles by Celts from Central Europe occurred at a comparatively late date (c. 500 b.c.). The Iberian peninsula remained predomin- antly non-Indo-European till Roman times, and in modern Basque there still exists a survival of pre-Indo-European speech. The Eastern limit is indicated by the fact that before the two Asiatic migrations (Tocharian and Indo-Iranian) Indo- European must have been bounded to the East by an early form of Finno-Ugrian, and there is some evidence of contact be- tween these two families in the primitive period. There is reason to believe that the original centre of Finno-Ugrian expan- sion lay between the Volga and the Urals and this forms the extreme limit beyond which Indo-European was not to be found in the early stages of its history. This leaves the central portion of Europe extending from the Rhine to Central and Southern Russia, and it is probable that by the time of the Indo-Iranian migrations the larger part of this area had long been occupied by various Indo-European dialects. It is not possible to define the original Indo-European home- land in terms any narrower than these, nor is it desirable to try, since those who have attempted to do so have usually suffered from misconceptions about the nature of ‘ Primitive Indo- European ' and about the time when the earliest divisions began. The evolution of the Indo-European should not be re- garded as being on a par with that of the Romance languages from Latin. In the latter case the various languages are derived from a single unitary language, the language to begin with of one city. But in the case of Indo-European it is certain that there was no such unitary language which can be reached by means of comparison. It would be easy to produce, more or less ad infinitum a list of forms like Skt. nabhi Gk. 6 fjuf>a Aos* ' navel ', which although inherited directly from the primitive IE period, and radically related are irreducible to a single original. In fact detailed comparison makes it clear that the Indo-European that we can reach by this means was already deeply split up into a series of varying dialects. It is from this point of view that the question of the r split- ting ' of Indo-European should be regarded. It has not been 12 SANSKRIT AND INDO-EUROPEAN uncommon to find in works of general history or linguistics a conception that somewhere about the second half of the third millennium b.c. a single undivided Indo-European occupying a comparatively restricted area being taken by a series of migra- tions to the various countries where IE languages are later found, after which migrations the various individual languages were evolved. But it is now becoming clear that by this period the various members of the family must have already begun to assume their historic form. For instance when the Indo- Iranians first set off on their migrations from Europe, very likely about 2000 b.c. as is often suggested, they took with them not Indo-European which they subsequently proceeded to change into Indo-Iranian, but the Indo-Iranian which we can reconstruct, which had already assumed its essential features in the original European homeland. It is clear that once the migrations began over such wide territories all opportunities for unitary development of Indo-Iranian must have ceased, and since, as we have observed there quite undoubtedly was at one time a unitary development of Indo-Iranian, this must have taken place before any migrations began. What applies to Indo-Iranian must apply with equal force to the other members of the family. We have already remarked on the deep divergencies between the various European mem- bers of the family, and this can only be accounted for by pushing back the period of original division to a period much earlier than is usually assumed. If there ever existed a unitary Indo- European which spread from a restricted area, this lies long behind the earliest period which can be reached by any com- parison. ' Primitive Indo-European ’ must be regarded as a continuum of related dialects occupying an extensive territory in Europe (very likely the major part of the area indicated above), dialects which already before the period of the great migrations had begun to assume the character of separate languages. §3. Divisions of Indo-European The question of the early Indo-European dialects has been the subject of considerable study and some useful results have been acquired. It is possible to form a fair idea of their distribu- tion in the period preceding the emergence of the individual SANSKRIT AND INDO-EUROPEAN 13 languages. The most striking and important early dialect dis- tinction is that which separates the sate*»-languages from the em/wra-languages. These two groups are so named from the way they treat IE h in the word for ' hundred ’ (IE *krpt6m ) . The ctfwfton-languages preserve it as such (Lat. centum , Gk. €#caroi/, Ir. cet f Toch. A. kdnt) ; in the satam-languages it is changed to some kind of sibilant (Skt. iatdm , Av. satem, Lith. Siihtas , O. SI. suto). Similar changes occur in the case of IE § and gh. The languages participating in this change are Indo- Iranian, Balto-Slavonic, Armenian, Albanian. Since this fea- ture is so wide-spread, and since it occurs without any variation of the conditions in all the languages concerned, it must be assumed that the change took place in the Indo-European period, before the dispersal of the several languages, and that it affected a group of contiguous dialects within the Indo-Euro- pean area. The unity of these dialects, and of the languages derived from them is further confirmed by the fact that the loss of the labial element in the IE series k w , g w , g m h (e.g. Skt. kd- f who ? Lith. kds as opposed to Gk. tto-6*v, Lat. quo-d t Goth, has) is characteristic of precisely this same group of languages. Before the discovery of Tocharian and Hittite it was common to regard the centum-satem division as a division between Western and Eastern Indo-European, and it was customary to regard the centum - languages as a united group like the satem - languages. This was never altogether satisfactory, since not only ts Greek cut off from the Western IE languages by the intervening safcm-language Albanian, but also because apart from this it displays no special similarities with them, but rather with the satera-languages. The discovery of the new languages, which were unmistakeably centum- languages, made it quite impossible to speak of an East-West division any longer, and also made it clear that there was no unitary centum- group. The centum-la.ngua.ges are alike only in preserving original k, g, gh as occlusives, and it is a commonplace of lin- guistics that common preservations are not necessarily a sign that dialects or languages are closely related. We may there- fore substitute a division of the Indo-European dialects into : I. A central group which can be equated with the satem- languages, and is characterised by the innovations mentioned above. 14 SANSKRIT AND INDO-EUROPEAN II. Four peripheral dialect groups surrounding the central group, namely (i) West Indo-European comprising Italic, Celtic and Germanic ; (2) Greek, which however has special relations with the central group ; (3) Eastern Indo-European which has survived as ' Tocharian ' ; (4) Hittite and other IE languages of Asia Minor which separated earliest from the original IE stock. The historical distribution of the IE languages corresponds on the whole to this, but in the case of Sanskrit migrations at a comparatively late date took it to the extreme East of the Indo- European domain. Before this period its ancestor, primitive Indo-Iranian must have held a fairly central position, being directly in contact with the other dialects of the satem-group, and having to the East of it that form of Indo-European which eventually turned into the dialects A and B of Chinese Turke- stan. Its position can further be determined by the specially close relations which are found to exist between it and Balto- Slavonic. Since the Balts and the Slavs are not likely to have moved far from the positions in which they are to be found in their earliest recorded history, the original location of Indo- Iranian towards the South-East of this area becomes highly probable. The Western group of Indo-European languages consisting of Italic, Celtic and Germanic, is distinguished by certain com- mon features in grammar and vocabulary, which indicate a fairly close mutual connection in prehistoric times. These ties are particularly close in the case of Italic and Celtic, even though they are not sufficient to justify the theory of common Italo-celtic. The connections of Germanic with the other two groups are less close, but they are quite definite. At the same time it has some special affinities with Slavonic, and further with the central group in general (e.g. absence of the medio-passive terminations in -r). There is an almost complete absence of special features com- mon to Indo-Iranian and Western Indo-European. All that has been pointed out so far consists of certain common elements of vocabulary which have been largely eliminated in the rest of Indo-European. These words are in many respects highly inter- esting and important, but they consist entirely of ancient Indo- European words which have been preserved independently by two groups which otherwise have no special connection. Such sz^ SANSKRIT AND INDO-EUROPEAN 15 words are : Lat. credo , Ir. cretim , Skt. srad-dha- * believe 7 ; Lat. rex , Ir. ri 1 king 7 ; Skt, raj-, rAjan-, Ir. rtgain ‘ queen 7 ; Skt. rajni, Ir. rtge ‘ kingdom 7 ; Skt. rajya- ; Lat. ius ‘ justice 7 , iustus, Ir. nisse ‘ just, righteous ’, Skt. yos t A v.yaos' rightness, purity 7 ; Lat. ensis ‘ sword 7 , Skt. asi- ; Lat. res ' property ' ; Skt. rai - ; Ir. bro ' millstone 7 ; Skt. grAvan - (also, differently rr/ * formed Goth, qairnns, etc.) ; Ir. gert ' milk 7 , Skt. ghrtd- ‘ ghee 7 ; Ir. aire (gen, s. airech) ' chief, noble 7 ; Skt. aryd ^<?^ Ary a- ' master, lord, noble, Aryan More dubious is the old equation Lat, flamen, Skt. brahman- ‘ priest Many of these words are connected with religion, law, etc., and the fact that they are preserved in these two branches alone is due to the highly conservative tendencies which characterised the societies concerned. They do not imply any close connection between the original dialects on which the languages are based. Greek shows little sign of close connection with any of the other centum- groups. On the contrary its closest connection appears to be with the s^tew-languages, particularly with Indo- Iranian and Armenian. It is sufficient to glance through a comparative grammar of Sanskrit to see that the correspond- ences between Sanskrit and Greek are much more numerous than those between* Sanskrit and any other language of the family outside Indo-Iranian. This is particularly so in the case of verbal inflection. The fact that the two languages are recorded from such an early period is partly responsible for this state of affairs, but it is by no means entirely so. Some of the common features involved are of late Indo-European origin, and must be regarded as common innovations, and not as cases of the common preservation of ancient forms. For instance the Indo-European languages have no common form of the genitive singular of o-stems. The form -osyo which is common to Greek (-oio, ov), Armenian (-oy) and Indo-Iranian (Skt. - asya , Av. -ahya) has no more claim to antiquity than Italo- Celtic -i or the Hittite form {-as, <os) which appears to be identical with the nominative. In fact the great variations in case suggest that the various forms have developed in the late Indo-European period when the language was already widely divided into dialects. It is therefore important evidence of close prehistoric connection. Likewise the augment is found only in Greek (e^epe), Indo-Iranian (Skt. dbharat) and Armen- ian [eber), with traces of Phyrgian. Since there is no reason to l6 SANSKRIT AND INDO-EUROPEAN believe that it ever existed as a regular component of the verbal inflection in those languages in which it is not recorded, its development in Indo-Iranian, Greek and Armenian must be regarded as a common innovation of the closely related dialects on which they are based. The elimination of the r- endings of the medio- passive in Sanskrit and Greek is a significant common characteristic. Phonetically Sanskrit and Greek show a common treatment of the sonant nasals (IE n t ip) , replacing them by the vowel a. In view of the close connection that exists between them in other respects this is unlikely to be a matter of chance. In other respects, e.g. in the matter of prothetic vowels Greek seems to be closest to Armenian, and there are also some remarkable coincidences of vocabulary between them. The fact that Greek shows more signs of close connection with the sataw-languages Armenian and Indo-Iranian than with any other is in striking contrast to the absence in it of the dis- tinctive sound changes of the satdm group. We must assume that the IE dialect on which Greek is based was originally in the closest contact with the central dialect group, but that this contact was severed at a period preceding the satdm sound changes. The most striking thing about the two Tocharian languages is that they have no special connections whatever with Indo- Iranian, the only other Asiatic family. They are no closer to Indo-Iranian — in some respects they appear more different — than to languages far to the West like Italic and Celtic-. This is in accordance with the fact that the parent dialect of Indo- Iranian was originally a central dialect, and as such would have with a dialect on the Eastern periphery, from which Tocharian is descended, no more in common than with the percursors of Italic and Celtic on the extreme West. Neither have they any special relations with any other of the individual groups of Indo-European. The prevalence of the middle terminations in r in Tocharian does not indicate any close relationship with Italo- Celtic on the one hand or with Hittite on the other, but merely a type of inflection that was characteristic of early Indo- European, but was tending to be reduced or eliminated in the later period in dialects of the central area. Attempts to find other evidence of connection with one group or another have been singularly lacking in results. The two languages have become much altered from the original Indo-European. The SANSKRIT AND INDO-EUROPEAN *J old system of nominal inflection has to a large extent broken down, and the percentage of words in the vocabulary for which it is possible to find satisfactory etymologies is comparatively small. At the same time some features of the two languages have an ancient aspect which suggests that they are derived from a comparatively early form of Indo-European. This would imply a comparatively early migration in the case of Tocharian, and such an assumption accounts best for the great difference between Tocharian and Indo-Iranian. We must assume that an Eastern Indo-European dialect group had for centuries existed in isolation before the comparatively late migration which took Indo-Iranian to Asia from the central Indo-European area. The separation of Hittite and the languages allied to it from the main body of Indo-European must have taken place earliest of all. This is the only way to explain the great differ- ences which exist between it and the type of Indo-European that has been reconstructed from the previously known mem- bers of the family. The most striking feature of Hittite is the preservation of h, which has elsewhere disappeared. In addition to this the language deviates from the usual type in many other respects. In the formation of nouns the percentage of consonantal stems, and in particular the old neuter types in l and r alternating with n, is much greater than in the standard types of Indo-European, The feminine gender is undeveloped. The inflection of nouns is much simpler than in the type of Indo-European represented by Sanskrit, and there is no reason to believe that this is due to losses on the part of Hittite. Above all, the conjugation of the verb differs widely from the system reconstructed largely by the comparison of Sanskrit and Greek, which at one time passed for primitive Indo-European. Con- sideration of these facts has led some scholars, notably E. Stur- tevant, to separate Hittite from the Indo-European family proper, and to postulate an earlier Indo-Hittite from which Hittite on the one hand and Indo-European on the other are separately descended. The majority of opinion is against this extreme view and it seems more satisfactory to speak of Early and Late Indo-European, rather than of Indo-Hittite and Indo- European. It has already been pointed out that the dialect divisions of Indo-European go back to a period long antedating the migration of Indo-Iranian. Even though the separation of l8 SANSKRIT AND INDO-EUROPEAN Hittite must have been very early indeed, it need not have preceded the beginning of these dialectal divergences of Indo- European. Certainly there was no united Indo-European in the late period, which the Indo-Hittite theory demands. It is true that much of the evolution which has taken place in Indo- European outside Hittite, and which must be placed in the period following the separation of Hittite, is evolution common to all the branches (e.g. the development of the feminine) , but this is easily understandable as long as the various dialects remained in contiguity. The important difference now is that instead of thinking simply in the terms of Primitive Indo- European we may now distinguish Early Indo-European of the time previous to the separation of Hittite, and Late Indo- European characterised by certain developments which can be determined, in which different dialects evolving in common were gradually beginning to assume the character of different lan- guages. §4. Indo-Iranian and Balto-Slavonic The satem-languages, apart from Indo-Iranian are only known from times much more recent than most of the centum- lan- guages. Further there is the possibility that some ancient members of this group, notably in the Balkan and Danubian regions, have disappeared without record. It is therefore not possible to form a precise idea of the position of Indo-Iranian within the satpm group as a whole at an early period. The only thing that emerges clearly is that there did at one time exist a special relationship between early Indo-Iranian and those dialects of Indo-European which developed eventually into the Baltic and Slavonic languages. Since this is important for the location of the early home of Indo-Iranian, the evidence may be given in some detail. Phonetically the most noteworthy common feature is the change of s to / {> Slav, ch) after k } r , i and u in Indo-Iranian and Slavonic, and after r also in Lithuanian . 1 This is unlikely to be a matter of chance, since the conditions under which the change takes place are so closely parallel. The conclusion which must be drawn is that at one time the two branches were in close geographical proximity, and that this innovation affecting IE $ 1 For examples see p. 79. SANSKRIT AND INDO-EUROPEAN 19 established itself over a limited area comprising Slavonic and Indo-Iranian, but excluding the rest of Indo-European. The fact that the change appears only to a very small extent in Baltic demonstrates that the Baltic group was to a certain extent autonomous of Slavonic even at this early date. Another change which has occurred in both groups is that of k to £ before the vowels £, t. 1 This however seems to be a case of parallel independent development. In Old Slavonic the paradig- matic alternation occasioned by this change, and by the second Slavonic palatalisation, remains in full force (e.g. Nom. S. vluku, Voc. vlu£e f Loc. vluce). Such alternation has been elim- inated in Sanskrit even at the earliest period, and it is unlikely that it could have maintained itself in Slavonic over the very long period that it would be necessary to assume if the change in Slavonic had been so ancient. In grammar a fair number of special features common to both groups can be enumerated, though there are also some notable divergences. The most important of these latter is the existence in Balto-Slavonic in common with Germanic of an element -m- which appears regularly in place of the -bh- which is familiar from Sanskrit and other IE languages (e.g. Dat, abl. PI, Lith. vilkdms, SI. vlukomu, Goth waif am : Skt. vfkebhyas). This is an ancient Indo-European divergence cutting across the usual dialect divisions. Another idiosyncrasy of Balto-Slavonic is the use of the old ablative to form the genitive singular of 0-stems : Lith. vilko, O. SI. vluka. In spite of these divergences there are many special gram- matical features uniting the two groups. The more important of these may be briefly enumerated : A. Nominal Inflection : (1) Nominative without r of r- stems, Skt. maid 1 mother svdsd ‘ sister ' ; O, SI. mati , Lith, moil, sesuo. (2) The locative plural in -su (as opposed to -at in Greek) is found only in these two groups : Skt. vfkesu, O. SI. vlucechu. (3) The Dual inflection is closely similar, containing a good deal that is not found elsewhere, e.g. Skt. Nom. D. bale, yug £ , namam, mdnasi, aksl , sunU ; O. SI. zene, ize, imeni , tehsi, oci (Lith. ak i), syny (Lith. sUnu), Gen. D. Skt. tayos , dvayos : O. SI. toju , dvoju . (4) A similar development in the singular stem of feminine nouns in -d : e. g. Instr. Skt. tdyd , senayd : O. SI. tojg, rQkojg, Loc. Skt. sdndydm , Av. haenaya : Lith. rankoje . (5) 1 For examples see p. 76. 20 SANSKRIT AND INDO-EUROPEAN Close similarity in the declension of i- and u- stems, as illustrated by equations like Dat. S. Skt. sundve : 0, SI. synovi. B. Pronouns and A dverbs : (i) Common characteristics in the form of the personal pronouns, e.g. Nom. S. in -om t Skt. ahdnt , O. SI. azu , nasalised accusative, Skt. mini, 0. SI. Gen. S. Av. mana, 0. SI, mene (as opposed to Skt. mama). (2) The extended stem of the demonstrative pronoun, etc., in certain cases, e.g. Dat. S. Masc. Skt. tasmai f O. Pruss. kasntu stesmu, O. SI. tomu, Fem. Skt. tdsyai , O. Pruss, stessiei . (3) Preference for the interrogative stem k w o- as opposed to the stem k w i-, Skt, ka- f Lith. kds. (4) The possession of certain common pro- nominal stems, e.g. Av, ava~ f O. SI. ov& t Skt. Av. ana Lith. ands, O. SI. onu . (5) Various adverbs, Skt. kuha } Av. kuda

• where 0. SI. ktide, Skt. kadd ' when ? iada 1 then ', Lith.

kadd y tadd , Skt. nd 1 like Lith. net , Skt. bahis ‘ outside O. SL bezu ‘ without J f Skt, vind, O. SI. vune ‘ outside ’, O, Pers, (1 avahya ) radi 1 on account of {that}', O. SI. (logo) radi. C. The Verb. In the conjugation of the verb features special to Indo-Iranian and Balto-Slavonic are not remarkably com- mon. This may partly be due to the fact that Slavonic (and to a greater extent, Baltic) is only recorded late, and in the verbal inflection is less conservative than it is in the nominal inflection (e.g. loss of the perfect and the middle). Points of note are (1) similarities in the s- aorist, e.g, vrddhi of root (SI. vesti : vesu } Skt. vdhati avdksam) and termination -om of isg. (as opposed to Gk. a) ; (2) The future in syo- is found with cer- tainty only in Indo- Aryan and Lithuanian : Skt. ddsyami ' I will give 1 Lith, duosiu ; (3) The causative is well developed in both groups, and many identical forms can be quoted, e.g. Skt. bodhdyati ‘ he awakens O. SI. buzdg, buditi. In the sphere of vocabulary Indo-Iranian shares with Baltic and Slavonic a considerable number of words which are not found in the other Indo-European languages. These corres- pondences are much more numerous than those which can be discovered between Indo-Iranian and any other member of the family, and they supply important evidence for the early con- nection of the two families. There is for instance no common Indo-European word for ' goat Sanskrit ajd- is connected with Lith. ozys , but parallels are absent in other IE languages. Greek and Armenian which go together in this case, as frequently have a similar word (a?f, arc) , but one that cannot be united SANSKRIT AND INDO-EUROPEAN 21 with it according to the laws of IE phonology. Another word is peculiar to the Western IE languages (Lat. haedus, Engl, goat , etc.). The distribution of these words corresponds roughly to the dialectal division sketched above, and illustrates the impor- tance of vocabulary in the study of this question. The deriva- tive ajina- ' skin ' corresponds to O. Sh azmu , jaztno, and in both languages an original meaning ‘ goat’s skin ' has been widened to the meaning 1 skin ’ in general. There is also a class of words in which the root is common to many IE languages, but the particular suffix found in Indo-Iranian is found else- where only in 'Slavonic and Baltic. Examples of these are : Skt. phina - ‘ foam O. Pr. spoayno , Lith. spdine, as opposed to the Western IE words with w-suffix (Lat. spurna, Engl, foam), Skt. dak§ina - ‘ right (-hand) 1 corresponds exactly to 0. SI. desinu, Lith. desine , whereas various different suffixes appear in other languages (Gk. Lat. dexter , Goth. taihswa). Similarly Skt. grivd ‘ neck ' and O. SI. griva ‘ mane ' correspond exactly in formation, but can be compared only as far as the root is concerned with Gk. hiprj ‘ neck’ {J g w er). The n- suffix of Skt. majjdn- reappears only in Slavonic (O. SI. mozdanu) and Baltic (O. Pruss. muzgeno). A form corresponding precisely to Skt. misrd- ‘ mixed ' appears only in Lith. mtsras . Another feature which deserves consideration is the existence of special meanings common to the two groups. The meaning ‘ wake ’ is common to Skt. budh- and the related Balto-Slavonic words, but is not found in other languages. The meaning ' write ’ of Iranian (ni-)pi$- recurs in O. SI. pisati . Of the remaining words which are peculiar to Indo-Iranian and Balto-Slavonic the following are the most important: Av. spdnta - ‘ holy O. SI. svqtu, Lith. fventas ; Skt. savyd- ' left Av. haoya- } O. SI. sup ; Skt. barhts ‘ bed of Kusa grass Av. bar9zis ‘ cushion, pillow O. SI. blazina ' cushion ’ ; Skt. krsnd- ‘black’, 0. SI. crunu, O. Pruss. kirsnan ; Skt. bhdra- 1 fight, battle cf, O. SI. borjo * fight ’ ; Skt. ostha- ' lip O. SI. usta, O. Pruss. austin 1 mouth J ; Skt. avatd- ‘ spring, well Lett, aviiots ; Av. var?sa - ‘ hair V O. SI. vlasu, Russ. volos; Skt. gin- ‘mountain’, Av. gatri-, 0. SI. gora ‘id’, Lith. girici; gire 1 forest ’ ; Skt. tusnim ‘ silently Av. tusni- ‘ silent O. Pruss. tusnan ; Skt. tucchyd- 1 empty Khotanese ttussaa-, O. SI. lusti, Lith. iiiscias ‘ id ’ ; Skt. dddhi (gen. dadhnas ) ‘ curds O. Pruss. dadan ; Skt. pdyas ‘ milk Av. 22 SANSKRIT AND INDO-EUROPEAN paema, Lith. penas (with varying suffixes) ; Av. xsvid - 1 milk cf. Lith. sviestas 1 butter 1 ; Skt. dngdra - 1 coal ’, O. SI. gglt, Russ, ugol* ; Skt. bradhna - ' yellowish, light-coloured 0. SI. bronii ' white ' ; Skt. drbha - f small, child cf. Russ, rebenok ‘ child 1 ; Skt. vratd- 4 vow Av. urvata O. SI. rota 4 oath 1 ; Skt. dndd- 4 egg, testicle ’ (Kalasa ondrak 4 egg '), O. SI. j$dro ' testicle * ; Skt. pdmsu- 1 dust ' Av. pasnu -, 0. SI. pesuku 'sand 1 ; Skt. dhana ‘com, grain', Pers. ttana, Lith. duona ‘bread'; Skt. syamd syavrf- ‘dark-coloured’, Lith. semas 4 grey Skt. sdndra- 4 thick, viscid cf. O.SI. krivinyje, sjadry krovnyja ' thickened, congealed blood ' ; Pers. raz 4 vine , vine-tendril O. SI. loza ; Av. fsarama-, Pers. sarm ' shame O. SI. sramu ; Skt. sraraa- 4 lame O. SI. chromu ; Av. hcima-

• the same Pali sdmam adv. 4 self, of oneself O. SI. samu

‘ self ’ ; Skt. vispdti - 4 head of settlement or clan Av. vTspaiti -, Lith. viespats 4 lord 1 ; Av. sardta- 4 cold ', Lith. iia/tas ; Skt. 4 vegetables, greens Lith, ‘ green fodder ’ ; Skt. saphara - 4 Cyprinus sophore Lith. sdpalas ‘ Cyprinus dobula ’ ; Skt. iakund- 4 (large) bird 0. SI. sokolu 4 falcon ? ; Skt. sctpu- 4 drift wood Lith, sapai ; Skt. bhanga- 4 wave Lith. bangd. Among verbs which are common to the two groups we may mention Skt. hdvate 4 calls Av. zavaiti , O. SI. zovetu ; Skt. svit- * to be bright, white Lith. sviteti, 0. SI. sviteti ; Skt. bhi-, bhayate 4 fear O. SI. bojg sg, Lith. bijaus ; Skt. pru$-, prusnati 4 sprinkle 0. SI. prysngti ; Skt, dham-, dkmd- 4 to blow O. SI. dump, dgti , Lith. dumiu, dumti ; Skt. bryih-, brmhate 4 (elephant) trumpets Lith. brenzgn, branzgu 4 to sound, make a noise ’ ; Skt. muc- 4 to release, Lith. mitnkit , miikti 4 to get loose ’ ; Skt, gf-, grndti 4 praises O. Pruss. girtwei 4 to praise Lith. giriu, girti. The list of common words and other features which are special to the two groups is clearly impressive, and the whole of the material must be referred to the period of Primitive Indo-lraniari. When on the contrary we look for signs of special contact between Iranian itself and Slavonic (or Baltic) we find that there are practically none. It is true that some of the words that are listed above are found only in Iranian and not in Sanskrit, but it is equally possible to point out others in which the reverse is the case. Furthermore if we take such a word, e.g. Av. spjnta-, 0. SI. svgtu, Lith. sventas, it is immedi- ately clear that the form of the Baltic and Slavonic words is SANSKRIT AND INDO-EUROPEAN 23 such that they cannot be derived either from the Primitive Iranian form (*svanta-) or the Primitive Indo-Iranian (*£vanta-) t but that all the words must be referred to an earlier satsm form (*svcnto~). Attempts to find examples of Iranian loanwords in Slavonic have been singularly unsuccessful. There is a Russian word sobdka ' dog ' which is plausibly derived from Median onaKa (Herod.) but the word is not pan-Slavonic, and it remains quite obscure by what means the word has reached Russian. In the case of Russ, topor ' axe J _, Pers. tabar, w T e are dealing with a migratory word of uncertain origin. Iranian origin has been assumed for SI. suto ' ioo ' because the form of the word does not agree with Slavonic phonology, but neither is it the form we would expect to be derived from Iranian satzm (which should give so*-). There is a remarkable coincidence between the Slavonic word for ‘ god * (O. SI, bogu) and O. Pers. baga but in view of the complete absence of other loanwords it is better to see in these w T ords a case of common inheritance. This absence of Iranian influence on Slavonic is surprising in view of the repeated incursions of Scythian tribes into Europe, and the prolonged occupation by them of extensive territories reaching to the Danube. Clearly at this later period the Slavs must have remained almost completely uninfluenced politically and culturally by the Iranians. On the other hand at a much earlier period (c. 2000 b.c.) before the primitive Aryans left their European homeland, Indo-Iranian and the prototypes of Baltic and Slavonic must have existed as close neighbours for a considerable period of time. Practically all the contacts which can be found between the two groups are to be referred to this period and this period alone. §5. Indo-Iranian and Finno-ugrian During the same period there is conclusive evidence of con- tact between Indo-Iranian and Finno-ugrian, a neighbouring family of non-Indo-European languages. This latter family consists of three European languages which have attained the status of literary languages, Finnish, Esthonian and Hungarian, and a number of now minor languages which are spoken by a small number : Lapp, Mordwin, Ceremis, Zyryan, Votyak, Vogul, Ostyak. Of these Vogul and Ostyak are now found to B 24 SANSKRIT AND INDO-EUROPEAN the East of the Urals, but are considered to have moved there from the West. These two, with Hungarian form the Ugrian sub-group, and are distinguished from the rest by certain com- mon features. The Hungarians moved from the region of the Volga to the territory they now occupy in the ninth century. In Siberia there are several Samoyede languages which as a group are related to Finno-Ugrian. The two families are classed together as the Uralian languages. Even before the Indo-Iranian period there is evidence of contact between Indo-European and Finno-Ugrian. Certain remarkable coincidences (e.g. Lat. sal * salt Finn, suola ; Skt. mddhu 1 honey ', Gk. ^ U 9 v : Finn, mete- ; Skt. ndman-, Gk, ovofia 1 name ' : Finn, nime-, Goth, wato * water etc. : Fi. vete~) have long since attracted attention, but there is lack of agreement as to how exactly they are to be interpreted. One theory is that the two families are ultimately related, but the available evidence is not sufficient to establish this with any certainty. On the whole it seems more probable that the coincidences, insofar as they are not due to chance, are the result of mutual contact and influence in the early prehistoric period. 1 Evidence is both more abundant and easier to interpret when it comes to early Indo-Iranian contacts with Finno-Ugrian, Here it is possible to point out a considerable number of words in Finno-Ugrian which can be shown to have been borrowed from Indo-Iranian at this stage. The most important of the Finno-Ugrian words which have been ascribed to Indo-Iranian are as follows : (i) Finn, sata * ioo Lapp, cuotte , Mordv. sado, Cer. siiSo, Zyry. so, Voty. &u, Vog. sat , $ai t Osty. sot, sdt, Hung, szdz : Skt. satdm , Av. satrm. (2) Mordv. azoro , azor * lord Voty. uz’ir, Zyry. ozir * rich ' ; Vog. oter, dter ' hero ' : Skt. dsnra, ' lord ', Av. ahura- ‘ id (3) Finn, vasara ' hammer Lapp. v&ler t Mordv. vizir , uzer : Skt. vdjra - ' Indra's weapon Av. vazra- ' club, mace (4) Finn, porsas, Zyry. pori , porys , 1 Borrowings are likely to have occurred in both directions, and usually it is difficult to decide which family has been the borrower. As an example of a probable loan from Finno-Ugrian we may quote Engl, whale, O.N. hvalr , O. Pruss. kalis: Av. kara- ‘ mythical fish living in the Rar)ha { -Volga) : Finn, kola ' fish ' etc. The restriction of the meaning indicates that the IE languages are the borrowers, and it is likely that Iranian and the northern IE languages have done so separately. SANSKRIT AND INDO-EUROPEAN 25 Voty. pars , paris ‘ pig ' was ascribed to an Aryan *parsa- ( — Lat. porctts) and this is now attested by Khotanese pd'sa -. (5) Finn, or as 1 (castrated) boar Mordv. ures 1 id ' : Skt. vardha -, Av. vardza- ‘ boar (6) Finn. Mordv. aiar, Cer. vodar 4 udder 7 ; Skt. Udhar " ' id * ; (7) Finn, ora , Mordv. uro, Hung, dr ' awl* : Skt. ard ‘ id ’ ( — OHG ala, etc.); (8) Hung, ostor ‘ whip Vog. osier , Cer. wostyr : Skt. dstrd, Av. as/ra ' whip r (J~aj- ‘ to drive ’) ; (9) Hung, arany ' gold Vog. sureri, saren, Mordv. sirne, Zyry. Voty. zarm : Skt. hiranya Av. zaranya- ; (10) Finn. ara> 4 value, price Hung. a>, etc. : Skt. arghd *, Osset, ary ' id ' (Lith. alga, etc.) ; (11) Finn, sisar 1 sister Mordv. sazor, Cer. iuzar : Skt. svdsar-. Av. x Y aykar - ; (12) Hung, 4 beer Voty. sur, Vog. sor t Osty. sar : Skt. swra 4 strong drink Av. hurd ; (13) Finn, sarvi ' horn Mordv. iuro, Cer. sur, Lapp coarwe , Hung, szarv : Av. srvd ‘horn' ( = Gk. xepas, etc.); (14) Vog. suorp , ior/j * elk ' : Skt. sarabhd- 4 a kind of deer ' (from the root of the last) ; (15) Mordv. sed f 4 bridge ' : Skt. sdtu-, Av. haetu - ; (16) Mordv. v^rgas ' wolf Zyry, vtirkas : Skt. vfka Av. vdhrka - ; (17) Zyry, Voty, /wrm 4 grass ' : Skt. ; (18) Zyry. i/orA 4 kidney J : Skt. vrkkd-, Av. vzrthka- 4 id 1 ; (19) Vog. /as ‘ stranger ' ; Skt. ddsd- 4 non-Aryan, slave ' ; (20) Hung. vdszon ' linen ' : Skt. vdsana- 4 garment, cloth (21} Fi. mehi- Idinen 1 bee Mordv. mekS, Cer. Zyry. Voty. mus , Hung. mih : Skt. maks-, mdksd , mdksikd 4 bee, fly Av. maxst 4 fly ' ; (22) Fi. siika-nen ‘ beard of grain, etc. 1 , Mordv. iwva, Cer. iw, Zyry, s« : Skt. swAa- 4 id 7 ; (23) Mordv, sava, seja 4 goat 7 : Skt. chiga The detailed problems raised by these and other comparisons are not without complications, but certain general conclusions emerge clearly. Most important of all is the fact that, taking the words as a whole, the primitive forms which have to be assumed after a comparison of the Finno-Ugrian forms, are identical with those which have been reconstructed for primi- tive Indo-Iranian, and are free of any of the later sound changes which are characteristic of Iranian on the one hand and Indo-Aryan on the other. This is quite well illustrated by the first word which represents a primitive form sata- (the Indo- Iranian and Sanskrit form) and not sata - (the Iranian form). The characteristic Iranian change of s to h is uniformly absent (3 Mordv, azoro , n Mordv, sazor , 15 Mordv. sed etc.). Like26 SANSKRIT AND INDO-EUROPEAN wise characteristic Indo-Aryan changes such as of zh, jh to h are not to be found {5 Finn, or as, etc.). There is therefore not the slightest doubt that the period when these borrowings took place was the primitive Indo-Iranian period, and it appears probable that the seat of this primitive Indo-Iranian must have been in the region of the middle Volga and the Urals for this contact to have been possible. One point that is noticeable when looking at a few of these words is that the change of Indo-European l, l to Aryan r, r has already taken place (7 Finn, ora, 9, Vog. sareri , etc., 16. Mordv. vargas ). This is a change which is complete in Iranian, but in- complete in Indo-Aryan. That is to say that there were dialects in early Indo-Aryan which preserved IE l (not /), as well as those (the Rigvedic) which agreed with Iranian in this respect. The Finno-Ugrian forms show that this feature must have already been widespread in the earlier, Indo-Aryan period, and the existence of r-forms in the Aryan of the Near East cor- roborates this. It cannot however have been universal, for in that case no /-forms would have been found in Sanskrit at all. It is usually quite clear that these words have been borrowed by Finno-Ugrian from Indo-Iranian and not vice versa. We have equivalents of the words in other IE languages, and before being borrowed into Finno-Ugrian they have undergone the changes characteristic of the Aryan branch. Even where an Indo-Iranian word has no actual equivalent in the other IE languages, its structure and the possibility of deriving it from a known IE root will often show it to be an old inherited word. For instance Skt. vdjra Av. vazra~ f is formed with the well- known suffix -ra (IE -ro), and can be derived from the IE root which appears in Gk. (F)aywfu 4 break, smash There are however a few words in the above list where it is not possible to be certain in this way. Nothing like the Indo-Iranian word for ' bee ' (No. 21) is found in any other IE language, and this makes it more likely on the whole that in this case the Indo- Iranians have adopted a Finno-Ugrian word. Similar considera- tions apply to Nos. 22 (Skt. suka-) and 23 (Skt. chaga -). There may be further examples of Finno-Ugrian words in Indo- Iranian, but the matter has never been investigated from this point of view. As plausible equations we may mention : Skt. kapha- ' phlegm Av. kafa - , Pers. kaf 'foam, scum: ^ SANSKRIT AND INDO-EUROPEAN 27 Hung, hah 1 foam, froth, cream Veps. kobe ' wave, foam Sam. (Kam.) khdwil 1 foam ' ; Skt. kiipa 1 pit, well 1 : Fi, kuoppa 1 pit Lapp guoppe y Cer. kiip , Voty. gop , etc. ; Skt. saldkd c/u*a ' splinter, etc/ : Hung, szilank ' chip, splinter Fi. sale , 3. saleen c ‘ id etc. In cases like these, and others could be added, no IE etymology has been found for the Sanskrit words. Since it is certain that we must assume long contact between the early Indo-Iranians and the neighbouring Finno-ugrians, and since there is no reason why the movement of words should have been entirely one way, we should consider Finno-Ugrian to be a likely source of Aryan words in cases like the above where striking similarity in form and meaning is found. §6. Aryans in the Near East The earliest recorded traces of the Aryan peoples come neither from India nor from Iran, but from the Near East. The presence of Aryans in this area is recorded principally in documents of the Mitanni kingdom of North Mesopotamia during the period 1500-1300 b.c. The list of royal names pre- served in a variety of cuneiform documents has a distinctly Aryan appearance, even though their interpretation is not absolutely certain in all cases. The names of these kings are as follows : Sutarna , Par£asatar> Saussatar, Artaddma, Ariafumara, Tnsratha, Matiwdza, i.e. in Indo-Aryan form Sutarana - (cf. Ved. sutdrman -), Prasdstar - ‘ director, ruler * 5 auksatra- ‘ son of Suksatra- (?)', Rtadhdman - (nom. Rtadhdmd) V.S., Rtasmara- ‘ mindful of right *Tvisratha -, cf. V. tvesdratha- ‘ having rushing chariots *Mativdja- £ victorious through prayer In addition there are found in private documents from this area written in Assyrian a number of proper names of local notables which can be interpreted as Aryan, e.g. Artamna, Bardasva, Biryasura, Purusa, Saimafura, Saiaivaza , i.e, Rtamna - ' mindful of the law V drddhdsva- * son of Vrddhdsva- Vtryasura -

• hero of valour Purusa- ' man, male Ksemasura - ‘ hero of

peace or security *, Sdtavdja- 4 who has won prizes’ (Bvr. cf. v. Vdjasdti -) . This was a period of the expansion of Mitanni influence in the surrounding territories. Consequently we come across rulers of neighbouring principalities having similar Aryan names, and this extends as far as Syria and Palestine. The dearest examples of Aryan names among these are Suvardata : 28 SANSKRIT AND INDO-EUBOPEAN

• svarddta- ‘ given by heaven Satuara : *Satvara- , a stem

bearing the same relation to Skt. satvan - 4 powerful, victorious : a warrior as does Skt. isvard - ' lord 1 to Av. isvan- ; Aria - manya : Rtamanya- 1 thinking on the law Biridasva : Vrdd- haiva - 1 possessing large horses Biryawdza : Fwyauo/a- 4 having the prize of valour Indarota : Indrotd- (RV) 4 helped by Indra Subandu : Subandhu The contemporary 7 Hittite kingdom had close relations both of peace and war with the Mitanni kingdom, and some of the documents from the Hittite capital provide important evidence for the presence of Aryans in the Mitanni country. The most interesting of these documents is a treaty concluded between the Hittite king Suppiluliuma and the Mitanni king Matiwaza (c. 1350 b.c.). Among the divinities sworn by in this document there occur four well known Vedic divine names. They are Indara, Mitra$(il) , Nasatia(nna) t Uruvana$s{il) , which stripped of their non-Aryan terminations are unmistakeably Ved. Indra-, Mitra Nasatya and Varuna-. It is clear that not only Aryan language, but also Aryan religion in a form closely resembling that known from the Rgveda, was current in this region of the Near East during this period. The introduction of the horse to the countries of the Near East which took place during the early part of the second miUenium b.c. seems to have been due mainly to these Aryans. The usefulness of this animal in war soon made it popular in the neighbouring kingdoms, among them the Hittites, Among the archives of the Hittite capital there exists a treatise on the care and training of horses. This is WTitten in the Hittite language, but the author, who had charge of the royal horses was a Mitannian called Kikkuli. Furthermore some of the technical terms used in the work are Aryan words. These are aika vartanna , ter a v° f panza v° t satta v° and navartanna (hapL for nava-v 0 )-— Skt. eka-vartana- 4 one turn (of the course) and likewise for the numbers 3, 5, 7 and 9. The existence of these loanwords in the Hittite text show's clearly the priority of the Aryans in this field. In addition to the above evidence there are a few Aryan traces among the documents of the Kassite dynasty of Babylon (c. 1750-1170 b.c.). The Kassites- themselves were invaders from the East, from the Iranian plateau, and their language, of which something is known, has no connection whatever with SANSKRIT AND INDO-EUROPEAN 29 Aryan or Indo-European. Nevertheless in a list of names of gods with Babylonian equivalents we find a sun god Surias (rendered Samas) which must clearly be identified with Skt. surya~. In addition Manillas the war god (rendered En-urta) has been compared with Skt. marut-, though here some diffi- culty is caused by the fact that the Skt. word always occurs in the plural. Among the kings of this dynasty one has a name which can be interpreted as Aryan : A bir atlas : abhi-ratha- 4 facing chariots (in battle) The existence of Aryans in this area was unsuspected until the discovery of these Aryan names in cuneiform documentsTand a long discussion has proceeded for many years concerning them. Even though the material is small, it can nevertheless be con- cluded that a significant migration of Aryans in this direction had taken place at this early period. Although the chief centre of the Aryan influence, as far as our records go is the Human state of Mitanni, what w r e find there is an Aryan dynasty ruling over a Hurrian population, with no evidence of any sizeable Aryan settlement. The kings of the Mitanni belonging to the dynasty with the Aryan names use Hurrian as their official language, and it is considered unlikely that in this situation they would have retained their original Aryan speech for long. The question that then arises is to decide from what base the conquest of the Mitanni state had proceeded, since such a base, with an Aryan population, is necessary to account for it. A clue to the answer to this problem is probably to be found in the fact that Aryan influence is found not only among the Hurrians, but also among the Kassites, a people originating in the mountainous regions of Western Iran. A settlement of Aryans in North-Western Iran, to the East of the Hurrian country, and to the North of the Kassites, w r ould account for both of these developments. Since the first discovery of Aryan traces in the ancient Near East discussion has proceeded as to whether these Aryans are to be connected with the I ndo- Aryans or the Iranians, whether they are to be identified with the Pro to- Aryans from which those two branches are descended, or whether they are to be regarded as a third branch of Aryans beside the other two. The predominant opinion at present is that they are to be connected specifically with the Indo-Aryans. This conclusion is founded partly on linguistic grounds, and partly on the fact 30 SANSKRIT AND INDO-EUROPEAN that the Aryan gods mentioned in the above mentioned treaty are specifically Vedic gods. The linguistic argument for this conclusion is illustrated by the word aika- ‘ one ' which corre- sponds to Sanskrit eka- wehereas Iranian has aw a- with a different suffix (cf. p. 258). Likewise the name of the sungod SuriyaS corresponds to Sanskrit Suryas , whereas Iranian shows no such form, but only the base (hvar Vedic svdr) form which sUrya - is derived. As far as the treaty gods are concerned, they are prominent in the Rgveda, and are indeed all mentioned together in one hymn (RV. 10. 125. 1), but they cannot be shown to be all Proto- Aryan or Proto- Iranian, and there are in fact some reasons for believing that they were not so. If the Aryans of the ancient Near East are to be connected specifically with the Indo-Aryans, then we must assume that they were two migrations proceeding from the same source, a massive one into North-West India and a smaller one in the direction of North-Western Iran. Since the Vedic Aryans certainly entered India across the mountain barrier separating the subcontinent from North-Eastern Iran, we must conclude that the closely related section of the Aryans who appear in the Near East started from the same source, taking a route which was to be followed early in the next millennium by the Medes and Persians. The chronology of these two movements corresponds very closely. The Aryan invasion of N.W. India can be dated by the end of the Indus civilisation, for which no doubt they were largely responsible, and it may be assumed to have taken place in successive waves over a considerable period of time. In the ancient Near East their presence is established by 1500 b.c. at latest, and the process of migration and settlement must have occupied a considerable period before that. § 7 . The Emergence of Indo-Aryan The pre-history of the Aryan language of India takes us far from the North-West India of the Vedic period both in space and time. Comparison with other languages renders possible a reconstruction of linguistic history which is nowhere directly recorded, and establishes as a fact important migrations and movements of peoples which otherwise would be unknown to history. It has also been possible to say something definite. SANSKRIT AND INDO-EUROPEAN 31 T though naturally within fairly wide limits, about the origin of these movements, and about their chronology. The distribu- tion of the IE languages suggests that their origin is to be sought in Central and Eastern Europe. The special relations of Indo- Iranian with the s«tew-group of languages, and with Balto- Slavonic in particular, together with evidence of contact be- tween it and Finno-Ugrian in the Primitive Indo-Iranian period, point to its original location in Central Russia. From there the movement was eastward and southward, with the result that Central Asia became for a time the home of the Aryans. There is evidence that the division into the two branches, Indo-Aryan and Iranian, had already commenced at this early period. The Indo-Aryan group was the first to move south, first into eastern Iran, and then into India on the one hand, and into western Iran on the other. The second wave of migratiqn was that of the Iranians, who established themselves first in eastern Iran, thereby cutting off the Indo- Aryans to the east from the Proto- Indoaryans to the west. Eater the advance of the Iranians west- wards resulted in the submergence of the latter, but their original presence there is attested by these documents from the Near East. Chronologically there is not much direct information to rely on. The earliest and most important data are those relating to the presence of Aryans in the Near East from 1500 b.c. on- wards. This is an important pointer to the period of the migra- tions, which to judge by historical analogies are likely to have taken place during a limited period of time. The first half of the second millennium b.c. which would seem to be indicated by this evidence as the general period of the migrations is one which agrees comfortably with all the general considerations which can be adduced. The next direct information about the Aryans refers to the Iranians. The presence of Medes and Persians in Iran proper is attested in the Assyrian annals from the ninth century b.c. onwards, and it is unlikely that they had occupied this area in any force for very long before this period. For the Indo-Aryan invasion of India no direct evidence is available. Nevertheless the very great similarity between the Vedic language and the earliest Iranian precludes any long period of separation between the two, and makes it impossible that the age of the Vedic hymns can be pushed back to the third or fourth millennium b.c. The average rough guess which places 32 SANSKRIT AND INDO-EUROPEAN the period of the Indo-Aryan invasions c t 1700-1400 b . c . and the period of the composition of the Rgveda c. 1200-1000 b . c . is not likely to be many centuries out, either one way or the other. There is some linguistic evidence to show that the Indo- Aryan invasion took place in successive phases, and not in one simultaneous movement. There are dialectal differences be- tween the Vedic language of the North West and the later classical language of Madhyadesa. The most striking of these is that the Vedic language turns l into r whereas the classical language, to a large extent, preserves the distinction between r and l. This Vedic feature is characteristic of the whole of Iranian, and furthermore it can be traced in the Aryan of the Near East and in some Aryan words in Finno-Ugrian. Clearly the fact that the more easterly dialects of early Indo-Aryan have avoided this change indicates a comparatively early separation from the main body, in comparison with the Vedic dialect which has undergone this change in common with the rest of Aryan before being introduced into India. Certain features of the Kafiri languages of the North West indicate important dialectal divergencies of ancient Aryan at a time preceding the invasion of India. In some ways these languages stand half way between Indo-Aryan and Iranian. They agree with Indo-Aryan in retaining s which Iranian changes to h, but with Iranian in the treatment of the two palatal series (e.g. zim f snow ' : Skt. himd-, jd- * kill ' : Skt. han). In this respect they form simply an intermediate dialect group, as might be expected from their position between the two main groups. On the other hand in their treatment of the sound which appears in Sanskrit as s they have preserved a form which is more archaic than anything found elsewhere in Indian and Iranian (c in cuna - 1 dog due 1 10 etc.). This can only be satisfactorily explained as the isolated preservation of a very ancient dialectal feature within Indo-Iranian. The same considerations apply to the absence of cerebralisation of s after u in words like dos * yesterday ' and musd ‘ mouse The change of s to s (>Skt. s) under specified conditions is, as we have seen, so ancient as to be shared by both Indo-Aryan and Slavonic, but it seems that some peripheral dialect of Indo- Aryan must have escaped it in connection with u, and it is from this source that the Kafiri forms are derived. The evid- ence would suggest that the Aryan dialect which preserved SANSKRIT AND INDO-EUROPEAN 33 these archaisms was the very first to reach the borders of India, and that later successive waves of Indo-Aryan invaders con- fined it into a narrow space in the mountain valleys of the North-West frontier, where it has survived in isolation to this day. The history of In do- Ary an begins with the first introduction of Aryan speech into India, but between this event and the composition of the first recorded document of Indo-Aryan, the hymns of the Rgveda, a considerable period must have elapsed. This is clear from the fact that in the text of the Rgveda itself, although historical allusions are not uncommon, there is no reference anywhere to the fact of the migration, nor any definite indication that it was still remembered. Linguistic reasons also compel us to assume such a period, since the number of lin- guistic (mainly phonetic) changes that have taken place since the common Indo-Iranian stage is considerable. No doubt the beginnings of dialectal cleavage go back to the Indo-Aryan period, but there is no doubt that the bulk of the characteristic changes of Indo-Aryan and Iranian respectively have taken place after the complete separation of the two groups, that is to say, after the Aryan invasion of India. Some of the more important changes that affected Indo- Aryan during this period may be briefly listed : (i) /A and zh become A ( — Ir. j and z), (2) j and i are confused as j ( = Ir. j and z ) , (3) a single group ks results from the two combinations k- -s and s -K9 ; these are kept apart in Iranian, (4) Aryan voiced groups of the type gzh, bzh are replaced by unvoiced As, ps (Skt, dipsa - : Av. diwza-), (5) Aryan z is elided in all positions (Skt. medhd cf. Av. mazdd ), (6) Elision of Aryan z before d gives rise to cerebral d (nidd-) and this, in conjunction with other combinatory changes {as tan, vit f katana-) is the beginning of a new series of consonants previously foreign to Aryan, as well as to the rest of Indo-European. (7) s ( s ) is elided between two consonants (< dbhakta , s-aor.). (8) All final consonant groups are simplified and only the first remains (Skt. vdk : Av. vdxs ). (9) A tendency begins to weaken the aspirates dh and bh to h (ihd ' here ' : Av. iSa, but Pa. idha has retained the older form). (10) The Aryan diphthongs at, au are turned into the simple vowels e } 5 . This list of changes is impressive enough, and of great import- ance for the future history of Indo-Aryan, and a reasonable 34 SANSKRIT AND INDO-EUROPEAN length of time must be assumed for their completion. At the same time we have the impression that the period of fairly rapid linguistic change preceded the Vedic period. With the estab- lishment of a recognised literary language, and a tradition of education associated with it, this rapid evolution was stopped as far as classical Sanskrit is concerned. The phonetic changes that distinguish Classical Sanskrit from the Vedic language are negligible in comparison with those that took place in the immediate pre- Vedic period. On the other hand the popular language, developing soon into the Prakrits, continued to show this tendency to rapid change. In particular it is interesting to note that the type of change seen in the examples listed above is similar to that of the later Middle Indo-Aryan changes. The assimilation of consonant groups in final position is the be- ginning of a process that affects all consonant groups. The development of voiced aspirates to h, which is general in the case of /A, zh and sporadic elsewhere, is continued in Pali and Prakrit. The cerebral consonants once created become more and more prevalent. From the first Indo-Aryan is affected by certain characteristic tendencies to change which continue to be influential in later periods. These changes which set in from the beginning were rapid, and in the language of the people continued to be rapid. It was only the standardisation of San- skrit at a very early period by organisers of Brahman civilisa- tion, that arrested this development, in the case of the classical language, before it had proceeded too long, and thereby pre- served for us a form of language which in most respects is more archaic and less altered from original Indo-European than any other member of the family. CHAPTER II OUTLINES OF THE HISTORY OF SANSKRIT §i. The Vedic Language and the Classical Language About the pre-history of Indo-Aryan, both in India where it emerged as an independent form of speech, and outside India through the successive stages of Indo-Iranian and Indo-Euro- pean, much can be deduced, and deduced with certainty with the help of comparative philology. But of all these stages of the language no direct record is preserved. The historical period of the language begins— probably, as we have seen, round about the period 1200-1000 b.c,— with the composition and compila- tion of the Rgveda. From this time the literary tradition is con- tinuous and uninterrupted, and the gradual development of Indo- Aryan, through the various stages until the period of the modern languages is reached, can be followed in detail. During this period great changes have taken place, and their operation has been continuous throughout the whole period. By all this change and development Sanskrit has been affected only to a small extent. From the beginning; from the time of the composition of the Vedic hymns and the establishment thereby of a recognised literary language, there was a strong tendency among the Brahmins, the guardians of this literature and of the religious and social system that went with it, to pre- serve the language against change. This applied not only to the preservation of the sacred texts themselves, which have been handed down with scrupulous accuracy by oral tradition, or to the composition of literary works on ancient models, but also to the language of everyday speech among the Brahmins, and in the royal courts w r ith which they were always closely associated* This led to a growing divergence between the language of the educated classes and that of the people, which was subject to a fairly rapid alteration in the direction of Middle Indo-Aryan from an early period. 35 36 OUTLINES OF THE HISTORY OF SANSKRIT At the same time the language of the elite did not remain without change, in spite of all the influence of conscious con- servatism. The classical language as fixed by Panini (fourth cent, b.c.) is a noticeably younger form of language than that found in the Vedic texts, though much less altered from it than the spoken language of the masses, which is known slightly later from the inscriptions of Asoka. We have in fact up to this period two parallel developments of Indo-Aryan occurring side by side in different strata of the community, slow and gradual change in the dominant Brahman community restrained by education and a literary tradition, and beside it a rapid evolu- tion among the mass of the population unhindered by education and tradition. With Panini's work Sanskrit in its external form became finally stabilised and no more change was allowed. From then on the history of Indo-Aryan is the history of Sliddle Indo-Aryan in its various phases (Pali, Prakrit, Apabhramsa) and then of Modern Indo-Aryan. In this evolution Sanskrit took no part, but remained as it was fixed by Panini at a period long antedating the bulk of the classical literature. The differences between Vedic and Classical Sanskrit affect to a very small extent the phonetic structure of the language, and in this respect the contrast between Classical Sanskrit and Early Middle Indo-Aryan is most striking. They are more noticeable in the field of vocabulary and grammar, though here also they are comparatively restricted in scope. Phonetically, apart from some dialectal phenomena such as d , dh for l, lh, and the replacement of r by / in certain words, the differences are mainly concerned with innovations of Sandhi. This is noticeable for instance in the case of -iy- -uv- of the Veda which are normally replaced by y, v. The rule has been applied to the accepted text of the Rgveda, so that for instance what is written tanvds is, from the evidence of the metre, to be pro- nounced tanuvas . Among Vedic peculiarities of final Sandhi we may note that s is only inserted between final n and initial t when it is etymologically justified (sarvdms tan, but varsman tasthau, maghavan tava , ajagmiran U), and that before vowels the terminations of the acc. pL -an, in, tin appear as -dm, itfir and ufhr (sargatk iva, paridhitfir ati ). In such cases later San- skrit has regularised the Sandhi by the analogical extension of a form that was originally justified only in a certain context (e.g, -an, -in, -tin was the regular phonetic development before OUTLINES OF THE HISTORY OF SANSKRIT 37 voiced consonants and in final position, and from there its use was extended to cases where a vowel followed). Apart from such comparatively minor changes Classical Sanskrit pre- serves the basic phonetic structure of the Vedic language intact. The tendency to change is a good deal more noticeable in the morphology, and in Classical Sanskrit the wealth of forms pre- valent in the earlier language is considerably reduced. Nominal stem-formation shows a reduction in variety in the classical language, and with the disuse of certain suffixes whole classes of words so formed tend to become obsolete. For instance, the suffix -yu is productive in the Vedic language producing not only primary derivatives ( ydjyu - ' pious *) but also a number of secondary formations connected with denominative verbal steins (i devayu - ‘ devoted to the gods vdjayu - ' eager to win '). After the early Vedic period it ceases to be productive, and in consequence most of the words so formed went out of use. Only those that were common enough to survive as individual words remained : manyu- 1 anger ', dasyu - ‘ robber etc. Examples could be given of the same tendency over the whole field of nominal stem formation. In nominal composition the Vedic type of governing com- pound seen in examples like bharddvaja- 1 carrying off the prize etc., became early obsolete. In other respects we see not a diminution in nominal composition, but a steady extension in its use. The members of a compound are rarely more than two in number in the Veda, and the conditions under which they may be formed are limited. As the language advances greater freedom is observed both as to the number of members in a compound, and as to the type of syntactical construction allowed to be so expressed. Finally the stage is reached where compounds of any length may be formed almost without restriction, the whole character of the literary language being thereby changed. In this respect the later classical language goes far beyond anything that would have been countenanced by Panini. There is considerable simplification and modernisation in nominal declension. Older forms of inflection in a-stems such as instr. sg. in -a (viryd beside mryena) and nom. pi. nt. in -d (i bhuvand beside bhuvandni) are given up in favour of the new formations. At the same time some innovations of the Vedic language (nom. pi. m. -asas t instr, pi. -ebhis beside -is, - ai $ ) are discarded. One type of inflection of i- and w- stems ( dvyas , 38 OUTLINES OF THE HISTORY OF SANSKRIT krdlvas) is abandoned, though one type o f special neuter inflec- tion is preserved (* varinas , mddhunas) and a new type of feminine inflection is introduced from the t- stems ( gdtyas , dhenvis). Of the two types of inflection of I- stems, the vrki type is abandoned in favour of the devi type with some influence of the former on the latter (nom. pi. devyds as opposed to devts of the Vedic language), and some isolated survivals (nom. sg. laksmts). The irregular vocatives in -s of the van- and vant- s terns (bhagavas, etc.) are abandoned. Endingless locatives of the type aksdn are abandoned in favour of the fully inflected forms (aksdni or aksni) . The Vedic locative formations from the personal pronouns in -e (asme, etc.) disappear. In the dual the number of cases that can be formed from these pronouns is reduced from five to three by the elimination of the nominative and ablative forms (avdm, ynvdnt ; dvdt.yuvat ). In the conjuga- tion of the verb the classical language simplifies considerably the complicated morphology of the earlier language. The alter- native termination -masi of the ist pers. plural is abandoned, and likewise the long forms of the 2nd plural in - tana , -thana. That form of conjugation in the middle which is characterised by the absence of t in the 3rd singular and the termination r in the 3rd plural (< duke , dnhrd , impf. dduha, dduhra) is abandoned. The old imperatives in -si disappear. The 5- aorist is enlarged in the 2nd and 3rd singular to produce forms more easily recog- nisable (dnaisit for dnais) . The root aorist is confined to roots in long a and the root bhu. Pluperfect forms are eliminated. The most important loss in the verbal structure is that of the subjunctive. This mood is very common indeed in the Veda, and also later till the close of the Brahmana period, but by Panini’s time, apart from forms of the first person incorporated in the Imperative, it had fallen quite out of use. The use of the unaugmented forms classed as Injunctive is confined to con- structions with the prohibitive mi. Modal forms outside the present system cease to be used, as also participles from aorist bases. In place of the great variety of infinitive forms in the Veda, only one, that in -turn is used in the classical language. Similarly old variant forms of the gerund (e.g. in -tvt, - tvdya ) and gerundive (e.g. in 4 va t -enya) disappear. An important difference between the Vedic and the classical language lies in the treatment of the prepositional prefixes attached to verbal roots. In the classical language the prefix OUTLINES OF THE HISTORY OF SANSKRIT 39 stands immediately before the verbal form with which it is compounded. On the other hand in the Vedic language its position is quite free, and it may be separated from the verb by several words, or, on occasion, come after it. This freedom was characteristic of Indo-European, and elsewhere the tendency has generally been to associate the prefix more closely with the verb as time went on. There is the same difference in this respect between Homeric and classical Greek as between Vedic and classical Sanskrit. Finally there are changes ill vocabulary. This has already been noticed in connection with disuse of certain types of nominal stem formation, but it applies equally to the whole field of vocabulary. A number of old Indo-European words which are current in the Veda are no longer used in the classical period. Such are dtka - 4 garment 1 (Av. a8ka~), dpas 4 work ’ (Lat. opus), dndkas 4 juice of soma plant 1 (cf. Gk. dvdos 4 flower ' ?), dma- 4 strength ’ (Av. ama-), drvant - ‘ steed ’ (Av. aurvant - 4 swift '), avatd - 1 spring ' (Lett, avuols ), ddhra- 4 mean, lowly 1 (Av. ddra-), apt - 4 friend, ally * (cf. Gk. ? faios 4 kind r ), isird- 4 vigorous, strong ’ (cf. Gk. Upo$ 4 sacred Irma- 4 fore- leg * (Lat. armus, etc.), ustj- 4 a kind of priest' (Av. usig -) , rsvd - 4 high ’ (Av. zrzsva-), kravts- 4 raw flesh J (cf. Gk. Kpea$), gdtu - 4 way, course ; abode ’ (Av. gdtu -), gnd 4 wife of god ' (Av. gmd 4 wife Gk. ywtf, etc.), cdnas - 4 pleasure, satisfaction P (Av. canah-), cyautnd- 4 deed, enterprise ' (Av. syaoBna-), janas 4 race '■ (Gk. ycV o?, Lat, genus), jdni- 4 woman ' (Av, faint -), j iiny a - 4 under the care of, entrusted to ' (Sogd. zynyh , Khotanese ysimya-, whence Central Asian Prakrit jkeniga 4 id f ), jrdyas - 4 expanse, flat surface 1 (Av. zrayak - 4 lake '), Utau- 4 sieve ' (for

• titanu-, cf. Gk. StaTTaoj 4 sift ', etc.), tokd - 4 offspring tokman -

4 offshoot ' (Av. taoxman - 4 seed ’), tvaks - 4 to be active, ener- getic ' (Av. Owaxs-), dasmd-, dasrd - 4 accomplished, clever ddmsas- 4 wonderful deed * (cf. Av. dahma - 1 instructed datjra- 4 accomplished day hah- 4 cleverness Gk. SeSae 4 taught aSon/s- 4 uninstructed 4 instructed ', etc.), dam a- 4 house ’ (Lat. domus , etc.), d&nu - 4 moisture ’ (Av. ddnu 4 stream Osset, don), ddivds - 4 worshipper drapsa- 4 banner ' (Av. drafsa-), ndkt - 4 night ? RV 7. 71, 1 (Lat. nox, etc. ; ad- verbial ndktam remains), ndhus- 4 neighbour nima- 'half* (Av. naema-), pan - 4 to praise, extol pasty d 4 habitation pitu- 4 nourishment, food r (Av. pitu~ 4 id Lith. petus 4 midday 40 OUTLINES OF THE HISTORY OF SANSKRIT meal etc.), bhrsti- 4 spike ' (Av. °bar9sti~, cf, Engl, bristle , etc.), pajas- 4 surface 1 (Khot. pdysa-, Sbgd. p’z 4 face '), mr&ika- ' mercy J and related words (Av. mdrdzdika-), mdrya - 1 young man ' (cf. Gk. /xetpaf ' lad ’ and the Marianni of the Mitanni documents), miyidha - ' sacrificial offering 1 (Av. myazda-), midhd- "reward, prize* (A w mizda~, Gk. puo66^ t etc.), ydhu-, yahvd- 4 young, youngest, latest * (Av. yazu-, f. yezivi ' id *), yds 4 welfare, rightness * (Av. yaos, Lat. ius) , rodasi * the two surfaces (of heaven and earth) (Av. raodah raoda- 1 face Pers. toy 4 id ’), vddhri- 1 castrated ’ (Gk. iOpls ‘ castrated ram *), vasna- 4 price, value * (cf. Lat. venum , Gk. dwo?, etc., Hitt, ussania- ‘ to sell ’), v&ja- 4 prize, booty ', vtdu- r strong, firm ', ven- 4 to long for ', sdma- 4 hornless * (cf. Gk. K€fids 4 young deer Lith. hnulas 4 without horns '), stprd 4 moustache sund- 4 welfare iyeta - 4 orphan 1 RV. i. 71. 4 (cf. Av. sue , 0. SI. situ, etc.), sap-, sdparya - 4 to attend to re- ligious ceremonies * (Gk. enuj 4 attend to Lat. sepelio 4 bury *), sas - ' to sleep * (Hitt. sel-) stigh- 4 to step, stride * (Gk. crreix w > Goth, steigan , etc.), syond - * soft, gentle haras - ' heat ' (Gk. dlpos ' summer Arm. jer "warm weather*), hary - "to be pleased, exhilarated * (Gk. x<uptu ‘ rejoice ’). The existence of homonyms frequently results in the sup- pression of one of such pairs. The early Vedic language pos- sessed dsura- 1 4 lord * (Av. dhura-) and dsura - 2 ' demon '. Only the latter is in use from the later Vedic period onwards. Similarly of the pair ari - 1 ' devoted, trustworthy * (whence drya- t drya-, cf. Hitt. ara~ 4 friend, ally etc.) and ari- 2 4 enemy* (<*ali-, cf. Lat. alius, etc.), only the latter is pre- served. Vedic kdru - 1 ‘ singer * (kf- 4 to celebrate cf. Gk. KTipvt f, Kapvi 4 herald ') yields to classical kdru- 2 ' artisan * (kr- 4 to do, make), and Vedic rajas 1 f space * (raj- 4 to stretch out cf. Lat. regib, etc.) is abandoned on account of the com- ' petition of rajas 2 ' dust, dirt Similarly of the pairs parusd- 1 ‘ light grey ' (Av. pourusa-, ct. Engl, fallow, etc.) and parusd - 2 4 knotty, rough * (pdrus-, pdrvan- 4 knot r ), pdyu- 1 ' protector * and pdyu- 2 ' anus phalgu- 1 4 reddish, pink phalgu - 2 1 hollow, without substance the homonym listed first ceases to be used in the later language. Changes of meaning naturally occurred over so long a period. Many of these occurred in the natural growth of the language. For instance vdhni- in the Veda means simply ' carrier * and it is OUTLINES OF THE HISTORY OF SANSKRIT 41 applied to Agni in his capacity of carrying the oblations to the gods. Later it means ' fire * in general by a perfectly natural extension of meaning. The term dasyti- is primarily ethno- logical, meaning the non-Aryan inhabitants of India ; later it acquires the meaning 4 robber, brigand Similarly ddsd- * slave * was originally a tribal name (cf. the Dahae of Central Asia) t and the same may apply to siidrd- ' member of the fourth caste since a tribe with this name is known to have existed in N.W. India from both Indian and Classical sources. In other cases the change of meaning in the later language is due simply to a misunderstanding of the Vedic word. This is the case with kratu- 4 sacrifice ' as opposed to Vedic krdtu - ' wisdom, insight Here there is no change of meaning, but simply a failure to understand properly the meaning of the Vedic texts. In classical Sanskrit mdtansvan - means ‘ wind ' ; originally it meant the divine being who discovered fire by the method of rubbing two sticks, and also Agni himself (from

• mdtaris- 1 fire-stick etymologically equivalent to Lat.

matrix) ; the change of meaning can only be due to the fading of the old mythology in the popular mind. Vedic ktldla - meant a certain milk preparation (cf. Khowar kildl 1 a kind of cheese ') ; its use in classical Sanskrit to mean ‘ blood ’ is due to a mis- understanding of the old texts. False popular etymology in the case of the old words dsura - ' demon ' and risita- 4 black 1 led to the creation of two new words. Since the initial «- in these words was falsely inter- preted as the negative a sura- ‘ god 4 and sita- ‘ white 1 were created as their opposites. The above examples show that the losses in vocabulary during this period of the history of the language were consider- able. As elsewhere the Indo-European heritage of In do -Aryan was steadily reduced with the passage of time. One result was that many of the old words of the Veda ceased to be understood in later times. The difficulties that ensued gave rise at an early period to a special school of interpretation (nirukta-) of the Veda. Collections of difficult vocables were made and attempts to explain them on an etymological basis were made. These labours were summed up in the work of Yaska, who may be roughly contemporary with Panini. These attempts at inter- pretation were successful to only a limited degree, and it emerges quite clearly there existed no reliable tradition as to 42 OUTLINES OF THE HISTORY OF SANSKRIT the meaning of many Vedic words, with the result that the authors were frequently reduced to guessing. The same applies to the later commentatorial works culminating in the great Bhasya of Sayana. In modern times the labours of scholars, equipped with greater resources than the ancients, have done much to reduce the field of uncertainty, but even now there remains over a considerable amount of material which defies certain interpretation. In contrast to the losses of the old vocabulary, classical San- skrit has acquired a large number of new words from various sources. These gains far more than counterbalance the losses, and the vocabulary of classical Sanskrit is one of the richest known. Of course there are many words which appear first in the later language and at the same time belong to the most ancient layer of Indo-Aryan. The absence of such words from the older texts is partly accidental, since, extensive though they are, the Vedic texts do not contain the whole linguistic material of pre-classical Sanskrit. Partly also it is a question of dialect ; the widening of the horizon in the case of later as opposed to Vedic Sanskrit led to the inclusion of Old Indo-Aryan material which may not have been current in the dialects which underlay the early standard language. For instance the adverb parut ' last year' is not recorded before Panini, but it is an ancient IE word as is shown by the Greek equivalent Trcpum. Similarly the related pardri ‘ year before last 1 is, we may be sure, absent from the early texts only by accident. There are many words which must be ancient because their formation is of an ancient type, e.g, vipula - ' abundant 1 from the root pf- ' to fill In this case the preservation of IE l , elsewhere not found with this root, suggests that its absence from the early Vedic texts is a matter of dialect. A large number of the new words are fresh formations based on the existing stock of roots and formatives. As long as the suffixes of derivation retained their living character, there was ample scope for the creation of new terms as occasion demanded. This was particularly so since it was combined with the facility of compounding verbal roots with prepositional prefixes, and in this way terms could be created at will for any conceivable need. From the root kr- 4 to do ' alone, by means of the suffixes of derivation, and with the help of Some score of prepositional prefixes, many hundreds of words were manufactured, whose meanings cover every field of practical and theoretical expression. OUTLINES OF THE HISTORY OF SANSKRIT 43 The vocabulary was further enriched from outside Indo- Aryan itself. The pre-existing vernaculars made a sizeable con- tribution to the Sanskrit vocabulary. This influence is strongest, it seems, in the case of Dra vidian words and that can be identified with certainty as Dravidian run into several hundred. Though a few are found already in the Vedic language, the majority do not become current before the classical language. A smaller proportion was provided by the Kolarian languages. Some words were introduced from outside India, e.g. from Iranian (varabdna-, * breast plate ') or from Greek (hard - hour See further Chapter VIII. Even when all these new words have been accounted for there remains a considerable number of words in classical San- skrit whose origin is unknown. Most were no doubt originally desi words in the Indian terminology, and since the linguistic complexity of pre-Aryan India must have been greater than anything that now appears, we should not be surprised to find so many words whose origin remains unexplained. Such in brief are the main changes which took place in San- skrit between the early Vedic and the classical period. In the pre-classical literature this evolution can be traced in its suc- ceeding stages. This literature, which is devoted entirely to religion and ritual, falls into three main sections. I. The Sam hit as of the Rgveda, Samaveda, Yajurveda and Atharvaveda. II. The Brahmanas, prose texts devoted to the mystical in- terpretation of the ritual. IIL The Sutras, containing detailed instructions for per- forming the ritual, of which the Srautasutras deal with the great public sacrifices, and the Grhyasutras with household ritual. The periods which are conventionally assumed for the com- position of this literature are (i) Samhitas 1200-800 B.C., (2) Brahmanas, 800-500 B.C., (3) Sutras 600-300 b.c. In the absence of any definite information, such chronology rests mainly on guess-work. On the other hand, the relative chronology of the succeeding strata can be established beyond all doubt by means of linguistic data contained in the texts themselves. The linguistic changes summarised above took place gradually and the language of the succeeding phases of the literature becomes steadily more and more similar to the classical norm. The gulf that separates the language of the 44 OUTLINES OF THE HISTORY OF SANSKRIT Rgveda .from classical Sanskrit is very much greater than that separating the language of even the earliest prose texts from it. The later Samhitas can be shown on the basis of language to be later in date than the Rgveda, and within that work itself the tenth book is known to be the latest for the same reason. In the same way a chronological distinction can be made between earlier and later Brahmanas. Since the Brahmanas are in prose, their language may be taken as reasonably representative of the spoken language of the upper classes in the later Vedic period. It still retains pre- classical features., such as the use of the old subjunctive, but already the majority of the old Vedic forms have fallen into disuse. By the time of the composition of the Sutras the lan- guage has reached in all essentials the stage at which it was codified by Panini. In all the Sutras it is possible to find grammatical forms which do not conform strictly to the Paninean rules. But in contrast to the earlier literature these forms are not as a rule archaisms. The difference is rather that their usage is somewhat more lax and careless than that allowed by the strict formulation of the grammarians, and in this respect they accurately reflect the spoken usage of the period of Panini himself and of the period immediately pre- ceding. Their language is based, not like the later classical Sanskrit on an established and traditional grammatical system, but on that same spoken language of the educated Brahmins, which is the source of the grammatical system of Panini, These texts are very important in linguistic history : they stand side by side with Panini as an independent authority on the living Sanskrit language during the period immediately preceding its final codification. It is here, and not in the later literature, that we must look for a living illustration of the lan- guage that Panini established in its final form. §2. Old Indo-Aryan The Sanskrit language, in its Vedic and Classical form, had, as already observed, a definite geographical location. In the very earliest period this lay in the Punjab, but the centre soon moved eastward to the countries of Kuru and Pancala, and there it remained during the whole later Vedic period. Certain dialectal divergencies between the language of the Rgveda and that of the later literature — notably the use of l instead of Vedic OUTLINES OF THE HISTORY OF SANSKRIT 45 r — can be ascribed to this. In addition there existed dialect areas containing features which find no place in Sanskrit* Quite early in the Vedic period there were extensive settlements of Indo- Aryans to the East (Kosala, etc.) and to the South (Avanti, etc,). The spoken language outside the area which was the home of classical Sanskrit differed in certain respects from the spoken language of this area. The term Old Indo-Aryan is sometimes used as alternative to Sanskrit, but this is incorrect, since there were other dialects of Indo- Ary an in addition to those on which Sanskrit is founded. The term Old Indo-Aryan should be used for the whole body of Indo-Aryan during the early period, and Sanskrit is not co-extensive with this. Of the non-Sanskritic dialects of Old Indo-Aryan no direct remains are preserved, and there would not be much to be said about it, if it were not for the fact that in the later Middle Indo- Aryan dialects a fair amount of material exists which cannot be explained out of Sanskrit, Vedic or Classical, but only out of equally ancient, but different forms of Indo-Aryan such as have been referred to above. A complete collection of such material has never been put together, but enough evidence is available to demonstrate the one-time existence of non-Sanskritic dialects of Old Indo-Aryan. The differences involved were not very great (as compared for instance with the early dialects of Iran- ian), but they are sufficient to be worth taking into account. Among the phonetic features we may note primarily the change of final -as to -e in contrast to its treatment as -0 in San- skrit and the later Prakrit dialects of the central area. This was a distinguishing feature of Eastern Indo-Aryan, but examples are a^Q found in the extreme North-West. In sure duhiti ' daughter of the sun * one dialectal form of this type is pre- served in the IJgveda. In place of Sanskrit k$ Middle Indo- Aryan forms sometimes show jh~ t jjh- t ggh in cases where Iran- ian has the sonant combination yz (Pkt. jharai ‘flows'. Pa. paggharati , SkLMsar-, Av. y£ar- t etc.). There are dialectal vari- ations in the treatment of ancient rH (f) ; OIA ur in place of Skt. ir is attested in some cases : Pkt. junria- * old * <*jurnd~ (: Skt. jirnd-), tuha- ‘ ford *<*turtha- (: Skt. tirthd-)* In some dialects ancient ~zd- was replaced by dd - instead of by single -d- with compensatory lengthening as in Sanskrit, e.g. Pa. nidda- : Skt. ni&d- ‘ nest In Pa. idha * here 1 a more ancient form of the word is preserved than in Skt. ihd, Pkt. sidhila46 OUTLINES OF THE HISTOKY OF SANSKRIT ' slack ' (out of *sithila-) and Skt. sithird- ( °la -) are parallel in- dependent developments from earlier *sr third- (srath- 4 to loosen '). A number of grammatical differences can be observed, though the loss of so much of the old inflection in Middle Indo- Aryan has eliminated much of this. Forms of the third plural atmanepada like Pa. vijjare 4 are seen ' ( > *vidyare) indicate that such r- endings were more extensively used in some Old Indo- Aryan dialects than in Sanskrit. Pa. jigucchati ‘ is disgusted * (Skt . jugapsate) shows i- reduplication of the desiderative in the case of roots containing the vowel -u- which is absent in San- skrit but known to Old Iranian (Av. cixsnusa- ‘ desire to please’) Pa. hardyati 4 is angry ’ continues an Old Indo-Aryan form bearing the same relation to Skt. hrnite as Vedic grbhdydti to grhnati. In viheseti 4 injures vihesd 4 injury ' « *vihesayati ,

• vihesd, J'hims-) Old Indo-Aryan forms unknown to Sanskrit

are represented. In Pa. sabbadhi 1 everywhere ’ an old adverbial termination is preserved which has a parallel in Gk. -Ot. Pa. kdhdmi 1 I will do ' represents an Old Indo-Aryan unit future

• karsydmi as opposed to Skt. karisyami . Pa. sdmam 4 oneself ',

which is unknown to Sanskrit, is the equivalent of Av. hdma- t O. SI. samii. Participial forms like mukka- 4 released runna- 4 weeping ' continue old formations in - na, as opposed to the Skt. formations in - ta . The Ardha-MagadhI participles in -mina seem to represent an ancient Indo-Aryan variant of Skt. - mana . The participle dinna - 'given' implies an old formation re- duplicating with i as in Greek ( *didami : Gk. S/Sayu )> The parti- ciples in -1 tdvin (vijitdvin- 4 who has conquered ') are an ancient formation equivalent to the -tavant- participles of Sanskrit. In etase 4 to go ' we have a Vedic type of infinitive not elsewhere found. The absolutives in -tuna (Pkt. °una), gantuna , etc., differ in apophony from the Vedic forms in - tvdna . Difference in apophony is frequently observable in stem formations : e.g. supina - ' dream '“Gk. vnvos as opposed to Skt. svdpna- with guna ; garu- 4 heavy ' has guna of the root as opposed to Skt. guru - ; compare in the same way turita- 4 hastening thina-

• slothful * ( < *stina-) with Skt. tvaritu sty ana-. Nominal stem

formations unrepresented in Sanskrit are not uncommon, e.g. Pa. nahdru - 'sinew', theta - 'firm' theva - ' drop ’<*sndru- t +$theta~, *$tepa- (stip- 4 to drip ') ; Pkt. mahana - 4 brahmin 9 (lit. ' great one '), cf. V. mihina - ' great OUTLINES OF THE HISTORY OF SANSKRIT 47 A number of words with IE etymology appear first in the later stages of Indo-Aryan: e.g. Kharosthl Prakrit ala s flour Hindis 1 id ’ , cf. Pers. drd from the root in Gk.aAca) ‘ togrind ' ; Pa* latnaka- ‘ inferior, WTetched V Pj. lava ‘ maimed cf, Engl. lame, Russ, lorn, etc., Pkt. (Apabhramsa) tilra- 1 cheese cf. Gk. TVpOS. From evidence such as this we can form some idea, fragmen- tary though it is, of the dialectal variety of Old Indo-Aryan. It is necessary to bear this in mind so that the evolution of Sanskrit can be seen in its proper perspective. The form- ation of a standard language implies a rigorous process of selec- tion and exclusion. In all spoken language there is continuous variation from area to area and from class to class. Sanskrit was based on the spoken language of the higher classes of Madhyadesa, influenced by the older sacred language of the Rgveda which had originated further West. This was the centre of propagation of Brahmin religion, in its orthodox form, and of a fixed standard language which was the property of the Brahmin community in whatever part of Aryavarta they re- sided. Panini speaks occasionally of differences in speech be- tween the Easterners and the Northerners. But these are always trivialities. Sanskrit as a spoken language was essenti- ally the same over the whole of North India, and from an early period also in the Deccan. Under the surface there were dialectal differences which for the earliest period can be dimly perceived, and which come out into the light of day during the next stage of the language. Middle Indo-Aryan. §3*' The Grammarians The importance of the grammarians in the history of Sanskrit is unequalled anywhere in the world. Also the accuracy of their linguistic analysis is unequalled until comparatively modern times. The whole of the classical literature of Sanskrit is written in a form of language which is regulated to the last de- tail by the work of Panini and his successors. Grammatical interest in India arose in the first place in con- nection with the necessity of preserving intact the sacred texts of the Veda. It was of the utmost ritual significance that every word used in the recitals at the sacrifices should be pronounced absolutely correctly. Among the means by which the correct 48 OUTLINES OF THE HISTORY OF SANSKRIT transmission of the Vedic texts was achieved was the Pada- patha, in which each word of the text was repeated separately. To do this correctly, as it is done in the main, involved the be- ginning of grammatical analysis and, since it involved the resolution of Sandhi, phonetic analysis. The phonetic teaching necessary for the correct recitation of the Vedas is embodied in the Prdtisdkhyas . There are several of these attached to various Vedic schools, and they deal with the subject in great detail and with accuracy. They are a very important source for our knowledge of ancient pronunciation. It is disputed whether any of these texts in their present form are earlier than Panini, but in some form or other instruction of this sort must be as old as the Vedic schools themselves. Later works dealing with phonetics are the Siksds which exist in large numbers and contain valuable observations. Difficulties in the interpretation of the Vedic texts owing to the obsolescence of words led to the beginnings of lexicography. The earliest work of this kind, the Nighanfu consists of lists of difficult Vedic words, of divinities, etc., drawn up for the use of teachers. The commentary on these by Yaska, who is probably not far removed from Panini in time, contains the earliest systematic discussions on questions of grammar. Here we find the parts of speech already distinguished as ndman r noun sarvandman- f pronoun dkhydta - ' verb upasarga - * preposi- tion * and nipata- * particle The derivation of nouns by means of krt and taddhita affixes has become a well established theory, and an interesting argument between Sakafayana and Gargya is reported as to whether all nouns can be derived in this way from verbal roots. The former maintained that they could, and in spite of the cogent arguments on the other side advanced by Gargya, this was the theory that generally held the field in Sanskrit grammatical theory. It is a fact that a larger propor- tion of the Sanskrit vocabulary is capable of such analysis than is the case in most languages. The date of Panini is most commonly fixed in the fourth century b.c, which is in accordance with the native tradition which connects him with the Nanda king of Magadha. Nothing is known of his life except the fact that he was bom in the extreme North-West of India at Salatura. His Astadhydyi which fixed the form of Sanskrit grammar once and for all, consists of some 4,000 aphorisms of the greatest brevity. This OUTLINES OF THE HISTORY OF SANSKRIT 49 brevity is achieved by the invention of an algebraical system of notation of a kind not found outside the grammatical schools. The system is so idiosyncratic that it could not possibly have been invented there and then by one man and imposed immed- iately on all his colleagues. It is clearly the growth of many centuries and Panini is to be regarded as the final redactor of a traditional Vydkarana who superseded all others on account of his superior comprehensiveness and accuracy. Many of the pre- decessors of Panini are in fact cited in the text, but the merits of his own work condemned theirs to early oblivion. The brevity which the Sutra style aimed at and achieved was due to the fact that all instruction was still oral and dependent on memory. It implies also from the very beginning the exist- ence of a commentary {vrtti), also oral, in which the examples were contained. When this was first written down is not known, but the earliest existing commentary on Panini, the Kasika , dates from a thousand years after his time (c. a.d. 700). A ganapdtha containing lists of words referred to in the Sutra by citation of the first word in them followed by - ddi , and a dhdtupdtha, containing a list of verbal roots, formed essential parts of his system. The Sutras of Panini were supplemented and to some extent corrected by Katyayana at a date not long after the composi- tion of the A stadhydyi itself. These notes ( Vdrttika -) are of the same brevity as the original work, but were fortunately soon made the subject of an extensive commentary ( Mahabhdsya ) by Patarijali. His date is fortunately known through contemporary references, notably to the Suhga king Pusyamitra and to an invasion of the Bactrian Greeks, which fix him definitely in the second century b.c. Later grammatical works exist in abundance, and many diverse schools arose, but none of them have any independent authority, being completely derivative from Panini. The earliest is the Kdtantra which arose about the Christian era, and whose author Sarvavarman is said by tradition to have been connected with the Satavahana dynasty of the Deccan. The work aimed at introducing the study of correct Sanskrit to a wider public than the educated Brahmins for whom Panini and his immediate successors had written. Of later works mention may be made of the Grammar of Candra (a.d. sixth century) which achieved great popularity among the Buddists, and the 50 OUTLINES OF THE HISTORY OF SANSKRIT Jainendra Vyakarana { c . 678) which was composed on behalf of the Jains. Later the polymath Hemacandra produced also for the Jains the Haima Vyakarana. In addition, a number of minor systems are known which were popular in various localities, but which have nothing original to contribute. The object of all these later grammars was to present the material contained in Panini in a form comparatively easy to assimilate, and in this respect they performed a service to very many who were not equal to the arduous task of mastering the original text itself. How useful they were is shown by their continuous popularity. They contain little that is original since for them there existed no other source from which they could draw except the work of their illustrious predecessor. To Panini the main source of his work was the living speech of himself and his contemporaries. It is the merit of his gram- matical system that by means of the Sutra and commentary, and by such subsidiary compilations as Dhdtupdtha, Gana~ pdtha , etc., the vast bulk of the contemporary linguistic usage was incorporated, analysed and codified in the teachings trans- mitted from teacher to pupil in the schools of the Grammarians. The rapid process of linguistic change that took the vernaculars through the various stages of Middle Indo- Aryan enhanced pro- gressively the value of this codification. It is characteristic of Ancient India that the founders of schools and doctrines should be exalted to semi-divine status and regarded as omniscient. In the case of Panini this was more justified than in other cases since he had direct knowledge of the living Sanskrit language of the fourth century b.c. which is the source of all his statements. As a result of his labours and the labours of his school this form of language was accepted as a standard throughout the long period that remained of the classical civilisation of India. As the gap between this and the vernaculars grew continually wider, the usage of the speakers and waiters of Sanskrit grew more de- pendent on Panini, and his authority more absolute. Panini's grammar was based on the language of his contemporaries, and conversely the language of Kalidasa and his successors is based on the grammar of Panini. The Sanskrit of the classical litera- ture was a living language in the sense that it was written and spoken by the educated in preference to any other, but at the same time it was a language that had to be learnt in schools by OUTLINES OF THE HISTORY OF SANSKRIT 51 means of an arduous discipline. It was a prerequisite for all men of letters of the period that they should know by heart the Astadhydyt, and evidence of this dependence appears continu- ally in their works. As time went on, the cultivation of classical Sanskrit also came to depend on the Kosas or lexica. Apart from the Vedic nighantus lexicography is a later growth in India than grammar. The extant lexica are mostly late and are compilations out of earlier works. Amarakosa , the earliest existing, has not been accurately dated, but it is put approximately in the period a.d. 600-800. Earlier works are known and sometimes quoted, but not preserved. These works are in metre and intended to be learnt by heart, a practice which in the traditional schools has continued to this day. When this first became an essential re- quirement of a literary education is not clearly known, but certainly for the later period of Sanskrit literature we may assume that the writers were so equipped. In spite of their late date and, in general, unscientific method, the lexica are of considerable value, since they preserve a large number of words which are not recorded in available texts. At the same time careless copying and inaccurate transmission has created some ghost words, which careful comparison of the various lexica may remove. §4. Epic Sanskrit The bulk of the classical Sanskrit literature was composed at a period very much later than the fixing of the language by Panini. An earlier period in literary and linguistic history is represented by the two great popular epics, the Mahabharata and the Ramayana. It does not seem that either of these two works reached its final form until well after the Christian era, but the tradition of epic recitation goes back into the Vedic age. The Mahabharata in particular was a long time in forming, and a good deal of what is incorporated in the final recension may claim an earlier date. We have therefore in the Epics extensive documents of Sanskrit belonging to a period nearer to Panini than the classical, literature in the narrow sense. They provide also evidence of the wide popularity of one type of Sanskrit literature among the masses of the people, since these works were reserved for no 52 OUTLINES OF THE HISTORY OF SANSKRIT special or cultivated audience, but intended for public recital to the population in general Their popular character is evidenced by their language. This is Sanskrit definitely enough as opposed to the contemporary Middle Indo-Aryan, but it is a Sanskrit which frequently violates the rules which Panini had laid down and which were always observed in the more orthodox literary circles. Among the common deviations of the Epic language a few characteristic types may be quoted. The dis- tinction between the active and middle forms of the verb, which was still fully alive in Panini's time, and for which he caters in some detail, is beginning to be blurred in the Epic. Active forms are used for middle and vice-versa, and even the passive verb sometimes takes active endings (sruyanti ' are heard etc.). There is some confusion between the gerunds in -tvd and -ya, and the rule of Panini which restricts the former to uncompounded and the latter to compounded verbs is not always observed. Unaugmented preterites occur, a character- istic which is also found in the Veda, as well as in early middle Indo- Aryan. Conversely the augmented forms are occasionally found with the prohibitive particle (ntd . . . agamah 1 do not go '). The particle md is not used exclusively with the unaug- mented aorist according to rule but indifferently with impera- tive (md bhava) optative (md bruydh) future (md draksyasi) and so on. The tenth class and causative verbs make a middle parti- ciple in -aydna (< codaydna - as opposed to correct codayamana-) a usage to which metrical convenience has contributed. The care- ful rules of Panini concerning the use of the alternative forms -atl and - anti in forming the feminine of present participles are not strictly observed. The distribution of set and an it forms fre- quently does not conform to rule. These and other irregular forms correspond to what is found in early middle lndo-Aryan, indicating that Epic Sanskrit is a later form of Sanskrit than that of Panini. No pre-Paninean forms are found in the Epic, which means that although the epic tradition goes back to the Vedic period, and although the Mahabharata story was familiar to people before Panini's time, even the earliest portions of the present text must be distinctly later than him. Since for centuries the transmission of the epic stories depended on oral tradition, and not a fixed oral tradition like that of the Vedic schools, it is not surprising that a circle of stories originating in the Vedic period should in their final form OUTLINES OF THE HISTORY OF SANSKRIT 53 appear in a language of a much later date with no archaic forms preserved. The recitation and transmission of the Epic legends was not the business of the Brahmans, but of the Sutas, a class of royal servants whose duties had originally included that of charioteer. It was natural that their language should be of a more popular nature than that of the educated classes par excellence , the Brahmans. At the same time it is interesting that all along, in spite of the competition of Prakrit, Sanskrit was cultivated in much under circles than in the priestly schools for whom Panini's work *was intended. Outside the brahmanical schools the knowledge of grammatical theory must have been ele- mentary to say the least, and in the early period at least the knowledge of Sanskrit on the part of the epic reciters must have depended primarily on usage and not on formal instruction. From this arose the tendency to approximate the language to some extent to the prevailing type of Middle Indo-Aryan. Later when the gulf between the two became greater formal instruction in Sanskrit became a universal necessity, but by this period the epic style and the epic language had already established itself in its own right, and linguistic features such as those mentioned above were accepted and retained. The language of the Epics served also as a model for the lan- guage of the Puranas, of which the earliest core dates to the same period. It is continued in the numerous later compila- tions, and further in a variety of sectarian dgamas , etc. Lin- guistically these compilations are not of great interest, except occasionally in the matter of vocabulary, and many, particu- larly the later ones, testify to the deficient education of their authors in grammar. §5. The Sanskrit of the Classical Literature The special characteristics of classical Sanskrit arise from the fact that most of the literature dates from a period very much later than the period in which the form of the language was fixed. If Kalidasa is to be dated c , a.d. 450 a period of no less than eight hundred years separates him from the grammarian Panini. The work of Kalidasa stands almost at the beginning of the body of classical literature which is preserved, and the greater part of this is separated by more than a millennium from the regulator of the language. This accounts largely for the 54 Or T LINES OF THE HISTORY OF SANSKRIT artificiality of style and language which is not absent from the best authors, and which in some is exaggerated beyond reason. The literary gap in the period immediately preceding arid succeeding the Christian era is due to the loss of the bulk of the pre-Kalidasan literature, since it is known that kdvya in all its forms was actively practised during all this time. The earliest Sanskrit inscriptions (e.g. of Rudradaman, a.i>. 150) show the existence of a developed Sanskrit Kavya. Patafijali (c. 150 b . c ,) quotes some kdvya fragments and mentions by name a poet Vararuci. His own work is a valuable example of the prose style of the period, and it enables us to form a picture of early Paninean Sanskrit at a time when it was still a fully living language. The works of Asvaghosa who flourished under Kaniska (a.d. 78 -f) preserved in Nepal (and fragmentarily in Central Asia), though long forgotten in India, have survived by fortunate chance, as the sole examples of Sanskrit Kavya literature in its earlier phase. The gap, only partially to be filled, between Panini and the classical literature, is responsible for certain changes in style and usage, which have affected the language of the latter, in spite of strict adherence to the rules of grammar. These changes comprise certain losses and also a number of innovations. Of the losses the most important was that of the old system of accentuation. This was still in full force in the time of Patanjali and it must have continued in being for some time after that, but by the time the bulk of the classical literature was composed it had certainly disappeared from ordinary use. Certain of Panini's grammatical forms though recognised were not in practice used. Already Patanjali remarks that forms of the second plural of the perfect like usa, ter a, cakra are no longer in use, their place being taken by the participial forms usitdh , tirndh, krtavantah . Later a good deal else was tacitly ignored. There are many constructions and idioms taught by Panini which are not recorded in the later literature (anvdje- or updje-kr ' to strengthen ’, nivacane-kr ' to be silent ', etc.), and many others which have obviously been employed by the later authors as evidence of their grammatical learning (e.g. in N aisadhacarita , darsayitahe, first person of the periphrastic future middle). There are losses in vocabulary and such words as anvavasarga- 4 allowing one his own way niravasila- ' excom- municated ' and abhresa- 'fitness, propriety 1 are no longer OUTLINES OF THE HISTORY OF SANSKRIT 55 used. In particular the Ganapatha contains numerous terms which are found nowhere else, and since this text was handed down without meanings for a long time, it is often impossible now to discover the meaning of such words. The old distinction in meaning between the three past tenses (Imperfect, Aorist, Perfect) was not normally observed. The Aorist, though culti- vated by the learned, seems to have gone out of common use, 1 The middle perfect participles in - ana are entirely disused, and the active participles in -vas appear only rarely. The innovations of the later classical Sanskrit affect mainly syntax and vocabulary. The most striking syntactical develop- ment is the increasing tendency to use compound words and the increasing length and complexity of the compounds used. In the earliest Sanskrit the use of compounds is not noticeably more predominant than in the Greek of Homer. In the lan- guage of Panini’s day there were still strict rules and limitations in the formation of compound words, as is clearly evident from his own statements and examples. In the later language they are formed without restriction (e.g. any adjective' may be so construed With any noun, as opposed to the original arrange- ment by which this could only be done when the term had a special significance, krsnasarpa- ' cobra J , etc.), and not infre- quently in direct contradiction to Panini's rules (e.g. jagat- kartar - ' world-creator ' against P. 2. 2. 15-16). But the main thing is that there ceases to be any limitation to the number of members a compound may contain, since compound words treated as units may be compounded with further words, and by a process of accumulation long complexes are built up in which the syntactical relation of the members is expressed without recourse to inflection. This practice is not only at vari- ance with the earlier usage and with Indo-European usage in general, but is also obviously incompatible with any form of popular speech which can have prevailed in India during the period This linguistic development is a purely literary de- velopment, and it is a sign of the growing artificiality of the Sanskrit language as the difference between it and the ver- nacular Middle Indo-Aryan grew wider. 1 The hero of the drama Padmaprabhrtaka (c. second or third cent, a.d,) asks a grammarian who speaks pedantically to use ordinary Sanskrit {vydvahdrikd bhd$a). The pedantry which is illustrated consists in the liberal use of aorists and desideratives. 56 OUTLINES OF THE HISTORY OF SANSKRIT Another syntactical development affects the verb, but this is based on popular usage. Of the past tenses the aorist, with the amalgamation of some imperfect forms, survived in Early Middle Indo- Aryan, but by the time of the later Prakrit all traces of the old preterites have disappeared. Their place was taken by passive constructions with the past participle passive, and it is from this usage that the preterites of modem Indo- Aryan derive. The tendency is also reflected in Sanskrit litera- ture, and the passive construction becomes gradually more predominant. It had obviously the advantage of simplicity, since the complicated verbal inflection of Sanskrit could be dispensed with, and in works of deliberately simple style like the Hitopadesa it is evidently chosen for this purpose. For active use the participle in - tavant is adapted to serve as an alternative to the past tense : krtavdn ' he did The nominal phrase in which the meaning is expressed by the juxtaposition of subject and predicate, without any verb becomes increasingly popular. This is particularly so in the philosophic literature, and since that language also favours long compounds, we may find long passages of exposition in which the only grammar consists of a few case inflections of abstract nouns. The vocabulary of Sanskrit was on the whole remarkably stable. Nevertheless it is possible to collect from the later literature a considerable body of words wdiich do not appear in the earlier period. In some cases it may be an accident that they are not recorded earlier, but even making this allowance, there must remain a fair number of new words. Increases in vocabulary derive from the following sources : (i) They could be created, wdien required, on the basis of existing Sanskrit roots, prefixes and suffixes, and by the forma- tion of new compounds with special senses. {2) In the course of time some Prakrit w r ords were adopted into Sanskrit, though proportionally the number is never very large. No certain examples of this kind appear in the Vedic language, but they begin to appear in small numbers in the Epic and classical period. Words of this type are bhatta-, bhaitdra - bhattaraka - ' master, lord ' (Skt. bhariar -), nata - 1 actor ' (Skh nft- * to dance ') and dohada-, dohala- 4 morbid desire of a pregnant woman 1 which occur (more commonly) beside the regular Sanskrit daurhrda Skt. udu- 4 star ' derives from a Prakrit uduvai - * moon 1 misinterpreted as meaning 4 lord of OUTLINES OF THE HISTORY OF SANSKRIT 57 the stars ’ though it is actually derived from a Skt. rtupati-

• lord of the seasons In the medical texts kotha - 4 a form of

leprosy ' occurs beside kustha- 1 leprosy ' of which it is a Prakrit development. Among other words originating in Prakrit we may mention kola - * breast, lap * beside the original kroda- * id khuddaka - ‘ small ' beside ksudraka-, vaiyavrtya- with Prakritic -v-, more commonly used than the original vaiydprtya- ‘ busi- ness or commission entrusted to one vicchitti- 4 carelessness in dress or decoration * which probably derives from viksipti and oja- 1 odd (of numbers) 1 which in later texts tends to replace the original aynja-. In the case of some words taken over from Prakrit there are nc Sanskrit originals, e.g. avahittha- ‘ dis- simulation chatd 4 heap, mass ' (Pkt. chadd) etc. Some words are disguised by false Sanskritisation ; e.g. karpata- ‘ragged cloth " is taken from Pkt. kappada - which itself represents

• kat-pata - ‘ inferior doth Likewise Pkt. ludda- ‘ hunter '

(ultimately identical with the name of the god Rudra-) is Sans- kritised as lubdhaka^ .as' if derived from lubh - 4 to be greedy', and Pkt. paraddhi- 4 hunting ' (from rabh- ‘ to attack ') is Sanskritised as paparddhi - meaning literally ‘ evil gain (3) The Greek and Iranian invasions of India from the North- West resulted in a limited number of loanwords from these sources being admitted into Sanskrit. These are dealt with in Chapter VIII. (4) Sanskrit received a considerable number of words from the substrate languages, Austro- Asiatic and Dravidian, parti- cularly from the latter. These are also dealt with in Chapter VIII. (5) The.tHerm.i^i is applied to those words in Prakrit which are derived from no Sanskrit equivalent. The number of such w r ords which can be explained out of Dravidian or some other source is comparatively small and will probably always remain so. They become still more abundant in the Modern Indo- Aryan period and present a philological problem which is not easy to solve*- On the whole classical Sanskrit avoids such words, but a number are incorporated, and in particular the Jain writers have adopted a fair number. §6 . Sanskrit and Prakrit During the whole period of its existence Classical Sanskrit had beside it as competitor Middle Indo-Aryan in its various forms, not only as a spoken language but also as a language of 58 OUTLINES OF THE HISTORY OF SANSKRIT literature* In the early period this competition was much more important than it was later ; though it appears paradoxical at first sight, the Sanskrit language only reached its full develop- ment as a language of culture and administration at a time when it had ceased to be a mother tongue. The rise of Middle Indo-Aryan as a literary language coin- cided with the foundation of the new religions of Buddhism and Jainism round about 500 b,c. The founders of these religions deliberately chose the vernacular — the dialect of Magadha in the first instance — as the vehicle of their teaching. In the third century b.c. Asoka had his inscriptions engraved in various local dialects and ignored Sanskrit. It follows that the language of administration of the Mauryan empire was also in Middle Indo- Aryan, and not as universally the case later, in Sanskrit. If this process had not been reversed Sanskrit might have yielded place to the younger language, but quite the reverse happened and from the end of the Maurya period a steady process set in which resulted in Sanskrit becoming the predominant language of literature, culture and administration. The epigraphical tradition established by Asoka continued for some centuries. Until after the Christian era the vernacular language alone was used for epigraphical purposes, and this means that business and administrative documents — all of which were written on perishable materials and have not survived — were composed in the same language. After the Christian era Sanskrit too begins to appear in inscriptions, at first in competi- tion with Prakrit, and finally in exclusive use. The inscription of Rudradaman (a.d. 150) marks the victory of Sanskrit in one part of India. In the South Prakrit remained in use longer and was not finally ousted by Sanskrit until the fourth or fifth century a.d. Eventually the use of Prakrit was discontinued entirely and from the Gupta period to the Mahommedan in- vasions Sanskrit — admittedly often incorrect Sanskrit — re- mained in exclusive use* The linguistic revolution in epigraphy is paralleled in other fields. The early Buddhist scriptures were exclusively in Middle Indo-Aryan. Towards the beginning of the Christian era a change took place, and the northern Buddhists adopted Sanskrit instead. A^vaghosa (c. a.d. ioo) is a master of polished Sanskrit, and that he should choose this language as a vehicle of propaganda is an indication of the ascendancy which SanOUTLINES OF THE HISTORY OF SANSKRIT 59 skrit had achieved at this time. Here also we may observe that Sanskrit established its ascendancy first in the north. The Thera vadins of South India and Ceylon remained faithful to Pali* The Jains were slower in making a change than the Buddhists, They were the most conservative of Indian sects and up to the time of the final constitution of the present canon of the Svet- ambaras (at the council of Valabhl in a. d. 526) they used Prakrit exclusively. But even they turned to the use of Sanskrit in the succeeding period. At the same time they continued to culti- vate Prakrit seriously, beside Sanskrit, at a time when in other literary circles the traditional Prakrit was being employed as little more than a literary exercise. In these fields we may observe the transition which led to the predominance of Sanskrit. Elsewhere lack of material makes a clear picture more difficult. In poetic literature there was under the Satavahanas and their successors an active tradition of lyrical poetry in Maharasfri of which fragments are preserved in the anthology of Hala. At the same time the major poetic works of the early period were in Sanskrit. The Mahabharata and the Ramayana have an importance in the literary history of India which nothing in Prakrit could even remotely approach, and they were the productions of a period when to judge by inscriptions Prakrit had almost superseded Sanskrit in everyday use. Thus it is obvious that the inscrip- tional evidence gives a very one-sided picture of contemporary linguistic conditions. Outside the sectarian religions Sanskrit was always, even when the use of Prakrit was most flourishing, the primary literary language of India. The growing predominance of Sanskrit as opposed to Prakrit in the period succeeding the Christian era can be attributed to two reasons, one ideological and one practical. In the Maurya period the heterodox religions of Buddhism and Jainism had attained such influence as to threaten the existence of the old Brahmanical order. In the succeeding period, beginning with the usurpation of Pusyamitra (c. 188 b.c.), a reaction set in and there began a gradual decline of these systems in the face of victorious orthodoxy. This change in the religious atmosphere was reflected in language, and Sanskrit, associated with the traditional Vedic religion gained ground at the expense of Prakrit, whose cultivation was mainly due to the activities of the unorthodox sects. 60 OUTLINES OF THE HISTORY OF SANSKRIT The practical reason was that Sanskrit offered a united lan- guage for the whole of India. In the early Middle Indian period the differences between the various local vernaculars were not so great as to preclude mutual understanding, but even at this period Asoka found it necessary to engrave his edicts in three different dialects. With the progress of time the differences between the local dialects grew greater, so that Sanskrit became a necessary bond for the cultural unity of India. Furthermore the Prakrits were unstable and subject to continual change through the centuries. Any literary language established on the basis of a vernacular rapidly became obsolete. The tradi- tional Prakrits in the later period were as artificial as Sanskrit, and did not have the advantage of its universal appeal and utility. For such reasons alone Sanskrit was the only form of language which could serve as a national language in Ancient India, whose cultural unity, far more influential and important than its political disunity, rendered such a language essentiaL The relation between Sanskrit and Prakrit in the classical period is admirably illustrated by the Sanskrit Drama. Here it is the convention that certain characters speak Sanskrit and others speak Prakrit, and the usage of the drama no doubt accurately represents the actual practice at the _time. The use of Sanskrit is fairly narrowly limited to the highest classes of society, namely kings, ministers, learned Brahmans and so on. Women, with few exceptions, speak Prakrit, and also children, showing that it was everybody's first language. Furthermore, Prakrit is spoken not only by all the lower classes, but also predominantly by the wealthy and influential class of merchants and bankers. The comic figure of the vidusaka , an unlearned Prakrit-speaking Brahman, shows that not all members of this class were capable of mastering the strenuous discipline necessary for the acquisi- tion of Sanskrit. Only the earliest dramas, of which Mrcchakatika is the best surviving example, reflect living usage in this way. In the greater number of extant dramas which belong to a later period (a.d. 500-1000), the composition is according to tradition, and the Prakrit becomes merely a transmogrified Sanskrit composed according to the rules of the grammarians learned by rote. In this period the vernacular had advanced much further on the road to Modem Indo-Aryan. OUTLINES OF THE HISTORY OF SANSKRIT 6l §7. The Sanskrit of the Buddhists and Jains The adoption of Sanskrit by the Buddhists, and later the Jains, widened the field of Sanskrit literature, and the Sanskrit language which was thus adapted to new needs did not remain unaltered in the hands of these authors. The Sanskritisation of Buddhist literature is particularly complicated since it took place gradually and beginning by a compromise between San- skrit and the Middle-Indian dialects of early Buddhism ended in the adoption of pure classical Sanskrit. We may distinguish between the following types of Buddhist Sanskrit : (1) Mixed Sanskrit. This language was used by the Maha- sarighika school. In it the original Prakrit appears half San- skritised, the words being in the main restored to their Sanskrit phonetic form while the Prakrit grammar is largely retained. For instance Pa. bhikkhussa, gen. sg. of bhikkku 4 monk ' (Skt. bhiksu-) is not replaced by a regular Sanskrit bhiksos , but is mechanically changed to bhiksusya . It may be assumed that for a period, in certain circles, such a hybrid language was actually employed by those who wished to employ the superior Sanskrit language but were not able to master its grammar. (2) The Sarvastivadins of the north-west adopted proper Sanskrit from an early period. The old canonical works were translated into Sanskrit, and fragments of them are preserved in this form. The language of these works has of necessity incor- porated wholesale the vocabulary and syntax of the original Mag- adhi, but allowing for this, and for some false Sanskritisations which are to be expected, it is free from the barbarisms of (1). (3) We must distinguish from (2) works of the same school which were not translations but which were independently composed at a period much later than the canonical literature. The stories which were inserted to enliven the matter of the Vinaya-pifaka, and which are collected in the Divyavadana, Illustrate best this type of Sanskrit. Though it fails often enough to satisfy the canons of Paninean grammar, the style is admirably clear and lucid and not an unwelcome change to the laboured artificiality of some of the classical prose. The vocabu- lary is characterised by the use of many vernacular and pro- vincial words, many of which turn up again in Modern Indo- Aryan (e.g. lardaya- ‘ to load * : Hi. Iddnd ), and which are not found in the older types of Buddhist language (e.g. Pali). 62 OUTLINES OF THE HISTORY OF SANSKRIT (4) The use of pure classical Sanskrit, associated with all the characteristics of the Kavya style, is seen in the works of A&vaghosa and his successors. Such works are distinguished from other works of classical Sanskrit literature only by the use of Buddhist technical terms. Likewise the works of the logicians and philosophers follows the style of similar orthodox works in Sanskrit, with the addition of the terminology peculiar to the Buddhists. The Jains resisted longest the use of Sanskrit, and only began to take to it in the second half of the first millennium a. d. During this period Prakrit only gradually gives way to Sanskrit, but in the end Sanskrit establishes itself here as elsewhere . The Sanskrit of the Jains is influenced by the language of the earlier Prakrit literature in the same way as the Sanskrit of the Buddhists. In vocabulary it draws more extensively than con- temporary classical Sanskrit on vernacular sources, and words familiar later in Modem Indo-Aryan are often first recorded here. § 8 . Sanskrit in Greater India The expansion of Indo-Aryan was halted in South India by the native Dravidian languages (Tamil, Telugu, Kanarese) which in course of time established themselves as literary lan- guages. Nevertheless the influence of Indo-Aryan in this region was at all times powerful, and it is evident in the vocabulary of these languages from the earliest records. They were earliest influenced by Prakrit, which was the administrative language of the Satavahanas and their immediate successors. Inscrip- tions extending as far south as KancI show that all the Telugu- Kanarese area was governed by Aryan dynasties whose mother tongue was Prakrit. The intruding Indo- Aryans were not numerous enough to impose Indo-Aryan as the spoken language of the area and after about a.d. 400 the Prakrit inscriptions cease. Sanskrit replaced Prakrit, as elsewhere, for purposes of administration and culture, and as a spoken language it was replaced by the native Dravidian. At the same time the native Dravidian began to be cultivated, Kanarese from c . a.d. 450 and Telugu from c . a.d, 650. The Prakrit influence in these languages, dating from the earlier period, is rapidly overlaid by extensive borrowings from the Sanskrit vocabulary. In their early classical form these languages draw on Sanskrit wholesale, and the process was continued in the succeeding periods. At OUTLINES OF THE HISTORY OF SANSKRIT 63 the present time a considerable and essential part of the vocabulary of these languages is Sanskrit. In the Tamil country of the extreme South Indo-Aryan influ- ence was weakest. Tamil was the earliest Dravidian language to be used for literary purposes, and it was to begin with com- paratively free from Aryan influence. In the later period the influence of Sanskrit increases, but never on the scale that is found in its two northern neighbours. Ceylon received its Aryan language through colonisation from Northern India. In addition Buddhism established Pali as a literary language. At a later period still the cultivation of San- skrit was introduced, at some periods on quite an extensive scale. The spread of Buddhism was responsible for the introduction of Indo-Aryan linguistic influence into large regions of Central Asia. At one time a form of Prakrit served as the administra- tive language of the kingdom of Kroraina in Chinese Turkestan. Buddhist Sanskrit texts were current over a wide area, and works long lost in India have been recovered in recent years in Central Asia. Under Buddhist influence the native languages of this area began to be cultivated, notably the Iranian Khotanese, and the two closely related Indo-European languages which go by the name of Tocharian. The vocabulary of these draws abundantly on Sanskrit or Prakrit sources. On the other hand Tibetan which became Buddhist from the seventh century on- wards resisted foreign linguistic influence, and by what must have been a considerable tour de force , the whole Buddhist vocabulary was rendered into native Tibetan. This had been done at an earlier period by the Chinese where differences of script and language rendered any other course impracticable. The influence of Sanskrit was equally extensive in countries to the East and South-East. In Burma there is early evidence of the influence of Sanskrit Buddhism. This was replaced (a.d. eleventh cent.) by a religious reformation which estab- lished Theravada Buddhism as the official religion and with it Pali as the language of religion. Further East there were Hindu colonies in South Siam (Dvaravatl), Cambodia (Kambuja) and Annam (Campa). Abundant Sanskrit inscriptions dating from the third century A.D. onwards remain to show the importance of Sanskrit in these areas, and its influence was felt on the native languages when they came to be cultivated. Even 64 OUTLINES OF THE HISTORY OF SANSKRIT today Siamese is drawing on Sanskrit for its technical vocabu- lary. At the same time Hindu culture spread to Indonesia and in Java, Sumatra and Bali Sanskrit literature was cultivated. The native languages came strongly under the influence of Sanskrit and Sanskrit culture-words remain widely current in the area today. The classical language of Java abounds in San- skrit words, just as its literature draws its inspiration from San- skrit models. The Mahommedan conquest of Java (a.d. six- teenth cent.) put an end to Hindu dominion in thearea, but the in- fluence of the preceding centuries was too deep to be eradicated. §9. Writing in India The art of writing was late in making its appearance in Aryan India. It had existed before the Aryan invasion in the Indus civilisation, but it perished along with this civilisation. During the period when the Vedic civilisation was being built up no form of writing was employed in India, and in its absence the technique was evolved of preserving intact the Vedic literature by means of oral tradition. Even when writing was introduced this oral tradition persisted in the various departments of know- ledge and it continued as a basic feature of Indian education and culture down to modem times. It is not known when the alphabet was first introduced into India. So far as preserved records go it is only attested from the third century b.c. when the two alphabets, Kharosthi and Brahmi, appear fully developed in the Asokan inscriptions. The Kharosthi alphabet, which is written from right to left, is confined to the extreme North-West of India, to that part of the country which in preceding centuries had been part of the Persian dominions. It is an adaptation of the Aramaic alphabet which was employed in this region in the Achaemenid period, and it w T as probably evolved towards the close of this period. It continued in use in the same area, and in some adjoining parts of central Asia. down to the fourth century a.d., after which records in it cease. The Brahmi alphabet, which differs from Kharosthi in being written from left to right, is the source of all later Indian alphabets, as well as of those in countries abroad which formed part of the area of Indian cultural expansion (Burmese, Siamese, Javanese, etc.). It is also derived from some form of the OUTLINES OF THE HISTORY OF SANSKRIT 65 Semitic alphabet, but the exact source from which it is adapted and also the period remain uncertain. It is suggested that it may have been introduced from the South Semitic area by means of the trade routes to the ports of Western India, and the period most commonly assumed is about 500 b.c. The work of adaptation was considerable since it involved not only the addition of vowel signs, but also the changes and additions necessary to express adequately the Indian conson- antal system. The perfection with which the task was accom- plished was consequent on the labours of the ancient Indian phoneticians whose achievements have already been mentioned. In spite of this, use of writing was only slowly adopted in the Brahmin schools, and in the early period its function lay primarily in business and administration and only secondarily as an instrument of literature. For this reason all the earliest records preserved are in Prakrit, and Sanskrit documents only appear later. It is unlikely that much literature existed in manuscript form before the second century b.c. The early Brahm! alphabet was comparatively uniform and served for the whole of India outside the small area where Kharosthi was in use. After the Christian era local variations were intensified and Brahmi developed into a variety of regional alphabets differing from each other as much as they had changed from the original form. The structural principles of the alphabets always remained the same but the individual shapes of the letters W'ere subject to endless variation. In North India the alphabet gradually evolved into what is now known as Devanagari. With the introduction of printing this alphabet was adopted generally for Sanskrit, but before this period Sanskrit manuscripts were written in the various regional alphabets of the localities where they w^ere produced, e.g. Sarada in Kashmir, Bengali, Oriya, Telugu-Kanarese, Malay- alam and, in the Tamil country, Grantha. The commonest material used for writing in India was palm- leaf. The exclusive use of this prevailed in South India down to modern times. The characters w r ere incised on this material by means of a stylus and the ink rubbed in afterwards. In the North, particularly in Kashmir, the inner bark of the birch was used on which the letters were WTitten in ink. This method was also used in the North for palm-leaf manuscripts, and the dif- ferences between the Northern and Southern alphabets is largely 66 OUTLINES OF THE HISTORY OF SANSKRIT occasioned by different methods of writing. As a result of the perishable nature of these materials really ancient Indian manuscripts are rare. The oldest are those that have been discovered, in a more or less fragmentary condition, in the dry soil of Central Asia* The bulk of Sanskrit literature is preserved only in manuscripts belonging to the last few centuries. CHAPTER III PHONOLOGY §i. Indo-European Consonant System The comparative study of the phonetic systems of the existing IE languages makes it possible to reconstruct, with a reasonable degree of certainty, the phonetics of the parent language* On this basis a systematic historical account of the Sanskrit phonetic system can be provided in which the various stages of development in the prehistoric period can be distinguished in respect of their relative chronology* Developments may be severally characterised as : (i) Changes in the Indo-European period ; (2) Changes common to Indo-Aryan and Iranian only ; (3) Changes peculiar to Indo-Aryan, which have occurred after its separation from Iranian. In sketching the phonetic development of Sanskrit we shall indicate, as far as possible to which of these three periods the various changes belong. The following reconstruction of the IE consonantal system has been generally adopted by comparative philologists ; Surd Sonant Surd Aspirate Sonant Aspirate Labio-velar k w h g* g w h Occlusives Velar k kh g gh (Stops, * Palatal k kh A g i h Plosives) Dental t ih d dh .Labial P ph b bh Nasals : m , n , y ; Liquids : l, r ; Semivowels : y, v Sibilants : 5, z ; Doubtful p: , d. The reconstructions are of two kinds. In the first and com- monest case the phoneme postulated for Indo-European occurs in a number of the existing languages in which it has continued unchanged ; in the second and rarer case the phoneme assumed for Indo-European is nowhere preserved as such, but it is deduced by comparison of the forms derived from it. Naturally 67 68 PHONOLOGY there is the greatest certainty in the case of the first class, but even the pure reconstructions of the second class are, with few exceptions, established beyond reasonable doubt. In the following cases an Indo-European consonant is pre- served unchanged in Sanskrit and in other languages : p : pane a * 5 * * Gk. irevre ; pdtati 4 flies ’ : Gk. ttctctcu ; dpa 4 away, from 1 : Gk. airo : sdrpati ' crawls ' : Gk. epnet Lat. serpit. t: tanu- ' thin', Gk. raw Lat. tenuis ; trdyas 4 3 J , Gk. rp€lg t Lat. ires ; variate ‘ turns ’ ; Lat. verto . d : ‘ 10 Gk. SeVa, Lat. decern ; dirghd- 4 long ' ; Gk. SoAi^ds, O. SI, dlugu ; ‘ I know Gk. otS a, Lat. video . k : kravis - ' raw flesh Gk. Kpeas ; kdksa- 1 armpit cf. Lat. Co##. g : ytigdrn 4 yoke Gk. Lat. iugum ; ' to cover Gk. ordycu. n : 4 name Lat. nomen ; ndva- 4 new Gk. veos, Lat. novus ; ndbhas ' cloud *, Gk. v^o? ; ddnam 4 gift Lat. donum , O. SI. tfaiift. m : mat dr- 4 mother ', Lat. ma/er ; ma me’, Lat. me ; ' mouse V Lat. mils, O. SI. my si ; ddma- 4 house Gk. Sopog, Lat. domus. I : lubh-, lubhyati ' desire, covet Lat. hibet, Goth, liu/s, 0. SI. ljubu 4 dear laghu- 4 light, swift Gk. iXaxvs, Lat. levis. r : rudhira- 4 red, blood Gk. ipvdpos, Lat. ruber r&j-, rajan- 4 king Lat. rex , Gallic °rix ; bhdrati 4 bears Gk, <f>4pw, Lat . fero, Goth, baira . y : yuyan- 4 young man Lat. iuvenis ; ydkrt 4 liver Lat. iecwr ; yus- 4 broth, soup Lat. ids , 0. SI. jacha. w (t/) : vac- 4 speech Lat. vox ; vdhati 4 carries V Lat. vehil ; ndva- 4 new Lat. novus ; dvi- 4 sheep ’, Lat. ovis. s : sdna- 4 old ’, Lat. senex , Ir. sen ; sdnti 4 they are’, Lat. sunt ; sunu- 4 son ’, Lith. sunus, Goth, sunus ; dmsa- 'shoulder’, Goth, ams ; dsthi 4 bone ’,.Gk. oariov, Lat. os, ossis. In cases like the above the reconstruction of the IE forms presents a minimum of problems ; reconstruction in the full sense is not necessary since the phonemes in question are widely preserved. They are not preserved in all languages (e.g. Engl, thin : Lat. tenuis ; Welsh hen, Ir. sen), but a study of all the available evidence leaves little doubt as to which languages PHONOLOGY 69 preserve the original sound. In other cases change has been more widespread. There are instances where the original IE sound is preserved only in one language, others in which the sound, which theory demands for the parent language, is pre- served nowhere at all. Even in these cases it is possible to fix the original sound with reasonable certainty. §2. The Sonant Aspirates The sonant aspirates which it is normally believed Indo- European possessed are preserved as a class by Sanskrit alone. Elsewhere they are changed in various ways ; in Iranian, Slavonic, etc., the aspiration is lost ; in Greek they are changed into the corresponding surd aspirates, in Latin (and the other Italic dialects) into fricatives. Examples of this series are as follows : bh : Skt. bhru - 4 brow Gk. oppvs, 0. SI. bruvi ; bhr&ta

• brother V Gk. <f>pdrr]p 'member of a phratry Lat . f rater,

O. SI. brairu , 0. Ir. brdthir ; bhdrati ' bears Av. baraiti, Gk. <f>epaj, Lat. fero, Arm. berem, O, SI. berg, Goth, baira , O. Ir. berim ; ndbhas ‘ cloud, sky Gk. vdpos, O. SI. nebo, Hitt.

• — ► nepis.

dh dhd~, dddhdti * to place Av. daSditi , Gk. Wflij/u, Lith. dcti ; dhumd - ‘ smoke 0. SI. dymu, Lat. fumus ; mddhu ‘ honey, mead Av. madu, Gk. fieOv, O. SI. medu, A.S. medu, O. Ir. mid ; vidhdva ( widow cf. Gk. rjldeos f young (unmar- ried) man O. SI. vtdova f widow Lat. vidua , O. Ir. fedb. gh : stigh- f to stride J , Gk. a tgixoj, Goth, steiga ; meghd - ' cloud cf. Gk. ofLix^V’ O. SI. migla , Alb. mjegule. In the case of the guttural series the sonant aspirates have undergone changes in Sanskrit in common with the other members of the series. These will be detailed below. Although the sonant aspirates are preserved in Indo-Aryan alone among the IE languages, there is. little doubt- that they should be attributed to the parent language, since no other type of phoneme can account so simply for the various develop- ments that appear. The theory, prevalent in some quarters, that in these cases we are dealing with a series of original IE fricatives, has nothing to recommend it. Although Sanskrit preserves the sonant aspirates as a class, it does not preserve them all unchanged. The special developPHONOLOGY 7 0 ments of the sonant aspirates belonging to the two guttural series will be treated below. The dental and labial sonant aspirates are normally preserved as in the examples given above, but in some cases, even in the earliest period dh and bh are weakened to h, an anticipation of their later fate in Middle Indo-Aryan. dh : hitd- 4 placed ' (dhd-) } also -dhita- in the Veda ; -hi, termination of the 2 sg. impv., also -dhi, Av. -hi ; the verbal terminations of the 1st dual and plural atmanepada, - vahe , -vahi, - vahai ; -make, - mahi , -mahai, ci. Av. -maide, -maidi ; ihd 4 here Pa. idha, Av. iha ; sahd 4 with *, Vedic also sadha- in cpds., Av. haha ; lohita rdhita- 4 red cf. rudhird- ; rdhati 4 climbs Vedic also rddhati ; nah- 4 to bind cf. ppt. naddha - ; snuh- 4 to drip cf. Av. snaod bh : grab- 4 to seize Vedic also grabh- ; kakuhd- 4 high # be- side kakubhd - 4 id kakubh- 4 peak An aspirate was not allowed to remain in Sanskrit when an aspirate followed. The effect of this rule in grammar is seen in reduplication where the corresponding unaspirated sonant is used — dhd- f dadhau , bhd- t babhau f han - : jaghdna. The same rule is observable in Greek (dvyvKU) : reOvrjKa) ; it is not how- ever an Indo-European feature inherited in common, but a phenomenon that has occurred independently in each language. In Greek this de-aspiration did not take place until the sonant aspirates had been turned into surds, and consequently the unaspirated surd is the result. We find therefore in these cases an initial surd of Greek corresponding to an initial sonant of Sanskrit. Thus the IE root bheudh- 4 to perceive ' produces on the one hand Gk. Trevdoum (through *(f>€v 0 -) and on the other hand Skt. budh The same correspondence is seen between Gk. retxoy 4 wall * and Skt. dih- ' smear, cement with earth deht 4 rampart * ; similarly Gk. v evdtpos 4 father-in-law 4 : cf. Skt. bandhu- 4 relation bandh- 4 to bind Gk. 7 rrjxvs 4 arm ' : Skt. bdhu-, Gk. 7 TvOfjLTjv 'bottom' : Skt. budhnd -. In other Indo- European languages no dissimilation of this kind took place, and whatever phoneme corresponds regularly to an IE sonant aspirate appears also in this position ; e.g. from IE dheigh- (Skt. dih-) Goth, deigan 4 knead Lat. Jingo , Osc, Jeihuss 4 muros When a sonant aspirate came immediately before final s or t the aspiration was lost at an early period, e.g. in Skt. ddhok , 2 and 3 sg, impf. of duh - 4 to milk 1 for earlier *ddhoks t *adhokt } PHONOLOGY 71 ddhdk , 3 sg. s-aor. of daA- ' to bum for *adhak£t ; likewise in the nom. sg. of root stems °dhuk 4 milking*, °dhruk 'injuring' from °dhuk$, °dkruks. In their earlier form these roots had two aspirates, *dhugh-, *dhagh-, *dhrugh-. In the above combina- tions the second aspirate was lost before the rule of aspirate dissimilation set in, and consequently the aspiration of the initial consonant was preserved in these cases. An exactly parallel development took place in Greek : QpL£ 4 hair pL rplx€s, rpipw

• I nourish fut. Opeipco. From the standpoint of the individual

languages it appears that in such cases the aspirate of the final is thrown back, but it becomes clear from comparative study that there were originally two aspirates of which the first was preserved from dissimilation in these circumstances. The treatment of internal combinations of sonant aspirate + t or s was different, and will be described below (§ 14 , 16 ). Here we may note only that the aspiration in class, dhipsati , etc., as opposed to Ved. dtpsati is an innovation due to analogy. §3. The Surd Aspirates The surd aspirates appear only in Indo-Iranian with any frequency, and even there they are much less common than the corresponding sonants. Of these very few indeed can be estab- lished as Indo-European by direct comparison. Such com- parisons are available only in Greek where there appear a few aspirates corresponding to the surd aspirates of Sanskrit, and in Armenian. Elsewhere, apart from a few traces in Latin (e.g. funda 4 sling ' : Gk. <j<f>tv&6vr)) this series has been con- founded with that of the unaspirated surds. The following examples will serve for illustration : ph : Skt. sphiirj Gk. crpapaydopai ; sphyd - ' wooden ladle cf. Gk. oprjv ; phala- 4 ploughshare Pers. supar ; sphurdti 4 throbs, quivers Lith. spiriu, Gk, mrai pw, Lat. sperno, cf. also Gk. < 7 <f>aLpa 4 ball ' ; phdna- 4 foam Osset, fing, O. SI. pena, 0. Pruss. spoayno ; iaphd- ‘ hoof Aw safa th : vdttha * thou knowest Gk. olaQa ; stha~ ' stand Gk. iorrjfu ] sthag - 1 to cover Gk. areyto ; rat ha- 4 chariot *, Av, raOa cf. Lat. rota 4 wheel 1 ; prthu - 4 broad Av. pn^Qu- t Gk. irXaTvs, without aspiration, but cf. TrXddavos 4 flat board for making cakes 1 ; prthu ka- 4 young animal Arm. orth 4 calf Gk. iropris , 7Toprai ' heifer without aspiration, but cf. na pOevos 4 maiden ' with aspiration. 72 PHONOLOGY kh : sahkhd- ' shell Gk. Kvyxos ; iAkhd r branch Arm. tax t Lith. sakd ; khA- ' well Av. xL In contradistinction to the sonant aspirates which constitute an ancient and primary element of the IE phonetic system, the surd aspirates are a late creation, and recent advances in IE theory have made it possible to show how they came into existence. Apart from some possible cases of spontaneous aspiration in combinations with s (Skt. sthag-, etc.), their origin can be attributed to a combination of IE h with a pre- ceding unaspirated surd. The nature of this h which has lately figured largely in IE studies will be examined below. For the present it will be sufficient to remark that it has disappeared in all IE languages except Hittite, but when immediately pre- ceded by p t t, k, it has had the effect of aspirating these con- sonants. Thus the aspiration which in Skt. sthd- * stand ’ (IE sta, i.e. staH-, i.e. ste h 2 ) arises in the first instance from the reduced form of the root, IE srii- (Skt. tistha<titi-stn-eti) t and is generalised from there. In Greek on the other hand (urrrjfu) the unaspirated t proper to the strong form of the root has pre- vailed. In Sanskrit path - ‘road’ (Instr. pathA , etc.) <patn~, H is a suffix which appears in its guna grade as a (<aii) in nom. s. panthds. In the latter case th appears only by analogy from the weak cases, as is confirmed by the Av. form panta (original alternation pantaii-patn-). The root prath- 1 to ex- tend ' and its derivatives [prthu- ‘ broad etc.) contains two incorporated suffixes (IE pl-et- H-), the combination producing the surd aspirate in Sanskrit. This h- suffix, without the inter- vening t- suffix appears in Hitt, palhis * broad ", and in Lat. planus . In Gk. TrXarvs , no k- suffix is present, and there is therefore no aspiration. In Skt. rdtha- f chariot 1 we have a nominalised adjective ( *ratii-d ‘ wheeled 1 with a common change of accent) based on the d (-ah) stem which appears in Lat. rota ‘ wheel * (rotan). A corresponding aspiration of sonants by h is possibly a factor to be considered, but not many examples have been found. Such an instance may appear in Skt. smdhu - ' river ' as compared with the root syand - ‘ to flow ", where the form of the noun may be easily explained by the presence of an h- suffix before the final u- suffix. It is believed that one type of IE h (h 3 ) affected a preceding surd differently, by voicing it, in Skt. pibati = 0. Ir. ibid PHONOLOGY 73 • 4 drinks ' ( sfpo -, i.e. pen it pres. *pi-pn^-eti) ) but this appears to be the only example. Incidentally it may be observed that b, apart from a case like this, or as an allophone of p (Skt. upabdd cf. pad- 4 foot etc.) was of extreme rarity, if not altogether absent in Indo-European. §4. The First Palatalisation When we come to the so-called guttural series (Labio- velar, Velar, Palatal, see above), we find that these sounds have been very considerably altered in Sanskrit. To begin with the Palatal series, the treatment is illustrated by the following equations : k : Skt. ivdn - 4 dog Av. span - ; Gk. kv cov, Lat. cams ; iatdm 4 100 Av. sat&n, Lith. fimtas : Gk. exarov, Lat. centum , O, Ir. cet f Goth, hand ; idle 4 lies ' : Gk: Keircu ; idnkate ' hesitates, doubts ' : Lat. cunctor ; vis - 1 clan, settle- ment Av. vis-, O. Pers. vid-, O. SI. vist 4 village Alb. vis 4 place 1 : cf. Gk. ohtos 4 house Lat. vicus ; sroni- 4 buttock Av. sraoni-, Lith. flaunls : Lat. clunis , 0. Ic. hlaun ; £ru- 4 to hear irdvas - 4 fame *, Av. sru- t sravah O. SI, slovo 4 word ' : Gk. *Aeoy, kAuw, Lat. clueor, etc. ; daddrsa 3 rd sg. pf. 4 he saw ' : Gk. h&opKt ; did - 4 to point out ' : Gk. belKWfju, Lat. dico, etc. g: Skt. jdnu 4 knee 4 , Av. zanu ; Gk. yow, Lat. genu: jar ant- 4 old Arm. cer 4 old man ' : Gk. yepwv ; jdmbha- 4 tooth 0. SI. ZQbu , Alb. Semp 4 id ' : Gk, yoppos 4 nail Engl, comb , Toch. A kam, B keme 4 tooth ’ ; jus- 4 to enjoy jdsa- 4 enjoyment Av. zaosa - 4 id ' : Gk. yevopuai 4 taste Lat. gustus , Goth, kiusan- 4 chose 4 ; rajata- 4 silver Av. wvzata-, Gk. apyvpos , Lat. argentum , Ir. arga/ 4 id ' ; 4 drives Av. azaiti ; Gk, ayou, Lat. ago ; bhdrja- 4 birch Lith. bdrzas, O. SI. &reza : AS. fi^orc. gh ; Skt. himd- 4 snow hemantd - 4 winter Av. zima-, Arm. 4 snow Alb. dimm , Lith. imi, 0. SI. ma 4 winter' : Gk. Lat. hiems , 0. Ir. gim-red 4 winter 4 ; amhu- 4 narrow dmhas- 4 distress, difficulty 0. SI. azw&w 4 narrow ' : Gk. ayx^ ‘ throttle Lat. ango, angustus, O. Ir. cum-ung , Goth, aggzfltts 4 narrow 4 ; lih- 4 to lick Arm. lizum , Lith. /ezrii, O. SI. /tzari : Gk. Aei'xto, Lat. /mgo, O. Ir. ligitn, Goth. bi-laigon 4 to lick '. 74 PHONOLOGY It will be observed that ultimate Indo-Aryan and Iranian developments differ from each other (Skt. £ f j, h — Av. s, z, z). This is because the changes that have occurred have taken place in two phases : (i) a previous common Indo-Iranian de- velopment £, z , zh, (2) the change of these to the actual forms in Indo-Aryan and Iranian after the separation of the two groups. The first change is common to Indo-Iranian and the languages of the satzm group (see Chapter I) and took place within the Indo-European period. Later there were various special de- velopments in other languages too, which in the case of Slavonic happen to be identical with those of Iranian. The intermediate forms s , i, zh will explain most of the Indo- Iranian developments, as they will the Balto-Slavonic, but it is likely that before complete assibilation there was an affricate stage. Evidence of this is preserved in the Kafiri dialects, which occupy an intermediate position between Indian and Iranian. The treatment that occurs in Kati due 1 10 ' and cut

• empty ' for instance (Skt. dasa, sunyd~) seems clearly to reflect

a form more ancient than that of Sanskrit. We may therefore postulate an earlier Indo-Iranian (and the same will apply to the satftn languages generally) series 6 , /, jh (or/', d d f h to be distinguished from 6 , J, jh of the second palatalisation). This being so it becomes possible that (1) Skt.y of this series, with its renunciation, has develop^Jirectly^ouFof j, through with the other affricate £ ; without a sibffant stage z f and (2) that Old Persian 9 f d which appear in place of s, z in the rest of Iranian ( dard - # year Av. sarzd-, Skt. iarad- ; dauStar -

1. friend Av. zaoSa- ' enjoyment Skt. jdsa-, jostdr- ; dasta -

' hand Av. za$ta- t Skt. hasta -) have developed directly out of such affricates and that there is therefore no common Iranian treatment. affricate p confusion §5. Treatment of the Labio velars The languages of the satzm group all agree in another feature, namely in the loss of the labial element in the IE series k w f g w , g w h. In the centum languages the series was to begin with pre- served, but later subject to various developments, of which the commonest is the substitution of pure labial occlusives. Leav- ing aside cases which have been affected by the second palatal- isation, the treatment of the labio-velars may be illustrated by the following examples : PHONOLOGY 75 k w : Skt. ka- 1 who ? 4 kim ' what 4 , Lith. kds, O, SL ku-to : Gk. TiV 4 who 7 to$€v 4 whence 4 , Lat. quis ; ydkrt 4 liver Av. ydkarp ; Gk. ^ap, Lat. tVtftr ; krtndtni ' I buy 4 , O. Russ* krinuti ; Gk. inplap^v 4 I bought ', Ir. crenim, Welsh prynaf ; kfmi- 4 worm Lith. kirmele : Ir. cruim, Welsh />ry/; rindkti leaves p f 4 left 4 , Arm. lk 4 anem 4 1 leave 4 he left Lith. lekii ‘ I leave 4 : Gk. AetTroj, Lat. linquo , etc. g w : Skt. gam- gacchati r go ’ : Gk. pdcrtca), Lat. Goth. qiman ; gtf- (nom. s. gatis) 4 cow 4 , Arm. kov, O. SI. gov§do : Gk. jSouff, Lat. £>es, O. Ir. bo ; Skt. gna 4 (divine) woman, wife Av. gznd, O. Pruss. genna , Arm. kin, pi. kanaik* : Gk. yvvrf, fiava, Ir. ben , Goth, qind ; guru- ‘heavy 4 Gk. fiapvs ; gwr- 4 to lift up, heave 4 , Av. gar- r to throw 4 : Gk. fldAXw ; grdvan- 4 pressing stone 4 : Ir. bro, gen. broon Welsh breuan 4 mill- stone g w h : Skt. jagkdna 4 slew 4 : O. Ir. geguin, cf. Gk. ihrepyov, povos ; arghd- * value, price 4 ; Lith, alga : Gk. aXpdvoj 'earn' ; gharmd - 4 heat Av. garpma -, O. Pruss. gorme : Gk. depfxos 4 hot 4 , Lat. formus ; laghu - 4 light, swift 4 : Gk. iX appos, Lat. levis. As can be seen from these examples the sat pm languages have uniformly abandoned all trace of the labial element in these constants, and this, in conjunction with their sharing in the first palatalisation, provides strong evidence that they are derived from a single dialect group within Indo-European. Both changes belong to the late Indo-European period and not to the separate evolution of the various languages of this group. The element w which is attached to these velars differs from full IE w in that it is not convertible to the vowel u when the succeeding vowel is elided : e.g. Skt. hdnti 4 slays 4 : ghnanti 4 they slay 4 from IE g w henti : g w hnonti ; similarly Gk. €? repvov, etc., as opposed to usual treatment of w as seen in such cases as svdpna- 4 sleep 4 : suptd - 4 asleep It is possible, however, that this applies only to the late Indo-European stage. There exist some survivals, particularly in Hittite, in which this element is seen to be vocalised, after the manner of w proper. Corresponding to the Sanskrit verbal forms above Hittite has 3 s. kuenzi ; 3 pi. kunanzi and corresponding to Skt. nakt -, Gk. wJf, Lat. nox 4 night 4 (IE *nok w t-) it has nekuz mehur 4 night time Other instances of this treatment are seen in Toch. A. kurydr PHONOLOGY 76 ' commerce * : Skt. krt- ' to buy 1 (IE *k w ri), in Toch. A. kukdl ' chariot Gk. kvkXos 1 wheel 1 : Skt. cakra-, Engl, wheel (+k w ek w los) and in Gk. yvmj ' woman 7 : Skt. gnd (*g^nd) r Examples like these seem to reflect an earlier sta ge of Ind o- European, "when the labial elemen t in connection with these velars was equivalent t o ordinary §6. The Pure Velar Series This series has been invented to account for those cases in which k, g t etc., of the centum - languages are not palatalised in the stf/^m-languages, and they are devoid of the labial element whose influence is so marked in the case of the labio-velars. Examples quoted are such as the following : k : Skt. kdksa- 4 armpit cf. Lat. coxa, OHG. hahsa ; Skt. kravis- * raw flesh Lith. kraiijas , O. SI. kruvl ' blood Gk. Kpeas, Lat. cm or ; Skt. krntdti ' cuts 1 (J~ker4~), Gk. Kelpw ; karkata - 4 crab Gk. KapKivos, Lat. cancer ; krndtti ' spins 7 , cf. Gk. 4 spin Lat. colus 4 distaff 7 ; ahkd - ' bending, lap cf. Lat. uncus , Gk. oyKos ; °kulva- ' bald Lat. calvus ; kavt- 4 a wise man Gk. ko4 <0 ' take notice of Lat. caveo . g ; Skt. sthag- ' to cover Gk. crrlyoj, Lat. lego ; tigmd- 'sharp', Gk. onypurj 'puncture 7 ; agas- 'guilt 7 , Gk. ayos ugrd- r strong 7 , Av. aogar 9 ; ' strength 7 , Lat. augeo, augustus . gh : stigh- 4 to stride Gk. orelx o> Goth, steiga ; dlrghd- ' long O. SL dlugu, Gk. SoAi^s-, Hitt, dalugas ; megha- ' cloud *, Lith. migla , Gk. The difficulty that arises from postulating a third series in the parent language, is that no more than two series (Lat. quis : cams t Skt. kds : svam) are found in any of the existing lan- guages, In view of this it is exceedingly doubtful whet her three distinct series existed in Indo-Europeam The assumption of the third series has been a convenienceior the theoreticians, but it is unlikely to correspond to historical fact. Furthermore, on examination, this assumption does not turn out to be as convenient as would be wished. While it accounts in a way for correspondences like the above which otherwise would appear irregular, it still leaves over a considerable number of forms in the satem-languages which do not fit into the framework. Such are Skt. klam- ' to be tired 7 beside it am- ' id cf, Gk. tcXapiapos 1 weak, slack and ruc-jruk - 4 to shine 7 beside rusdnt ' bright 7 , cf. Gk. Ac vkq$ ' white etc. Examples of this kind PHONOLOGY 77 are particularly common in the Balto-Slavonic languages : Lith, akm& ' stone Skt. dsman cf. Gk. cuv ; klausyti ' to hear ', cf. Skt. srdsamdna-, Toch. A. klycs - ; Lith. pekus, O. Pruss. peckn , Skt. pdsu ' domestic animal Lat. pecu ; O, SI. svekru * father-in-law Skt. svdsura Lat. socer ; 0. SL creda * collection, herd Skt. sdrdha - ; Lith. smakra ‘ chin Alb. mjekre, Skt. stndsru - ' beard Ir, smech 1 chin Clearly a theory which leaves almost as many irregularities as it clears away is not very soundly established, and since these cases have to be explained as examples of dialect mixture in early Indo- European, it would appear simplest to apply the same theory to the rest. The case for this is particularly strong when we remember that when false etymologies are removed, 1 when allowance is made for suffix alternation, and when the possi- bility of loss of labialisation in the vicinity of the vowel u is considered (e.g. kravis-, ugrd not many examples remain for the foundation of the theory. § 7 . The Second Palatalisation After the completion of the changes characteristic of the satzm - languages the parent dialect of Indo-Iranian possessed the two series k, g, gh and s, z, zh (or 6, /, jh ). The latter re- mained essentially unchanged till the end of the Indo-Iranian period. The former underwent the following alteration. Be- fore the vowels $ (later changed to a in Indo-Iranian) and t and before the semivowel y, k, g and gh developed respectively into the affricates c, / and jh, of which the last was later altered into Iranian / and Sanskrit h respectively. k : Skt. ca ‘ and Gk. re, Lat. que ; catvdras ' four 0. SL cetyre , Lith. keturi , Gk. reacrapes, Lat. quattuor sdcate ' assoc- iates with', Av. haiaiti , Gk. h rercu 'follows', Lat. seqiiitur, Ir. sechithir ; panca 1 five Lith. penki, Gk. 77cWe, Lat. quin- que ; pdcati ‘ cooks O. SI. pecetu , but 1 sg. pekq ; cakrd - ' wheel A.S. hweohl, cf. Gk. kvkXos, Toch. A. kukdl ; cdru- ' a particular vessel A.S. hwer 1 kettle cf. O. Ir. coire , Welsh pair 'id.' g : Skt. jivd- ' alive, life jfvati ' lives Av. jfvaiti, O. Sl. zivu ' alive L Lat. 1 nvus, Gk. fiios ; jyd ' bowstring Lith. gijd 1 For instance Skt, kola- ' black ' is from Dravidian and not connected with Lat. caligd, etc. ; the Aryan root hayi - (Skt. hanya ' girl ‘ , etc.) cannot be con- nected with Gk. Acatvoj ' new because its primary meaning is quite certainly

• little ' and not ‘ new PHONOLOGY

78

• thread Gk. f}i6s ( bow * ; jdtu ' gum, lac A.S. cwidu , cf.

Lat. bitumen ; °jdni- ‘ wife ', Goth, qens ; rajant ‘ night Gk, cpefAvos, ipefiewos * dark gh : $kt. hdnti ‘ slays Av. jainti, as opposed to Skt. jaghdna 1 slew 0. Ir. geguin , Hitt, kuenzi r slays ' ; arhati * is worth Av. arpjaiti, as opposed to Skt. arghd- * price Lith. alga ‘ re- ward Gk. aX<f>dva) ; haras - ‘ heat Gk. Qepos * summer ' (: gharmd-, etc. above) ; ddhati 1 bums Av. dazaiti : nidaghd- ' heat of summer Lith. degu * I burn The second palatalisation took place fairly early in the de- velopment of Indo-Iranian, before the change of l to a which distinguishes this branch from the rest of Indo-European. Similar changes appear in some other languages of the satam group, e.g. Slavonic (detyre 1 4 zivu * alive ') and Armenian (jerm ' warm ') but these appear to have occurred independently and later. §8. The Two Palatal Series in Indo-Aryan In Indo-Aryan the distinction between the two palatal series, which is fully preserved in Iranian (s, z, z : £, /, ]) is retained only in the case of the surds (s' : c). On the other hand the sonants, both unaspirated and aspirated, are confused with each other as j ( — z and/) and h ( = zh and jh) respectively. But the distinction between the two remains effective in many ways in the grammatical system, because according to their origin both j and h are treated in two different ways in various con- texts. In declension and inflection the rules of sandhi operate differently according to the different origins of j and h . This may be illustrated from the formation of the participle in -ta from the two types of root respectively. / : (a) yaj- ‘ sacrifice ' (Av. yaz-) : is id-, srj- ‘ to let go * (Av. ham-) : srstd mrj - * to wipe J (Av. mar^z-) : mrstd-. (b) nij- to wash p (Av. nae]-) : niktd - ; bhaj- ' to distribute * (Av. ba]-) ; bhaftta- ; yuj- ' to join ’ (Av. yao]-) : yuktd -. h : (a) vah- 1 to carry * (Av. vaz-) : udhd-, lih- ' to lick * (Av. raez-) : Udhd- ; sah- ' to overcome ' (Av. haz-) : sddha-, sodhd (b) dak- ‘ to burn ' (Ir. da]-, Av. dazaiti) : dagdhd -, druh - ' to injure, betray ' (Av. dru]-) : drugdhd-. In this, as in many other respects the distinction between the two series remains active in Sanskrit grammar. In the case of the second palatalisation, as opposed to the PHONOLOGY 79 first which operated in all conditions, there exists an alternation in the roots affected between palatalised and non-palatalised forms, depending on whether the vowel following was origin- ally I or o. This is seen in perfects like jaghina * slew ' and jigdya ‘ conquered ', as contrasted with the present tense forms hdnti and jdyati. Similar alternation is not permissible in the case of roots whose j and h belong to the first palatal series, e.g. jajana 1 begat A (jan- : Av. zan-) and juhava * called ' (hu : Av. zav-). To begin with the distribution of palatalised and non- palatalised foijns must have depended entirely on the nature of the succeeding vowel, and consequently an alternation be- tween the two must have been active in the paradigms of noun and verb. In the parallel palatalisation of Slavonic such alter- nation exists, e.g. between vluku 4 wolf ’ nom. s. and vlule voc. s., and between pekg ( I cook' and pecetu 'he cooks'. Since the natural tendency of linguistic evolution is to smooth out such irregularities (as is done later in Russian, etc.) it is likely that the Slavonic palatalisation did not long precede the beginning of the literary tradition. On the other hand in the case of Indo-Iranian the change had taken place early enough for the working of analogy to become widely effective. Varia- tion in the paradigms of noun and verb after the Slavonic style has been eliminated, except as between vocalic and consonantal suffixes (loc. s. vact, loc. pi. vdksu). On the other hand the alter- nation remains active between different nominal derivatives (i bhoga - : bhoja -, etc.). At the beginning of a root alternation between palatal and guttural remains active only in the case of a few roots as those quoted above. Mostly it is eliminated, and in this respect Sanskrit shows a greater tendency to innovation than Old Iranian ; cf. Skt, akar 3rd sg. root aor., Av. lord{ (early Aryan acarKekert ), and agamat, a-aor. Av. In the latter case the proper name Jamadagni- (' who goes to the fire ') preserves the earlier, pre-Vedic form. In the reduplica- tion of the perfect, etc., the alternation always remains, based on the fact that the vowel of the reduplicating syllable was originally e (jagdma<g v, eg u, dme J etc.). There are a few instances in Sanskrit where j of the older palatal series alternates with g in the formation of nominal derivatives, e.g. sdrga- 'emission' (srj-, Av. ham-) ' yaga-

• sacrifice ' (yaj- t Av. yaz -). The guttural here cannot be 8 o

PHONOLOGY original, and it is due to the analogy of the palatals of the later series operating after the two had fallen together in pronunciation* §9, Developments of s in Indo-Iranian and Slavonic In Sanskrit it is the rule that the dental sibilant must be re- placed by the cerebral after k, r, r, i and u. By a similar rule in Iranian $ (>Ir. h) is replaced by s. Further in the Slavonic languages s is usually replaced by ch in these conditions, and this ch represents an earlier s. In Lithuanian a similar change is found, but only after r . Examples are : After k ; (Skt. ks, Ir. xs, Slav* ch<kx<ks) Skt. vaksyami, Av. vaxiyd ; Skt* ksudrd- 4 small cf. O. SI. chudu ' id 1 ; Skt. araiksam , s-aor. of ric- 1 to leave ' (IE leik w -), Gk. eAen/ra ; cf. O. SI. techu, s-aor. of tekq ' I run After k : Skt. s : (Skt. ks, Ir. s, Slav, s), Skt. dksa- 4 axle Av. asa -, O. SI. osi, Lith, asis, Gk. agcov, Lat. axis ; Skt. daksina- ' right (hand) Av. datina-, O. SI. desinu , Lith. desinl, Gk. Sexto's 1 , Lat. dexter, etc, ; Skt. taks - ‘ to construct in wood (as a carpenter), Av. tas-, O. SI. tesati, Lith. tasj/ti, Gk. rexvr) ' art ' ( *Uksnd ), tIktwv ' carpenter ' ( ! Hdkstdn ), Hitt. taks- 4 to join ", Lat. texo 4 weave ’ ; Skt, maksu 4 quickly ', Av. mosa, Lat. mox. After r (and r) : Skt. vdrsman - 'summit vdrstyas - ' higher O. SL vrichu 4 summit Lith. virsus ' upper part ’ ; mrs- 4 to overlook, forgive ', marsa - ‘ forbearance ', Lith. mirsti 4 to forget marsas ‘ forgetfulness Toch. A. mars- 4 to forget ' ; dhrsnoti 4 dares O. Pers. adarsnaus 4 he dared cf. Gk. Qapvos f boldness ', Goth, gadars 4 dare ’, Engl, durst, etc. After i : Skt. pis- ‘ to pound 0. Slav, pichati 4 to knock, strike Lat. pinso ; trisu , loc. pi. of iri- 4 three O. SL trichu ; asvesu, loc. pi. of diva- 4 horse ", Av, aspaesu, cf. O. SL vlucechu lyluku 4 wolf ') ; visa- e poison Av. visa-, Lat. virus, Gk. toy. After u : jus - 4 to enjoy josa- 4 enjoyment Av. zaos-, cf. Lat, gusius, etc, ; mils- 4 mouse O. SL my si, Lat. mus, etc. ; suska- 'dry', sosa - 'drying up 1 , Av. huska - ‘dry’, O. Sl. suchu, Lith. sausas, Gk. avos ‘ id As can be seen, the parallelism between Indo-Iranian and Slavonic is not absolutely complete, because they differ in the treatment of that s which follows IE k (>Skt, s, Slav. 5 ). But in all other respects they agree, and the correspondence is too PHONOLOGY 81 close for such changes to have taken place independently. In Lithuanian the same kind of development is observed, but only in connection with r, which points to a remoter contact than was the case between Indo-Iranian and Slavonic, Cerebralisation of $ does not take place in Sanskrit when r or r immediately follows : e.g. visra- 4 bad-smelling (meat) cf. visa-, etc., Av. vaesa- ‘ corruption ’ ; list as, tisfbhis, tisfn&m , from tri- 4 three 7 ; gen. s. usrds from usar - 4 dawn sisrate from sar - 4 to go In Avestan there is no such restriction, e.g, tisrd, nom. pi. fem. cf. dri- ‘ three §10. The So-Called Mobile s Indo-European s when it formed the first member of an initial consonant group, was an unstable sound, and Hable to disappear under conditions which it has not been possible accurately to define. Forms with and without s are found side by side in the various languages, as illustrated by the following examples : Skt, tdnyati 4 thunders 7 , Lat. tondre : Skt, stanayitnu - 4 thunder cf. Gk. arevoj, 0. SI. stenjg , etc. ; Skt. tdyu - 4 thief O. SI. tatu 4 id Gk. r^raco, Hitt, tdya- 'steal': Skt. stend- ' thief ', stdyu-, stay ant- 1 etc. ; Skt. tf tdra 4 star 7 : Skt. stf-, Av. star-, Gk. aarrjp, etc, 4 id 7 ; Skt. tij- 4 to sharpen tigmd- 4 sharp ' : Gk. onypri, etc. ; Skt. tud- 4 to push Lat. tundo : Goth, stautan 4 id 7 ; Skt. phena- 4 foam O. SI. pena : O. Pruss. spoayno , with variant suffix Lat. spuma : Engl. foam ; Skt. plihdn- 4 spleen ' : Av. spzrzzan-, Gk. anX^v, etc. ; Skt. pdsyati 4 sees r : spas - 4 spy Lat, specio ; Skt. khanj- 4 to be lame ' : Gk. aicdCw ; Skt. phdla- 4 plough-share 7 : Pers, supdr ; Pers. fih 4 oar ' : Skt. sphyd- 4 wooden ladle ' ; Av. (vl-) xad - 4 to break up (earth) 7 : Skt. skhad- 4 to smash to pieces cf. Gk. oKehdwvpi ; Skt. nava- r sneeze 4 : Germ. niesen : Engl, sneeze (neu-s- : sneu-s-) ; Skt. lavand- 4 salty, salt 7 ; cf. Lat. sal , nihakd 1 fog nihara- 4 mist, dew 7 : snih - 4 to be moist etc. There is no perfectly satisfactory theory to account for this variation which affects aU Indo-European languages. Most probably it is the result of some kind of external sandhi affecting initial s- in the Indo-European period. It seems fairly clear that the phenomenon is due to loss of initial s, and if this is so the theory that would regard the 5 as the remains of some kind of prefix is out of the question. 82 PHONOLOGY §11. The Reconstructions p t ph,ih In a certain number of words etymologically connected, chiefly between Indo-Aryan and Greek, an $ (£) in the former appears to correspond to r or 8 in the latter. The commonest examples of this interchange, which has caused considerable difficulty, are the following : Skt. tdksan - carpenter ' : Gk. t€kt<a>v ; Skt. fksa- 4 bear Lat. ursus : Gk. apKro$, Ir. art ; *- Skt. ksan - * to wound ’ : Gk, fcrcivu) 4 kill ' ; Skt. ksi- 4 to dwell Av. Say- : Gk. Kripevos ; ksi- ‘ to possess # : Gk. Krdofiou ; raks- 4 to injure 4 : Gk, ipcyOio ; ksam- 4 earth ’ : Gk. I ksar- 4 flow ' : Gk. ^Oeipta 4 perish ' ; ksi - 4 to de- stroy ' : Gk. (f>8iva). It has been customary to assume a set of IE fricatives to account for these correspondences, namely p , ph, dh, b ut it is certain that th ese creat ions are without serious fou ndatio n, since in the case of some of them Hittite evidence has shown that a quite different kind of explanation is necessary, and in view of this the rest are naturally suspect. Correspond- ing to the Greek and Sanskrit words for 4 earth ' Hittite has tekan- and Toch, A tkam from which it may be deduced that the original form of this word was *dheghom or, with elision of the radical vowel, *dhghom-. In Greek there has been metathesis of the initial consonant group; elsewhere we find elision of the first member of the group: Lat. humus , OIr. zam- } etc. Like- wise for 4 bear ' Hittite has (in all probability) hartagas , from which it can be seen that there has been the same transposition in Gk. apKTos. The Sanskrit forms of these words have de- veloped from original unmetathesised form changed according to the rules of Sanskrit historical phonology. An original H rtkos 4 bear with loss of h ; rtkos, develops through the stages rt$a-> rtsa - to fksa-. The development of the word for 4 earth ', though more complicated as containing a voiced aspirated group, follows the same line in principal: dhghom- >dzham- > d zham- isam-> ksam-. In the case of Skt. ksi , Gk. the original root dhg^hi- produced Indo-Iranian djhi- which was modified to dzhi- and then treated as the initial group of the word for 4 earth In the case of some words Sanskrit has elided the first member in such groups, so that an original *dhghyes 4 yesterday 1 (cf. Gk. transposed) appears as hyas ( <ghycs ). The correspondence of Skt. syend- 4 hawk ' and Gk. Iktwos is of the same nature. PHONOLOGY 83 Not all the examples are to be accounted for in this way. In some cases there is suffix variation. Skt. dksi ' eye ' contains original -s- which is not to be compared with Gk. r in oktoAAos on the assumption of original p. Suffix variation should probably also be seen in Gk. t€kto>v 1 carpenter ' (*teks-ton) as opposed to Skt. taksan §12. Treatment of r and l In Iranian IE r and l appear indiscriminately as r. 1 In the language of the Rgveda this is predominantly the case. In Classical Sanskrit both l and r are found, but their distribution does not correspond exactly with that of Indo-European. In certain Eastern dialects of Indo-Aryan (notably in the inscrip- tions of Asoka and in the Magadhi of the Drama) only l is found. The treatment of IE / in Sanskrit is illustrated by the following examples : (a) l becomes r : rinakti ‘ leaves 1 : Lat. linquit ; sroni-

• buttock ' : Lat. clunis, Lith. slaunts ; sarpis ' butter Toch.

A. sdlyp, of. Engl, salve ; at aini- ' elbow ' : Gk. u>Xh nj, Lat, ulna ; fravas 1 fame ' : Gk. kAco.?, 0. SI. slovo ' word ' ; gdrbha- 1 embryo 1 : Gk. ScXpvs ' id aSeA^o'r * brother ’ (cf. sodara -) ; cakrd- ' wheel ' : Gk. kvkXos ; parasu - f axe ' : Gk. TreXeKvs ; piparti * fills # : Gk. mfjLTrXrjfu ; pur - 1 city 9 : Lith. pills , Gk. TroAtF ; sri- ‘ to lean ' : Gk, kXIvo> ; surya- * sun Lat. sdi. (b) l remains : lubhyati * covets ' : Lat. lubei ; palild - ' grey- haired ' : cf, Gk. -n-oAto?, ttzXitvos, etc. ; °kulva - ' bald ’ ; Lat. calvus ; palava- 1 chaff J : O. SI. pleva t Lat. palea ; palvala- ' pond ' : cf. Lat. palus ' swamp ' ; plihdn- ‘ spleen ' : Gk. otttAtjv, Lat. lien ; dala- 1 portion ' : Lith. dulls ; kldman- 1 lung * : Gk. nXeu/juav. In comparing the Vedic with the Classical language we notice : (i) that in a number of words the latter has / where the former has r, and this normally in cases where / appears in other IE languages, e.g. laghu- ' light ' v. raghu Gk. iXaxys, Lat. levis ; plu- ' to float v. pni- y Gk. vXiw ; lip- 4 to smear v. rip-, Gk. aXclpoj ; iih- ' to lick v. rih Gk, Aei'x&j ; (ii) that a con- siderable proportion of the classical words which preserve IE / 1 There are a few exceptions in Modem Persian and occasionally elsewhere : Pers. USian 4 to lick Skt. rih-, tik, Gk. A ; lahrt 4 soft Skt. lab ’ lip ’ : Lat. labium ; Oss. said 1 cold * : Lith. Sdltas. PHONOLOGY 84 are not found in the text of the Rgveda , either by accident, or because their meaning was of such a nature that they were not likely to appear in a text of sacred hymns (e.g. plusi- 4 flea J : Arm. lu, Alb. pVest, cf. Lith. blusd) ; (iii) that some derivatives which have become isolated from their roots preserve IE l even when it is normally replaced by r in the corresponding roots : Sloka- * verse J vipula- 4 great, extensive ' {pf~, piparti fill J ). The explanation of this apparently complicated treatment is fairly simple. The dialect at the basis of the Rgvedic language lay to the north-west, while the classical language was formed in Madhyadesa. The original division must have been such that the Western dialect turned l into r in the same way as Iranian (being contiguous to Iranian, and at the same time probably representing a later wave of invasion), while the more easterly dialect retained the original distinction. It was in this latter area that Classical Sanskrit was elaborated, but it was not evolved as a separate literary language, distinct from that of the Veda ; on the contrary it developed as a modification of the old sacred language of the Vedic hymns. The latter was always the foundation of the literary language, but since after the earliest period (and this excludes most of the later tenth book of the Rgveda), the centre of its cultivation shifted eastward to Madhyadesa, in its further development it was subject to the continuous influence of the dialectal forms of this region. So in the case of the distribution of r and l many of the basic words of the vocabulary retain always the form established by the Vedic literature, but in other cases /-forms based on the dialect of Madhyadesa replace them. In cases where the word in ques- tion is not found in the Vedic text, and where therefore there existed no established literary tradition, the Eastern form with original / almost universally appears. The treatment of IE r is different in that in the vast majority of cases it continues to be represented by r in all periods of the language, e.g. rudhird- 1 red, blood ', Gk, ipvdpos ; jdrant- ‘ old Gk. ytpojv ‘ old man ' ; rat- ' property Lat. res , pari 4 round Gk. 7 rtpl ; variate 4 turns Lat. vertitur, pdrdate 4 breaks wind Gk. Trephtrai ; pdrsni- 4 heel ', Gk. nr^pva, Goth, fairzna ; sru- ' to flow’ ', Gk. peaj ; tiaras n. pi. 4 men J , Gk. avtpes, a vdpes sdrpati 4 crawfls Gk. Lat. serpd ; rtij-, rdjan - 4 king Lat. rex ; rdtha - 4 chariot Lith. rdtas PHONOLOGY 85 ' wheel Lat. rota ' id ' ; vita- ' man, hero Lith. vtfras, Lat. vir, etc. On the other hand instances of l in place of IE r are comparer lively rare : lohita - ' red ' (also rohita Av. raoiSita cf. rudhird -) ; ' suitable, enough ’, v. aram, cf. Gk. apaploKio ; palayate ' flees 1 (£>ar# with i- ‘ to go '). The- number of such examples is too small to justify the assumption of an /-dialect to account for them. Such an /-dialect does in fact occur later in the Magadhan Prakrit, but it was limited to a small area, and this Prakrit cannot account for forms with / out of r which occur in the later Vedic literature. It is also to be noted that in some cases where a change />r has been assumed (e.g. lup - compared with Lat, rumpo) it is more likely that / is original. §13, Indo-European h Of late a new phonetic element has entered into accounts of Indo-European as a result of the discovery of Hittite. In this language there appears a sound h which was unaccounted for in the normally prevailing conception of IE phonetics. It is found in basic IE words and must therefore be attributed to Indo- European. Since it is absent in the corresponding words in all the other languages, they must be presumed to share a common change by which it has been lost, and to represent, in this respect, a more advanced state of Indo-European than that preserved in Hittite. Common examples of h are : Hitt, eshar ' blood ' : Skt. dsrk , Lat. user, Gk. Zap, Toch. A. ysdr ; hadtai *- bone ' : Skt. dsthi, Gk. 6vt4ov, Lat. os ; hunt- ' front ' : Skt., inti * (in front of), near', Gk. dvr 7, Lat. ante ; harki- 1 white ' : Toch. drki, Gk. apyos, Skt. drjuna -, etc, ; pahhur * fire ' : Gk. nvp ; pahs- * to protect ’ : cf. Skt. pd - ' id etc. ; ishai ishiya-

• bind 1 ; Skt. sydti 4 binds ' ; newahh- ‘ renew ' : Lat. novdre ;

palhis ‘ broad ' : cf. Lat. planus, etc. For Indo-European the symbol h, used by H. Pedersen is the most convenient (IE

• pelH- etc.).

In some instances H disappears without trace (dsrk ' blood ') but in others its effects survive. It is clear that the long vowel in Lat. novdre results from the combination of a short vowel + h, a combination which remains in Hittite, and the same can be assumed in the case of Lat. planus (plan-, varying in apophony from Hitt, palh -). Skt, sydti ' binds ■, from the Hittite evidence, stands for *snydti f of which sh- is the root in its weak form, and 86 PHONOLOGY yd the suffix of the fourth class. The simple root with guna appears in the aorist (dsat) showing the same development of an to d. Hitt. paM- * to protect ' is enlarged by an s- suffix, and when this is removed we see the same correspondence ah : d between this and Sanskrit pd -. Since the nominal and verbal suffixes d are identical, the same development an>d (as in novdre) is to be assumed in the case of the feminine suffix a (Lat. nova, Gk. via, Skt. ndvd), and this implies a similar de- velopment in the case of the long vowels % and u in nominal and verbal derivation, since the compound suffixes yd and vd must in the same way be derived from i + an and u + an, and the corresponding weak grades are for i- H and u- H. Another effect of h, observable in languages other than San- skrit, is the coloration of a succeeding vowel by h, producing notably a change from e to a. For instance the root which appears in Sanskrit as kri- ' to buy ' is to be set up for Indo- European as k™ri H-, and in Greek an original aorist form ek w rin- eto appears as iirplaro, the original presence of H being indicated by the a instead of e of the termination. Similarly from the root krun- (Skt. kru- in krurd-, etc.) an original esjos stem *kreunos appears in Greek as an -as-- stem, rcpias. The confusion of the .vowel qualities has eliminated such variation in Sanskrit, but there remain a number of other circumstances in which the presence of h can be detected, notably : (i) The older theory assumed a vocalic r and / (written f r , p and in other ways) before a following vowel in certain cases to account for correspondences like Gk. fiapvs ‘ heavy Skt. guru-, Goth, kaurus 1 id But there was no clear reason why the rule that these phonemes appear as vowels in intercon- sonantal position, but as consonants before and after vowels, should not be valid in this case. A restoration g w f nit- showing that the f originally occupied an interconsonantal position, accounts for all these developments. In most IE languages where f develops into a vowel (varying from language to lan- guage) + r, this combination remains before a vowel when h disappears, and similarly in the case of l. In Sanskrit the pro- cess is somewhat different, since here vocalic r normally re- mains, but when the loss of h would leave it before a vowel, its place is taken by the combinations it and ur. Iranian, which is usually so close to Indo-Aryan, differs markedly on this point, showing ar where Sanskrit has ir or ur . Examples of such words PHONOLOGY 87 are Skt. liras ' across ' : Av. taro ; sir as- 4 head Av. surah-, cl. Gk. Kapd, Kaprjvov ; pur as 4 in front J , Av. pard, Gk. napo$ ; puras n. pi, 1 cities cf. Lith. pills 4 fort 5 (guna in Gk. ttoXis) ; puru - ' much Av. pouru- (lx. paru-) ; hiranya - 4 gold Av. zaranya- ; gin- ‘ mountain Av. gairi cf. Lith. giria 4 forest 1 (guna in O. SI. gora 1 mountain ') ; girdti , gildti 4 swallows tirdti ‘ crosses, overcomes ', kirdti 4 scatters etc. The varia- tion between i and u in these cases depends on the preceding consonant ; preceded by a labial, or in some cases by an old labio-velar, u appears, elsewhere i is normal. (2) The combination rH also gave rise to a special develop- ment when followed by a consonant. In this case ir } ur appears in Sanskrit, but in Iranian predominantly ar. So we have Sirs an- ‘ head ' beside sir as (krusen- : krues-), tirnd- and kirnd- beside tirdti, kirdti , etc., and with u, purnd- 4 full ' (after labial). Other words with ir , Hr of this origin are : irntd- ‘ arm, foreleg Av. arzma O. Pruss. irmo, Lat. annus , Engl, arm ; urdhvd- 4 upright, high Av. vrdhwa-, Lat. arduus ; urnd 4 wool f , Av. varzna, Lith. vilna ; urmi- ' wave Av. varmii-, A.S. wielm ; urvdrd 4 cultivated land Av. urvara 4 cultivated plant ' <*rHvara, cf. Lat. ardre 4 plough Gk. dpovpa 'ploughed field etc. ; dirghd- ‘ long Av. dardya-, O. SI. dlugu ; pfirva - 1 former Av. paurva O. SI. pruvu ; bhtirja- 4 birch ' (Lith. bdrzas , etc,, with different grade). To account for these de- velopments original long sonant liquids were set up (f, /), and these could have conceivably existed at an intermediate stage (rn>f>ir, ur), the development being parallel to that of in, uii to i , u. In the same fashion the long sonant nasals which were postul- ated may be replaced by nu and mu. From the root san(ii) 4 to win ' the Skt. participle said- develops regularly through *santd- from snttto-, with regular weak form of root. In the case of mu the nasalisation is preserved (or reintroduced), but the original presence of H is clearly enough indicated by the long vowel : ddnta- ‘ tamed ', sdntd- 4 appeased from dam{y, sam( h)-. (3) In the Vedic language -ya- after a light syllable is pro- nounced as one syllable if it is simply a combination of y + a : dvya- * belonging to a sheep kavya- ‘ wise vdnya- ' of the forest ', havya - * oblation When on the other hand it goes back to -t Ha (a suffix parallel to -ira, - isa , etc.) ya is pronounced -iya ; ddmiya- 4 belonging to the house ', rdthiya- 4 relating to a 88 PHONO! OGY chariot jdniya - ‘ relating to the people udaniya- ' watery r The two types are of course confused in the later language, and the difference revealed by the Vedic metre, is simply explained when it is realised that there are two different suffixes, (i) i + a, (2) i + H+tf. The declension of the stems in t and u ( <tH, mh) where the suffix always retains its syllabic value before a vocalic ending (gen. s. vrkiyas, tanuvas<°tuas ) °unas) confirms this quite clearly, since the corresponding genitives of dm- ‘ sheep ' and mddhu ' honey 1 (avyas, mddhvas ) show always the consonantal value of y and v. (4) Most significant of all, traces of the original nature of h are preserved in Sanskrit in cases where it was immediately preceded by an occlusive. Here the combination occlusive + h may produce an aspirated occlusive. As already stated, it was in this way that the whole category of surd aspirates arose in late Indo-European. Examples of this have already been given. Examples of sonant aspirates arising in this way are seen when an aspirate in Sanskrit appears to correspond to a non-aspirate in other languages, or when a final non- aspirated occlusive of a root appears with aspiration in a derivative. In these cases original suffixal h is responsible for the aspiration ; e.g. mdhd- 4 great ' : Gk. tUyds. Here the root is followed by the suffix <zh(>a), which appears in its weak form (-H-) in the gen. sg., and this h being in immediate contact with the preceding g causes aspiration (meg-H.-ds>meghd$>niahds) , and from such forms the aspiration is extended to the whole declension. The same thing has taken place in duhitdr- ‘ daughter ' (dhug- h- itdr-) : Gk. dvydrr^pj ahdm 4 1 # (egn-dm) : Gk. £yu> (egou), sadhastka- ‘ seat, abode 1 [sed-n-es-) : sad - 4 to sit slndhu- 1 river ' : syand - f to flow Before the discovery of Hittite there existed in Indo-Euro- pean studies a 4 Laryngeal Theory which, since it received partial confirmation from the new Hittite evidence, has come to be generally adopted in recent years. Briefly stated in its most popular form the theory maintains that there existed three laryngeals, which in this notation would be represented by Hj, h 2 , h 3 . The original long vowels of Indo-European (as opposed to those long by vrddhi), result from a combination of a single guna vowel e with the several laryngeals, so that from en lt en % and 6H a , e (e,g. dhe - 4 to put '), d (e.g. std 4 to stand ') and 6 {e.g. do r to give ') are respectively derived (i.e. the roots PHONOLOGY

• 9

are originally dhe h 3j ste h 2 , den 3 ). The theory further maintains that when preceded by these three laryngeals this same guna vowel takes the form e, a and o respectively (es- * to be ' : h ; anti ' in front 1 : w^enti ; ost(h)i- * bone ' : ii z esti). Hittite provides some positive evidence in support of this theory, but it is incomplete, and in certain respects contra- dictory. We have already quoted instances showing the de- velopment of the guna vowel -f h to d and of the change of e to a when preceded by h. On the other hand there are difficulties : although h appears where the theory demands it in hunt- : Gk. avrl it is absent in appa : Gk. wno where the theory equally demands it. In Hittite there is only one h and it is a long way from this three or even four demanded by the theory. It is not therefore surprising that the theoreticians differ considerably in the details of their exposition. For the purposes of Sanskrit grammar the question of the plurality of h is fortunately of little significance, because the variation of vowel quality (a, e, o) t with which it is bound up, has ceased to exist in Sanskrit, For all practical purposes it is possible to operate with a single, undifferentiated h, and that will usually prove sufficient. Another aspect of the Laryngeal theory should be briefly mentioned. From the beginning it has been involved in the theory of Indo-European ' Shwa ' (2). In the Laryngeal theory it is replaced by a vocalic version of the laryngeals (h with three varieties). As a result of this the laryngeals themselves com- monly receive the notation It will be pointed out below that the hypothesis of an Indo-European f is without justifica- tion either in the framework of the laryngeal theory or of any other. Indo-European h is not capable of vocalic function and when left in interconsonantal position through loss of the associated guna vowel it is in Sanskrit elided : e.g. Skt. dadmds, dadhmas from da, dhd . §14. Combinations of Occlusives The following changes in combination are inherited from Indo-European : (1) A sonant is changed into a surd when immediately fol- lowed by a surd : yuktd - * joined ' : yuj- 1 to join yugdm ‘ yoke', cf. Gk. ^€vkt 6 $ : ^vywfju ; patsu , loc. pi. of pad - ‘ foot *, cf. Gk. rrooaL, irooi. Conversely a surd becomes sonant when folPHONOLOGY 90 lowed by a sonant : Skt, upabdd- 4 trampling on ' : pad - 4 foot cf. Av. frabda - ' fore part of the foot Gk, eirlpSai 4 day after a festival ' ; dadbhis instr. pi. of dant- } dat - 4 tooth * ; abjit- ' conquering the water * : dp - 4 water (2) In the case of the combination sonant aspirate followed by -t- the whole group is voiced and the aspiration attached to» the second consonant ; thus from dah- ' to burn 1 (from dagh- by the second palatalisation), budh- ' to understand 1 and labh- 4 to receive the participles in -id are dagdhd-, buddhd- and labdhd In the older Avestan language a similar development is ob- served, though the aspiration as always in Iranian has been lost: aogzdd 4 said 1 from *augdha t i.e. Aryan aitgh- (Av. aog-)+ta, cf. Gk, €i ?xo/iou ; ubdaena- 4 woven ' from Aryan vabh The later Avestan substitutes combinations of type (1) above even in the case of the original sonant aspirates : aoxta 4 said druxta - f betrayed ' (draog- : Skt. druh-) dapta - * deceived 1 (dab- : Skt, dabh-). In the same way in Sanskrit dhatti 4 places ' has been substituted for * daddhe ( -- Av. dazde) which would be the regular combination of dadh- te. Elsewhere in Indo- European innovating forms of this type have completely re- placed the old type of combination : e.g. Gk. ckto? : i rtvvrisy cf. Av. apaitibusti * not noticing as opposed to San- skrit buddhi-. (3) Dental combinations in Sanskrit normally conform to the above rules : vetti 1 he knows ' from vid-, rttddhd- 4 ob- structed 1 from rudh- + td, etc. On the other hand Iranian sub- stitutes the sibilant 5 or 2 in these positions : v crista 4 thou knowest ' : Skt. vdttha ; hastra- 4 session ' : Skt. sattrd - ; ni-uruzda- ‘ locked up ’ : Skt. ruddha-. The Greek treatment agrees with Iranian : otcrda 4 thou knowest 7 rv<rn,s ' informa- tion cf. Av. apaitibusti : Skt. buddhi-. In the Western IE language -ss- results from the combination : Lat. ob-sessus (sedeo), 0, Ic. sess 4 seat A tendency to modify the dental combinations is therefore wide-spread. It is assumed that in Indo-European a sibilant was inserted in these cases (t*t f t s th r d £ d , d z dh). Since all interconsonantal sibilants are elided in Sanskrit an IE voitHha would produce Skt. vettha, and at the same time it accounts for the Iranian and Greek forms. In the case of the voiced combination we find two kinds of treatment in Sanskrit, on the one hand the usual type ruddha- , vrddha -, etc., and on the other hand some ancient PHONOLOGY 9 1 forms testifying to the existence of z instead of d as in Iranian : debt impv, • give 1 beside daddhi, cf. Av. dazdi, and dheki 4 put both with e out of earlier az according to the rule below. Either this is a case of dialectal divergence, or the type dehi (<dazdhi) represents the regular phonetical treatment which has been re- placed in the majority of cases by new analogical formations. §15. Combinations Involving the Palatal Series The second palatal series is simple in the matter of consonant combination, since all that is involved is the retention of the original guttural before a consonant, which then combines according to the rules given above ; vac- ‘ to speak ' : uktd -, yuj- 4 to join ' : ynktd yoksydmi , dah- 4 to burn ' : dagdha Combinations of the old palatal series are much more com* plicated. In the early Indo-Iranian period s, z and zh were changed to / and z before dental occlusives (with aspiration and voicing of the occlusive in the case of zh according to the rule given above for dagdha -, etc.) : the resulting sibilants were identical with those that arose from Indo-European s, z after i, u, etc., and their subsequent history is the same. In Sanskrit $ became cerebral s and cerebralised the following dental and z , after undergoing the same process, was elided leaving cerebral d : e.g. vdsti 4 he wishes ' (1 ms-) Av. vasti, Hitt, wekzi , cf. Gk. ckwv ptc. 4 willing J ; astau ‘ 8 1 (cf. asiti- ' 80 ’), Av. asta, Lat. octo , etc. ; Skt. mrdtka- 4 pardon Av. mzrdzdtka- (mrg-d t cf. Skt. mrj- ‘ to wipe away 1 and Pers. dmurztdan ' to pardon ') ; Skt. udhd 4 carried 1 for *uzdha- < uzdha- [yah- 1 to carry 1 from vazh -< IE vegh-) ; Av. g&rvzdd *' complained 3 sg. aor. mid. from gar 9 z- = Skt. garh- ; Skt. Udhi ' licks ' ( lzh-) t etc. The same change was liable to take place in contact with other consonants : cf. Av. fsumant- 4 possessing flocks ' ( pasu Skt. paid-) ; Av. frasna - 4 question Skt. prasnd- ; Av. vtzibyd, Skt. vidbhyds (vis-). In these combinations there is not complete agreement between Indo-Aryan and Iranian, but in the case of the latter example at any rate an Aryan *vizbhyas is attested. The Sanskrit development through vizbhyas to vidbhyds is exactly the same as that seen when z (z) was out of IE s, e.g. in viprudbhyas for *vipnizbhyas ( viprus - 4 drop '), When immediately followed by s these palatals appear as k in Sanskrit in intervocalic position ; and the treatment is therePHONOLOGY 92 fore the same as that of the second palatal series : vdksi ' you wish * (vas-) like vaksyami ' I will speak ' (vac-). This is not a case of preservation of IE k (cf. Hitt, wek- ' wish ') but of its restoration. This is known for the following reasons : (1) Iranian continues to distinguish the two types of com- bination derived from IE guttural + s and from IE palatal f s : (a) vaksyami , Av. vaxsyd (wek m ~) ; ksatrd- 1 sovereignty 1 : Av. xsadra- ; ksap - * night 1 : Av. xsap- t cf. Gk. ; bhaks- ‘ partake of, eat 1 : Av. baxS- ‘ distribute ’ ; tvaks- ‘ to be active 1 : Av. Owaxs-, (b) kdksa- ' armpit 1 : Av, kasa ; fksa - ' bear ' : Av. arasa- ; ddksina - ‘ right ’ : Av. dasina cf. Gk. Sefioy, etc. ; maksu ' quickly ’ : Av. mosn, cf. Lat. mox rdksas- ' injury ' : Av. rasah- ; tabs- ' to construct in wood ' ; Av. tas-, cf. Hitt, taks- ' join etc. ; ksudh- ' hunger ' : Av. sitSa ; ksi- ‘ to dwell ksStra- 1 field ' : Av. si-, soidra- ‘ settlement (2) The evidence from Iranian that there were two combina- tions in early Indo-Iranian both represented in Sanskrit by ks is further confirmed by evidence provided by Sanskrit itself, In cases where the group is followed by t it gives k or s according to its origin : (a) dbhakta 3 sg. mid. s-aor. of bhaj- f cf. 1 sg. dbhaks-i ; cf. also bhaktd - ' food ' : bhaks - ' to eat ‘ ; (b) 3 sg. tdsti, caste from taks- s caks- ; nlrasta- ( castrated ' from niraks-, etc. When the group is final the sibilant is elided according to the general rule. When the k goes back to the guttural series it invariably remains : vdk nom. s. from *vdks, cf. Av. vdxs . When on the other hand the old palatal series is involved, although there are some instances of k ( °drk , °sprk, nom. sg. to °drs- t c sprS‘) the normal and regular treatment is -t : vit * settle- ment ' : vis- ; vipdt ' the river Beas 1 : vipds- ; spat ‘ spy ' : spas- ; r&t ' king * : rdj-, cf. Lat. rex ; °vat ' carrying 1 : vah-

• to carry etc. Here the anomaly of the A-forms is explained

by the dissimilatory influence of r in the vicinity. It is clear from this evidence that, where Sanskrit has a single combination ks, there were originally two different combina- tions. What immediately preceded ks in the prehistoric period ot Indo- Aryan where the palatal series is involved is made clear by the forms of the nom. sg. quoted above. Just as nom. sg. vdk is derived from earlier *vdks by regular loss of the final sibilant, in the same way vit is derived from *vits. At a period which probably did not very long precede the beginning of the rePHONOLOGY 93 corded tradition this was changed into ks and thus confounded with original ks. The change is seen in the loc. pi. viksu (later supplanted by an analogical vitsu) as opposed to the nom, vit(s) . Since the simplification of final consonant groups pre- ceded this change, the cerebral, that is to say half the original combination, is preserved in the nom. sg. It is necessary also to go beyond this ts since even from the point of view of Sanskrit this will not explain castd , etc. (*catte would have resulted). In this connection the sandhi of two sibilants should be compared. There are instances of $ + $ be- coming ts : vatsyami , dvdtsit from vas- ' to dwell and of s + s becoming ks : dvgksi ‘ you hate ' from dvis Here again forms of the nom. sg. — dvit viprut - — show that there was an inter- mediate stage ts (which is obviously what would be expected in the case of this combination). In both cases ts (~>ks) may be derived from 55 (Aryan ss). The treatment of the palatals before s is on the lines of their treatment before dental occlusives. Just as palatal s + dental i produce the cerebral group st, so palatal £ + dental 5 produced ss which then, in precisely the same way as the original sibilant combinations, became ts and finally ks. In Iranian the development was somewhat different. Here / + / out of Aryan s + s result in single i. This is in accordance with another rule for the sandhi of sibilants by which one can stand for two when they come together : cf. Skt. dsi for as + si. When palatal c is preceded by 5 the latter is changed to the palatal sibilant £ : e.g. sascati 3 pi. reduplicated present of sac - ‘ to associate with When s is followed by s the two lose their identity and are merged as cch : ducchund ‘ misfortune 1 from dus + sund- 1 prosperity Since the same sandhi results when £ is preceded by a dental ( pacchas from pad - ' foot ' 4 - suffix -sas) we may see here the same tendency to occlusion as in the other sibilant combinations noted above. In tuccha - ' empty * (for

• tu£ya- t cf. Khot. tus£a-<*tusya-) and kacchapa - ‘ tortoise '

(cf. the proper name Kasyapa ~) we have the same development of a group -££- which has resulted from an early ' prakritic ' assimilation. From the standpoint of Indo-European Skt. ch (cch) results from an original combination sk ( skh ) ; and in these cases Iranian has s : chid- ‘ to cut Av. said-, Gk. vxi^lo, Lat. scindo ; chayd * shade Pers. sdyah, Gk. ukU ; gdcchati * goes Av. PHONOLOGY 94 jasaiti , Gk, fidafcw ; prcchdti ‘asks', Av. pzrzsaiti, Lat. poscit . We must assume that in these cases Aryan s and s were assimi- lated to ss which was then treated as above. It should be noted that Skt. ch is different from the other consonants among which it is classified in that it is always a long or double con- sonant. It is a matter of indifference whether ch or cch is written, though it is customary to use the former at the be- ginning of a word and the latter in the middle. §16. Combinations Involving Sibilants Indo-European s became z when followed by a sonant occlusive. This z became i in primitive Indo-lranian under the same conditions that s became s (Skt. s), In Iranian z and i are preserved : Av. hazdydt 4 would sit down ’ {sad-, Ir. had-), mizda- 4 reward 4 , Goth, mizdo. In Sanskrit they are eliminated in the following ways : (a) Before unlike consonants 2 and z (Aryan z) are replaced by d and d respectively : ddga - 4 branch ’ : PahL azg ; madgu- 4 a water bird majj- 4 to dive ’ « *madj-<*mazj-) ; Lith. mazgoti 4 dive 1 ; instr. pi. usddbhis , mddbhis from usds- 4 dawn mis- 4 month 1 ; viprudbhis, instr. pi. of viprus - 4 drop (b) Followed by dental d, dh, z is elided and a preceding vowel a is changed to e : edhi 2 sg. impv. 4 be 4 : Av, zdx {as-) ; seddr for *sazdur 3 pi. perf. of sad - 4 to sit 4 : cf. Av. hazdydt ; nddistha- 4 nearest 4 : Av. nazdista * ; medha 4 wisdom 4 : cf. Av. mazdd 4 wise 4 ; miyddha- 4 food offered to the gods/ : Av. myazda- ; ddhvam 4 sit ’ 2 pi. impv. from as- ; sasadhi from sds- etc. This applies also to cases where 2 developed from original d : cf. dehi, dhehi above. A different treatment is seen in addhi adv. 4 certainly 4 : Av. O. Pers. azdd . (c) Before d, dh , z (which may be derived from IE $ or from the old palatals, § 15) is elided with cerebralisation of the fol- lowing consonant, and compensatory lengthening of a short vowel : ntdd- 4 nest Lat. nidus, Engl, nest (*nizdo- from ni + sed-}; midhd- ‘reward 4 , Av. mizda-, 0 . SI. mizda, Gk. /itaflo?, Goth, mizdo ; duldbha - 4 difficult to deceive 4 [dus + dabh Vedic l for d) ; pid- 4 to press 4 (cf. pis - 1 to pound '), hid- 4 to injure 4 (cf. hints- 4 id ’) krid- ‘ to play 4 (cf. ON hrista 4 shake 4 ) for pizd -, hizd-, krizd- ; astodhvam 2 pi. mid. s-aor. of stu - 4 to praise 4 ; mrdika- 4 mercy 4 (first syllable metrically PHONOLOGY 95 long in the Veda) : Av. mzrvzdlka- ; udhd- ‘ carried ' : vah- (<uzdha-, i.e. uzh+ia), ledhi "licks' ( <lezdhi f i.e. lezh + ti). A preceding short a may be either lengthened (tddhi<lazdhi for taks + dhi , dsadha- : sah -), turned into o (vddhum 4 to carry ' : vah - ; sodha 4 sixfold ' : sds) or turned into e : trnddhi 4 shatters ’ from the present base trndh- of trh-. Occasionally z is represented by d even in this position : dididdhi, mimiddhi , ririddhi from dis- 4 to point tnih- 4 to urinate ', rih- 4 to lick J ; saddhi beside sodha. The same varia- tion is seen in Pa, nidda kiddd which reflect a different dialectal treatment in Old Indo- Aryan. Between consonants 2 disappeared without trace : jagdhd- 4 eaten ’ <*jagzdha- (ja-ghs-ta-), as also did s (cf. dbhakta above). The combination sonant aspirate +s was in the Indo-Iranian period treated in the same way as the combinations of sonant aspirate -f 1 noted above. That is to say, in intervocalic posi- tion gh + $ gave gzh t and so on. Avestan preserves such voiced combinations though the aspiration as elsewhere is dropped : aoyzd 4 you said' (<augzha t i.e, augh --h sa) r diwzaihydi 'to injure' ( <dibzha i.e. di(d)bh + sa). Sanskrit has the surd combinations ks, ts, ps in these cases, but these have replaced original gih, dzh , bzh . The absence of aspiration in Vedic aduksai 4 milked dipsati 4 desires to injure coming under the general rule (§ 2) of the dissimilation of aspirates, presupposes forms like dhugzha-, dhi{d)bzha- where the rule could operate. On the other hand in final position, where these groups were surd and de-aspirated from the beginning (Aryan c dhuks nom. sg.) , there is never any loss of aspiration in Sanskrit, Furthermore there are a few cases in Sanskrit where jh, jjh appear instead of ks where such a voiced combination is in- volved : jdjjhat - 4 laughing ' (reduplicated formation from has-), nirjhara- 4 waterfall containing the root which normally appears as ksar- ( = Av. yzar-). These are Prakritisrns, and further examples are quotable from Middle Indo-Aryan : Pa. Pkt. jhdma- 4 emaciated r : Skt. ksdmd - ; jhay- 4 to burn ' : Skt. ksay - ; jhlna- 4 exhausted : Skt. kslnd Pali has also jagghati 4 laughs 1 with ggh instead of the more usual treatment -jjh-. In all these cases voiced combinations of the type pre- served in Av. yzar-, etc., are to be assumed, and the difference between these forms and the normal ks of Sanskrit is indicative of dialect variation in Old Indo-Aryan. PHONOLOGY 96 Something has already been said about combinations of sibilant 4- sibilant. There are three types involved which differ in respect of the date of their operation : (1) By an old IE rule s + $ could be represented by a single $ : dst f you are Av. ahi, Gk. et (IE esi out of es + si) ; amhasu loc, plur., Av. qzahu (dmhas ‘ distress ' + su). (2) By a rule specific to Indo-Aryan, but one whose opera- tion lay mainly in the prehistoric period, s + $ became ts and s 4* s became ks (through */s) : vatsydti , dvdtsit from vas- 4 to dwell ' ; jighatsu- 4 hungry ' from ghas- 4 to eat ' ; dveksi 4 you hate ' from dvis 4 to hate When these combinations are final only the first element remains, and in the case of the cerebral combinations, since the loss of the final sibilant took place during the stage *ts, this appears as t : nom. sg. ukhasrat ' dropping from the pot ' (srams-), parnadhvat 4 shedding leaves ' [dhva{m)s-) ; °dvit 4 hating viprut 1 drop (3) Neuter nouns in -as, -is, - us make their loc. pi. in - ahsu , - ihsu , -uhsu (optionally - assu , - issu , -ussu). This is the latest type, and it is patently imitated from the external sandhi of the nom. sg., as has happened also in the bh- cases ( manobhis , havirbhis, etc.). §17. The Cerebrals In the cerebral series {t, th, d } dh, n t s) Indo-Aryan presents an innovation as opposed to the rest of Indo-European. This somewhat infelicitous name, a mistranslation of Skt. mur- dhanya- t dates from the very earliest days of Indo-Aryan philology, and has stuck through long habit. Phonetically ‘ retroflex 1 or ' retroverted 1 more adequately describes these sounds which are distinguished from the dentals in that the tip of the tongue is turned back to the roof of the mouth. They are characteristically Indian sounds, and were certainly acquired by the Indo-Aryans after their entry into India. At the same time their use spread to the more easterly of the Iranian lan- guages, those bordering on the Indo-Aryan area (PaSto, Khotanese, etc.)* Cerebrals are also found abundantly in Dravidian, and they are certainly ancient in that family. They are also found prevalently in the Mun<Ja languages, but since they appear to be absent in Savara, a member of the family less affected by external influences than any other, they may not be original in that family. Since it is only in India and PHONOLOGY 97 the immediate vicinity that an Indo-European language has developed such sounds, and since it may be safely assumed that an early form of Dra vidian possessing such sounds was spoken over large portions of India prior to the advent of the Aryans, the influence of Dravidian may be held to be responsible to some extent for their emergence. At the same time, in native Indo-Aryan words they are explicable entirely out of the com- binatory changes that affected certain consonant groups. Most of these have been mentioned and can be classified quite simply : (i) Originally dental t, th became cerebral when preceded by 5 (Aryan s) which in this position may either rep- resent IE s (§9), or be a modification of Aryan palatals s, z (>Skt. j : IE k t g) : vrsti - ' rain ' {vrs- : cf. Gk. epoq, itpaq ' dew Ir. frass ‘ rain shower '), visit ' wishes ' (vas-, cf. Hitt. wekzi ), dstrd ‘ goad 1 (aj - * to drive * : Lat. ago , etc.) ; (2) Origin- ally dental d } dh , became cerebral when preceded by z (Aryan z of the same twofold origin as s) ; since in this case the sibilant was elided the resulting cerebrals d , dh (l, Ih in the Rgveda) came to stand alone in intervocalic position : exx. nldd-, iidhd-, etc., see above ; (3) The occlusion of the first part of the group ss (which may be for s + s or s + s) produced ts ; finally the t came to stand alone, the simplification of the consonant group in this position {dvit, vit , above), while intervocalically the group de- veloped further to ks ; (4) Originally dental n became cerebral n under wider conditions, namely when preceded in the same word by s y r or r, except when a palatal or dental intervened : k dr ana- * cause etc. In addition to these rules by which cerebrals developed there are others which have been more controversial, but which can now be regarded as established. A notable case is the change of IE / followed by dental to cerebral, commonly referred to as Fortunatov's law. Though long opposed, this rule is to be accepted. Examples of this phonetic change are as follows: pata- 'cloth' <*palta- : Arm. (<Ir.) pariah 'veil', Npers. pardah, OSlav. platino ' linen cloth etc. ; patala- ' fold, layer, stratum puta- 1 fold, thing folded ’ <*paltala- } *pulta-: cf. the IE root pel - ' to fold ' in Gk. SittAos- ' twofold Lat. duplus , duplex , Eng. fold, etc.; sphatika - * crystal sphuta- 'clear 1 : cf. the IE root sp(h)el- 1 to be bright ’ appearing in Sanskrit also in sphul-jsphur- r to glitter sphulinga - ' spark ' ; hdtaka - 'gold': OS 1 . zlato , Russ, zoloto ; jathdra- 'belly': Goth. PHONOLOGY 98 kilpei ‘womb 1 ; kuthdra - ‘axe 7 : cf. kulisa- ‘id' and Lat. culter ‘ knife etc.; jada- ‘ cold, stiff, numb, dull ' <*jalda- : cf. Lat. gelidus, etc.; adhya - ‘ rich J : cf. rdh- ‘ to prosper a root which originally had l (Gk. do}xai etc,); puna- ‘stake, stipulation, hire, wages ' : Lith. pelnas ‘ gain, profit, earnings ' ; kuni- ‘ lame in the arm' <*kulni cf. Gk. kuAAo? {<kulnos); pasdna - ‘stone, rock': Germ Pels, etc.; bhas- ‘to bark', bkds- ' to speak Lith. Wsas ' voice Germ, bellan , etc. 1 In addition to these types of combinatory change, there has also been a considerable amount of spontaneous cerebralisation in Sanskrit, and it is possible to list a fair number of words in which a cerebral represents an original dental without any combinatory change being involved. Some instances of this kind have long been recognised, particularly where later San- skrit has cerebral as opposed to dental in the Veda: e.g. at- ‘ to wander di- ‘ to fly 7 , nadd- r reed as opposed to Vedic at-, dl- f nadd-. Other cases such as mani- ' jewel ' (Lat. monile) and sthtind ‘ column ' were also early recognised, and the cerebral was put down to Prakritic influence. Recently, however, it has become clear that this process has occurred in Sanskrit to a much greater extent than previously admitted, and that it is in most cases a genuine Sanskrit and not a Prakrit phenomenon. The following further examples illustrate the spontaneous change to cerebral: avatd- ‘ hole in the ground ' as opposed to Vedic avatd- ‘well'; kuntha- ‘blunt': NPers. kund, Bal. hunt] pinda- lump, mass': Arm. (<Ir.) pind ‘compact'; pandita - ‘ wise panda ‘ wisdom 7 : NPers. pand ‘ good advice ' ; kiita- ‘ hammer ' from *kuta- } cf. the root in Engl, hew, Lith. kduju

• strike OS1. kovati ‘ forge, hammer ', etc. ; koti- ‘ tip, point ' :

Lat. cautes, cautis ‘ sharp, jagged rock ' ; kadevara- ( > kalevara-) ‘ corpse ': Lat. cadaver . A small number of Sanskrit cerebrals are of Prakrit origin, e.g. bhata- ‘ soldier ' ( <bhrta-), nata - ‘ actor cf. nrt- ‘ to dance bhattdraka- ‘ lord cf. bhartr and atani- ‘ tip of bow ' beside earlier dr ini. There are some which occur in Dravidian loan- words: eda- ' goat', kuti- ‘hut, house', kathina- ‘hard', guda- ‘ bad. ', for which see Chapter VIII. In late Sanskrit some words with initial cerebral appear, which cannot be explained out of either Indo-Aryan or Dravidian: such are ! The change did not take place in words in which l was changed to r : e.g. urnd 1 wool ' (Lilt vilna, ch.) f jartu- * womb ’ as opposed to jdthdra - etc. PHONOLOGY 99 tika ‘ commentary damara- ' uproar ' and dhakkd ' a large drum ' . §18. Miscellaneous Changes The phonetic changes undergone by a language are for the most part subject to general laws, but when all has been done to elucidate these some exceptions will remain. For instance one can hardly doubt that Skt. kiia- f hair ' is the equivalent of Av. gaesa - ‘ id since an associated u- stem is available in both cases (Av. gaesav- : cf. Skt. kesav-a-). Contamination with kesara- ' mane ' (cf. I at, caesaries) may explain the change in Sanskrit. Such sporadic changes are found more abundantly in the case of certain consonant combinations. The combination pi normally remains when l does not become r [plu- ‘ swim, float ') but it is changed to kl in kloman- * lung ' as opposed to Gk. Trhevfjiaiv, Lat. pulmb 1 id * (the original meaning was

• swimmer, that which floats '), and in viklava- r distressed * as

opposed to viplava -, vipluta In the case of trp- ' to steal ' IE klep (Gk. Goth, hlifan , etc.) has been altered to tlep : a confusion of the groups kl and tl is common the world over. A similar change of the occlusive, also sporadic, is found in the case of the group -in-. It remains normally (rain a-, pdtni) but in the feminines of certain adjectives in -ita it appears changed to -kn- : dsikni, pdlikni {dsita- * black palitd- ‘ grey-haired ') Later examples of this tendency are seen in Pkt. savakkt (be- side savattx) ' co-wife Panj. saukkan, and in Panj. arak ' elbow 1 (aratni-). The sibilants are liable to certain changes when in proximity to one another. Initial s was changed to s in Sanskrit when i followed in the next syllable : svdsura - 1 father-in-law ' ; Av. x y asura- t Gk. €Kvpos, Lat. socer ; smasru - 1 beard ' ; cf. Lith. smakrd , Ir. smech * chin The same assimilation in the reverse order is seen in said- ‘ hare 1 for *sasa- : cf. Khotanese saha- f Germ. Hase, Engl. hare. On the other hand s is preserved in the roots sds- and ia$~ because the change was impossible in forms like i&sti, iasta - (but cf. iaiana- for sdsana - in the N.W. Prakrit). There is also a change of s to s when s follows : iu$ka- f dry ' : Av. huska-, ilaksna - * soft Pers. laSin r cf. Gk, Aa yvos, Xayapos, Lat. laxus, Engl, slack ; Ms- ‘ to adhere slesman - ' phlegm cf. Engl, slime , etc. Since in cases where a final s has disappeared an initial s is re100 PHONOLOGY placed by $, e.g. sat 'six' (*sats<*sats) and vird-s&t 'over- coming men', it is likely that suska-, etc., are the result of secondary dissimilation of an intermediate *suska-. This rule does not apply where the system of related forms is strong enough to prevent it, e.g. vdsu-, vdstyas vdsistha-, though even here isolated forms with s are handed dow r n. In Vedic ksumdnt — Av. fsumant - (:pasu-) we have an isolated example of a change w 7 hich has parallels in Iranian (cf. Khot. ksdrma- ‘ shame ’ —fiar^ma-}. More or less isolated cases of the reduction of three consonant groups in Sanskrit appear in stdna - ‘ breast ’ : Av. fit ana-, Pers. fiistdn Toch. B pescane ; hr add- 4 lake ' : cf. Av. yzrdd- to flow 7 ; and tvdstar- ' n. of the divine architect ' = Av. Ow dr 9 star-. From the Indo-European period there was a certain instab- ility about r, v, y as the second members of initial consonant groups. This accounts for equations like Skt. hhanj - * to break Ir. bongaim, Skt. bhuj- ' to enjoy Lat. fungor on the one hand and Lat. frango, fruor on the other. Similarly in the case of v Skt. sas, Lat. sex, etc., are opposed Av. xsvas, etc., and Skt'. ksip- 4 throw- ' to Av. xsvaew-, §19. Final Consonants In final position consonants and consonant groups receive in many respects special treatment. This was true also in the pre- historic period ; for instance the aspiration in nom. sg. °dhuk as opposed to its absence in aduksata is due to the fact that at an early period the combination gh 4- s became -ki in final position, whereas intervocalically it became gzh. But the tendency to special development in final position has become much stronger by the historical period, and its features anticipate in some re- spects the later Prakritic developments of Indo-Aryan. Of the occlusives only the unvoiced series p , /, t, k are allowed to stand in absolutely final position, and in their place the corresponding voiced series b, d, d , g are substituted before voiced consonants and vow r els. The sonantisation before initial vowel is a special characteristic of Sanskrit, and it anticipates the voicing of all intervocalic surds in later times. Consonant groups were drastically reduced during the period immediately preceding the historical record, and in this respect Vedic contrasts remarkably with early Iranian. Here too the PHONOLOGY IOI same general tendency was at work which later resulted in the assimilation of all consonant groups. With few exceptions (e.g. nom. sg. urk from tirj- * vitality ') not more than one con- sonant may stand at the end of a word, however many were there to begin with. This had serious results in some aspects of the morphology, and led to some grammatical innovations. Thus the terminations are lost in the case of the second and third persons singular of the root and s-aorists, and the s of the s-aorist suffers the same fate in these persons when preceded by a consonant, so that the formations lose their grammatical clarity. On account of this the root aorist comes to be abandoned in Classical Sanskrit except in the case of roots in long d t and new extended formations are provided in the case of the s-aorist (anaistt for dnais). The weakest of the final consonants was s. In final position this is weakened to the breathing h ( visarga ). In sandhi the same change occurs before k, p and the sibilants. Preceded by d it is elided before voiced consonants and vowels. The same thing happens when it is preceded by a, but here the -as is in most contexts replaced- by o. In the non-Sanskritic dialects of Old Indo-Aryan there was an alternative development of - as to -e. An example of this is found even in the Rgveda (sure duhitd ‘ daughter of the sun ') ; later it is a characteristic of the Eastern (Magadhi) Prakrits, and examples are also found in some of the Kharosfhi inscriptions of the North-West. These developments of final -as began in the Indo-Iranian period, and in Avestan likewise -6 is the common representative of final -as, beside which there is a dialectal variant -5 corresponding to the -e of Magadhi, etc. Final s is preserved only before t , th, while before c, ch it exists in the modified form s'. When preceded by i, u, s became originally s (§ 9 ) which would normally become Sanskrit s. In place of this s, before voiced consonants and vowels r is substituted. The intermed- iate stage was presumably 2 and in this case there is complete difference of treatment of a phoneme at the end of a word from iis treatment internally. This external sandhi was extended to the sandhi of compounds (durdama-, etc.) but old forms like diildbha - (dud°-<.duzd°-) and kdrulatl show that this is not original but analogical. Likewise the sandhi of neuter s- stems in declension (havirbhis, havihsu ; . manobhis , tnanahsu) is in imitation of the external sandhi. 102 PHONOLOGY Final r was weakened in much the same way as s. Finally it is represented by visarga (punah), but it is retained when pre- ceded by a and followed by a vowel (punar dgacehati). Else- where it behaves in sandhi exactly like s. It even becomes s before t {punas lam) a development which is certainly analogical rather than phonetic. In the case of uninflected forms with final r preceded by i or u it is impossible to tell from any sandhi context whether the word originally ended in s or r ; thus though we may conclude that the original form of the adverb meaning * outside ' was bahtr because of Pa. Pkt. bahira- * ex- ternal ' (: Skt. bdhya-) derived from it, its form cannot be phonetically determined from Sanskrit itself, and the stem is usually given as bahts . This weakening of final r had no doubt a good deal to do with the abandonment of a large number of the old neuters in r. They were already in decline, but the phonetic weakness of final r no doubt hastened the process (the stem udhas ' udder ' beside tidhar is due to the tendency to con- fuse s and r in final position). VOWELS § 20. Correspondences The correspondences between the vowels of Sanskrit and those of other IE languages may be gathered from the following table, in which the examples are quoted after the presumed IE original vowel ; a : Skt. djra- ' plain Gk. iypos 4 field Lat. ager , Engl. acre ; dpa 4 away, from Gk. a tto, Lat. ab ; dnli 4 opposite, near Gk. avn, Lat. ante 4 before ' ; dniti ‘ breathes ', cf. Gk. avc/xo? ‘ wind Lat. animus ; dyas 4 metal Lat. aes ; nas - 4 nose ' (instr, nasd y etc.), OHG nasa , 0. SI. nosu ; hamsd- 4 goose Gk. Lat. anser, Germ, gans ; yaj- 4 to sacri- fice, worship yajnd - 4 sacrifice Gk. aytoy, ayvoy 4 holy ' ; iad- 4 fall Lat. cadit. e : Skt. bharati 4 bears Gk. pepa, Lat. fert, 0. Ir. berid ; asti 4 is Gk. Ion, Lat. est ; dsva- 4 horse Lat. equus ; sdna- 4 old Lat. senex, 0. Ir. sen ; sddas 4 seat Gk. ISoy ; paiu- 4 domestic animal Lat. pccu> O. Pruss. pecku, Lith. peku$ t Goth, faihu ; mddhu * honey, mead Gk. p.48v, AS. medu ; madhya- 4 middle Lat. medius, Gk. pAoos. o : Skt. dvi~ 4 sheep Gk. oiy, Lat. ovis ; pdti - 4 husband, PHONOLOGY 103 lord Gk. moms, Lat. potis ' able ' ; upas- ‘ work Lat. opus ; anas- ‘ wagon Lat. onus 4 burden ' ; dsthi 4 bone Gk. oareov, Lat. os ; ddma - 1 house Gk. Sd^o?, Lat. domus, O. SI. domu ; daddrsa 4 saw Gk. SeS opKf. vfkas nom. sg. ‘ wolf Gk. Ad*o?. d : mcitdr- ' mother Lat. mater ; bhratar- ‘ brother Lat. /rater ; svddii- 4 sweet Gk. aSd?, ^Sds, Lat. suavis ; dsthdt ‘ stood J , Gk. €crrd, eV?] ; bdhii- 4 arm J , Gk. ; sna- ‘ to bathe Lat. ndr? ; asVd ' mare Lat. e : Skt. rAj-, rdjan- ' king Lat. ; mds- 4 month Gk. firjv, Lat. mensis ; sdmi - 4 half- Gk. y/u-, Lat. semi- ; mi 4 not J (prohibitive), Gk. py , Arm. mi ; pitd 1 father Gk. irar-qp. 0 : as- mouth ’, Lat. os ; vdk 110m. sg. * speech Av, vdxl t Lat. vox ; dsu- 4 swift Gk. <b/<v$, Lat. odor 4 swifter ' ; pa - ‘ to drink J , Gk. tuLvw, Lat. potus ; napdt 4 grandson J , Lat. nepds ; pit nom. sg. ‘ foot Gk. (Dor.) 7 ubs ; vfkas nom. pL ' wolves Goth, mdfos. 1 : Skt. imds 4 we go J , Gk. tfiev ; vidmd 4 we know Gk. ibpLw ; cf. Lat. video ; tisthdmi 1 1 stand J , Gk. larypu ; divi 4 in heaven Gk. AiFL ; rincdnti 4 they leave Lat. UnquunL i ; Skt. jivd- 4 alive Lat. vtvus ; vlrd- 4 man, hero', Lith. 1 rfras ; pivan - 4 fat ', Gk. ttiwv. u : Skt. sritid- 4 heard Gk. kAi/tos ; rudhird- r red Gk. Ipvdpos, Lat. ruber ; snusd 4 daughter-in-law 0 . SI. snitcha , Gk. vvos, Lat. nurits ; uddn- 4 water Gk. v&cop, °aro^ f Lat. unda 4 wave it : Skt. dhunid - ‘smoke', 0 . SI. dymu, Lat. furnus ; bhru- 4 brow Gk. o<f>pv$ f AS. bru ; pit - ‘ to be rotten pdti- 4 putre- - faction J , Gk. 7 tvOw, Lat. pus t puteo, Goth . fills. -v >1 Q c * ai : Skt. edkas 4 fuel Av. aesrna-, Gk. a Wa> 4 burn devdr - ' brother-in-law Gk. h&rip (<SatF7jp), Arm. taigr f Lat. levir. ei : Skt. dti ‘ he goes Lith. eiti, Gk. elm ; heman 4 in winter hemantd - ‘ winter J , Gk. x € ^ lwv > Alb. dimen; devd- 4 god J , Lith. devas, Lat. divus, Osc. deivai 1 divae ' ; dehi 4 embankment, wall Gk. reix o?, Osc. jcihiiis 4 muris oi ; Skt. veda 4 I know Av. vaeba, Gk. ofSa, Goth, wait ; te 4 those Gk. roL ; bhares 4 you should bear J , Gk. <f>tpais. ait : Skt, ojas- ' strength J , Av. aogan ' id cf. Lat. augustus ; sosa - ‘ drying up ', Lith. sausas 4 dry Gk. ados' id AS. star. eu : Skt. bddhdmi 4 I observe Gk. 7T€d^op.at ‘ find out, learn ’ ; osati 4 burns Gk. €vu> ' burn, singe Lat. iiro ; josati Gk. ycdo/x at ‘ taste Goth, kiusan ' choose ^ 'l enjoys r7 j 104 PHONOLOGY ou : Skt, lokd- 1 space, room, world Lith. laukas ' plain Lat. lucus ' grove * ; bodhdyati * he awakes* (trans,), Lith. pa-si-baudyti ' to awake oneself *, O. SI. buditi ' to wake, rouse * ; sunds gen. sg. of sunu - ‘ son Goth, sunaus , Lith. sunaus . di : Skt. dat. sg. fern, sindyai, devyat, Gk. Lat. equae , etc.

Skt. araiksam, s-aor. of rxc- ' to leave cf. Gk. ZXeupa.

di : Skt. instr. pi. vfkais , etc., Av. daevais, Gk. Avkois* ; dat. sg. idsntai 1 to him Av. aetahmai, cf. Gk. iTrrru), etc. du : Skt. naus ' ship Gk. vavs, cf. Lat. navis, eu : Skt. dyaus ' sky Gk. Zzvs ; dyauksam ‘ I joined *, cf, Gk. i^ev^a. du : gaus ‘ cow Gk. ftovs ; ; astau * 8 Goth, ahtau. Sonant Liquids and Nasals : r : Skt. prcchdti 'asks', Lat. -^oscr/ {<*^ 0 rscx/), OHG forscon ; pitfsu loc, pL of pitdr - ‘ father Gk. Trarpdcn ; ty/te- ' turned Lat. versus, vorsus ; mr/a- ' dead ’, cf. Lat. mortuus , wors, Lith. ' to die 0. SI. sumriti ‘ death *. I : Skt. mrdu- ‘ soft Lat. mollis, cf. Gk. dpaASvvto ' soften, weaken ' ; prthu- ' broad *, Gk. nXarvs ' flat ' ; ' w r olf *, Av. vahrka Lith, vilkas , Goth, wulfs. n : Skt. raa/a- f thought, considered *, ma/f- ' thought, idea * (man-), Gk. a vrofiaros ' of one's own accord Lat. commentus , mens, mentio , etc. ; Aa/a- ' slain ' (Aan-), Gk. ^aro? (: <f>ovo$ t etc.) ; asf- ‘ sword Lat. ensxs ; nima 4 name ', Gk. ovopa, Lat. nomen, Hitt. Idman ; a- ' not ' in djnata- ' unknown Gk. ayvcura?, Lat. ignotus , 0. Ir. ingnad. ip : Skt. ' hundred Gk. Ikcltov, Lat. centum, Goth. Awnrf, Welsh can/, Lith, Hmtas ; gtf/x- * going ', Gk. /Jam?, Lat. in-ventio , Goth, gaqumps ; abhrd- ‘cloud*, Av. awra~, Lat. xm&er ; so^/a ' seven Gk. in ra, Lat. septem. §2i. Notes on the Vowels The most characteristic distinguishing feature of Indo- Iranian as opposed to the remaining IE languages is the posses- sion of only a single vowel a corresponding to the three vowels a, e, o elsewhere, and likewise in the case of the long vowels, a corresponding to d, e, o. It is clear that this uniformity is due to a special Indo-Iranian development, since the other languages PHONOLOGY 105 are in substantial agreement with each other in the distribution of the vowels a , e, 0 . Furthermore the palatalisation of the velar series which occurs in lndo-Iranian before a only when it corresponds to e in the other languages (ca — Lat. que, etc.) testifies to its existence in these positions in the prehistoric period of Indo- Iranian. The confusion of a and 0 is found also outside lndo-Iranian, in Germanic, Slavonic and Hittite. It is not possible to say for certain whether we have here independent parallel development in the various language groups, or whether th is fus ion of 0 and a is an ancient dialectal feature of Indo - European. Certainly in the case of lndo-Iranian and Slavonic, which show other signs of special affinity, the possibility of an ancient common change is deserving of consideration. The change e to a o n the ot her hand is found only in lndo-Ira nian, and it is one of the most characteristic features distinguishing this family from the rest of Indo-European, The lndo-Iranian development of the sonant nasals (to a) is the same as that of Greek, and it is one of the several features that links these two branches. Sonant nasals as such are found nowhere, but have been reconstructed for Indo-European from theoretical considerations. The sonant liquids have in the same way been replaced in most languages by combinations of vowel + r or /. Only lndo-Iranian preserved the vocalic r, which repre- sents also original vocalic l. In Sanskrit there exists only one case of vocalic l , namely the root kip - 4 to arrange Because of Vedic kfp Av. kshrp- 4 form, body which are usually compared with Lat. corpus it is generally considered that this / is of secon dary origin , but this is not altogether certain. Neverth? less as a general rule Sanskrit is much more consistent in turn* ing / into r in its vocalic form than in its consonantal form. On the basis of sonant f (which is attested in lndo-Iranian) and the sonant nasals can be safely reconstructed. They occur in the same conditions, that is to say by the suppression of the associated guna vowel which leaves them to function as vowels, and their treatment in various languages is similar. Thus we have for f in Gk. ap{pa), in Balto-Slavonic ir and in Germanic ur similarly for n Gk, a, B. SI. in, Germ. un. It is clear that the assumption of original sonant nasals is as much necessary to account for the variation in the associated vowel in the various languages as it is by the principles of apophony which are briefly noted below. io6 PHONOLOGY Among the vowels of Primitive Indo-European it has been customary to postulate the so-called ‘ shwa ' (2). This is based on such comparisons as Skt. pitdr - r father ' : Gk. mrrjp, etc. Skt. sthitd - ' stood 1 : Gk. oraros, etc. In such cases the 9 was considered to represent the reduced grade of the original long vowels, corresponding to the zero grade of the short vowels e, a t o. It was supposed to have become i in Indo-Iranian, and a in all the other IE languages. I have shown elsewhere 1 that this reconstruction is without justification, and that it was due to a faulty analysis of the Sanskrit words concerned. In these words the i is IE i and it is part of the suffix, not part of the root. Skt. sthitd- should be analysed sth-itd and its formation therefore differs from that of the related words, so that the phonetic reconstructions based on these comparisons become void. The same analysis is to be adopted in all the relevant forms : sth-iti- ' standing ’ (cf. snih-iti-) sth-ird - f firm ' (cf . sthdydn, stheman, Pa. theta-), aor. 3 sg. asth-ita 4 stood ' (cf. avad-i-ran, etc.), perf. I pi. dad-ima (contrast pres, dadmds), stan-i-hi 1 roar ' (cf. stanayitnu -, etc.), s-itd- 4 bound ’ (cf. sindti , sisdya , etc.), s-itd- 4 sharp " (cf. Ved. sisayd Av. saeni, etc.), krav-is- 4 raw flesh ' (cf. roc-ts- f etc,, and Lith. kraujas, etc.). It is also clear, and established by many examples in Sanskrit that in the zero grade the original long vowels are completely elided : e.g. in the present tense of dd and dhd t dadvds , dadmdq, datU f daisi , dadhvds , dadhmds , dhatse , dhatsva , etc. (likewise in Iran- ian. Av. dadomahi, dasia f daste, dazde, dadamaide, etc.) ; the same elision is found in the participles dattd r given 1 and °tta (devdtta- 4 given by the gods etc.) and in Av. ptar- 4 father ' beside p-itdr. If this 9 had been confined to the comparatively few words in which Sanskrit i appeared to correspond to a in the other lan- guages, it would never have acquired very great importance in Indo-European theory. It was due to its becoming a basic element in the early theories of apophony that it acquired such importance in the traditional theory of Indo-European. In the comparative dictionaries this 9 , so insecurely founded, appears in the utmost profusion in IE reconstructions, particularly in the case of the so-called disyllabic roots. Skt. i is also suffixal when it appears after such roots and the h which constituted the final element of the root is elided {tdritum<*tam-itum). The 1 TPS. 1949, pp. 22-61. PHONOLOGY IO7 theory of apophony was further complicated by the invention of original long diphthongs, possessing a weak grade n which was held to have developed into % (sometimes into -ay-), but there is nothing in the facts to justify the assumption of such long diphthongs or of the weak grades which are supposed to be de- rived from them. In addition a second ' shwa supposed to be a reduced grade of the short vowels wais introduced by certain authorities. As a result the theory of apophony, which, as will be seen below, is really of the utmost simplicity, became extra- ordinarily complicated. With the discovery of Hittite A, and the subsequent rise to popularity of the laryngeal theory, the main features of the old theory were transferred to the new. IE h was identified with the old shwa ( 2 ), and it was believed that all its varieties could function in a vocalic as well as a consonantal function like the liquids and nasals. It has even been common to use the sign 9 to indicate IE h in its conson- antal function (%, ? 2 , 2 a ), and the whole presentation of the laryngeal theory has continued to be vitiated by the original error of the invention of * shwa Needless to say the objections that apply to * shwaT* in the old form of the theory apply to it with equal force in the new. There is no satisfactory evidence to show that H in stay of its varieties could function as a vowel and it is certainly never represented in Sanskrit by i. The effects of IE h on the vowels have already been noticed. By the restoration of h a very considerable simplification of the vowel system is achieved. (i) The long vowels d t e t 0 (>Skt. a) may be long through vrddhi, in which case they have developed out of the short vowels a, e f 0. But there is another series of long vowels which are long by nature, e.g, the d, e, 0 in std- 1 to stand dhl- f to place ’ and do - ‘ to give ’ (Skt, sthd -, dhd- t da-). In such cases the laryngeal theory analyses the long vowel into short vow r el + several varieties of h (dheu lt ste h 2 , deu 3 ) the quality of the vowel being determined by the following laryngeal. Thus in all cases long vowels are of secondary origin. (ii) The varieties of guna vowel are partly due to qualitative alternation in Indo-European. This was particularly so in the case of the alternation ejo (</>epo> : ^dpo?). But some cases of 0 are left over which have been considered to be original (dcrreov ' bone ') and a can only rarely be put down to vocalic alterna- tion (Lat. quater, etc.). Cases of ' original * a, and 0 according to IC-f w ' 108 PHONOLOGY the laryngeal theory go back to and h 3 followed by the guna vowel which was in itself undifferentiated (h gnU 4 front Hit t. haul Gk. etc., u^est- 4 bone Hitt, hcistai , Gk. otrreW, etc.). Thus we are reduced to a single original guna vowel, conventionally written which is the state of affairs to which Indo-Iranian again returned at a later period as a result of special developments of its own, A few words of caution should be added in illustration of the fact that the laryngeal theory has not yet acquired a completely satisfactory form. It is never possible to be certain for instance that the vowel o is original, since alternating e- forms may be missing by accident. Furthermore there exist some oja alterna- tions which the theory does not altogether account for. As re- gards original a the absence of any h in forms like Hitt, appa ' away ' can only be explained away by making the theory un- comfortably complicated. It must be admitted in such a case that the actual evidence available does not allow us to go any further than IE apo. (iii) IE i r u have in all cases developed out of Hi, rm. The special developments of r, /, n, m followed by h have already been outlined. In this way the old reconstructions of long sonant liquids and nasals can be dispensed with. These simplifications effected, the IE vowel system is reduced to very few primitive elements. There is only one purely vocali c element t o begin with, which may be written e . The develop- ment of three varieties [e, a , o) and of the corresponding long vowels can be explained on the basis of the effect of laryngeals and of vocalic alternation. In addition there are six elements which may under certain conditions (between consonants, initi- ally before, and finally after consonants) function as vowels — t, «, f , /, 19, ip — but elsewhere (between vowels, etc.) function as consonants — -y, v, r, l ,n, m. As regards diphthongs it should be noted that the second element is consonantal, and that from the point of view of Indo-European it would be more consistent to write eyti 4 goes gews - 4 taste etc. §22. Quantitative Alternation: Apophony The purely vocalic element (Skt. a, IE a, e> o) was subject to a quantitative gradation of the following type. It could be elided in any syllable, radical or suffixal, or alternatively it could be K* c-y* --w 1 PHONOLOGY 109 lengthened. In other words any syllable may appear in the normal grade (a), the strengthened grade (d), or the zero grade. This gradation is of fundamental importance in Sanskrit grammar, and its importance was fully recognised by the Indian grammarians. They gave the name vrddhi to the strengthened grade and guna to the normal grade. The weak or zero grade they did not name because they cpnstructed their grammatical system in such a way that they started from the zero grade as the basic grade and from this they derived the guna and vrddhi grades by two successive processes of strengthening. The com- parative philologists differ from the Indian grammarians in that they regard the guna as the normal grade and from it derive the vrddhi and zero grades by the opposite processes of strengthen- ing and weakening. The operation of this gradation may be illustrated by a few examples : (1) Normal grade : sddas 4 seat sacate ‘ associates with padds, gen. sg. of pad - ' foot ghas- ' to eat J , dabhnoti 4 injures hdsati 4 laughs ^ (2) Extended grade : sdddyati 4 causes to sit rdtisicas nom. pi. ' associating with liberality pidam acc. sg. * foot ghdsd - ' fodder J , dddbhya- 4 that cannot be injured hdsa- 4 laughter (3) Zero grade : t sedur 4 they sat ' < *sazdur f cf . Av. hazdydt ' would sit sdicaii 3 plur. ' they associate upabdd- 4 trampling under foot a-dbh-uta- 4 wonderful' (literally ‘ that cannot be harmed, impregnable of divine beings), jaksiti 4 eats J (be. ja-ghs-i-ti) , jdksati 4 laughs ' (ja-hs-ati, cf. Vedic jdjjhati and Pa. jaggff&ti for different treatments of A + $). The same three grades apply to all suffixal elements. Thus in the case of the n-suffix we normal grade (guna) in voc. rijan f loc. r&jani zero grade in gen. sg. rdjnas, extended grade in acc. sg. rajdnam from the stem rijan - ' king The same grada- tion applies to all suffixal elements. Fundamentally this alternation a/djze to is all there is to the system of apophony. Some complications are caused by the combinations of a with semivowels, etc., and by some phonetic changes. These may be briefly summarised as follows : (i) When a is lost the semivowels (y t v) assume their vocalic form in the appropriate phonetic context : ydjati 4 sacrifices ' ; ijya 4 sacrifice ' ; vdpati 4 sows ' ; uptd - * sown When the semivocalic element comes second, i.e. in the diphthongs, the no PHONOLOGY original Indo-lranian alternation ai , ai, i, an, au , u is modified in Sanskrit to at, e , i ; aw, 0, « of which e and o ceased to be diphthongs in pronunciation. The guna and vrddhi grades acquire the alternate forms ai, au, e, ojdy , dv, ay, av according as a consonant or vowel follows. Exx, Normal grade ; jdtum ‘ to conquer ' r jdyati ‘ conquers srotnm 4 to hear srdvana- 4 hearing Strengthened grade : ajaisam ' I conquered 1 (s.aor.), jigdya ' he conquered ' (perf.), dsrattsam * I heard susrava ‘ he heard Zero grade : jitd- 1 conquered srutd- ' heard (2) The liquids r , / were vocalised under the same conditions. Though l has been mostly merged with r Indo-lranian preserves the original sonant pronunciation, so the apophony remains simple. Guna : kdrturn 'to do 1 ; Vrddhi : cakara ' did ’ ; Zero : krtd ‘ done * jcakre 3 sg. perf. atm. ‘ did (3) The nasals were likewise capable of functioning as vowels, but here the situation is complicated by the change, in Indo-lranian as in Greek, of the sonant nasals to a. The series is therefore (1) an, am, (2) an, dm, (3) ajn, ajm, e.g. (1) gdmana- ' going hdnti ' slays (2) jag dm a ‘ went ', jaghana * slew (3) (a) hatd- ' slain gatd - 4 gone (b) ghndnti ( they slay jagmur 4 they went (4) Long d is sometimes original, that is to say it appears in the guna position, e.g. in the roots dha 1 to place id f to give and sthd 4 to stand It is elided in the zero grade like the ordinary guna vowel, e.g. dadhmds r dadmds . We have seen that this d is for aw (or in the IE system l, d, 0 are for ew x , en 2 , en t ). Thus we are dealing with the ordinary guna vowel in this apophony, and h which cannot function as a vowel but is elided in such positions. (5) The combinations in, 11H resulted in t, u, while in com- bination with corresponding diphthongs (-mi-, -euu-) the H disappeared without trace. So there arises an apophony cl > ojii, etc., beside the normal diphthongal apophony : (1) ndtum ‘ to lead ' jndyati f leads ’ ; homan ' invocation ' jhavana - ' id (2) anaisam 4 I led ' jndyaka- 1 leader 1 ; juhdva 4 called (3) nitd - ' lead hutd- 4 called (6) Sonant r followed by h resulted in Ir, ur, while in the corresponding guna grades it disappeared. In these cases we have the weak grade lr, ur in apophony with ar, dr : (1) tdrtum PH ONOLOGY III ‘ to cross pipdrti 4 fills ' ; (2) tarayati 4 causes to cross ' ; (3) tirnd- 1 crossed purnd - ‘ filled (7) When n and m were followed by H the result in Sanskrit is in the first case a y in the second case -an- (examples occur only before t). Hence the apophony sandfi : said- ; damyati , damdyati, ddnta (8) There are some deceptive cases where no real apophony is involved. An example is pd-tum 4 to drink 1 ; p-itd - ' drunk Here the i of the second form is suffixal and therefore cannot be in apophonic relationship to the radical d of the first form. (9) Roots consisting of more than two consonants admit of two types of guna grade : (1) variate 4 turns etc., (2) trdsati 4 is afraid etc. Usually roots belong to one or the other type and keep to this in the guna grade (and in vrddhi which follows the guna in this respect) but double forms occur in some cases, e.g. from drs- 4 to see " we have the series ; Weak grade drstd guna 1 darsdyati, guna 2 draksyami , vrddhi 1 ddrsanika vrddhi 2 adraksam . (10) Final vrddhied forms terminating in semivowel, liquid or nasal may lose this final element : sdkha 4 friend ' (acc. sdkhdyam , stem sdkhi -), dsmd 4 stone J (acc. dsmdnam , stem dsman-), ddtti 1 giver*' (acc, ddtdram , voc. ddtar p dat. ddtre, etc.). The tendency is found elsewhere in Indo-European (Lat. sermo , etc.) but nowhere as consistently as in Sanskrit (e.g. Gk. TiaTjjp, Lat. pater beside Skt. pita). This vocalic gradation was connected with the Indo-European accent. In Sanskrit the connection between alternation of grade and alternation of accent is clear from many examples : imi ‘ 1 go ' : ini as ‘ we go ' ; srotnm ‘ to hear J : irutd - r heard 1 ; dsti 4 is 1 : sdnti 4 are ’ (Lat, est, sunt) ; hdnii 4 slays * : ghndnti 4 they slay ' (Hitt, kuenzi : kunanzi) ; dtti 4 eats ' (ad-ti) : ddnt - ‘ tooth ’ ( ( eater '). From such examples it is clear that the zero grade is due to the unaccented position of the syllable, and that the guna grade is properly the grade of the accented syllable. There are of course many examples in Sanskrit, as in other languages where accent and apophony do not agree, e.g. Skt. vfka- 1 wolf fksa- 4 bear tfna- 1 grass vipra - ‘ sage, brahmin These however create no difficulty since it is known that in many cases the position of the accent has changed in course of time. This is obviously the case in the examples quoted since they are all in origin adjectival formations (e.g. 112 PHONOLOGY tfna - ( <*trnd -) is ' what pierces cf. trndtti) and it was the rule that such formations were suffixally accented. It is also very common in Sanskrit for nominalised adjectives to throw back the accent on to the first syllable. The application of the above accent rule in its full rigidity would allow only one guna syllable in any word. The words quoted are of that type, but the majority of Indo-European words, in any language, are not so. This is mainly due to two reasons. Firstly, when inconvenient or grammatically less clear forms would result, the elimination of the unaccented guna vowel was resisted, or if eliminated it was quickly restored. So we have as the gen. sg. of pad - ( foot 1 not *bdas which would have resulted from the rule, but padds with guna vowel in un- accented position. The existence of ddnt- * tooth ' (' eater cf. sdnt- ‘ being ’ : as-) beside ad ant- ' eating ' gives us one clear case where a guna vowel in unaccented position has been re- stored by analogy. Secondly the nature of the Indo-European accent underwent a change during the later Indo-European period. It had the power to reduce neighbouring unaccented syllables for a certain period of time, and then, in later Indo- European it ceased to have this effect. Consequently forms like those quoted above which show the full effects of apophony must be considered as belonging to the most ancient stratum of Indo-European. But after the accent ceased to have the effect of reducing adjacent syllables, Indo-European was creating new formations in abundance, a faculty retained by the indi- vidual languages particularly in their early stages. The very numerous formations of the type yajata- ' adorable dariatd- ' worth seeing devdsya ' of the god etc,, etc., had their origin in this later period when the accent had ceased to have the power to influence the vocalism of the surrounding syllables. §23. Qualitative Alternation: Metaphony There existed in Indo-European also a qualitative alterna- tion of the guna vowel, and this is well preserved in most branches of the family : e.g. Gk. Adytu ‘ I say 1 : Aoyos- ' word 1 ; Lat. lego ' I cover ' : toga * gown ' ; Russ, vezu 4 I carry * : voz 1 cart, load * ; Engl, sing : sang . The alternation affects both the guna vowel, as in the examples above and its vrddhied ex- tension (Gk, SoTrjp : Sturiop ' giver *). In Indo-Iranian this alternation has entirely disappeared owing to the confusion of PHONOLOGY H3 the vowel qualities a, e, o in a. Consequently this Indo-European alternation has no significance for Sanskrit grammar, and it de- serves brief mention only because the student of the compara- tive grammar of Sanskrit will meet it in the material cited from the related languages. This alternation, like the quantitative alternation is clearly connected with the Indo-European accent. This is evident from the juxtaposition of such forms as Gk. Saipuov, Sat^tovo? on the one hand and 7 Toi/njv, TroipLtvos on the other. The rule is clear that e is the normal grade of a syllable which bears the accent and has always borne the accent (Gk. eon, eiros , Wos, etc.). Accentual changes and the workings of analogy have to some extent contrived to obscure the picture but this central fact remains beyond doubt. An example of the working of analogy may be mentioned ; the termination of the genitive singular appears in some languages in a form that represents IE -es, in others in a form that represents IE -os. Since this termination was sometimes accented and sometimes unaccented, we may reasonably assume that the two forms were originally differ- entiated according to accent. Later in the individual languages one form was generalised, sometimes the -es form and some- times the - os form being chosen. The fact that the IE accent should have two quite different effects is bound up with what has been said above about the accent. The elision of the guna vowel was frequently resisted for morphological reasons, or if eliminated it was restored. Such retained or restored guna vow'd s w r ere then, possibly at a later period, affected in a different way by the accent, so that 0 appears in place of e. Or again the main accent of a word may have changed with the result that the vocalism of the syllable which lost the accent was altered. For instance the numerous words of the type represented by Gk. Salfiojv belong to a class (agent-nouns) which was originally suffixally accented. The type of formative -nop, -pcov, etc., beside older -r^p, -p-jv seems to have come into existence as the result of such an accent shift. §24. Sanskrit and Indo-European Accent The last two sections illustrate the importance of the part played by accent in Indo-European. In dealing with the morphology the accent is an indispensable element, without PHONOLOGY 114 which no proper grammatical analysis can be made. For this reason the accent will be continually under discussion during the succeeding chapters. Here a few general remarks will suffice. The full technical details of the Vedic and early classical accent of Sanskrit, and of the various methods used to denote them, are somewhat complicated but the main principles are as follows. Each word had normally one accent whose position varies from word to word. Any syllable from the first to the last may bear the accent (e.g. dpaciti ' retribution dhdrdyati 1 holds/, namasydti 1 respects ' and aparahnd- ‘ afternoon ’ are accented on the first, second, third and last syllable respec- tively). No simple set of rules can be given to determine on which syllable of a word the accent will fall. Certain words were enclitic by nature and never bore the accent. These are such particles and pronominal forms as ca ' and md 1 me me * of me etc. Elsewhere the accent might be dropped in certain circumstances. (1) In the vocative a noun lost its accent except at the beginning of a sentence, when it was accented on the first syllable regardless of its natural accent. (2) The finite verb in the main clause of a sentence is unaccented unless it appears at the beginning of a sentence, in which case it retains its natural accent. In dependent clauses it retained its accent whatever its position. In this case a verbal preposition is most commonly compounded with the verb and loses its accent, e.g. prd gacchati ‘ he goes forward yadi pragdcchati ‘ if he goes forward The accent so indicated is termed by Panini udatta- 1 raised ' and the rise was one of pitch or musical tone. The main accent affected also the pronunciation of the following syllable, since the return of the voice to the normal level was effected during the enunciation of this syllable. The accent of the syllable immediately following the udatta is termed svarita- and it is described by Panini as a combination (samdhdra-) of udatta and anudatta. That is to say it begins at the high pitch of udatta and descends in the process of utterance. There exists also an independent svarita which arises secondarily out of the contraction of iya to yd, etc., in which case the main accent of the word is the svarita. This is a post- Vedic development since the metre of the earlier texts shows that the contraction had not yet taken place. PHONOLOGY 115 The main accent affected the pronunciation of the preceding syllable. This was pronounced lower than normal and it is termed by Panini sannatara-. The remaining unaccented syllables were termed anudtitta-. Thus out of one main accent of a word there arose four different varieties of pitch : udatta anuddtta-, svarita-, sannatara -. Since however all this variation is dependent entirely on the main accent, only that needs to be noted, as above. A separate notation is needed for the inde- pendent svarita (vrkyds, tanvds , etc.) but even that may be dispensed with for the Veda if the words are transcribed accord- ing to the pronunciation [vrkiyas, tanuvas ), The complications of the accent detailed above were respon- sible for an unnecessarily complicated system of notation adopted by the Vedic schools. According to the usual system, that adopted in the Rgveda for instance, the principle is to mark the syllable preceding the udatta, the sannatara, with a sub- script line, and the dependent svarita following the udatta by a vertical stroke above. The udatta itself is left unmarked. This achieves the same purpose in a less convenient manner than the method adopted in modern transcription, and by some Vedic schools. The modern recitation of the Rgveda follows the notational system to the extent of pronouncing the sannatara lowest and the svarita highest musically of syllables and ignor- ing the udatta altogether. This is a secondary development although it may be old, and at variance with the teachings of Panini which are in complete agreement with the findings of comparative philology. The system of accentuation described above has for centuries been totally extinct in spoken Sanskrit as it has in all forms of Indo-Aryan derived from it. When exactly the accent died out in ordinary spoken use it is impossible to say with certainty. It was certainly a living thing in the time of Patanjali and even later than Patanjali, Santanava treated of the subject in his Phitsutra . According to the author of the Kdsikd commentary (1 c . a.d. 700) the use of accentuation was optional in the spoken language, which probably means that in practice it was no longer used at this time. On the whole it is unlikely that the use of accentuation survived long after the Christian era. In Middle Indo-Aryan we may take it that the change occurred much earlier, at the very beginning of anything that could be called Middle Indo-Aryan. PHONOLOGY il6 The old Indo-European accent was lost, at some time or other in most Indo-European languages just as in Indo-Aryan. Accentual systems derived directly from Indo-European are found only in Greek and Balto-Slavonic. They may also be de- duced from the phonetic developments classified as Verner’s law for an early stage of Germanic. The existence of accurate information about the accentuation of two of the oldest mem- bers of the family, Sanskrit and Greek, is of the utmost value for the understanding of Indo-European. A comparison of the accentuation in those languages in which it is preserved reveals basic agreement, though to a greater or lesser extent all languages have innovated in detail. The position of the accent in Indo-European for instance is fre- quently established by the correspondence of Sanskrit and Greek : e.g. bhdrati ' bears 7 : Gk. <f>ipti ; srutd- * heard 1 : Gk. kAuto? ; guru- ' heavy 1 : Gk. papvs ; vacas * word ’ : Gk. tiros, etc., etc. In other cases they differ showing innovation on one side or the other : mdtdr - 4 mother 7 ; Gk, pL-qrqp ; bdhu- 4 arm : Gk. nfjxvs, etc. The tendency to innovate is also evi- dent from the frequent disagreements between accent and apophony, whether in individual languages (Gk. *8pev ‘ we know 7 as opposed to the more original accentuation of Skt. vidntd) or in all (Skt. vfka-, Gk. Xvkos, etc.). In the latter case the innovation is of the Indc-European period. By means of comparison of the individual languages, by the study of apo- phony so intimately bound up with accent, and by the proper understanding of the part played by accent in the morphology, it is possible to form a clear and accurate idea of the Indo- European accent. The details are part of the morphology and will be found in the chapters concerned. The nature of the old accent in Sanskrit and Greek is known from the technical descriptions handed down and partly in the case of Sanskrit from the traditional recitation of the Veda. It was in both languages predominantly musical, and not a matter of stress. This is confirmed by the fact that in both languages metre is completely independent of accent, depending solely on the length of syllables. From this agreement it is deduced that the same kind of accent prevailed in late Indo-European. But, as we have already seen, there must have been a change be- tween early and late Indo-European in this respect. Earlier the accent had the power to reduce the neighbouring syllables. PHONOLOGY 1X7 indicating a strong stress element. In the later period this power was certainly lost and this agrees with what is known about the accent of Sanskrit and Greek. Beside the normal acute accent Indo-European possessed under certain circumstances a circumflex accent. This is clear from the agreement between Greek and Lithuanian, e. g. cir- cumflex accent in gen. sg. fem. Gk. deas, Lith. gerds, gen. pi. masc. Gk, 9eu>v, Lith. gent, instr. pi. masc. Gk, deols, Lith, vilkais, as opposed to acute accent in norn. sg. fem. Gk. 0ea. Lith. gero-ji ( gera ). In such cases Sanskrit has the ordinary udatta accent'as elsewhere, and it does not, as Indo-European did, distinguish between the two types of accent. The inde- pendent svarita which came to exist in Sanskrit as a separate type of main accent is, as we have seen, a post-Vedic creation and unconnected with differences of accent type in Indo- European. Nevertheless traces of the old circumflex have revealed them- selves in the Veda from a study of the .metre. In certain cases the metre makes it clear that a long d is to be pronounced di- syllabically, e.g. gam , dydm as gaam } dyaam, and the ter- mination of the genitive plural -dm as -aam. In such cases the corresponding Greek forms frequently have the circumflex accent, and this gives reason to believe that metrical pecul- iarity of the Veda is the effect of the circumflex accent of Indo- European. THE FORMATION OF NOUNS §i. General Remarks The Sanskrit nominal stem may coincide with the root, as happens in a minority of cases, but usually it is derived from it by the addition of a suffix. These suffixes are very numerous and are inherited from Indo-European. They are not, as occurs in some languages (e.g. Engl, man-ly, man-hood) derived from what were originally independent words, but are in every case analysable into their component parts, that is to say the indi- vidual consonants or semivowels of which they are composed. These primary elements include nearly all the available phonemes, but the ones most commonly used are r, n, s, t, yji , vju, m, h and h. They may appear either with guna, i.e. preceded by the thematic vowel (-ar, -an, -as, IE er-, en-, es- f etc.) or in their weak form (- r , -n, -s). The thematic vowel itself may appear as a suffix but naturally, since elsewhere it is always a question of the guna grade of a consonantal suffix, only in final position ( bhav-an-a -, udr-d, etc.). The IE primary suffixes could be added either to roots or to words already end- ing in another suffix : e.g. the suffix -as is added to the root in vdcas - ‘ speech to a base having the suffix )i in rcknas- 1 inherit- ance, property the suffix t in srotas- ‘ stream * and the suffix v in fiivas- ' fat Since the root itself could originally function freely as a noun, that is to say was a word in the full sense, there is no difference in principle between primary and secondary derivation of this kind. A suffix could be added to any word, whether it already had a suffix or not, and the nature of the process was precisely the same. The result was that owing to the very large number of possible combinations of the primary elements, the number of these compound suffixes in all IE lan- guages is very large, and the complexity of nominal stem formation in Sanskrit and the allied languages is entirely a matter of the multifarious combination of a comparatively small number of primitive elements. THE FORMATION OF NOUNS IX 9 In the descriptive and synchronistic grammar of Sanskrit the various suffixal combinations are treated as units, which is what they have in fact become in the course of the develop- ment of the language. For the historical and comparative treatment of the subject a more radical approach is needed. Here it is necessary, as is done in the following pages, to start from the single, primitive elements, and in the exposition to build up the whole suffixal system from them in the way that it had developed in the prehistory of the language. Between the original simple suffixes, as so analysed, no dis- cernible distinction of meaning or function can be found. In some ways they have no meaning. Thus an ancient IE word wet ‘ year ! appears in Hittite as such (also in Sanskrit reduced to -ut in parut f last year ') ; in Greek it appears with the suffix - os as {F)ctos, witho ut anything being ad ded to the meaning ^ Likewise in Sanskrit neuter action nouns with suffixes (-0$, etc.) do not differ fundamentally in meaning from roots used in the same sense : dvls- ‘ hatred 1 : dvdsas ‘ id/, etc. Of course when several words derived from the same root with different suffixes appear side by side, differences of meaning between these words usually develop, but this is a matter of idiom and nothing to do with the ultimate nature of the suffixes as such. What applies, to the simple suffixes applies equally to the compound suffixes. In the examples quoted abdve the suffixes -tas t -nas and -vas are used in precisely the same way as the simple suffix -as. Hittite has a series of suffixes -sar t -tar, -mar, -vat (with variant n-stems) making neuter action nouns of exactly the same nature, and this accurately reflects Indo-European usage. In the course of time specialisation of usage in the case of various suffixes has developed in all languages, but this is secondary and it is possible in some cases to show how it has come about . 1 The most important distinction in nominal derivation in early Indo-European was not between the different suffixes simple or compound, but in a difference of accentuation accord- ing to which a word formed with the same suffix functioned either as an action noun or agent noun/adjective. Accented on the root it was an action ^noun and neuter , accen ted on the suffi x itjwas an a gent noun or adjective and originally of the so-called ‘ common gender *. The system is preserved to some extent in Sanskrit and is exemplified by such doublets as brahma n. 1 E.g. in the case of the comparative suffix- tara, see p. 149. 120 THE FORMATION OF NOUNS 4 prayer ’ : brahmd m. ' priest ydsas n. 4 glory 1 : yasds - m. 1 glorious The Sanskrit examples are not very numerous, and are only found in the case of a small number of suffixes ; they are in fact the last remnants of a system dying out. In earlier Indo-European on the other hand the system was of very great extension and importance, and it is fundamental to the under- standing not only of the formation of nouns but also of their declension. The ^hematic vowel s tands apart from the other suffixes in many way s . Its original fun ction seems to have be en to produce age nt no uns orjtdjec tives from the j^ arious prim iti ve neut er ^tK^^nouns, e.g. udr-d ‘ otter ' : Gk. vSajp * water It was in fact an u alternative method to the above in the formation of such nouns. This , is it s_ normal use in Hittite, which indeed ignores the method indicated above. "The numerous neuter t h ematic st ems which are only enlargements of simple consonantal steins (Skt. anjan-a- n. : Lat. iinguen , etc.) appear to be a later develop- m^l^ajQil^i^Li CT o re d by Hittite. In the descriptive grammar of Sanskrit nominal derivatives are divided into two major classes, primary and secondary, in the terminology of the Indian grammarians krt and taddhita. The' former comprises all those formations which are derived directly from a root by means of a suffix (e.g. vdcas ‘ speech * from vac-) and the second those which are derived from the basis of nouns already made (e.g. dsvavant- ‘ possessing horses ’ from dsva- ‘ horse '). Cohvenient as this twofold classification is from the point of view of Sanskrit itself, it has no fundamental or ancient significance from the point of view of Indo-European. For one thing the same suffix is found functioning in both ways, and when a suffix is found to function predominantly or even exclusively in secondary derivation, it is historically a case of secondary specialisation. The suffix -vant is normally a second- ary suffix in Sanskrit, but it is primary in such examples as drvant - 1 steed yahvdnt - ‘ young also in Av. bzzvant- 1 abun- dant zrzzvant- ' straight and in Hitt. dasSuwant - 1 strong It was as a primary suffix that this, like other suffixes which have become predominantly secondary in Sanskrit, first came into being. A historical account of nominal stem formation must therefore be arranged entirely according to the external form of the suffixes concerned. Secondly many formations which from the point of view of THE FORMATION OF NOUNS 121 Sanskrit are primary, are in origin secondary formation. Thus udrd - 1 otter ' which counts as a primary derivative (ud-rd-) is clearly in origin a secondary derivative meaning ' connected with water, water- animal ’ ( udr-a cf. Gk. vSwp n. ‘ water etc,). In origin all types of such thematic adjectival derivation are secondary, but owing to the obsolescence of the old neuter action nouns on which they were based they acquired the nature of primary derivatives. In so far as such suffixes re- mained living suffixes, they were employed as units in the later period in a primary way. Not all formations in -ra have the same history as udr-a etc,, but the type of derivation came into existence in this way. Again in the old IE dichotomy of the types brahma : brahmA we have in the second of these pairs a type of secondary deriva- tion. Logically and presumably historically the neuter action noun precedes the agent noun. The form brahma ' one connected with brahma * presupposes by its meaning the existence of the more primitive neuter. In the Sanskrit system such agent noun formations count as primary formations, and this is what the majority have become owing to the disappearance of the corres- ponding neuter types. This is illustrated very well by the agent nouns in -tar ( ddtar - ‘ giver etc.). Hittite has nouns in - tar but only neuter action nouns. It is clear that the relation be- tween the two types is the same as that between brahma and brahmA and that ddtar - was originally ‘ one connected with giving J corresponding to an obsolete n d&tar n. ' giving When the neuter type went out of use it became a primary formation connected directly to the verbal root. In the development of the system of nominal stem formation in the prehistoric period, certain general tendencies , will be observed, notably : (i) The^decline^of The neu ters. Whole categories of neuter nouns with ancient IE suffixes such as -er and -el have become almost extinct in most IE languages except Hittite ; but the letters r and l play a great part in IE nominal derivation, so that although the original types have disappeared, they have left great masses of further derivatives winch cannot be explained without them. In other cases the old neuter nouns have not dis- appeared, but have been transferred bodily to the masculine and feminine classes. This is particularly the case with stems in i and u, and the action nouns in -d } but it occurs frequently elsewhere* 122 THE FORMATION OF NOUNS (2) The g rowth of grammat ical gender. In the earliest period the threefold classification did not exist. There was no feminine and nouns were divided into two types, 1 neuters 1 and ' com- mon gender*, the latter so called because the masculine and feminine developed out of it. This is the state of affairs actually found in Hittite, and it is further confirmed by many survivals in other languages (Lat. ferens masc. and fern., SkL suvdsds nom. sg. m. and fern., etc., etc.). The feminine gender arose in the later period of Indo-European, and strictly speaking only then is it possible to speak of gender in the proper sense of the term (3) The great variety of possible suffixes that could arise from the various combinations of the primitive suffixal elements led necessarily to a process of selec tion, so that many combinations which are known to have existed have not survived to the Vedic period. Thus out of a series of suffixes forming neuter nouns, and based on the primitive suffixes -er and -en, namely IE -erjr, -merjmr, -werjur, -yerjir, -terjtr, serfs? and -enjn, menjmnj , wenjun , -yen jin, - tenjtn , - senjsn , only -men jinn remains as a" living suffix in Vedic in the formation of neuter nouns. The others are better represented in agent-noun and adjectival derivatives which is in accordance with what has been said above about the decline of the neuter formations. (4} Another feature of great importance is the growing use and extension of the them atic vowel ( Skt. a, IE ejo ) as a final suffix. It has been saffi^bove that the original use of this vowel as a suffix was probably to form adjectival derivatives, in which case it was accented. Later its use spread as an extension of consonantal stems. This tendency is well known in the further development of Indo-Aryan (Class. Skt. pada- ' foot * replacing pad-jpad etc.), but it had been actively at work for long in the prehistoric period. Thus Skt. dnjana- n. 1 ointment ' replaces an older consonantal stem which is preserved in Lat. unguen. As a result of this development thematic stems became by far the most numerous type both in Sanskrit, 1 and in other lan- guages which reflect the late IE stage. In Hittite, on the other hand, which reflects an earlier stage of Indo-European there is not such a great preponderance of a-stems, 1 In the Bgveda 45 per cent of all nominal stems end in -a. THE FORMATION OF NOUNS 123 §2. Root Nouns Root nouns are an ancient type very much in decline in the earliest recorded Indo-European languages. In Sanskrit they are preserved better, on the whole, than elsewhere. Such are pad- j pad- ‘ foot * : Lat. pes , pedis , v&c- * speech J : Lat. vox, raj - * king ' : Lat. rex . They are usually either masculine or feminine, but in the case of verbal roots functioning as action nouns they are normally feminine (dyut- f. ' brilliance etc.), The old neuter type of root noun (cf. Lat. mel,fel, etc.) is prac- ? ~ tically extinct. Of the few examples mention may be made of ^ sdm r welfare ' which is indeclinable, van - ' wood, tree ' which is probably neuter (like its extension vdna- n.) and as ‘ mouth ' : Lat. os. This type of stem was originally subject to the laws of apophony : Vrdd hi or Gu n a in the Nom . Sg^Gupa in Acc. Sg, and Nom.Pl .amd the weakened form of the root in th e other_ cases . This system is partially preserved in the case of some common nouns (n. sg, p&t , gen. sg. padas) but the general tend- ency is for it to be levelled out. In this levelling out any of the three grades may come to function throughout the declension, the weak form of the stem being normal in the case of verbal roots used as nouns ; (1) vac- ‘ speech gen. sg. vacds, (2) spas- ‘ spy nom. sg. spat, (3) fc- ' hymn nom. sg. fk } gen. sg, reds. Roots functioning as nouns may be used either 33 action or agent nouns lyn the "Tatte r case they ar e feminine) : druh- injurer (2) ‘injury'; dvts (1) ‘hatred', (2) ‘ enemy ' ; bhuj- (1) ‘ enjoyment (2) 1 enjoyer When used as the second member of compounds they have the latter function only. This type which has a moderate extension in the Vedic lan- guage, becomes more restricted later in accordance with the pre-existing tendency. That is to say with one exception — when such stems are used as the last memb erjpf comp ounds. In this case both in the Vedic and Classical language roots may be freely used as nominal stems. They are also used more widely in the Vedic language as infinitives of the type d rSd * to see ' than otherwise independently. Elsewhere as we shalFsee the V edic infinitive tends to preserve old nominal stems whi<; h h ave otherwise become extinct . Roots ending in short -i, -r cannot function as nominal 124 THE FORMATION OF NOUNS stem s. In circu mst ances where other r ootSjiQ so-thay Jake the. suffix -i : °jit- 4 conquering mil - 4 pillar stul- 4 praise °bhrt 4 bearing etc. §3. Thematic Stems based on the Root Stems with the thematic suffix -a can be formed on the basis of stems ending in all the other suffixes, and these are best treated in connection with the various types of stem to which the thematic suffix is added. The most simple type of thematic stem is that formed directly from the root. These stems may be divided into two classes according to the usual scheme : action nouns and the like with accent on the root, and agent nouns, etc., with accent on the suffix, final accented -d having here as elsewhere an adjectival function. Certain pairs are quotable where both types occur in connection with the same root : c6da - 4 goad ’ : coda- 4 instigator ' ; isa- 4 speed ' : esd- 4 speed- ing ’ ; vara- 4 choice 4 : vara - 4 suitor 1 ; ioka- 4 glow 1 : iokd- 4 glowing The same type of alternation is familiar also in Greek, 4 a cut ’ : to^o? * cutting etc. In Greek and in other languages which distinguish the vowels 0 and e the vowel of both root and suffix is 0. Original 0 is attested by Sanskrit in some cases where a guttural has not bee n palatalised ^ keta- 4 intention gdya- 4 property ghand- 4 striker, solid Such an arrangement can hardly be original since normally e would be expected in the accented and 0 in the unaccented syllable. It is likely therefore that we have here a phonetic compromise between the nominal and adjectival types. In Sanskrit there are distinct traces of a variation oje between the two types in the suffixal vowel in the case of roots ending originally in gutturals. The final guttural in these cases is usually preserved in the case of action nouns with radical accent and palatalised in the other type : e.g. bhdga- 4 enjoy- ment ' : bhojd - 4 bountiful ' ; toga- 1 disease ' : raja- 4 breaking, destroying ' ; soka- 4 glow, heat ; grief ’ : sued- 1 bright 4 ; yoga- 4 union 4 : a-yujd- 4 without an associate 4 ; drgha - 4 value 4 : arha - 4 worth, valuable ' (accent not quoted). The distinction is found in some cases where the accent is on the suffix in both types ; arkd- 4 ray 4 : area- 4 brilliant 4 ; rokd- 4 lustre 4 : roed- 4 radiant Here the accent of the action nouns has been secondarily transferred to the suffix. The variation THE FORMATION OF NOUNS 125 between the palatalised and non~palatalised form indicates an original variation in the quality of the suffixal vowel according to the position of the accent. In action nouns the radical syllable commonly h as guna- in agreemen t with the, related, languages : ddma- 4 house Gk? Solos', Lat. domus, Russ, dom (IE dem- 4 to build '). Nouns of this type are dya- 4 going, course ', hdva- 4 invocation tdra- 4 crossing vdda- 4 knowledge josa - 4 enjoyment ' and so forth. There is however another type, for which it is difficult to find parallels outside Indo-Iranian, with vrddhi of root and, paradoxically, usually having the accent on the suffix. These appear to have been formed on the basis of the vrddhied nom. sg. of root nouns, just as later Vedic pat nom. sg. 4 foot ' is ex- tended to pdda-. Typical instances are : bhard- * burden ' (cf. the vrddhi in Gk. <f><Lp 4 thief '), ddvd- 4 fire tdrd- 4 crossing sadd- 4 sitting sdvd- 4 libation vdsd - 4 residence * ; with radical accent, vara - 4 choice mdna- 4 opinion Irregular accent is found in the whole class of such nouns which are formed from verbal roots combined with a prefix : samgamd- 4 coming together, union abhidrohd - 4 injury etc. ; and in a minority of cases elsewhere : bhoga- 4 bend jayd - 4 victory java- 4 speed These irregularities show that a tendency to confuse the two types was beginning ; jayd- and java - for instance also mean 4 victorious 1 and 4 speeding * which is their original significance, but at a time when the im- portance of the old distinction was diminishing, they came to be used indiscriminately in both functions. Here, as elsewhere throughout the formation of nouns, the Vedic, accent is not original, but in spite of such exceptions the old system remains predominant enough for its principles to be clearly seen. The oldest type of apophony among agent noun/adjectives of this class is that which has the weak grade of the root due to the accented suffix : vrdhd- 1 increaser 1 (: vdrdha- 4 increase budhd- 4 intelligent ' (: bddha- 4 understanding '), sued- 4 bright * [sdka- 4 glow '), turd - 4 victorious * {tar a-, tdrd- 4 crossing ') f priyd - 4 dear krsa- 4 thin rued- 4 brilliant etc. More fre- quently the restored guna vowel appears : area- 4 shining dravd - 4 running yodhd - 4 fighter nadd- 4 roarer vadhd - 4 slayer etc. Like the action nouns they may also appear with vrddhi, and this is connected with the vrddhi of the nom. sg of the corresponding root stems : vdha- 4 beast of burden 1 (cf 126 THE FORMATION OF NOUNS the root stem °vdh-, nom. sg. °vat, acc. sg. °vdham) sdha- * victor- ious ' (cf. °sdh-), tidy a- 1 leader grdbhd- ‘ seizer etc. The agent nouns of this type are on the decline, and as hap- pens elsewhere in the same circumstances, many such stems are no longer used as independent nouns, but only as the last mem- bers of compounds : e.g. °add- ‘ eater 1 ( annddd - ' eater of food'), °gamd- 'going' ( duramgama - ‘going far*), Q gard- ‘swallower* (ajagard- ' goat swallower*, i.e. python), °ghnd-

• slaying ' ( goghna - ' slayer of cows *) . This is because the for-

mation which came to be normally used in making agent nouns was that in -tar , and this tended to oust other formations in ordinary free use. In contrast the tar- formations were not capable of being used in composition, so there is a dichotomy of the type dnnasyd ~ atti : annddd- ' eater of food In a small number of nouns of this type the accent has been secondarily transferred to the root. Such are vfka - ' wolf visa - ‘ server Such transference is common in nominalised adjectives through- out the system. The adjectives sdna- ' old * and ndva- 4 new * can be fitted into neither of the above classes. From the corresponding forms in other languages (Gk. v£os 4 new' Lith. senas 4 old etc.) it can be seen that they are distinguished from other thematic stems by having the radical <?-grade and from the adjectival type by having radical accentuation. This is because they are based on old root stems new-, sen-, which were adjectives by meaning from the beginning, and consequently the thematic vowel here is merely an extension of the stem, as in the action nouns, and not meaningful as in the usual oxytone thematic adjectival type. In Latin the root stem sen- is still used outside the nom. sg. ( senex , senem, senis). Both types of the above nouns are masculines. Feminine agent nouns such as are found in Greek (17 6&6s 4 way *, ij rpo<f>6$ 4 nurse *) are non-existent in Sanskrit, which in this respect is less archaic than Greek. The masculine gender of these action- nouns is in contrast to the neuter gender of the various thematic action nouns derived by extension from the various neuter suffixes to be mentioned later. This is because the root nouns, at least those ending in occlusives developed early the distinc- tion between nom. and acc. and used -s in the nom, sg. That is to say they were ' common gender*, and the thematic deriva- tives based on them automatically acquired the same gender, THE FORMATION OF NOUNS 127 Later they were specialised as masculines owing to their ex* temal form. In contradistinction there are a couple of neuter nouns vdna- 4 forest 1 and tana - ' offspring 1 which are extensions of root stems (van-, tan-) which had retained their neuter gender. Old neuter formations are found in yugdm * yoke ' (: Gk. £vyov t Lat. iugutn) and padam 4 step 1 (: Gk. tt&ov, Hitt. pedan). These are old formations, among the very few simple thematic neuters that can be traced to Indo-European. They will be discussed in connection with the suffix m (p. 172 ff.). §4. Neuter Formations with alternating rjn Suffix The suffixes r (which in Sanskrit may also represent IE l) and n must be studied together since they early became associated in a common paradigm in which the nom. acc. was formed by the r-stem, while the oblique cases w T ere formed on the basis of an n-stem. This ancient type of neuter noun is tending to obsolescence in the earliest Sanskrit, as it is in Greek and most of the other languages. InHittiteon the other hand, which presents here, as so often, a more archaic stage of Indo-European, the system is unimpaired. The system as found in Hittite contains simple rjn stems with this alternation, e.g. eshar 4 blood gen. sg. einas, also a series of compound suffixes formed by the addi- tion of these suffixes to stems in u t m, s, t, namely -war, -mar, -sar, tar , Examples are partawar ‘ wing gen. sg. partaunai , tarnummar 4 letting go, to let go", gen. sg. tarnummas (mm < mn), hannessar ' law, law suit gen. sg. hannesnas , paprdtar 4 uncleanness gen. sg. papr annas (nn<tn). This early system of neuter nouns exists only in fragments in other IE languages, but an abundance of suffixes containing r and n have these primitive neuter types as their ultimate source. There are a few simple neuter stems in r with alternating ^ n-stem in Sanskrit. Such are dhar * day gen. sg. ahnas (Av. ^ azan - 4 id ’), tidhar 4 udder ’ gen. sg. udhnas (Gk. ov8ap, ovdarog, Engl, udder , etc. ; there also appears to be in the Veda a second tidhar 7cold ; =Av. aodar- 4 id. 1 ). In these the suffix has the guna grade, but it may also appear in the weak grade, in which case it is strengthened by a further suffix. This is usually t: ydkrt "liver gen. s g. yaknds (Av. ydkar-, Lat. iecur, Gk. ^ } * fjirap, all without any/), idkrt 4 dung gen. sg. saknds, with a 128 THE FORMATION OF NOUNS ► £>>* 1 J variant -th in kdprth - 1 penis This additional t may be com- pared with additional t which in Greek strengthens the alternat- ing w-suffix in these nouns : ovOaros compared with Skt. 4dhnas, etc. We may also compare the fact that final radical -r is avoided and - 1 added in lokakft , etc. A suffix j {<g) is found in dsrk 1 blood ' (nom. sg., stem dsrj-), gen. sg. asnds (; Hitt. elhar , esnas, Toch. ysar, Gk. eap, Lat. assir). Thisg also appears in Lat. san-g-uis ‘ blood ' which like sanies * gore 1 is derived from the w-stem of this word with loss of initial vowel through apophony. Since this type is becoming obsolete we have occasionally de- fective nouns like vadhar ' weapon ' (Av. vadar -) not used out- side nom. acc. sg. The old alternating n-stem appears in the extension vadhand fern. r id Some stems even more obsolescent occur only as the first members of compounds : usar- ( usarbiidh - ‘ waking at dawn '), anar- ( anarvts - * seated on a chariot '), vasar-, (vasarhdn- * smiting in the morning '), vanar- ( vanargu - ' going in the woods cf. the deriv. vanar a- ' monkey ' ; n-stem in vdnan-vat ), sabar- ( sabardhuk : for savar cf: sdvana ‘ milked at the soma-pressing '). The stems mdhar 1 greatness ' and bhuvar 1 abundance ' appear only in liturgical formulas and in the compounds maharloka- and bhuvarloka- > but the instru- mentals of the corresponding n-stems, mahna (cf. Av. mazan- n.) and bhund are common in the Rgveda. The n-stem gdmbhan f depth ' appears only as endingless loc. sg., but a corresponding r-stem *gdmbhar is implied by the extension gambhdra- n. * id.'. Corresponding to Vedic loc. sg. rdjdni r under the direction of ' Avestan has nom. acc. sg. rdzarz ' rule, regulation The instr. ddni implies an old nom. sg. *ddr * gift and from these alter- nating stems the two extensions Gk. Stopov, O. SI. daru and Lat. donum , Skt. danam are derived. These instrumental forms, and also dsnd ‘ with a stone ' ( Av. a^an-) and prend f with affec- tion ' being isolated, have come to function as the instrumentals of the corresponding man- stems. The adverb avdr r down, downwards ' has the same formation as the above neuter nouns, but differs in its accent which corresponds to that in the endingless loc. sg. The same adverbial accent is found in pratdr , etc. below. Such adverbs also re- semble the locatives without ending in that they may optionally add the termination -i : Just as we have aksdni ' in the eye ' beside aksdn , so we have Skt. updri ‘ above ' beside Gk. vrrep 0 * 14 ; * THE FORMATION OF NOUNS 129 and Av. aSairx ' below ' beside the *adhar adv. which is implied by the adjectival derivative ddhara - * lower Neuter stems in -at not otherwise preserved form the basis of a small class of denominative verbs in the Vedic language : rathary dti 4 rides in a chariot fratharyati 4 becomes loose implying *rdthar nt. * riding in a chariot ’ and *srathar nt. 4 looseness ' ; cf. vadharydti 4 smites with a weapon ' beside vadhar. There are parallel denominative verbs from the corres- ponding n-stems : vipanyd- 4 to be wise, inspired ’ (cf. also ~ vipanyA, vipanyu and with - 7 - vipra- 4 inspired, wise '), bhuranyd - 4 be turbulent, agitated/ (r-stem in Lat. furor), etc. There are various secondary formations testifying to the exist- ence of old neuter r-stems. The curious formations dhaiirita - n. 4 horse's trot ' and ddhorana - m. 4 elephant driver which turn up in later Sanskrit can be explained as denominative formations on the basis of an old neuter noun *dhavar 4 running The Vedic vrddhied derivative jamarya- 4 earthly ' is based on a /

• jdmar 4 earth 4 corresponding to Av. zzrnar- 4 id.' ( j - as in jmd t (

jmris, etc.). A neuter *svdtar 4 whiteness, white spot ' is implied > by the derivatives ivaitarx 4 having an (auspicious) white mark ' l { (a cow) and svetra- nt. 4 white leprosy and the alternating / n-stem appears in the extension svetanA fern. 4 dawn ‘ / / Since Skt. r represents both r and / of Indo-European, IE J stems in 4, which functioned precisely as r-stems, cannot be ^ distinguished from r-stems in Sanskrit, except by comparison with other languages. Such a stem is found in svdr (suvar) c < w 4 sun ', gen. sg. suras (cf. Lat. sol „ Goth, sauil, etc.). The hetero- clitic declension which is absent in Sanskrit appears in other languages (Av, x v btg<*svan$, gen. sg., etc.). Some of the r- stems mentioned above appear by comparison with other lan- guages to have been originally Z-stems : mdhar- : Gk. fieya Ao- ; Av. zwnar- : cf. Gk. Lat. humilis , etc, ; Av. rdzar ? : Lat. regula . The denominative verb saparydti 4 serves, honours, worships ' is like Latin sepelio 4 bury 1 ( 4 honour with funeral rites ') derived from an old IE neuter *scpel 4 honouring, wor- shipping ' derived from the root sep- t Skt. sap- 4 honour, serve.' The /-extension to the suffix n, which appears in Gpeek (outfaro?), etc,, is absent in Sanskrit (tidhnas, etc.), but in Seme derivative forms an additional /-suffix is found which may be connected with the extension / of the Greek neuters : vasantd- 4 spring', cf. vasar p , veiantd- 4 pond ' ( > fvis4j t bhuvanti THE FORMATION OF NOUNS 130 4 causing abundance cf. bhuvar, bhund . Another example is stmanta - 4 parting of the hair ' as opposed to stmdn - ‘ id and the lexica give a word aimanta- 4 fireplace ' which is related in the same way to dsman- 4 stone Compare also the pair sakuni- and sakunti - 4 bord The /-extension appears regularly in the participles in -ant which will be discussed later.

§5. Compound Neuter Suffixes in rjn By adding the simple suffixes r and n to stems in u t m, s and / the suffixes -warjn, marjn f etc., which were so productive in Hittite in the formation of neuter nouns, were produced. In addition they could , be added to f-stems, and this ancient neuter type is preserved in the Latin passive infinitives, iitier, scribier, etc. Outside Hittite the compound neuter r-stems have become comparatively rare , the corresponding n-stems are better represented, and have tended to replace the r-formations in the nom. acc. sg. A neuter suffix - wer , -war is found outside Hittite in such examples as Gk. efSa p 4 food ' (*eSFap), SeXeap 4 bait Lat. cadaver r corpse Toch. B malkwer 4 milk 4 and the like. They are not uncommon in Avestan : sndvar- 4 sinew’, danvar - 4 bow karsvar- 4 region of the earth dasvar - 4 health sdx v ar - 4 design, plan vazdvar- 4 firmness The Iranian evidence shows that they had survived in reasonable abundance to the Indo-Iranian period, but in Indo-Aryan they had already become extinct by the earliest period. This was be- cause the tt-stem was generalised in all cases : nom. acc. sg. s?iava 4 sinew etc. Only the adverb sasvdr 4 secretly, stealth- ily ’ preserves the suffix in this form, with the usual adverbial ( = locatival) shift of accent. It implies a neuter noun *sdsvar which we may compare with Hitt, sesnvar 4 sleep ’ (ssszi 4 sleeps Skt. sdsti), the original meaning being 4 while people sleep In its weak form the suffix is preserved in the adverb muhnr ‘ suddenly, in a moment ’ (whence muhurtd- 4 moment ' ; also muku with simple u-suffix, for *mrhu cf. Av. nurdzu- g ‘short (of life)', Gk. Neuter nouns in -van are 1 dhanvan 4 bow *, 2 dhdnvan ' desert snavan- 4 sinew o )*■*$**& par van- ‘ joint ' (cf. Av. paourvainya - 4 linked sfkvan- 4 corner of the mouth and some sporadic occurrences in the Rgveda, pdtvan - 4 flight sanitvan- 4 acquisition ' and vivdsvan - 4 illumination In addition a small number ofdative infimtryes^ THE FORMATION OF NOUNS 131 are f ormed _jV'ith thi s suffix ; davdne * to give * (Gk. SoFevcu, Sourai) turvdne ' to overcome dhUrvane 1 to injure ' ; simi- larly in Av. vidvanoi ' to know A suffix compounded of i and r appears only in the adverb bdhir ' outside A few defective in-stems appear in the Rgveda in the instrum, sg., namely prathina , mahind, varina . These forms, in origin old neuter in-stems, have been attracted to the paradigm of the masc. man - stems praihimdn - f width mahi- mdn- 1 greatness 7 and varimdn- ‘ width and so preserved. The neuter suffix -in further compounded with v appears in sfkvin- nt. comer of the mouth The compound suffixes m-er t m-en alternated in the same way, but outside Iiittite, mer (mr) is rare and obsolescent. A fair number of examples can be collected from Greek (Avpap, Avparos ' impurity etc.) but they exist only as survivals. No examples are found in Sanskrit but their one time existence in Indo-Iranian is shown by adjectival derivatives like admar-d

• gluttonous 7 based on an old *ddmar * eating, food Such

forms have been entirely replaced by the extension of the man- stem to the nom. acc. sg. The neuter suffix man is the only one of the rjn suffixes that remained fully productive in languages other than Hittite. Examples are : (nom. acc. sg.) djma ‘ career, march ' (: Lat. agmen), karma 1 deed ’, cdrma * skin paksma ' eyelash pdtma

• flight brahma ' prayer bhdrma 1 maintaining, supporting ;

load ' (: O. SI. brem% ‘ burden ', Gk. pepfia), vdsma 1 garment ' (: Gk, ef/xa 'id', Lesb. Fe/i/xa), loma t roma ' hair ' (<J lu cf, lava - 'fleece, wool, hair 1 ), mdnma 'thought' (: 0. Ir, menme ' mind, understanding '), vdrma * protective armour vdrtma 1 course, way ' (0. SI. vremg ' time y ) sddma ' seat syuma ' thong, rein svddma ' sw r eetness A small number of dative infinitives are formed on the basis of this suffix : trimane ‘ to protect ', dimane ' to give dhar- mane ' to support bhdrmane ' to maintain vidmdne ' to know In Greek infinitives are also made with this suffix, much more abundantly, and including an archaic type without termination ; Igiev ' to go Softtv ' to give ' ; iSfievai ' to eat tfjLcvcu ' to go So/xeycu ' to give ', etc. The nouns of this class are primarily verbal abstracts (action nouns), but they show a strong tendency to acquire concrete meanings, as happens with other neuter abstracts : e.g. hdnma THE. FORMATION OF NOUNS 132 £ weapon ’ as well as ' blow tdrdma 4 hole mdrtna ‘ a mortal place, vital organ cdrtna 4 skin 7 , vdsma 4 garment etc. Like other neuter suffixes in -n this suffix is extended by t in Greek (gen. sg* owparos, etc.). A case of such extension is found in Skt. varimdt- 1 breadth 9 (inst. sg. varimatd RV. 1. 108. 2). The neuter suffix -t-er alternating with -t-en } which is so well represented in Hittite, has become exceedingly rare in other IE languages. Only isolated examples such as Lat. iter , Toch. ytdr 1 way 1 are quotable. In Sanskrit a solitary example of this kind appears to be preserved in RV. 6. 49. 6. ; jdgatah sthatar jdgad a krnudhvam, 4 may ye bring stability to the moving world Misunderstood by the redactors the form has been handed down without accent as if vocative of sthatar - ' stander and the passage has been rendered unintelligible. A restoration sthatar n. 1 stability 7 gives meaning to the line. Apart from this the neuter suffix -tar- appears in a small number of locative infinitives, e.g. dhartdri 4 to hold vidhartdri 4 to bestow In Avestan some dative infinitives are made with the same suffix : t 'id didr e 4 to look at 7 barzdre 4 to support These forms are interesting as showing that the neuter /'-stems were capable to som e extent of jbeing mflect ed t h r oi^ ^ of bemg_ nder what precise conditions this happened originally it is not now possible to say. Some adverbs appear with this suffix, having the usual change of accent : antdr ' inside 7 (: Lat. inter), pratdr 4 early ' (: Osc. pruter ), sanutdr 4 aside, apart 7 (cf. Engl, a-sunder). These imply old neuter nouns *an(ar 4 the interior etc. This type of adverb became very productive in Latin, aliter 4 other- wise 7 (cf. anydtr-a), breviter, leviier , etc. With additional suffix -d we get adverbs in -tra> e.g. sayutra ( in bed, abed and this suffix may, and normally does later, appear with a short vowel, anyatra 4 elsewhere dtra f here etc. Among the adverbs the Veda has some interesting collective formations : devatra 4 among the gods purusatrd 4 among men They may be explained by comparing Hittite forms like antuhiatar 4 the population, mankind 7 ( antuhsas 1 man 7 ). Similar neuter cob lectives *devatar , etc., lie behind these adverbial forms. A neuter suffix - tan is found only in naktdn- 4 night 7 (instr. pi, naktabkis). Alternating -r appears in Gk. vvktojp 4 by night 7 tp**' and Lat. nocturnus , This neuter suffix is found sporadically T THE FORMATION OF NOUNS 133 elsewhere (Lat. gluten ‘ glue etc.) and in Old Persian it is used to make dative infinitives ((artanaiy 1 to do etc.). ^ The compound suffix sarjsn , which is very common in Hittite, is more or less obsolete elsewhere. It is an extension of the neuter s-stems, and in Hittite the -arjn has been added so con- sistently that the simple s-stems have practically disappeared. In other languages the simple s-stems are well preserved so that it is unlikely that the sar/sn formation ever had the same exten- sion elsewhere as appears in Hittite. Nevertheless there are considerable traces of it. Adjectival derivatives of the type matsar-d ' exhilarated ' are based on such formations, and the coexistence of mandasdna - f id/ shows that there was the old rjn alternation. Similarly pusary a- * well-nourished r implies

• pusar nt. ‘ fatness, prosperity equivalent to Gk. Trvap ‘ beest

milk and the alternative n-stem is used in the masc. derivative Pusan- f nourisher (name of a god) V A few neuter stems in -sn- are preserved in the oblique cases of nouns which appear in the nom. sg. as simple 5 -stems, e.g., gen. sg. sirsnas, dosnds , nom. acc. sg. Ur as ' head dos ' arm A small class of locative infinitives is made on the basis of neuter stems in - san r : nesdni to lead parsdni ‘ to pass grnisdni 1 to sing isdni ' to emit ' (for is-sani with the old & - sandhi, p. 95). In Greek also this suffix forms infinitives. The common type ptpeiv is best explained in this way (*p€p€a€v,

• bheresen, the simple base without ending being used as in

iSfjL£v, etc.). §6. Action Nouns transferred to the Masculine It has been remarked above that the major distinction in Indo-European was between neuter action nouns accented on the root and masculine, originally common gender, agent nouns accented on the suffix. But in Sanskrit as in other languages not all nouns fit into this simple classification ; among the w-stems for instance with which we are dealing there are quite a number of masculine formations which cannot be classed as agent nouns, and which from the point of view of their meaning go rather with the neuter action nouns of the above type. Such 1 Originally *py-u$ar ( v r pi) as is seen by comparing vvap and Skt. ptyusa-. Similarly pus- J to thrive, prosper ’ is for *pyus~, originally a denominative formation like urusydti, etc. [*pyusydii) t THE FORMATION OF NOUNS 135 (3) The two adjectival forms pdmar-d- and paman-d- r affected with scabies ' show that there was an old neuter rjn stem which has given way to the masc. pdtndn -, and asman 4 gall stone ' attests an old neut. *asmar as opposed to the masculine n-stem which is always found elsewhere (Skt. diman Lith. akmuo ; Gk. aKfUDv ‘ anvil '). The same relationship exists between the Gk. adjective Repos’ * gentle, tame ' and Skt. sdmand - ' quiet The general tendency of the neuter to decline, and with it the decline of the old antithesis marked by accent and gender between action noun and agent noun made it easy for a type of masculine (and feminine) action noun to develop. To a certain extent also personification is responsible for the gender. Terms like daman - ' liberality onidn- ‘ favour (of the gods) ' and bhiimdn - ' abundance * are regarded in the Vedic hymns as divine powers in their own right. Disease (takmdn-) and sin [dhvasmdn-, pdpman-) are likewise regarded as active evil powers. At the same time there are some more mechanical transfers. The abstract formations in - imdn - (prathiman- ‘ width varimdn ‘ id etc.) take the masculine gender as a class, and there is probably a rhythmical reason behind this because the formations in -man- preceded by long i remain neuter ( vdrhnan - ' width hdviman-) call etc.). Masculine action nouns of the same type are common in Greek : rippajv ‘ boundary 1 beside repp a nt. J id ' (Lat. termen ), Xeipwv ‘ winter ' beside x e W a nt., Orjfxwv ' heap ' beside 6fjp a nt. / ^ ) (Skt. dhdman- nt. with different sense), Xeipiwv ' meadow XipLijv 'harbour', dSrjv 'gland' (Lat. inguen nt.). In Latin there are both masculine (ordo t sermo) and feminine ( mar go } legio , cupido) n-stems of this type. Owing to their early obsolescence the neuter r-stems have not undergone this transference with the exception of a single example. The defective stem usar- _( usr -) ' dawn ' is feminine (acc. pi. usr as) but was originally a neuter r-stem (cf. usarbudh § 4 ). §7. Thematic Extension of Neuter Stems in r and n The original func tion of the thematic suff i x was adjectival andjn this case it was ac cented : karan-d * doing etc. This conclusion!! indicated by Hittite which possesses such stems [veitaras 1 herdsman , etc.) but no thematic neuters like the THE FORMATION OF NOUNS I36 other languages. When we compare the thematic type of neuter, e.g. Skt. sanara- ' acquisition/, karvara- r act ' with the non-thematic stems in - ar , - var , etc., illustrated above, it is dear that the thematic suffix here has no grammatical function. This, coupled with the absence of such formations in Hittite is an argument for the secondary origin of this type ; clearly as between Lat. unguen and Skt. anjana- the former is the older formation. It is not difficult to see how such forms arose. There were agent nouns of the type brahman- beside brdhman- nt. but also from the earliest period another type of agent noun/adjective made by the addition of the accented thematic vowel (Gk. iarpos beside larv f p , Hitt, ve&aras beside Av. vdstar-). It was then natural and easy to create a neuter thematic type balancing the thematic adjectival type (kdrana- nt. after karand -, etc,). Thematic neuter stems corresponding to the various r- and n-stems listed above may appear either with the guna of this suffix (karvara- ‘ deed ’) or with the reduced grade (dhdrtra- ' support '). Both types occur from simple r-stems : (1) sanara- r gain tasara - ' shuttle panjara- ‘ cage ' ; framework of the *.*50 ribs (cf. pajra- adj. 1 fixed, firm Lat. pan-go , etc.), gambhdra- 1 depth uddra - ‘ belly It will be noticed, here and below, that there is a certain fluctuation in the accent of the tri- syllabic forms. Final accentuation, the characteristic of adjec- tives, is avoided, but the accent may fall on the penultimate 'syllable instead of on the root. (2) dgra- 1 point rdndhra-

• hole svdbhra - * pit There are a few substantives with final

accentuation but these are adjectival in origin, e.g. krcchrd - nt, ‘ difficulty but also krcchrd- adj. ‘ difficult riprd- nt. ‘ defile- ment ', but cf. Gk. Xcnapos ' greasy abhrd nt. ' cloud cf. Gk. appo$ f foam 1 masc. (nt. dtnbhas - ‘ moisture 1 beside which 7 there must have been *dmbhar, cf. Lat. imber) t kstrd - nt,

• milk 1 of uncertain etymology but from its accent of adjectival

origin. In sth-dla - ' place, ground 1 there is a thematic neuter forma- tion involving the suffix. A few old neuters in -var have been extended by the thematic vowel: karvara - 'deed', replacing earlier *kdrvar t gdhvara- ‘ hiding place pharvara- ' sowing, sowed field ' (*(s)phar - : 0 / t ^ ^ Gk. €T7T€ipiu) , catvara - ' quadrangle, cross-roads ' (from a neuter +catvar on which the adjectival catvaras ' four ' is based). -ntxrf'p* THE FORMATION OF NOUNS 137 With the tar- suffix we have a neuter formation in -tara in srastara - * bed of grass ' (fsrams-) and a fairly abundant series of neuters in -tra : dtra- " food ' (: atrd- m. ' eater'), kdrtra - ' spell ', ksitra- 4 field ' (: Av. $6i6ra- " habitation '), vdstra- 4 garment srotra- "hearing, ear* (: AS. hleopor "noise'), stitra- 1 thread * (cf. the Lat, agent noun sutor). In addition to such words which can be explained quite simply out of primitive neuter tar-stems, there is another series of neuters with the suffix - tra less easy to explain since they have paradoxically accent on the final syllable. Such are antra-, antra- (: Gk. *^7 ^ ^ evrtpa pi.), astra-, destrd - "direction', netrd- 'guidance', r astra - 4 rule, kingdom ', iastra- ' invocation sattrd- ' sacrificial session iastra - 4 command ', stotrd- " praise, hymn of praise ’, sthdtrd - 4 station ', hotra- 4 office ot hotar-, oblation potrd- " office of pdtar- ', nestrd- ' office or vessel of ndstar- '. A few of these forms may be explained as originally adjectival, e.g. antra - ‘ what is inside astra- ' what is thrown ', but the majority clearly cannot be explained in this way. They must be explained from a different point of view. There is a series of agent nouns in -tar denoting holders of professions and priestly offices, e.g. sdmstar- ' reciter ' hdtar - " sacrificial priest polar - ‘ purifier ’ and closely associated the ndstar- (prob. " sifter cf. Gk. vaicrjTTip A LKpirjT^p Hes,), Such nouns as a class have the nominal accent, i.e. on the root. The above neuters, hotra - ' office of hotar- , etc., have become, as far as their meaning is concerned, secondary derivatives from these agent nouns, thus reversing the position originally prevailing between neuters and agent nouns. In ordinary taddhita derivation there is a special rule in Sanskrit whereby the taddhita derivative is accented on the final -a if the primary formation from which it is derived is accented on a previous syllable : nairhasta- " hand- lessness atithya- ' hospitality ', saumanasd- " friendliness ' from nirhasta -, dtithi sumanas - (and vice versa, p&lityar from palitd- "grey-haired'). There are also examples from non- vrddhied formations, e.g. sakhya- 1 friendship ' from sdkhi-. This is a new way of using accent in derivation which Sanskrit has developed, and it is this system which accounts for the final accentuation in hotra-, etc. These formations are however not from the beginning taddhitas, but a subdivision of the old simple neuters in -tra which have been adapted for a special purpose and have had their accent altered accordingly. THE FORMATION OF NOUNS 138 There are a fair number of neuters with the gradation -atra, e.g. ndksatra - ' lunar mansion' (naks- ‘to reach 1 ), patatra- 4 wing vddhatra- ‘ weapon ' ; with penultimate accentuation, krntatra- * piece cut off d-dtra - ' gift ' ; with taddhita accent, ksatrd- ‘ sovereignty ' ; post-Vedic kalatra- 1 wife flbo A thematic extension of an old neuter sar - stem appears in tdmisra- * darkness ' (cf. Av. tadra-) ; with l patsala - ‘ way ' Un, In Hittite there is a neuter suffix in - an (not alternating with -at) making primary verbal abstracts, henkan * death ', etc. The same suffix appears in the Greek infinitives in -ev and sporadically elsewhere (Lat. unguen). In Sanskrit the thematic extension of this suffix has become exceedingly productive in the formation of neuter action nouns from verbal roots. They frequently stand in opposition to agent noujis having the same suffix but accented on the last syllable : karana - nt. ‘ deed ', karand- masc. ' doing Examples with radical accent are ailjana- ‘ anointing, ointment ', cdyana - ‘ heaping up ’, ddr&ana - ‘ vision f pdtana- ‘ fall bhojana - ‘ enjoyment sddana - ' seat etc., etc. ; from a non-verbal root, sdmana- r assembly Such words sometimes appear with weak form of root, e.g. bhuvana- 1 world and, when the root-vowel is a , sometimes with vrddhi bhdjana - ‘ vessel There is a tendency in the later language to associate these latter formations with the causative of the verb, e.g. tar ana- ‘ crossing tirana - ‘ getting (somebody) across, delivering Penultimate accent is sometimes found : vrjdna- ‘ enclosure, settlement 1 (also vfjana-), krpdna - ' misery J (: krpand - 4 miserable '), damsdna- 1 wondrous deed ', vesdna-

• service dh-dna - ‘ wealth r-dna- * battle ' (cf. Av. armu -

•id*.). Similar neuter formations are found occasionally from the compound n-stems : vayuna- ' delimitation, appointed time ' (: v&d ‘limit, time, etc.'), karuna- ‘deed ' ; drdvina - 1 property 57 ~ JfHl*# vajina - ' race, contest ' ; vdtwna - 1 wage ' (J vi), r-dtna- ‘ treasure ’ (rd- ‘ to bestow ’) ; mdtasna- ‘ lung The neuter suffix menlmn could be extended by the suffix -t (Gk. adifjiaros gen. sg., etc., cf. the Hittite infinitive forms in - manzi , - wanzi ). This complex could also receive the thematic extension, Lat, augmentum, strdmentum , etc. There is one such example in Sanskrit, iromata - nt. * fame ' (-mat- <*--mnt-) , cf. OHG hliumunt , Germ. Leumund. THE FORMATION OF NOUNS 139 §8. Masculine Formations of the Type brahman— from r- and w-stems The essential feature of this type is the suffixal accent as opposed to the radical accent of the neuters, and vrddhi in the nominative singular. There is some evidence that originally they were themselves capable of being used as adjectives with- out any change of form. Examples of this are found in many languages, e.g. Gk. paiKap ' blessed fidprvp ‘ witness Lat. uber in the sense of ' rich Hitt, kurur meaning both ' enmity ' and ' inimical, enemy In Sanskrit there are two formations which may be .compared with Gk. pLaprvp, although they have accent on the suffix, namely afitur- ‘ active in holy works ' and yantur- ' controller With the suffix -us we have forms like ndhus- ‘ neighbour ' and mdnus - ' man ' which are not in any way distinguished from, the corresponding neuter types. Such formations are, however, very much irf the minority, because Indo-European early developed this method of indicating the adjectival function of a stem by switching the accent. In some cases in Sanskrit the two types exist side by side, notably in the case of the suffixes - man - and -as-, but more often the old system has broken down. This is mainly due to the elimination of the old neuter types, which has left important classes of masculine agent nouns standing isolated. In Sanskrit there is an abundant class of agent nouns in -tar, kartdr- 1 doer etc. The nature of this formation only became clear with the discovery in Hittite of an archaic class of neuters in - tar . We have seen above that this, like similar formations, has left many traces in other languages and therefore must at one time have been widely prevalent. This means that the two types *kdrtar nt. * doing, action ’ and kartdr- masc. ' doer, agent ' must at one time have existed side by side, and this being so it becomes immediately clear that the relationship of the two types is exactly the same as that between brahman - and brahman- Skt. kartdr 1 doer ' is one connected with *kdrtar 1 doing ' and sthdtdr- * one who stands ' is similarly related to that sth&tar nt. which, as we have seen, is preserved in a somewhat disguised form in one passage of the Rgveda. Since this is one of the commonest formations in Sanskrit the citation of further examples may be dispensed with. A few words are necessary about the accent. Suffixal accent is proper 140 THE FORMATION OF NOUNS to this type and occurs most frequently in Sanskrit, but there is also a type with retracted accent associated with a curious syntactic distinction : ddtA vdsiindm but ditd vasum. It has been noticed above that specialist words with this suffix (hotar-, etc.) have also as a rule accent on the root. In Greek also there are two sets of forms, with suffixal accent, S orrjp ' giver parrjp 'goer 1 , derrjp 'established, and with radical accent, Sdyrcop ‘ giver etc. The first of these preserves the most ancient form, with reduction of the root consequent on the accentuation of the suffix. In Sanskrit the suffixal accent is preserved to a large extent, but apart from very few exceptions, e.g. drmhitdr- 4 one who makes firm ', guna is universal in the agent nouns. Its maintenance or reintroduction in spite of the basic law of apophony may be ascribed to the influence of the related neuters. This suffix is prominent in the formation of nouns of family relationship : pitdr- ‘ father ' (cf. Lat. pater , etc.), duhitdr- 4 daughter ' (cf. Gk. Bvydrrjp , etc,), mdtdr- ( mother ' (Gk. ^rrjp. Dor. fiaTTjp, OHG muoter, etc.) ; bhratar - ' brother ' (Gk. <fi par cop, pparrjp, <f>p&Trfp 1 member of a phratry Goth, hr d par, OHG bruoder , etc.), jimdtar - ' son-in-law ' (Av. zamdtar-, Alb. Sender)] yatar- ' wife of husband's brother ' (Gk. elvaripzs, Lat. ianitnces plur., O. SI. jgtry, Lith. jente) ; ndptar - 1 grandson ' (secondary substitute for ndpdt — Lat. nepos). Of these it is probable that the word for ' father ’ is an old agent noun (p-i-tdr- 4 protector ' from pd-(y)~ 4 to protect '), but in the majority of cases the etymology is too obscure for it to be pos- sible to say much with certainty. Its gradation is of the old type (cf. Gk, Sor^p, etc.) as is to be expected in such a word. Only duhitdr - agrees with pitdr- in accent and apophony ; the rest have both accent and guna of root with the exception of mdtdr-, and even here Greek has radical accent, which may easily be original in spite of the agreement between Sanskrit and Germanic. It is not unlikely that these contain some old neuters etc. : Lat. materies would be an extension of such a neuter) which were adapted when the gender-system developed. There is also the possibility of the analogical exten- sion of the suffix. This has certainly happened in Skt. ndptar- and probably in the unusually formed jamdiar- (cf. Gk. yajxfipo? son-in-law ' differently formed. The defective noun star - (instr. pi. stfbhis) , tar- (nom. pi. THE FORMATION OF NOUNS I4I taras) — Gk. aa-nrjp is formed with this masculine suffix ; the word has become so reduced that in the latter form only the suffix is left. The suffix -tar is the* only one of the f -suffixes that has be- come prolific in the formation of agent nouns. But both the simple r-suffix and the various compound suffixes which have been enumerated were capable of being used in this way and a small number of examples have survived. An example with the simple r-suffix is seen in ndr - 4 man, warrior ' (Gk. avrjp, Umbrian ner-, etc.).' A corresponding neuter *&ner is deduced from certain derivatives (Gk. 7 )vop€rj,

• irrjVO)p ] etc.).

The suffix -var is so used in catv&ras masc. pi. * four The corresponding neuter *catvar 4 square ' is not preserved, but there is a thematic extension of it in caivara- nt. f quadrangle, cross-roads Another example is devdr - 4 husband’s brother cf. Gk. har}p (for SacF^p), Lat. levir, etc. There is no example of such a formation with the wtfr-sufhx, but it will be noted below that karmara- 4 smith ' implies an earlier karm&(r). The suffix -sar appears in this class in the word svasar- ' sister" (sua- 4 one’s own’) with retracted accent, and in the numerals ti$ra$, cdtasras 4 three, four ' (fern.). Here the accusa- tive form has come to be used for the nominative ; the old nominatives *tisores f k w ete$ores are represented in Celtic (0. Ir. teoir , cetheoir). Lat. uxor 4 wife ’ is a noun of this class, being based on an r-ex tension of the s-stem which appears in Skt. ckas- nt. 4 home There seems to have been a tendency for this suffix, when used adjectivally, to be specialised in the formation of feminine nouns, and with the above we may compare the thematic derivative in Hittite ishassaras 4 lady ' (ishas 1 lord '). Similarly agent nouns and adjectives are formed with the accented n-suffixes, related in the same way to the neuter n-stems : tdksan - 4 carpenter ’ (: Hitt, taksan nt. 4 joining ’), raj an - 1 king ' (: rdjani nt. 4 under the guidance of alternating r- stem in Av. rdzar- nt.), pratidivan- 4 opponent at play vtbhvan-, vibhvan- 4 powerful There are a few old masculines of this formation no longer referable to verbal roots, yuvan - 4 young man 4 (cf. Lat. iuvenis , etc. ; yo-s-it 4 young woman, woman ’), svdn- 4 dog ’ (Gk. kvwv), Examples from other lan- guages are Av. spasan - 4 scout, spy vindan- 4 one who ac- quires Gk. 7T€vdTjv 4 spy dpyywv 4 helper Lat. edd 4 one THE FORMATION OF NOUNS 142 given to eating', etc. In addition this formation has provided in Greek the nom. sg, of active participles of the thematic type (<f>epu )v, An twv, alternating with -ovr- in other cases, as opposed to StSovs, etc.) and in Germanic it forms the basis of an adjec- tival declension. The accent is generally retracted in Sanskrit. It has already been remarked that this is usual in the case of words of adjectival origin which have become completely nominalised (e.g. rajan-). The accent is likewise retracted in the adjectival formations in -van, but here the weak grade of the root shows that this is not original. Examples are fkvan- 4 worshipping, praising ' (: Hitt, arkuwar nt, ‘ prayer '), druhvan- 4 injurious yudhvan - ' fighting subhvan- 1 beautiful stubhvan - ' praising pdtvan- 4 flying 1 (: nt. pdtvan- ‘ flight '), mddvan- * exhilarating, ex- hilarated jdsvan - ' famished Suffixal accentuation appears only in musivdn - ' thief Roots ending in i, u, r take the addi- tional suffix i before this suffix : krtvan- 1 active sutvan- 4 pressing sftvan - ' moving '. That these are based on an old set of neuters with alternating rjn stem is shown by the fem- inine. This is based on the j'-stem of the neuters, e.g. pivari fern. ‘ fat ' is derived directly from the neuter stem which appears in Greek as 77 tap, whereas the masculines are derived from the associated n-stem. Exactly the same distinction between masculine and feminine is found in Greek : rrluyv masc., imlpa fern. 1 fat Similar feminine formations in Sanskrit ar eyajvari ' pious sdrvarl ‘ night °sivan 4 lying ', °ydvar% ' going °ddvan 4 giving ' (e.g. Godavari * cow-giving name of the river). The suffix forms both primary derivatives, as above, and secondary derivatives. Such are rtavan- 4 righteous satydvan- 4 truthful maghavan- ‘ bountiful and svadh&van- 4 powerful A fair proportion of the secondary formations re- tain the adjectival accent : amatlvdn- 4 indigent ardtivdn - ‘ hostile rnavdn- 4 indebted srustlvdn- 4 obedient Their feminine is likewise in -van ( rtavan , etc,), indicating that there existed at one time also secondary neuter formations in varjn of the type *rtdvar 4 righteousness In the case of the suffix -man a number of pairs are found in the Vedic language with varying accent and meaning illustrat- ing the general principle of noun-formation in Indo-European : brahman- : brahman - ; sddman - ' sitting, seat ' : sadmdn- 4 sitter ' ; dhdrman - ' ordinance 1 : dharmdn- 4 ordainer ' ; THE FORMATION OF NOUNS 143 diman - 4 gift ’ : daman- 4 giver Other masculine agent nouns of this type are dayman- 4 breaker hhujman- * fertile soman - ' soma-presser There is one secondary formation in -man, aryamdn- ' friendly, allied J . In Avestan there is the same anti- thesis between neuter and masculine man^ stems in Unman - nt, ‘ care 1 : Unman- masc. ' caring for * ; zaeman - nt. ' wakeful- ness ' : zaeman - masc. ' wakeful Formations of this masculine type in Greek are ihpoiv 4 one w 7 ho knows ’ (: iSfiev-at inf.}, rArjfjiwv ‘ enduring rjyc{id)v 1 leader etc. ; in Latin an alimones nom. pL (> aliiponia) is quoted. Masculines with the suffix -san- are very few 7 : Pusdn -, name of a god (cf. pusaryd- above, § 5), vfsan - 'male uksan - 1 ox ' (Engl, ox, oxen , Welsh ych, pi .ychen, Toch. oAso). The last two involve extended roots with incorporated $-suffix ( vaks vrs-) and from that point of view may be classed as an-stems. A feminine stftt-stem appears in ydsan- 1 woman Masculine tan-stems are not found in Sanskrit, but appear occasionally in other languages : Av. marHan- ' mortal aiwixsaitan- ' dweller, occupier Gk. t 4 ktwv, yclroiv 4 neigh- bour '. The compound suffix -in- is one of the most productive adjectival suffixes in the language. It may appear in primary formations, arcin- * shining but it is used much more fre- quently in secondary formations with a possessive sense, asvin- 4 possessing horses dhanin - 4 wealthy paksin- 4 w 7 inged etc., etc. Such adjectives may be formed in any number from stems in -a and - d , in w r hich case the final vowel of the stem is replaced by -in-, and less frequently from other stems, e.g. svanin - ' keep- ing dogs savasin - ' strong The suffix has invariably the accent, which is proper to these adjectival types, but it is in- variably in the weak grade, which cannot be original, and this is associated with the formation of a new 7 analogical, form of nom. sg. in f. Originally there must have been vrddhi.in the nom. sg. and vrddhied forms of this formation are found in other languages (Gk. ovpavia)V€s , etc"). In Latin there are some feminine action nouns w r hicb use the same formation (just as bhumdn- 4 abundance J , etc., have taken on the form proper to action nouns, see above, § 6), legio, legionis, etc. These have likewise generalised the strong form. The original system with alternation of strong form in nom. sg. and w f eak form in gen. sg., etc., is preserved in Oscan which uses a weak form of the stem THE FORMATION OF NOUNS I44 in the oblique cases : dat. sg. leginei, etc. The use of this suffix in a specifically possessive sense is found also in Iranian, e.g, Av. pardnin- ‘ having wings but examples are comparatively few. A fair number of adjectives are formed with the accented suffix -vin : sragvin- * wearing a garland tapasvin- ' heated tejasvin - ' brilliant etc. This complicated suffix which is un- known outside Indo-Aryan, seems to be a contamination of the suffixes - van - and -in-. Beside it there is a rarer suffix -min (like -mant beside -vant), e.g. vdgmin - 1 eloquent gomin - * possessing cows svamin- ' owner, master ’ ( sva - * one's own *). §9. Adjectival Formations in -nt- It has been noted above that -n as a neuter suffix could in Indo-European take the extension -t. This appears regularly in Greek (v&o>p, vSaros etc.), and elsewhere there are traces of it, though not many (Skt, vdrimat -, etc.). The -t could also be added to the adjectival «-suffix, and the compound suffix so produced has proved more productive than the neuter -nt~. In Sanskrit it appears in the suffixes -ant, - vant and - mant , all of which are highly productive. The suffix -ant- appears in a small number of adjectives, namely brhdnt- f great mahdnt- * great rhant- ' small pfsant- 4 speckled ' and rusant- ‘ bright to which may be added the pronominal adjectives iyant- ( so much' and kiyant- f how much The first three have the proper adjectival accent, and in pfsant-, rusant- the apophony shows that the radical accent is unoriginal. Similar adjectives in Iranian are seen in bdrdzant- 1 high ' and mazant- * big These are related to neuters in the usual way (Av. barpzan- ‘ height mazan- * greatness ') but in this case the adjectival forms have received the t extension whereas the neuters have not. , These adjectives are sometimes referred to as being oTparti- cipial origin. This is obviously not so, since the specialisation of this suffix in participial use, though ancient, is nevertheless a secondary development. The common usage of the suffix in active participles had not been fixed at the time of the separa- tion of Hittite, because there the participles in -ant have a passive sense as opposed to the active sense which prevails in the rest of Indo-European : kunant- 1 slain J as opposed to Skt. ghndnt- 'slaying'. Both are specialisations out of a more THE FORMATION OF NOUNS T45 general sense ' one connected with slaying Such a general meaning is all that is inherent to begin with in any adjectival formation, and it is by adaptation that the special functions of the various suffixes arise. The non- thematic participles and those from thematic verbs which are accentuated on the suffix keep the suffixal accent : addnt- 4 eating tuddnt- 4 pushing etc. Elsewhere it conforms to the regular accentuation of the verbal stem : bhdrant- ' bear- ing jighdmsant - 1 desiring to slay etc. This accent is shifted to the suffix in the weakest cases, an ancient feature which has often been levelled out, and the same applies to the suffixally accented adjectives ; gen. sg. adatds, brhatds. In the nom. sg. the stems in -ant differ from the adjectival types in simple -n in that the case is denoted by the termination s and not by vrddhi. This is so also in Hittite and it seems that from the earliest period of Indo-European that can be reached the nom. sg. was normally expressed in this way in the case of stems ending in occlusives. The suffix -vant occurs in a number of primary formations which illustrate its origin from the compounding of simpler suffixes. Primary formations are : vivasvant- also vivasvant -

• brilliant ' (: simple w-stem in vivasvan - nt. ' brilliance ’ and in

the Av. derivative Vivayhana-) , idivant- ' numerous, all # (cf. idsiyas- 'more numerous" and s as ay a- 4 abundant '), drvant- 1 swift, steed ' (: drvan- r id '), fkvant- 1 hymning, worshipping ' (: fkvan - 4 id "), satvdni ■ * name of a tribe of warriors ' (: satvan- 4 warrior ’), yahvdnt 1 young, youngest’ (: yahu - 'id'), vivak- vdnt- 1 eloquent The existence of pairs like fkvan- : fkvant ; drvan- : arvdnt illustrates the fact that this suffix is a ^-extern sion of a simpler I'aM-stem. In yahu : yahvdnt the analysis goes further and a simple if-stem is left. In Avestan we find drzgvant- r wicked as opposed to Skt. druhvan - 1 id f , and an interesting treble series, zrzzu-, vrzzvan-, drnvant- ' straight 1 which shows how the compound suffix is built up step by step. Av. b9zvani- 4 abundant ' bears the same relation to Skt. bahu as Skt. yahvdnt - to yahu-. Such pairs are found also in Hittite : dassu-: dassuwant- 4 strong, healthy It is as a secondary suffix that -vant is most frequently used in Sanskrit : dsvavant- 4 possessing horses kisavant- ‘ hairy ", putrdvant 4 having a son ' and so on in unlimited number. The usage also occurs widely in Iranian, Av. zastavant - 'having I46 THE FORMATION OF NOUNS hands amavant- 1 strong etc., and, outside Indo-Iranian, in Greek : a pUi$, x apUvra (for ° Feis 1 , °Fcvra) 4 having grace, graceful IxOvoetg 1 abounding in fish etc. In the Rgveda there are occasional examples of non-adjectival formations in -vant ; for instance dsvavant- sometimes appears not as an adjective, but as an abstract-collective noun, e.g. 1. 83. i, divdvati prathamo gosu gacchati ' he goes first in (the possession of) horses and cows where the singular collective corresponds to the plural gosu. Such traces are valuable in that there was originally an old class of neuters in - vant related to the adjectives in -vant according to the usual principle. Second- ary formations with the neuter suffixes are known in Hittite (antuhsatar ‘ mankind from antnhsas , ' man etc.), and such are to be ascribed to Indo-European. We may construct on these lines a neuter *dsvavar ‘ collectivity of horses, property in horses * alternating in the way usual in the case of neuters with

• dsvavan- } or with extension dsvavant on the basis of which
• aivavdnt 4 possessor of horses would be derived in the usual

way. Another piece of evidence is got by comparing Av. karsivant - ' cultivator ' with Ski. krsivald - and kdrswana - ‘ id The alternation of suffix between the last two words can only be explained by the existence of an old alternating neuter

• karsivarjn t and from this Av. karsivant - has been derived in

the same way as dsvavant-. According to the usual system one would expect the original accent of the adjectives in -vant to have been on the suffix. In the secondary formations in Sanskrit this accentuation appears in nrvanU * manly padvdnt - ' having feet J and nasvdnt- 1 having a nose where the primitive stems are monosyllabic, and in some cases where the primitive stem is accented on the suffix (but never when this stem ends in -a or -a), agnivdnt- 4 having a fire dsanvdnt- 4 having a mouth J , etc. The primary formations have the accent only in a minority of cases. The same tendency to throw back the accent was observed in the adjectives in -van. The suffix -mant appears in very few primary derivatives, namely virukmant - 4 shining dyumdni- 4 bright ' (cf. dyumnd nt. ' brightness susumdnt- 4 kind ' (cf. susumnd- nt. ‘ kind, ness '), dasmdni- 4 glorious ’ (only dasmdt nt. sg. used adverbi- ally). The relation of dsumdnt - ' swift ' [asumdt adv.) to dsu- recalls that of yahvdnt- to yahu-, etc. Elsewhere it is used as a THE FORMATION OF NOUNS 147 secondary suffix in exactly the same sense as -vant. Occasion- ally the two suffixes are used after the same word, e.g. agnimdnt- beside agnivant-, but usually one only of the two suffixes is used in connection with each word. There are no absolute rules to say when each suffix will be used, except that - mant is regularly employed after stems in -u, pasumdnt- ' possess- ing cattle etc, (frequently also to avoid repetition of v, yavamant - 'rich in barley', etc.). This rule is interesting because much the same kind of rule is found in Hittite both in the case of suffixes related to this (Inf. arnummar ' to bring Supine wahnumanzi ' to turn (trans.) 1 as opposed to esuwar , asuwanzi from es - r to be etc.) ; and in the 1st person plur, of the verb (arnummeni ‘ we bring ' as opposed to epweni ‘ we hold ’, etc.). The accent of the adjectives in - mant follows the same rules that apply to the formations in - vant . §10, Thematic Adjectival Formations from r- and n- Stems There was an alternative way in Indo-European of making adjectives and agent nouns from the primitive neuter forma- tions, and it was equally commonly used. This was the addition of the accented thematic vowel. These derivatives have the same meaning as those just described, and the two types of formation often exist side by side, e.g. Skt. atrd - ‘ eater ' : attar- 9 id ’ ; Hitt, vestaras 1 herdsman ' : Av. vdstar- ' id ' ; Gk. larpos, IdTrjp ' physician ' ; Gk. £17 rpos ‘ executioner ' : Skt, ydtdr - ‘ avenger, punisher ' (cf, ydtand ' punishment, tor- ment ’ with -n- indicating an old alternating neuter). The accent is normally on the suffix, but it is occasionally transferred to the radical syllable: ddmstra - ' fang The type has prospered, and with the dying out of the bulk of the old neuter types, the suffixes -rd, etc., have come to have the appearance of primary suffixes. Formations of this kind are made on the basis both of the simple neuter suffixes ~(a)r, -{a)n and of the compound suffixes. (i) Examples of formations in -rd are : ugrd- ' powerful ' (: Av. aogar - ' strength’), usra- ' matutinal, shining like dawn ' ( vasar usar- ' dawn '), udra- ' water-animal, otter ’ (Gk. v 8 a>p, etc.), a-vadhra- ' not injuring ’ ( vddhar ' smiting ; weapon ’), ksudrd - ' small ksiprd - ' swift *, vakrd- ' crooked hasrd- 1 laughing etc. An old neuter alternating rjn stem is often THE FORMATION OF NOUNS I48 indicated by the existence of derivatives from the w-steirt side by side with these formations in - ra , e.g. beside vipra- 1 inspired ' (with retracted accent), vipanya , etc. cf. also ksiprd- 4 quick ' : ksepnu-, ksipanu - ; grdkrd - 4 greedy ' : grdhnu- 4 id ’ ; dhvasra- 4 dusty ' : dhvasani - 4 sprinkler (a cloud) 1 ; svitrd- 4 white 1 : ivetana 4 dawn etc. The accent is thrown back in only a minority of cases and these are mainly substantivised adjec- tives : sura- 4 strong man, hero ’ (Gk. a Kvpos 1 powerless '), ajra- 4 field ' {J~aj- ; original accent in Gk. ay pa? ; Gk. dyeiptn

• gather, collect ', is formed from a primary neuter *ager, cf. the

type ratharydti), vajra- ' club, thunderbolt ' (‘ smasher, crusher cf. Gk. ayvv/ju) vdpra- ' mound, earthwork * (Av. vajra - 4 snow ’), tumra- 4 humped ', of the Indian bull (cf. Lat. turned , tumor , etc.). In this, the oldest type, the thematic vowel preceding the r is eliminated on account of the following accent. There are also a smaller number of adjectives in which the -a is added without any such reduction ; dravard- 4 running patard- 1 flying nyocara- 4 suitable, agreeable ', dvara- 4 lower ' (: avdr), upara - 4 nearer 1 ; with vrddhi, vdnara- ‘ monkey ' (vanar° 4 forest ’), vdsara- f matutinal ' (vasar° 4 early morning ’). The same type with full vowel before the r appears also in other languages, Av, aSara- 4 lower urmsara- 4 fugitive * (urvaes-), Gk. eAev- Oepos 1 free etc. Such forms must have originated at a time when the accent had ceased to have the effect of reducing un- accented syllables. A parallel series in -Id was formed in Indo-European. In the Vedic language this suffix would also appear as -ra. Instances of -Id occur in Sanskrit, e.g. sukld- 1 white J (also sukrd-), sthula- ' thick ' (also sthurd -), gopald - 4 cowherd but it is never possible to be sure about the origin of l in Sanskrit. (ii) From the base -variur : With strong form of suffix, bhdsvara- 4 brilliant ' (*bhasvar nt. 4 brilliance '), tsvard - 1 lord sakvard- 4 strong adhvard - 4 sacrifice 1 (of adjectival origin from its accent ; cf. adhvdn- 4 way '), sthdvara- 4 stable ' (also sthdvand- showing old alternating neuter), nasvara - ‘ perish- able vyadvara - 4 a gnawing animal ' (: Gk. efSap nt.), nisadvard- 4 mud * ; with /-suffix inserted, ilvard- 4 going srtvara - 4 id jitvara 4 victorious The variant -vala appears in vidvald 4 clever J (cf. Gk. elSvMs 4 id '), palvala - 4 pond ' (cf. Lat. palus 4 marsh J ), and in a number of secondary formations, krsivalaTHE FORMATION OF NOUNS I49 ‘ cultivator ' (: kdrstvana - * id-'), urjasvala- ‘ strong parisad - ao/a- * having a parisad, king ' f asutivald - 1 priest who presses soma'. A vrddhied formation is seen in bharvard- 4 impetuous ' (cf. bhurvdni - * id.' and Lat. fervor , etc.). These formations often exist side by side with adjectives in -van : isvard- 4 lord ' : Av. i’saan- ; iivard- 4 going ' ; °itvan- ; sthdvara- * standing, stable saslhavan- 4 standing together ' ; ptvara - * fat ' : pivan - ‘ id This relationship is based on old alternating neuters, *l$varjn 4 authority etc. The weak form of the suffix appears in bhdsura - ' shining ', chidurd- 4 tearing ', bhangurd- 1 breaking bhidura- 4 splitting vidurd- 4 wise medurd - ‘ fat ankura- 4 bud, shoot ; swelling tumour 1 (: Gk. oyKvXos 1 swollen, proud'), ksura- ' razor' (Gk, %vpov nt., cf. O. SI. cesati 4 to comb etc.). In the nominalised svdsura- the accent is retracted, as commonly ; the accent of Gk. dfcupos is more original. Beside pdmsura - 1 dusty and madhurd- ‘ sweet ' forms with l appear, pdmsuld-, madhula without it being possible to say which is original. Beside smasrula- 4 bearded ' there is smasruna-. The weak form of the suffix appears as -vr- in tivrd- 4 intense '. (iii) A parallel suffix -ird appears in rudhira - ‘ red ’ (as opposed to simple -ra- suffix in Gk. ipvdpos , etc.), badhird- 1 deaf ' (, J~ba(n)dh - ' to bind, obstruct madird- 1 intoxicating isird- 4 vigorous ' (; Gk. Upos from simple r-stem), sithira- 4 loose rathird- 4 charioteer mtehira- 4 wise ' (with retracted accent), sth-ird - 4 firm sph-ird - ‘ fat rucira - * bright ' ; more rarely -ila, trdild- 4 porous salild 4 flowing nt. ' water ' beside sarira-, sithild - 1 loose ' beside iithird - ; with guna of suffix (like -vara- f vald) samusyald - 4 cohabiting ' ( sam + vas -). (iv) From the neuter suffix -mar there are a few such de- rivatives : admard- 4 gluttonous ' (implying *admar nt. ‘ eat- ing '), ghasmard 4 id srmard- 4 a swift moving animal ' (cf. Gk. oppalvto, denom. vb. from n-stem), pdmara- ‘scabby; miser- able ' (also pdmand- from n-stem), asmard- 4 stony ' (cf. aimdn above, § 6) ; with weak form of suffix dJiumrd- 4 grey ’ ; with l paksmald- 4 having (long) eyelashes 9 (pdksman- nt. ' eyelash '), slesmald - alternating with slesmand- 4 afflicted with phlegm ' (i ilesmdn - masc. * phlegm '), bhimald - ‘ fearful It was observed in dealing with the simple r-neuters that they might either appear with guna (udhar) or by the weak form of the suffix followed by the extension t [ydkrt ). Forms of the 150 THE FORMATION OF NOUNS latter type can be deduced from certain derivatives of mar- stems : kavmatha - ‘ workman 1 from *karmrtha- (*karmrt(h)- ‘ work • -fa; aspiration as in kdprth-), likewise narmatha- ' jester * and harmuta ‘ tortoise ’ ( *harmrt-a - ‘ an animal pos- sessing a *harmrt, i.e. ' roof or shell ', cf. harmyd 1 roof '). (v) A few such adjectives are formed from sar-stems : sapsara- ‘ attending on ' {J'sap-), matsard - 4 exhilarating; ex- hilarated. ' (cf. mandasand- from the corresponding san-stem), °rksard- ‘ hurting ; thorn ' (related to arsasdnd - ‘ injuring ’ in the same way as matsard- to mandasand-), samvatsard- 1 year ', dhusara- ' grey ', krsara - ' a confection of sesamum, rice, etc/, krcchrd- 1 painful, difficult' (if for *krpsra-), usra- 'bull' (for *ur$rd-, i.e. *vrsra -sr- alternating with the sun of vfsan -) . (vi) Based on the old neuters in - tar there-are adjectival form- ations in -tar a and -trd. By a secondary development the suffix -tar a has come to be specialised in the formation of comparatives, but there are a few old formations where this is not so, and where the original, more general function of the suffix is apparent. For instance asvatard- ‘ mule ' is an animal which partakes of the nature of a horse (*asvatar nt.) and in the same way Iranian kapautara- 1 pigeon ' (Mod. Pers. kabutar) is a bird characterised by bluish-grey colour (*kapautar nt.). In kdrotard ‘ filter, sieve ' there is a vrddhied formation based on neuter action noun *karotar ' sifting This root most commonly appears with i- extension (Gk. Kpivoj , etc., Ir. crlathar ' sieve ') but a u- extension, as here, is found in Goth, and-hruskan f avaKpivtw Other examples of this type are vatsatard - 1 yearling calf and with retracted accent sanutara- ' clandestine ' (: sanutdr adv.) and divdtara - ' diurnal The adjective antara - ‘ interior ' is derived from ant dr ' inside (Lat. inter, etc.) and this in its turn from IE en ‘ in ' + the neuter suffix - ter . In the same way Skt. pratara - (only in the adv. pratardm), A v. fratara- 1 being in the front ', Gk. 7rp6r€pos ' former ' are derived from pro- through an intermediate *proter * the front ' (adv. *prot£r ). In this way there arises a class of adjectives based on prepositions, such as Skt. avatar a- * lower ' (only in the adv. avatar am), uttar a- ‘ upper Av. nistara- 1 being outside Gk. tt porepo? ‘ former v7T€pr€pos ‘ higher ', etc. These prepositional formations have a comparative meaning ('higher, lower ', etc.) but this does not come from the suffix but from the nature of the base to which it THE FORMATION OF NOUNS 151 is attached. On the basis of these formations, and from similar cases like Lat, dexter , Gk. S compos 4 right, being on the right ' there was evolved for this suffix a special comparative meaning, with which it is added as a secondary suffix to adjectives in Indo- Iranian and Greek (rarely elsewhere ; 0. Ir. Ubrither, comp, of lebor 4 long ') : dmatara- 1 rawer cirutara - 4 dearer tavastara- 4 stronger etc. : Gk. wportpos 4 rawer Kov^orepos ‘ lighter etc. The fact that it is not widespread in Indo-European shows that this use of the suffix is comparatively late. There are a few adjectives and nouns of adjectival origin in -trd : air a- 4 eater 4 (for at-trd- from ad-), vrtrd- 4 enemy ; n. of a demon mitrd - 1 friend ; n. of a god putrd- 1 son ’ (cf. Oscan puk- lum 4 pucrum ’, Paelignian puclois, with -kl- <~tl- ; Lat .puer with simpler r- suffix) ; with retracted accent, ddmstra- * fang', tdrutra - 4 victorious vibhrtra - 4 who is carried about (Agni) johutra- ' calling aloud ', ustra- ‘ draught-animal, camel ’ (apparently from vah- with irregular sandhi, cf. Av. vastar - * drought animal ' — vodhar-) ; with vrddhi, jattra- 4 victorious 4 ; with gradation -air a, dmatra- 4 violent ydjatra - ' worthy of wor- ship (vii) With -nd- there are a number of nouns and adjectives, e.g. stend- 4 thief 4 , yajna- ‘ sacrifice ’ (Gk. ayvos ‘ holy, pure ’), ghrnd- 4 heat nagnd- 4 naked usnd- 4 hot 1 ; with radical accent svitna- ‘ w r hite The most common use of the suffix is to make participles from certain verbal roots (about seventy) ; bhinnd - 4 broken bhugna- 4 bent purna 4 full mldnd- 4 withered etc. With the gradation - and there are formed a certain number of agent nouns : karand - 4 active tvarand 4 hastening 4 , krosand- 4 shouting vacand- 1 speaking svapana- 4 sleeping etc, These are distinguished by their accent in the usual way from the corresponding class of neuter action nouns : cf. katana - 4 deed vdcana - 4 word In Germanic and Slavonic this forma- tion makes passive participles (O. Sh nesenu 4 carried Goth, fulgins 4 hidden '). The contrast in accent between kdrana- nt. and karand - masc. is also found in Germanic, where the infini- tive is the equivalent of this neuter type : Goth, filhan 4 to hide fulgins 4 hidden Though the old type of accentuation is preserved frequently in these adjectives (as above), the system was breaking down, and radically accented forms occur, particularly from verbs of the THE FORMATION OF NOUNS 152 first class with fixed radical accent : javana- 4 hastening ' (jdvati ) , dydtana - * shining ' (dyotate), etc. (viii) Corresponding to -vara and -urd thematic adjectival stems are made on the basis of the neuter van- suffix, with two gradations, -vana and - una . (a) vagvand- ‘ talkative susukvand - 4 shining ', satvand - ' warrior ' ; from prepositional bases pra- vand- f sloping forward, inclined ’, udvana - ‘ elevated (b) mi- thund- f paired ' (Av. miOwara- from alternating r-stem), sakuna- ‘ bird ’ ( Jsak, as prophesying the future), aruna- ' red ddrund- * terrible ' ; with radical accent drjuna- ‘ white ' (Gk. apyvpos ‘ silver 1 from r-stem ; cf. also Skt. rjrd- from uncom- pounded r-stem), plsuna- ' slanderous, treacherous ’ (cf. Gk. iTiKpos; ' bitter, inimical from simple r-stem), tdruna - r tender ' (cf. Gk. Tapfy w r ith uncompounded n- stem, repvs , uncom- pounded j^-stem), visuna- ‘ various ' ; with penultimate accent, dharuna- ‘ holding yaiuna- ' energetic Fuller types of gradation are found in occasional forms : -avana in sravana- ’ lame 1 (Lat. cl-au-dus, etc.), lavand 1 salty 1 ; nt, salt ( *slavana - : Lat. sal) ; -ona in srona-, slond - r lame syond- ‘ soft, agreeable durond- f house (ix) The suffix -ind parallel to -und appears in a few words ; vrjind- ‘ crooked harind * yellowish : deer antind- * over- powering asind - ' old '( J as -) , sakind- f strong ' ; with radical accent, ddksina- ‘ right With guna of the first element the com- bination appears as ena only in the feminine sdmidheni (fk) 1 con- nected with lighting the fire In Iranian the combination -aina is common : Av. izaena - ‘ made of leather drvaena - 1 wooden etc. In Sanskrit there are certain further derivatives from such a suffix, namely the gerundives in -enya : vdrenya- ‘ desirable Iksdnya- 4 worthy to behold etc. The gradatjon -yana (cf. - vand , etc.) is not found in Sanskrit, but it appears in Av. airyana- ‘ Aryan The full grade of both suffixes (- ayana ) is not found but certain patronymics with double vrddhi (Ddk- sdyana etc.) appear to be based on such a formation. The corresponding forms in Avestan (e.g. V ayhuddtayana-) are without vrddhi. (x) Adjectival formations from neuter man- stems are rare : nimnd - ' low ; nt. depth The neuters nrmnd - 4 manliness ', sumnd- 4 kindness ' and dyumnd- 4 brightness ' appear from their accentuation to be of adjectival origin. Formations of this type are commoner in other languages, e.g. Lat. alumnus THE FORMATION OF NOUNS 153 4 nursling * (cf, alimones, alimonia ), Gk. urtpcuvos ' hard the A vest an middle participles in -mna, yazzmna-, etc., and the corresponding Greek participles with guna (<f>ep6fievos, etc.). The two types differ in the same way as - vanaj-una , -varaj-Ura, stc. The corresponding Sanskrit participles with vrddhi will be treated below. (xi) There are a few r thematic adjectives based on the suffix ~$an : krsna - ‘ black ' (: 0. Pruss. kirsna 0. SI. crunti), ilaksnd- ' smooth akstta- 1 oblique ' (adv. aksnaya) , tiksna- 4 sharp krtsnd- all ' ; also a few substantives of adjectival origin : pyuksna- ' covering for a bow hdliksna - 1 a kind of animal ; a particular part of the intestines mrtsna- masc, nt.

• dust, powder desnd- nt. 4 gift 1 (' what is given J ). With

different gradations of suffix kardsna- ' arm vadhasnd- masc. or nt. * deadly weapon ' ; Pusand- beside Pusan „ (cf. satvand- and sdtvan duvasana- ‘ going far (or the like) (xii) Apart from cyautnd- 4 stirring nt. 1 exploit 1 (: Av. syaoOna -) the suffixes -tna and -tana are specialised in connec- tion with adverbs of time (cf. the similar use of -tar a in dlvdtara-): ntitana ntitna - 4 belonging to the present time pratnd- ' old sandtdna- * eternal adyatana 1 of today hyastana- ' of yes- terday ', etc. In the above examples we have a series of adjectives all formed in the same way by the addition of the accented thematic vowel to the various r- and n- suffixes. It has been pointed out that these suffixes were capable of taking the ex- tension -t, and there are a few adjectival forms which are based on such an extension. An example from an r-stem is muhurtd - 4 moment f from muhur (: Av. m&rzzu- 1 short of time). Re- ference has already been made to certain formations showing Prakritic tendencies, karmatha-, harmuta. From the «~suffix extended by t there are a number of thematic formations which to judge by their accent were originally adjectival, vasantd - ' spring ' (cf. vasar °), vesanta- 4 pond * {J’vis ‘where rainwater settles ') and with weak grade of suffix avoid - 4 well Based on the man- suffix there are hemanta- 4 winter simdnta - ' parting of the hair 1 (: siman- 1 id. boundary ') and asmanta - ‘ fire- place f ; on vanjun, sakunta- 1 bird ' beside iakuna - (also sakuni-, sakunti - with f-suffix, cf. sdkvan-, etc.). In pdrvata- 1 mountain which we may compare with Hitt, peruna per unant- 4 rock there is another variant of the weak form of 154 THE FORMATION OF NOUNS this suffix (- wn -), and, as often elsewhere, retraction of the accent. These formations are not very common and some of them from quite an early period were misunderstood as if they were compounds with anta- ' end 1 as second member. For this reason the variant forms, vesdnta-, simdnta- occur. It is not un- likely that some other apparent compounds of this type, e.g. kar mania- ' work, business ' (Pa. kammanta- ), which only occur in the latter form, are corruptions of this type. §n. Thematic Formations with Vrddhi of Suffix We have seen above that there exist two quite distinct ways of making adjectives and agent nouns on the basis of the primitive neuter suffixes. In addition there is a series of forma- tions which must be classified by themselves, since they par- ticipate in the characteristics of both the above types. They are thematic formations, frequently accented on the final syllable, but at the same time the suffix to which the thematic vowel is attached, has vrddhi, like the agent nouns of the type brahman - in the norm sg. In this respect they represent a cross between the two systems, and they appear to be thematic ex- tensions of formations of the brahman - type, based on the nom. sg. For instance we may explain Vedic karmara- ' smith r as follows. From the evidence of Hittite and Greek it is clear, as shown above, that the neuter man- stems were originally alter- nating steins with nom. acc. sg. in -mar. We have also seen that on the basis of all these neuter suffixes in r and n , simple and compound, adjectives and agent nouns could be made by the method illustrated by brahman Instances in connection with most of the suffixes were quoted. On this analogy we might expect on the basis of *kdrmar nt. (obi. base karman-) an agent noun *karm&r. Vedic karmara- is a thematic extension of such a form, and it has been already pointed out above that this tendency to thematisation, which is familiar from the later history of Indo-Aryan, had already been operating in the pre- historic period. Another formation of this type appears to be mdrjdrd- ' cat but they are exceedingly rare from r-stems. On the other hand . such formations^are, co mmon in the case of the R-sulfixes^ and m part icul ar they hay e_given ris e to a se ries oFmiddle^p^iciplesJa^Sa to which nothing exactly cor- ^ responds in the other languages. THE FORMATION OF NOUNS 155 From the simple w-suffix we have the suffix -and which is used in the formation of middle participles of the type adand- ( eating duhdnd- * milking, etc/ Skt. addnd - is a thematic formation corresponding to the non-thematic Lat. edo-dnis. A formation of exactly the same kind is Lat. coldnus, but such are exceedingly rare. There is final accent in the above examples, and in the perfect participles, bubudhand- ' waking etc. ; but in the reduplicating presents and in the desiderative, initial accent. Other formations of this type are rare : samdnd- 4 same, common a thematic extension of a masculine n- stem such as appears in Goth, sama (as sdmana- nt. is an extension of a corresponding neuter n-stem) and pur and- r ancient *. On the basis of the man-suffix there are middle participles in -amdna, ydjamdna- ‘ sacrificing etc. It is based on that form of the neuter men- stem which is attached toThe root with the thematic vowel : Gk. inf., etc. The vrddhied mas- culines corresponding to this type are represented by such forms as Gk. rjyefiwv ‘ leader inySt/wiv ' one who cares for Them- atic extension of such a type produces the Skt. type ydjamdna-. In the thematic conjugations to which this type of participle is attached, the stable verbal accent prevails. This type of parti- cipial formation is peculiar to Sanskrit, since the most closely related types of participle (Av. yaz9tnna-> Gk. <f>ep6p.evos) are differently formed, in the manner indicated above. As has happened in other cases it was by adaptation that such forma- tions acquired the status of. middle participles, and this adapta- tion seems to have been comparatively late, since such parti- ciples are known from only a small section of Indo-European. There is a parallel series of formations in - asand , most of which have the character of pseudo-participles. Such are : jr ay asand 4 far-extending namasdnd- 4 rendering homage a*! A«'(^)bAiydsdna- 4 fearing mandasdna- 4 rejoicing vrdhasand- 4 growing savasdna- 4 strong arsasdnd- 4 injuring sahasdna- 1 overpowering In many cases there exists a neuter - as - stem beside these formations (jrdyas-, ndmas- t savas -, sdhas-) . We have seen that the neuter as-stems were capable of taking the extension rjn. On the basis of the saw-stem produced by such an extension (*ndmasa?i-, etc.) these adjectives have been produced by the combined method of vrddhi of the suffix and addition of the accented thematic vowel. It will also be observed that they bear I56 THE FORMATION OF NOUNS the same relation to the infinitives in -sen (Gk, i.e.

• <f>ep€i rev, *bheresen) as exists between bhdramdna - and pipzixev.

To some extent they have acquired the character of participles, but the process of adaptation is incomplete. Unlike the parti- ciples in -mdna they are not integrated with any tense stem, and the practice of classifying them with the participles of the s- aorist was more of an emergency measure than a serious attempt at their analysis. In urdhvasand- ‘ erect ’ we find - sand used purely as an adjec- tival suffix. In Pa. rakkhitamdnasana- ' whose mind is guarded ' f a formation of this type is used to provide an adjectival termin- ation for a bahuvrihi compound. Suffixes of the same type are made on the basis of the other compound ^-suffixes : -avdna- in bhfgavana - ‘ shining vdsa- vdna- 4 possessing riches and in the proper names Apnavdna- and Pfthavdna - ; -aydna- in tUrvaydna- 1 victorious ' and H dr ay ana- n. pr. Here belong th e middle participles in - avdfi a -» v - made in the Epic language from tenth class and causative verbs^T cintaydna pdlaydna-, etc. Though not used in the Vedic lan- guage, nor allowed in the Classical, this formation could be an ancient dialectal feature, §12, Various Extensions of the r and n Suffixes The suffix -d is regularly used to make the feminine of the thematic adjectives classified above. In addition it appears in a number of independent formations. In ydsand (once yosdna) 4 woman ' and kanyand 4 girl 1 it appears as an extension of feminine «-stems (yosan- 4 woman Av. kaintn- 4 girl '). The formation k any aid which is also found shows that the fern.

• kanyan- on which kanyand is based was originally an alternat-

ing neuter. There are also words which may be of adjectival type though no corresponding masculine occurs, e.g. dstrd s goad ' (' driver '). In addition there are a number in which -a is simply an extension of old neuter r and n stems, adding nothing to the meaning, e.g, mitra "measure 1 (*mdtar + a). Others are surd 4 intoxicating liquor dhara 4 cutting edge urvdrd 4 cultivated land ' (Av. urvard * crop * : an old rjn neut. of Indo-European is attested by Ir, arbor , nom, acc. pL arbann )„ " vdgurd 4 net tdmisrd 4 darkness Mira r oblation '. The same type is formed on the basis of the n-suffix : tfsna 4 thirst THE FORMATION OF NOUNS 157 sdnd 4 army sthUna 4 post dhdnd 4 milk-cow The gradation - and is common, the words being accented either on the last syllable or the penultimate (for accent of ^-sterns see below, p. 191) : arhdnd 4 worth barhdnd ‘ might vadhand 4 slaughter vandnd 4 desire ' ; asand ‘ missile jar mid 4 old age 1 , dyotand

• brilliance rasand 1 rein hasana , With these action nouns

in -and compare the Greek type v&ovr} 4 pleasure There are a few such formations in -i and -u : rdtri- 4 night tandri - 4 sloth nabhanu - 4 spring Stems are frequently made by the addition of i and u to the r and n suffixes. (a) The suffix -i serves as an enlargement of r-stems in dngh-r-i masc. 4 foot ' (cf. 0. SI. noga 4 foot ' from different gradation of root) abhri- fern. 4 hoe dsri- fern. 4 edge, point anguri fem. 4 finger Adjectives are arcatri - 4 singing hymns dtri- 1 devouring bhuri- 4 abundant subhri- 4 beautiful jasuri- 4 exhausted dasuri- ' pious sdhuri - 4 mighty ' (cf. Gk. ixvpos, oxvpds ' firm ' with thematic suffix) ; nouns of adjec- tival origin, surd- 4 patron ’ (J ~su- t as the instigator of the sacri- fice), vddhri- 4 a castrated animal ’ (: vadhar ; Gk. Zdpis 4 id/). The suffix -i is in the same way added to the w-suffix in (fem.) sreni- 4 row sroni- ( hip sfni- {srni-) 4 sickle jurni- 4 heat jydnd- 4 loss gland- 4 fading ' ; (masc.) ghfni - 4 heat yoni d womb The suffix -ni is used to form a number of adjectives and nouns of adjectival origin. Such are asni- 4 eating vdhni- 'carrying' (later 'fire'), turni- 'speeding 1 , dharni- 4 sustaining ', preni- 4 loving pfsni- 4 speckled ' (cf. Gk. TTtpKvos with thematic suffix) ; of adjectival origin, agni- 4 fire '. With the gradation -and there are such nouns as (fem.) dyotand- 4 brilliance ', variant 4 track ', ardni- 4 firestick ' (ar- 4 to fit ', alternating r in atari - 1 door-leaf Adjectives are tardni- 4 swift cardni- 4 moving etc. Similar formations from the compound n-suffixes are : hrdduni - fem, 4 hail tuvisvani- 4 powerful ' {-vani as secondary suffix), aratni- masc. 4 elbow, cubit J , i$tdni r .e pithet of Agni, turvani- 4 overcoining bhurvdni- 4 agitated susukvani- 4 shining par sand- 4 carrying across saksdni- 4 overcoming carsand - ‘ active 1 ; no longer of clear derivation, vrsnd- 4 ram pdrsni - fem. 4 heel ' (Gk. rrripva, etc.). (b) The w -suffix in combination with r produces occasional neuter nouns, asru 4 tear ' (Toch. A. dkdr, plur. dkru-nt), imairu 4 beard and some adjectives, dhdru 4 suckling ' (Gk. 0i]Av? THE FORMATION OF NOUNS 158

• female "), bhiru- ' timid ", patdru - 4 flying " ; in combination

with n some adjectives and nouns of adjectival origin : dhrsnu- 4 bold ", grdhnu - ' greedy ", sunu- 4 son V dhenu- fern. 4 cow / bhdnu- masc. 4 light The combinations -inn and -snu were fairly productive in the formation of adjectives ; krinu- 4 active ", dartnu- 4 breaking ", dravitnu- 4 running ", pTyatnu- 4 reviling ", stanayitnu - 4 thunder ", kavatnu- 4 stingy, mean ' ; vadhasnu - 4 murderous ", jisnu - 4 victorious ", carisnu- 4 wander- ing", etc. The combination -vanu appears in vagvanu- 4 noise §13. The Suffix s The neuter suffix -as is better preserved than any of the other old neuter suffixes, and a larger number of words of this type have directly corresponding words in other IE languages than is the case with any other suffix. Such are : srdvas- 4 fame " (Gk. kAcos, Ir. clu 4 id/, O. SI slovo 4 word "), jdnas- 4 race " (Gk, y4vo$, Lat, genus), mdnas- 4 mind " (Gk. p4vos 1 spirit '), haras- 4 heat 1 (Gk. Qepos 4 summer "), ndbhas- 4 cloud, sky " (O. Sh nebo , Hitt. nepis-) f drsas - 'piles" (Gk. ZXkos, Lat. ulcus 'ulcer"), vdcas - ^ xe 4 word " (Gk. Fe 7ros t enos), sddas- ' seat " (Gk. IS 09 ), pdsas- 4 male organ ' (Gk. n cos), ddhas- 1 fuel " (Gk. aWo$ 4 burning "), sdhas- 'strength" (Goth, sigis 'victory"), dpas - 'work" (Lat. opus), anas - 4 waggon ' (Lat. onus 4 burden "), vdnas - 4 charm, desire " (Lat. venus ), Other examples of this very frequent suffix are tdpas - 4 warmth ", pray as- * pleasure ", tejas - ' splen- dour ", dohas- 4 milking ", katas- 4 deed ", hdsas 4 injury 1 {hints -) , etc. The normal type has the regular radical accent of neuters, also guna of suffix. Accent shift in declension (type ydkrt „ yaknds) has been abandoned, and the accent remains on the root throughout the declension. There are some variant types of gradation, namely ( 1 ) vrddhi of root in dgas- 4 sin ' (Gk. ayos), dpas- 4 work " (usually dpas-), vasas- 4 garment ", vdhas- 4 offer- ing ", pdjas- 4 side, surface " ; (ii) weak grade of root in uras - 4 breast ", stras - 4 head ", juvas - 1 speed ' (also javas-), mfdhas- 4 contempt ", dtivas- 4 offering " ; (iii) reduction of suffix in yds- 4 welfare " (Av. yaos-, Lat. ius), dos - 4 arm These variants show that the working of apophony was at one time active in these formations, although in most cases it has been levelled out. The type with weakened grade of root is interesting since it can only be explained out of original terminational accent in THE FORMATION OF NOUNS 159 the oblique cases. There are rare examples of this switch of accent in declension, e.g. bhiyds- * fear ' (transfer to fem.) instr. sg. bhtsd (beside bhiydsa ). Some twenty-five dative infinitives are formed with this suffix. These are sometimes accented on the root, dyase ‘ to go cdksase 4 to see dhiyase 4 to cherish but much more commonly on the suffix, reuse ‘ to praise cardse 4 to fare jtvdse ‘ to live dohdse ‘ to milk bhojdse 4 to enjoy sobhdse 4 to shine J , spurdhdse 4 to strive ’, etc. The origin of this anomalous accentuation, which is in contrast both with usual fixed radical actent of as- stems and with the original termina- tional accent of the oblique cases, is not at all clear. It may be noted that it recurs in other types of infinitive : davdne 4 to give vidmdne ‘ to know In a very few dative infinitives based on the 5-suffix terminational accent is found with reduc- tion both of root and suffix : jise ‘ to conquer siuse 4 to praise The normal locatival accent appears in updsi * in the lap ' (only this form), cf. aksdni , etc. The related adverbial accent appears in purds 4 in front liras 4 across ' and mithds 4 mutu- ally 1 ; cf. avdr , etc. There are a few instances of transference of gender in the case of action nouns in -as. Such are, (masc.) lavas - ' strength ' (as well as * strong 1 adj.), fem. jards- 4 old age ’, bhiyds- 4 fear ' and usds- 4 dawn ' (cf. Gk. Tjajs), These appear mainly to be due to personification. The transference involves the adoption of the adjectival accent (as in bhumdn- 4 abundance ', etc., above). This neuter suffix was capable from an early period of being extended by the addition of the neuter r- and ^-suffixes. Examples of this (sirsnds, dosnas, gen. sg., etc.) have been given above, together with derivatives from such stems (matsard-, mandasdna -, etc.). It could also be added to other suffixes, pro- ducing a variety of compound suffixes, e.g. : -las in rdtas 1 seed srotas 4 stream ' (simple l- stem in sravat) ; -nas in reknas- r in- heritance, property J , dpnas - ‘ wealth J , drnas- 4 flood ', and, preceded by i and t, drdvinas- 4 property pdrtnas- ‘ abund- ance ' ; -5A5 (repetition of the suffix) in daksas- * ability, dexterity ' (simple -as in daias-ydti) and pdksas- 1 side ’ (simple -as in pajas) ; -vas in pivas - 4 fat vdrivas- 1 expanse This latter combination normally appears in the weak form -us : l6o THE FORMATION OF NOUNS drus- 4 wound dyus- 4 length of life tapus - ' heat tdrus- ' victory ', ydjus - ' sacrificial formula vapus- ‘ beauty pdrus- 4 knot {of plants) dhdnus - * bow An alternative n-ex tension appears in dhdnvan- 4 bow 1 and parvan - ' joint and the simple western iyu nt. appears as well as dyu-s- (cf. also aru- in arun- tuda - ), In the same way 5 is added to the i-suffix to produce the compound suffix -is. Of these neuter nouns jyotis- 4 light and vydthis- 4 perturbation 1 are accented on the root according to the general rule. In the rest the original accent system has been disturbed, and they appear with accent on the suffix: arcis- ‘ flame ’ (also transferred to feminine), chardts- 4 pro- tection, cover barhts- 1 bedding, straw ’ (Av bardzis-), rods- /v T r ' < ‘ light vartis - 4 track socts- 4 flame sarpis - ‘ butter’ (Gk. ZXttos with simple 5-stem), havis- ‘ oblation kravts- 4 raw flesh ’ (simple i-stem in dkravihasta- 4 whose hands are not bloody cf. also kravyd-, Lith. kraujas , etc.). The neuters in -as can be turned into adjectives and agent nouns by the usual process of suffixal accentuation associated with vrddhi of the nom. sg. The neuter and adjectival types appear side by side in the case of dpas - ‘ work ' : apds - 4 active taras- ‘ energy, force ’ ; tards- 4 forceful ' ; ydsas- 1 beauty, glory ' : yasds- 4 beautiful ’ ; tydjas- 4 leavings something let go of ' : tyajds- 1 offspring ' ; mdhas - 4 greatness * : mahds - ** ‘great"; rdksas- ‘injury, damage’: raksds- ‘demon’; duvas- 4 worship ’ : duvds - ‘ worshipping Other examples of the adjectival formation are tavds - 4 strong tosds- ‘ bestowing dhvards- 4 deceiving yajds - ' worshipping ’ and vesds- 4 neigh- bour The same antithetic types appear in Greek : pev&os 4 falsehood * : pev^s ‘ false etc. From the compound suffix -vas there are some adjectival formations. The usual antithesis of the two types is seen by comparing vdrivas- nt. ‘ expanse ’ on the one hand, and okivds - C**- masc. ' accustomed to, familiar ' on the other. Formations of the same type are seen in midhvas- 4 liberal dasvas - 4 worship- u^^Ojping ’ and sahvds- 4 overcoming which inflect like perfect par- ticiples. On the other hand in fbhvas - ' skilful ’ and Sikvas- 4 id.’ are influenced in form and accentuation by the coexisting van- stems. There is a special connection between the adjectival suffixes - van and -vas in Sanskrit, because the latter is used to make the vocative singular of stems in the former (rtdvas, vibhdvas , etc. ; likewise of vantf-stems : rayivas , bhagavas, THE FORMATION OF NOUNS 161 sacivas, etc.). In Avestan we find this suffix used to form the nom, sg. of ^antf-stems, amavd nom, sg. of amavant- 1 strong Apart from the above examples the adjectiva l -yds has been having done jigivds- 4 having conquered tasthivds - * having stood babhuvds- 4 having been iusruvds- 4 having heard etc. These stems show an ancient apophony in declension, even though the original terminational accent in the oblique cases which caused it has been given up (gen. sg. tasihusas for

• tasthusds ) .

The adjectival ms-suffix was capable of taking the enlarge- ment -u (cf. vagvariu - above, a parallel extension of -van). - Examples are vibhavasu- * brilliant ’ and sactvasu- 1 powerful J based on the vas- stems which occur in the vocatives noted above. Compare also Pa. viddasu- 1 wise ' for *vi$vasu The specialisation of the adjectival - vas in the formation of these participles has resulted in its being separated completely from the corresponding compound neuter suffix which, as we have seen, usually takes the form -us. Consequently when adjectives are needed from these, it is done simply by adapting the neuters without change of form. As noted above (p. 138) this practice has parallels elsewhere and is old, although rare through the prevalence of the normal system. Examples are cdksus - ' eye ; seeing vdpus- ' marvel ; wondrous tapus -

• heat ; glowing ’ ; without corresponding neuters, nahus-

‘ ifeighbour mdnus- 1 man and with suffixal accent but not the corresponding gradation, vanus- 4 eager jayus - ' victor- ious ' and daksus- 1 flaming There is an adjectival suffix -yds which likewise underwent early specialisation and became totally divorced from the neuter suffix -is. This is used in Sanskrit, and in other languages for making comparative adjectives. Examples are : ndvyas- 4 newer p any as- 4 more wonderful bh-Uyas- 4 more rdbhyas- 4 more violent vdsyas- 4 better sdhyas- 4 more powerful sanyas - 4 older tdvyas- 1 stronger Similarly Avestan has spanyah - ‘ more holy tqiyah - 1 stronger dsyah- 4 swifter etc. This formation is the regular one in Avestan, but in Sanskrit it is much less common than a formation in which the -yas - is added no t directl y to the root, but to thej^opTu^ younger 9 (cf. gen. pi. kani-n-dm) t ndvtyas- 4 newer mrddtyas - 4 softer prdthiyas - 4 broader vdnyas - 4 wider 1 (cf. vdr-i-man I&2 THE FORMATION OF NOUNS ‘ width '), tdnyas - * very swift 1 (cf. tan-sani inf.), rdghiyas- ‘ swifter mdhiyas- ' greater ’, sthdviyas - ‘ stouter 1 (cf. Av. ranjyah mazy ah-, staoyah -). In both accent and apophony all these forms (excepting bhuyas - with weak grade of root) are Or reminiscent more of the^, neut er typ es (cf. vartman vanvas- : u ranyas). The reason for this is not clear, though it must have some significance in the question of their origin. Corresponding forms of the comparative exist in Italic and Celtic : Lat. senior , O. Ir. siniu (IE *s^nyos) ‘ older The e- grade of the root and the a-grade of the suffix indicate an accent identical with that of Sanskrit (sdnyas-). In Greek the same suffix appears in some cases : eAao-craj (*eAax~?/os-a) acc. sg., iXdooovs (*iax-yo(7-€9) nom. pi., but in other cases a -yon- suffix with simitar function appears : eAaomjy, IXdaoovo^, etc. In Greek, as in Sanskrit, a suffixal t may be inserted before the comparative suffix : f}Sia w ‘ sweeter In Germanic a compound suffix -is -on- is utilised : Goth, batiza ' better etc. The weak form of the suffix (-is-) could make adverbs with comparative sense, e.g. Lat. magis ‘ more It also forms the basis of superlatives which are made by the addition of the further suffix -tha (elsewhere -to) : kdnistha- * smallest, young- est ', jdvistha- ‘ most speedy nedistha- * nearest ' (Av. naz- dista-), asistha- ‘ swiftest 1 (Av. dsista -), ojistha- ‘ strongest 1 (Av. aojista-), gdristha- ' heaviest etc. As with the comparative the accent of the superlatives is invariably on the root and the apophony is according. Superlatives of the same formation are found also in Greek (iJStoroff = svadistha - ' sweetest ') and in Ger- manic (Goth .frumists ' first etc.). As with the other neuter suffixes, adjectives and nouns of ad- jectival type can be made by the addition of the accented thematic vowel to the neuter s-suffix. These are of the usual two types, an older type in -sd with reduction of the suffix on account of the following accent, and a more recent type in -asd formed from the neuter as-stems with guna of suffix. Unlike some other suffixes of similar structure, -sd never became very productive or developed independently to any extent. sd ruksd- * shining ' ( rocas - * light '), dyuksd- 1 bright ' (for

• dyutsd - after ruksd- ; cf. also dvdksam for dvdtsam , etc.),

prksa- f nourishment 1 ( pfks - * nourishment a reduced 6-stem which has been adapted as a fem. root noun), ghramsd- * heat of the sun', vatsd- ‘calf' (Gk. Feros 'year'), vrksd - * tree ' THE FORMATION OF NOUNS I&3 (vdlsa- 4 foliage ') , yaksd - 4 supernatural being ' ( yasas - 4 beauty '), s' usd- 4 strong ' ( sdvas - * strength '), paksd - 4 side ' (pdjas- 1 side, surface '), ruksd - 4 rough, dry ' (/ms-), drapsd - 1 banner ' (Av. drafsa - ; cf. drdpi - r mantle '), drapsd - 4 drop ' (for *drabiha- y cf. Gk. rpepaj, etc.), grapsa - and glapsa - 4 bunch ' ( *grabzha- t fgrabh-), saksa - 4 overpowering ’ (saAas- 'strength'), hamsd-

• goose ' (Gk. etc.). In a smaller number of words, chiefly

substantives, the accent appears on the root , dins a- 1 shoulder' (t™* (Lat. umerus with guna of suffix), utsa - 'spring' (Gk. tJSo? nt.), rksd- 4 star ' (arc- r to shine '), ddksa- ' clever ' (cf. dasas- ydti), gftsa- 4 dexterous, able J . - asd : vacasd- 1 eloquent anjasd- 1 straight arnasa - 4 flow- ing tamasd- 1 dark-coloured ', nabhasd- 4 cloudy ”, rajasd- 1 dusty rabhasd- r wild manasd - 4 wise upanasd - 4 being near a w r aggon vetasd- 4 reed camasd - 4 cup pivasd- 1 fat divasd- 4 day prapyasd - 1 swelling (with milk) Thematic extensions of neuter as-stems, such as are common in the case of the suffixes r and n t are rare if they exist at all. It is possible that ydvasa- (masc. and nt.) might be of this origin, but otherwise such neuter nouns in -asa as occur are oxytone, a fact which indicates their adjectival origin: an k asd - 4 flank avasd - 4 refreshment atasd- 1 bush p annas a- 1 abundance Adjectival derivations with vrddhi are ay asd- 4 made of metal vayasd - 4 bird manasd- 4 belonging to the mind ' and so forth. Similar adjectival formations from the is- and us- stems occur : tavisa- 4 strong bharisa - 1 rapacious mahisd- 4 great ; buffalo ' ; parusa- 4 knotty (as reed) ; rough parusd- 4 grey, speckled white ' (Av. pourusa - ; cf. Engl, fallow , etc.), arusa - 4 red ' ; with vrddhi, vapusd- ‘ wonderful The radical accent of ndhusa- 'neighbour', mdnusa - 'man' and vdpusa- 4 won- drous ' is connected with the fact that the corresponding us- stems are themselves used adjectivally without change of accent. Similar accent is found in tdrusa - 4 overcoming ' and pUrusa -, purusa- 4 man ' (related to puru- 4 man ; n. of a tribe ’ as mdnusa - is related to mdnu -), There are a few closely related formations in -Isa and -usa (iju + s + a), rjtsd- epithet of Indra, angusd - 4 hymn 1 and the neuters purls a- 4 rubbish kdrlsa - 4 dry cow-dung ' and pnyUsa- 1 beest milk ' (cf. pipyusi and Gk. nvos <*pyusos)> THE FORMATION OF NOUNS 164 There are a number of miscellaneous stems made by the addition of various suffixes to s-stems. Such are : bhtsd 4 fear mantsd 4 understanding savasi 4 strength ", tdvisi 4 id sarasi - ' lake ' ; plusi 1 flea " (Jplu-), dhdst- 4 abode ", sdnasi- 4 victor- ious dharnasi- 4 strong ", atasi- 4 beggar " ; ddksu-, dhdksu- 4 burning 1 ; bhujisya- 1 free " ; the rare infinitival forms avyathisyai 4 not to tremble “ and rohisyai ; masttska - 4 brain ' ; naristd "joking"; updstha- "lap" (cf. updsi), vanisthu - 'en- -cmC trails " (cf. Germ, wanst ; different suffix in Lat. venter). En- largement with the suffix -ti appears in a fair number of ex- amples : gdbhasti- 4 hand ", palasti - ' grey-haired * (cf. palitd - ' id ’), pulasti- 4 having straight hair ' (cf. pulaka- 4 bristling of the hairs of the body '), Agdsti n. of a rsi. This suffixal combina- tion is well developed in Slavonic, where, however, it makes abstract nouns (O. SL dlugosti 4 length ", etc.). It appears also in Hittite with the same function : dalugaiti 4 length On the other hand it appears occasionally in Latin in adjectival use, as in Sanskrit : agrestis 4 rural ", caelestis 4 celestial ", From is- and us- stems there appear formations of the same kind : ndvisti- 4 hymn of praise ", pdnisti- f admiration ", tatanusti- 4 spreading out §14. The Suffix t The suffix t existed with functions like the above simple suffixes, but as an independent suffix it has become much rarer. It also very rarely provides neuters, since the tendency was from a very early period to incorporate the stems ending in occlusives into the common gender system. Its original function as one of the primary neuter suffixes is seen most clearly when it serves as an extension of the neuter r- and n- stems, e.g. in Skt. sdkrt , ydkrt and in Gk. x € W a > gen. s g* X € W ar °s * winter " (but the corresponding -nt- stem in Hittite, gimmant- 4 winter ", is com- mon gender). Similarly the primitive suffix t on which the suffix -t-ar has been built may be presumed to have been neuter. Apart from this there remain in the various languages a few sporadic instances of a neuter suffix t: Skt. pfsat- "drop", upatdpai- 4 fever ' ; Gk. /xeAi (for *jueAtr) , Hitt, rnilit 4 honey " ; Gk. yd Aa, ydXaKTos, Lat. lac , lactis 4 milk ", Lat. caput 4 head The use of the simple suffix /, in the common gender, to make action nouns and abstracts, is fairly well developed in Hittite, e.g. kartimmiatt - ' anger ", duSkaratt - ' joy They are rare else- where, but where they occur such nouns have normally acquired THE FORMATION OF NOUNS 165 the feminine gender : Skt. ndkt - ' night Lat. nox , Gk. vv£ ; Lat. quies, salus r etc. Gk. Qefj, is-, x^P 1 *- The reason that such formations are so rare is that as a general rule they have been supplanted by extensions of the simple /-suffix, namely by ti in the case of action nouns, and by td in the case of abstracts. In Skt. dasdt ' decade 1 beside dasati- ' id.' we have an example of the unextended and ex- tended form side by side. An isolated avirat- ‘ absence of heroes ' (avtrate dat. sg. RV. 7.1. 19) represents a type that has otherwise been universally supplanted by the compound suffix td. A very small number of feminine nouns in -t appear in Sans- krit : sravat-, vahdt both meaning ‘ stream sascdt * obstacle’, vehdt- ‘ cow which miscarries Of these the last is presumably adjectival and the others could be. A masc. adjectival form appears in vaghdt- 4 worshipper This adjectival suffix appears elsewhere, cf. Lat. pedes 4 footman eques 4 horseman In ndpdt- 4 grandson ’ there is a formation with vrddhied suffix which may be compared with the similar formations analysed above. This stem is also interesting because it retains an ancient type of declensional apophony. A reduced grade appears in Skt. nddbhyas dat. abl. pi., Av. napto abl. sg,, nafsu loc. pi. (with reduction of the three consonant groups). There are other examples of the vrddhied suffix in adjectival forma- tions in other languages : Av. ravas-iardt- ' moving in the open country Gk. dpyrjs ‘ shining ’ ( apyrjr - or apyer- in the oblique cases, cf, the two types Sorrjpa and Troiftiva), ndvrjs ‘ poor yvpiv 77? ' lightly armed soldier’, Lat. aries ' ram ' (gen. -etis), AS. heele ' hero ’ (xalep), They differ from the vrddhied forma- tions involving the other suffixes only in that they take the nominative -s, which is the normal practice with stems in occlusives. Skt. paddti- 4 foot soldier ’ and patti- ' id.’ are both i- extensions of a /-stem ; the two different farms derive from the declensional apophony of the primitive stem. In one special case the suffix / remains a living formative in Sanskrit. It has been noted above (§ 2) that roots ending in the vowels i, u and r cannot, like other roots, function without any addition as nominal stems. Where other roots do so they in- variably add the suffix t : stut - 4 praise ’, samit - ' battle ', niyut- ' team vft- 'army, host rit- 4 stream mil- 4 post hrut-

• treachery ; foe These stems, like the root stems, may be

used either as action nouns (in which case they are feminine) or THE FORMATION OF NOUNS 1 66 agent nouns ; in the latter use they also appear most frequently as the latter members of compounds : devastut- ‘ praising the gods visvajit- ‘ all-conquering jyotiskft~ ' making light etc., etc. Like the root stems they have generalised the weak grade in declension. In this way the t- formations have come to form one system with the root nouns, since they are used in identical circum- stances and with exactly the same function as the root stems in the case of other roots. For this reason the Indian grammar- ians do not class the t which appears here as an ordinary suffix (pratyaya-) , but consider it to be a special addition or augment (i dgama -). The suffix i has acquired this character of augment in a number of other formations, notably in the gerunds in -ya ( °jitya 1 having conquered etc., as opposed to °drsya ' having seen '), and in the adjectival formations in van (kftvan- ‘ -active ' as opposed to ydjvan- ‘ worshipping '). In these cases too it appears to strengthen the roots ending in i, u and r. We shall see below that the same kind of development has taken place, and to a greater degree, with the suffix i. The compound suffix -it appears in a few examples, yosit- ' woman divit- 1 brilliance ' (whence divitmant- * brilliant ') sarit - ‘ river harit- 1 green, yellow rohit- 1 red These are accented on the suffix, but this does not agree with the apo- phony ; cf. the type socis-. The adjectives harit - and rohit-, which must originally have been accented on the root, were at one time stems w f hich could be used indifferently as* nouns or adjectives. Of these the adjectives hdrila- and rohita- are thematic extensions, and they preserve the original accent of the simpler forms, because they belong to that small class of adjectives which do not take the normal adjectival accent owing to the adjectival character of the stems on which they are based (cf. vdpus- t vdpusa- above). A suffix -ut appears in Marut - n. of the storm gods (*mar- ‘ to shine * : cf. mdrici - ‘ ray ' and Gk. fiapfialpco , afiapvGca>) , also in garmut - fem. ‘ a kind of grass and garut°- which is found only in the derivative garutmant ‘ winged The guna grade of this suffix -vat is employed in the middle cases of the perfect participles (vidvddbhyas, etc.) and the nom. acc. sg. nt. (vidvdt). In Greek it is used throughout the masculine (etSdk, ctSoros, etc.). A few adverbs are made with this suffix : sanat ' of old pradaksinit * moving to the right cikitvit * carefully '. SimiTHE FORMATION OF NOUNS 167 larly Av. paityaogzt 4 backwards etc. An adverbial suffix -tas has been abstracted from the ablative singular of Osteins. Originally the type daksinaids 4 from the right ’ was to be analysed daksinat-ds (cf. pradaksinit for the /-suffix in connec- tion with this word). With the growing obsolescence of the t- suffix, formations of this type came to be analysed daksina-tds, etc,, and the -tas thus abstracted became very productive in the formation of adverbs with ablatival meaning : mukhatds 1 from the mouth agratas ‘ in front sarvatas 4 on all sides ' f tdtas 4 from there ', pariias 4 around etc. Adjectives in - td arose in the usual way from the addition of the accented thematic vowel to /-stems ; so, prsatd- 4 speckled ' from pfsat - 1 spot, drop Since /-stems have mostly dis- appeared, such adjectives appear normally as independent formations. Some have the suffix in the form - aid , e.g. darsata- ' visible ’ (Gk. 0 Septet ros), yajatd - 4 to be adored bharatd , 4 (to be maintained), epith. of Agni, n. of a tribe rajatd- ' silver ' (cf. Av. 9r9zata-) t others in the form -td, trstd - 11 rough syetd- 4 white ', anapta - 4 not w T et ' (Av. napta- wet), dutd- 4 messenger ’, sutd- 4 charioteer ndpitd - 4 barber * (for *$napild- t cf. Pa. nahdpita-) ; in other cases it is preceded by some other suffix (i, u, etc.) : tigiid- ' sharp palitd- 4 grey-haired dmanyuta- 4 free from anger ddbhuta - 4 wonderful kapola - 4 bluish-grey ; pigeon Mention has already been made of its addition to the suffixes n and r ( vasanta -, muhurtd -). The colour words Aarita- and rohita- have radical accent due to the original use of thesimpleZ-stemsasadjectives. Radical accent appears in some other examples, dsita - 4 black ita - 4 speckled mdrta - 4 mortal The colour adjectives either substitute an n-suffix in the feminine : ent, syenl , hdrini, or add n with change of / to k : dsiknt , pdlikni. Here there are apparently traces of an old alternation corresponding to that of r- and n- stems. A part from t hese a dject jye s the suffix - id is specia liscdiiiThe als o in the other IE lan guages. There is a reduction both of root and suffix before the final accented d [bhrta-, as opposed to bharatd-) which is characteristic of the most ancient formations. These participles are very numerous and are formed from all roots except a small number which take -nd : srutd- 4 heard ' (Gk. kAvto?, Lat. in-clutus , Ir. cloth), srutd- 4 having flowed ' (Gk. pvros), tatd- 4 stretched (Gk. tch-os, Lat. tentus), hatdi68 THE FORMATION OF NOUNS 'slain' (Av. O. Pers. ]ata- t Gk. faro's*), gatd- 'gone' (Gk. /Jards* ‘ that can be traversed'), niktd- ' washed ' (Gk. avnrros 'unwashed'), usta - 'burnt' (Lat. ustus ), vrttd- ‘turned’ (Lat. vorsus, versus ) , drstd- 1 seen ' (AS. torht ‘ clear'), mista- savourv (of food)<Om’fo-, cf. Lat. mixtus. As elsewhere the specific function of this suffix is not inherent in it from the beginning but acquired by adaptation. The fundamental meaning of kluto - for instance, like that of any other adjectival form of the same type, is no more than r one connected with hearing In Greek the specialisation of mean- ing has not gone so far, since beside a passive sense, an active sense is frequently found : Si/mro's* r possible ' and ‘ able cwvctos f intelligible ' and ' intelligent etc. In Sanskrit an active sense is seen in siitd- ' charioteer ' (' driver i.e. ' one connected with ‘ driving ', from su- } suvati) and in ndpita- 1 barber The reason for the rarity of the simple suffix -t is that it has normally been replaced by compound formations in which a further suffixal element is added to the t. The commonest of these, which makes verbal abstracts is -ti. This i s the com- monest of all the -suffix es making ve rbal Abstracts or action nouns, and words formed with this suffix show less tendency to develop a concrete sense than is the case with other suffixes. T hese words are feminine, in contradistinction to the action nouns formed with the suffixes previously discussed, -which are neuter. In this respect they follow the simple ^-sterns which in most IE languages are feminine, and in Hittite common gender. The only traces of neuter ft’-stems that can be found are the pro- nominal forms kdti ‘ how many tdii ' so many ', ydti * as many ' which are such in form though they function differently. It is clear that those stems were among the earliest to break away from the neuter system proper to action nouns as the gender system developed. The process of the enlargement of Osteins by the addition' of the suffix -i is seen in such pairs as samit-, sdmiti- * assembly 1 ; ndkt-, ndkti- 4 night ' ; dasdt-, daidti- 4 decade The accentuation of these action nouns is subject to no rule. It may appear on the root as in fddhi - ‘ prosperity ', gdti - ' go- ing justi- * satisfaction dhfti - 1 firmness rati- 4 enjoyment v fddhi- 4 growth ', idkti- 4 power iinii- 4 peace or somewhat more frequently on the suffix : uti- 1 helping ksiti - * abode THE FORMATION OF NOUNS 169 juti - * speed dhtti- ' thought *, pusti - 4 prosperity *, bhakti- ‘ sharing *, Mrtf- ‘ maintenance *, rna/t- ‘ thought stuti- 1 praise This lack of rule in the accentuation is characteristic also of the simple - i stems, as will be noted below. In Greek the accent is normally on the root, but since the radical vowel in these formations, in Greek as in Sanskrit, appears always in the weak grade, it cannot have been there unchanged from the be- ginning. Compared with the general system— accent and guna of root for action nouns and vice versa for agent nouns — these formations present a striking anomaly, and it is not now pos- sible to say what particular developments in early Indo- European were responsible for this state of affairs. Examples of this formation are abundant also in Greek and a number of parallel forms can be quoted : dpaciti - ‘ retribu- tion *, Gk, olttot 1015 ' id * ; ksiti - 1 dwelling Av. siti- t Gk. ktiois 4 settlement 1 ; ksiti- ‘ destruction Gk. <f>6iats ; A-huti - ‘ obla- tion Gk. r pouring out * ; sruti - ‘ flowing Gk. pvais ; pluii- ' floating *, Gk. ttXvois ; gait- ' going *, Gk. pdcn?, cf. Goth. ga-qumps ; tati - 4 stretching, row Gk. ramj. In Latin they have been replaced by a still further developed suffix -tion made by adding the vrddhied w-suffix to the -ti. T^hese are feminines because the ti- abstracts on which they are based were feminine: mentio (Skt. maii-) iunctio (Skt.ywAri-,Gk. etc. The gradation -ati appears in a number of examples : amhait- ‘ distress drsati- ‘ appearance *, mithati - 4 conflict vasati- 4 abode *, paksati- 4 root of the wing A few datives of ri-stems are classed as infinitives : istdye 1 to refresh pttaye 4 to drink vttaye ‘ to enjoy sdtdye ‘ to win *, utdye 4 to help There are also a number of ti - stems functioning as agent nouns, mainly in the early language : jndti - ‘relation (: Lith. gentis ‘ id *), patti-, padati- 4 footsoldier addhdti * sage rati- 4 liberal *, dhuti - ‘ shaker *, sdpti- ' steed dhfsti - ‘ bold *, ptiti- ' putrid visit- * eager ' ; dmati - ‘ poor sthapdti 1 governor ; architect vrkdti- 4 robber r dmati- ‘liking to stay*, p-dti - ‘master *. 1 The position of the accent is subject to no rule, as is the case with the action nouns, and the two classes are not distinguished in the usual way. 1 From pa - ' to protect, govern That the i in this word is suffixal is evident from its absence in Gk. & 4 <rncn.va* ’’Therefore pdti~ is to c pd- (nr pa-, etc.) as vrkdti- to vfka -> THE FORMATION OF NOUNS 170 The adjectival formations are occasionally extended by the addition of suffixal n, abhimdtin - 1 insidious 1 ( abhimdti - ‘ id '), ratin ' liberal ' [rati 4 id '). Compare the similar extension in Latin in the action nouns. A suffix -hi is produced in the same way by the addition of u to the simple /-suffix. These are less numerous than the ti- stems, and morphologically less altered from the ancient system. There are for instance still a number of neuters pre- served : data ' division vdstu 4 abode 1 (Gk. (F)aarv ‘ city '), (j& vdstu 4 thing mdstu f sour cream 1 (cf. Gallo-Lat. mesga , Ir. " medg with different suffix). A neuter formation in -tu used adverbially appears in jatu 4 at all, ever These neuters have the regular radical accent, associated with guna, which is the characteristic of neuter action nouns. The same accent and guna appears also in the following masculines (this is the gender w T hich the non-neuter action nouns in -tu normally adopt ; con- trast the feminine ti- stems) : otu- ' weft tdntu- ‘ thread dhatu- ' element sdktu- ‘ groats situ- 4 bund, dam sotu- - c q 1 libation Occasionally suffixal accent appears : gdtu- 1 w r ay ', hetu- (but with guna of root) , pitii- 4 nourishment Guna of the first element of the suffix appears in edhatu - 4 welfare ', vahatu- ' wedding ' and kr-dtu- ' intelligence * (kf- ' to think, com- memorate '), There are very few feminines : vdstu - 4 morning sutu- ' giving birth pvAtu- ' life Agent nouns and adjectives are rare : mdntu- 4 councillor ’ and dhatu- ' suitable for sucking ’ do not have the proper adjec- tival accent, which contrasts with the formations in simple -u. Regular suffixal accent appears in tapyatu - * glowing ' and sisdsatu- ' desirous of obtaining The suffix -tu is a rich source of infinitives. These are regu- larly accented on the root which normally takes guna. They appear in the accusative, dative and genitive. (1) The accusative infinitive in -turn is the^only one used in The c lassica l language : kdrtum 4 to do gantum 1 to go ditum * to give irdiurft' to hear ', ndtum 4 to lead moktum ‘ to release etc., etc. In the Vedic language w r hich is rich in other kinds of infinitives, this formation is exceedingly rare, appearing in some five examples in the Rgveda, and in five others in the Atharvaveda . It is equivalent in form to the Latin supine, factum , itum, etc. (2) Over thirty dative infinitives formed with this suffix are THE FORMATION OF NOUNS X 7 X found in the Vedic language. They are invariably accented on the root which takes guna. (except sutave 1 to bring forth 7 be- side sdvitave) : Have 4 to go kdrtave ' to do ', datave 4 to give mdntave 4 to think ydstave 4 to sacrifice etc. (3) The ablatives occur less frequently : hdntos * from being struck etos 4 from going etc. The formation appears occa- sionally also in a genitive sense, kdrtos 4 doing 7 (with tnadhya), ddtos 4 giving ’ (with ise). In addition to these three types there also occurs occasion- ally in the Veda a type in -tavai : dtavai 4 to go hdntavai c to slay It has the anomaly of a double accent which has not been explained, and it is always followed by the particle -w. It appears to be based on a thematic extension of the gunated tu- suffix {*dtava- nt.) with the old form of the dative singular (Av. -di). The neuter suffix -tvd which must in origin be a thematic for- mation based on the /w-suffix, is frequent, and is used exclusively in the formation of secondary abstract nouns : amrtatvd- ' im- mortality devatvd - 7 divinity sucitvd- 4 purity ', satrutvd - f enmity etc. The accent of these secondary formations has nothing to do with the old system, but comes from the tendency to evolve a special taddhita accent for secondary neuters which has been noticed above. Avestan has such formations, though not abundantly (1 ratuQwa - ' office of rain- etc.), also a number of primary formations with this suffix, vadwa - * herd staodwa- 4 prayer Primary formations with this suffix are represented in Sanskrit by only a few forms and these are confined to Vedic, namely pdtva- 1 ram, wether *, and a series of words in which -pitvd is combined with various prefixes. The commonest of these are prapitvti - * going forth, time of going forth, morn- ing * and abhipitvd- * coining in, coming home in the evening, evening time The further extended suffix -tvand appears in the Veda in the same sense : mahitvand - 7 greatness sakhitvand - 'friendship etc. Though ignored by classical Sanskrit it turns up again in Middle Indo-Aryan (- ttana -), and has been preserved even to the modem period (Hi. -pan). There are some thematic adjectives based on the suffix -iu which are used as gerundives in the Vedic language : kdrtva- 'to be done *, jdtva- 4 to be won vdktva - 1 to be said sdnitva-

• to be gained hdntva - * to be slain Likewise in Avestan, THE FORMATION OF NOUNS

I72 ja6wa- 1 to be slain These are all accented on the root con- trary to the general rule for adjectives. Tho classical type in -tavya [kartdvya-, alternatively kartavya - ‘ to be done etc.) is made from the gunated /w-suffix with addition of the adjectival suffix -ya. In Greek yei another variant (-tewos) occurs,. 7TOl7}T€OS , etC. By yet another enlargement of the -tu suffix the gerunds in - tva are provided which are used with uncompounded verbs. These appear with the root normally in its weakened form, and with the accent on the suffix : istvA f having sacrificed krtvA ' having made gatvA ' having gone tirtvA ‘ having crossed drstva ‘ having seen pltva 4 having drunk sndtva 1 having bathed etc., etc. These formations have at first sight the appearance of being instrumentals of action nouns in -tu. As such their form would be in order, assuming they are ancient forms, since originally the accent was on the termination in the weak cases and this caused reduction of the root. The chief difficulty against such an ex- planation is the co-existence in the Veda of a gerund in -tvi. This formation is actually more frequent in the Rgveda than the gerund in -tv A- : krtvi 4 having done gatvi f having gone V bhutvi c having become etc. The latter formation cannot be explained as an instrumental or any other case ending of a verbal noun in - tu . Since the explanation of both forms must run on parallel lines, it follows that the forms in - tvi are also not case endings. The only explanation possible is that these are the suffixes d and i, and the fact that they function here in the same manner is in accordance with the close relation be- tween them elsewhere. So we must have here two compound suffixes used adverbially with the final accent that usually appears in the adverbial use of nominal stems (prdtdr, etc.). The suffix -td, an extension with the d - suffix of the simple t- suffix, is specialised in the making of abstract nouns from adjectives. As a primary suffix it is very rare, e,g. did 1 layer ", more common in Greek, ycverrf, etc. The usual type is repre- sented in the Veda by such examples as devdtd 4 divinity purusdtd 4 humanity bandhutd 1 relationship vasutd 4 wealthi- ness etc. In classical Sanskrit they are made freely from all adjectival stems: krsnatd "blackness', purnata 4 fullness V dirghatd 4 longness etc. The suffix appears with the same func- tion in other IE languages : Russ, polnota ' fullness * O. SL THE FORMATION OF NOUNS 173 crinota 4 blackness dlugota ' length OHG .fullida 4 fullness Goth, diupipa 4 depth Lat. iuventa ' youth This suffix could be strengthened by the further addition of suffixal t, producing the suffix -tat. Examples (found only in the Rgveda) are upardtat- 4 proximity dev at at- 4 godliness vrkatdt - 4 woliishness, murderousness sarvdtdt - 4 complete- ness The same suffix appears in Avestan ( haurvatdt - 4 whole- ness,’ etc.), Greek (fiapvnjs 4 heaviness etc.), where it com- pletely replaces -id as a secondary suffix, and Latin (civitds, etc.). Just as the simple suffix ~t could be extended by the addition of suffixal -i ( dasdt -, daiati-), so the compound could be ex- tended to -tdti. Examples are jyesthdtdti- ' superiority ’, devdtati - 4 divinity sarvdtdti - r completeness vasutati - 4 wealth', satydtdti- 4 truth sdntdti - 4 good fortune The last two may also be used as adjectives (like certain formations in -ti). § 15. The Suffix m The suffix m plays an important part in nominal derivation in Sanskrit and the other IE languages, It has already ap- peared in the groups -mar, -man, - mant , -min. It is also produc- tive of thematic adjectives in -md t and occasionally appears in other formations. The problem with this suffix is that, in con- tradistinction to all the others, it is hardly ever found as a simple, uncompounded suffix, although the various compound suffixes mentioned must have been formed, like other compound suffixes, by additions to just such a simple suffix. The only surviving examples appear to be the numeral stems saptd and data, Lat. septem , decern , IE *septm (earlier sdptm on account of apophony), ddkrp. Here the suffix appears in its weak form like the n- suffix in nama (Lat. nomen), etc,, and these numerals are of exactly the type as the primary neuter nouns. The frequency of m in various derivative suffixes shows that it must have originally been common as a primary neuter suffix. The reason that it does not appear as such, is that it has been replaced by other formations, and this has come about in two ways. The first process is illustrated by the juxtaposition of yugdm nom, acc. sg. nt. 4 yoke ' and yugmd - 4 paired Like other thematic adjectival suffixes yugmd - is to be analysed yugm-d and the consonantal w-stem on which it is based is none 174 THE FORMATION OF NOUNS other than the neuter yugdm. The m in the neuter yugdm was originally the w-suffix, but owing to similarity 'with the accusa- tive singular of thematic stems (originally all adjectival and of common gender), it came, by an easy process of adaptation, to be treated as a termination, with the consequence that forma- tions of this kind were turned into neuter thematic stems and declined accordingly. The neuter thematic type then became productive, particularly in forming extensions of neuter con- sonantal stems (- ana , -atra, etc., above). The relation that exists between yugdm and yugmd appears also between bhaydm ' fear ' : bhlmd - 1 fearful J and mddhyam nt. middle : madhyam-d- adj. 1 being in the middle The adv. sddam can be explained as a neuter m-stem in view of the deri- vative s adman- ‘ seat ' ; likewise dram , dlam ‘ fittingly, suit- ably ’ from the IE root ar- ‘ to fit J by comparison with deriva- tives like Gk. app6$. Ancient thematic neuters in IE are very rare, Skt. yugdm is shown to be ancient by the correspondence of Gk. t^vyov, Lat. iugum , etc. Another ancient word is Skt. paddm r step Gk. irihov, Hitt, pedan, which may be presumed to have originated in the same way, though direct evidence is lacking in this case. Gk. €pyov * work ' with the same rare and no doubt ancient apophony as ttsSov is to be classed with it. It should be noted that such primitive thematic neuters, which according to this theory are transformed m-s terns, are not only exceedingly rare, but they are the only class which provide certain word equa- tions between different IE languages. Thus the thematic neuters of secondary origin, namely (i) extensions of neuter consonantal stems and (2) the still later though numerous taddhita formations, are of later origin. It would be difficult otherwise to explain the absence of detailed agreement among these formations between the various languages. In this con- nection also we must note the complete absence of the latter two formations in Hittite. The rarity of the thematic neuter formations of the type paddm , yugdm shows that, although this was one way by which the primitive m-stems were removed, and although it provides the only plausible explanation for the existence of neuter thematic stems at all, this was not the only, or indeed the main way by which this was done. The other process which operated and which accounted for many was the extension of m-stems THE FORMATION OF NOUNS 175 by the addition of the r- and n- suffixes. Just as in Hittite the primary 5-stems were almost entirely replaced by formations of the type hannesarj-esnas , so over the whole of Indo-European with the exception of a few stems that were turned into a new type, the thematic neuters, the w-stems were replaced by the mar j man stems. Later in most languages the mar -stems were ousted by the man- stems. In Hittite the gen. sg. of neuter stems in -mar is in -mas (arnummar 4 bringing ', gen. sg. arnummas). It is usually said that there has been assimilation of mn in such genitives, but this cannot be proved, and the alternative theory is possible that such genitives are formed from the unextended m-stem. Here we may compare certain Vedic instrumentals, draghmd , rasmd , (draghmdn- 4 length ', rasmdn- 4 rein ') which in the same way may well be formed from the simple m-stem and not from the extended stem in -man. Thematic derivatives from man-stems frequently appear without the final element of this suffix : priyddhdma- r fond of home visvd karma- 1 accomplishing all ' containing the stems dhdman- and kdrman- ; cf. also dharma- masc. ‘ law ' : dhdr- tnan- nt. 4 id 1 ; 6ma- 4 way, course 1 : eman - r id ' ; tdkma- 4 shoot, offspring ' : tokman- ( id ' ; ydksma- f disease ' : yaks- man- 4 id ' ; darmd- 4 destroyer ' : darman- 4 id In these cases also there is no evidence to support the theory of a change mn to m f and the thematic type illustrated by these examples must have been based on the simple m- stem. The suffix -md makes a fair number of adjectives and nouns of adjectival origin : ajmd- 4 career, march 1 (Gk. 6y^o$ ; man- stem in rijman-, Lat. agmen), idhmd - 4 fuel ', grtsmd- 4 summer * ( gr-ts-md -, cf. guru- 4 heavy *) y gharmd- 1 heat ' (originally adjec- tival, cf. Gk. depfjios, Lat. formus 4 hot'), jihmd - 4 athwart, oblique tigmd- ‘ sharp ' (Gk. vnypos ‘ puncture ', n-stem in oriyjjia, nt.), dasmd- 4 wonderful ’ (dasmdnt- 4 id '), dhumd- 4 smoke ' (cf. mar- stem in dhumra -), narmd- 4 sport, pastime ' u (also ndrman- nt. 4 id f ), yudhmd - 4 fighter rukmd- 4 ornament vdmd- 4 charming ' ( J~van-) t sagma- 4 beneficial sydmd- 1 black ' (sydvd-, with alternative suffix), himd- 1 cold, frost *. The adjec- tive lima- 4 helper ' has anomalously the accent on the root, but the weak grade shows that this is not original. There are a number of thematic action nouns with the normal radical accent and masculine gender, presumably to avoid the repeti- tion of tn : dharma - 4 law bhdma - 1 illumination ksima- 4 reTHE FORMATION OF NOUNS ’I76 sidence ; security sdrma - ' flowing soma - f the pressed out juice of the soma plant ' (Av. haoma -), stoma - hymn of praise homa- ' offering The customary alternation of accent be- tween substantive and adjective appears in srdma- ‘ lameness ' and srdmd- ‘ lame The gradation -amd appears in dasamd - ' tenth ' (i.e. dasam-a-, Lat, decimus ; ddsa, decern ) p saptamd 'seventh' (after which astanid- 4 eighth madhyamd- 4 middlemost adhamd - ‘ lowest ' (Lat. infimus), paramd- ' furthest, highest From the ordinal saptamd- there is extracted tama which is applied to other numerals ( vimsatitama etc.). This is identical with the super- lative suffix - tama which presumably arose in the same way : dntama - * nearest, most intimate ' (Av. ant9ma-, Lat, intimus ), uttamd- ‘ highest ugrdtama- 4 strongest midhustama- ' most gracious mat Hama- ' most motherly rainadhdtama- ' most wealth-giving etc., etc. A compound suffix - ima appears rarely : agritna- ' foremost and in combination with other suffixes, krtrima - ' artificial ', bhidelima- ‘ fragile etc. Other combinations of suffixal m are -mi : (masc.) urmi- 'wave' (Av. varzmi-, AS. wielm ), rasmi- f rein, ray ’ ; (fem.) bhumi- ' earth ' ; -mi, (fern.) laksmi ' mark, sign surmi 1 tube J ; -md : him a ' cold season J , ksuma 4 flax §16. The Suffixes i and u These two suffixes, like the other suffixes, had in Indo- European a gunated form, ei, ei$, and a weak form, i, u , depend- ing on the position of the accent. They could also be vrddhied under the same circumstances as the other suffixes, and in general they develop on the same lines. The neuters were the most primitive type, and on the basis of these, adjectives could be formed by accenting the suffix, old neuters may be trans- ferred to the masc, -fem. class, a process which is very common in the case of these suffixes ; there are neuter thematic ex- tensions, and adjectives are formed by adding the accented thematic vowel. They combine with other suffixes in the usual way ; of these combinations the suffixes -ira, - in , -ina, - ri , -ni, - 1 var , - vara , -ura, -van, - vana , -una, -vant } - ru , -nu, -is, -us, - yas , - vas , - isa , -usa , -it, - ut , -vat, - ti , - tu , -ima, -mi have already been mentioned. Neuter nouns in -i are rare ; it is a type that early tended to become extinct, largely by the transference of such stems to the THE FORMATION OF NOUNS 177 masc.-fem. class. The few that remain are mainly defective. Of these sdtni 4 work ' is indeclinable ; dsthi 1 bone dksi 1 eye sdkthi l ' thigh 7 and dadhi 1 curd ' substitute an n-stem in declen- sion (gen. sg. asihnds , aksnds , sakthnds , dadhnds) ; vdri> ' water ' adds n to the stem in the same circumstances (gen, sg. varinas , cf. the same thing with neuter w-stems, and occasionally elsewhere, e.g. sir as } sirsnas) ; hirdi 1 heart 7 substitutes the root stem (gen, sg. hr das } etc.). There is one rare neuter with the com- pound suffix -vi, srkvi which inflects like vari. The neuters have not preserved an inflection in which the i- suffix is preserved throughout, in contradistinction to the neuter w-stems where such exists (mddhu, mddhvas). It can be traced however in the declension of dvi- ‘ sheep gen. sg. dvyas . This type of inflection contrasts strongly with the adjectival inflection (agnds, etc. ; in the end this becomes the only in- flection), and when the similar opposition in the w-stems is con- sidered (mddhvas ; sunds) it becomes clear that this first type of declension is that proper to the neuter nouns, as opposed to the adjectives. The difference is due to the different accentuation of the two types of stem, corresponding to that which we have found elsewhere. The declension of dvi- indicates that it was originally a neuter ; it has been transferred, on account of natural gender, but has retained some of its old features in de- clension. Greek has the same inflection in this word, and retains the more ancient alternating accent : oiy, otoy. There are occasional neuter /-stems which have come to be used as adverbs : sdci ' with prdti ' against 7 (pr-ati, Astern) . Often original neuter /-stems have been replaced by various extensions : ndbhya- ' nave ' retains the old neuter gender of the /-stem which in ndbhi- ' navel J has been replaced by the feminine gender. An old neuter *mddhi 1 middle * (in Av. maihydirya- * belonging to the middle of the year * for *madhi - ydrya-) is replaced by mddhya - nt. (originally madhyam -, m- stem, cf, above) ; likewise krdvi ‘ raw flesh 7 (dkravihasta-) is enlarged to kravya - (kravy&d-, cf. Lith. kraujas), and alterna- tively by addition of the s-suffix to kravis Similarly the neuters arch - ' flame J , rods - ' light 7 and socts - 4 flame 7 are s-extensions of old neuters in -/, but the /-stems arch masc., sod- fern., ruch, rod- fem. have been transferred from the neuter to masculine or feminine gender. This and similar evidence makes it quite dear that neuter I78 THE FORMATION OF NOUNS action nouns in -i were originally common. Further evidence comes from another quafter. At a time when these formations were readily made, the whole category of s-neuters was incor- porated in the verbal conjugation, and served to form passive aorists of the third person singular. As such they are well re- presented in Indo- Aryan and Iranian : tdri, jdni , darsi , pAdi, sddi ; with augment dkdri, etc. Similarly Av. srdvi ‘ is heard etc. The accent is on the root as regularly in the case of neuter stems. Neuter w-stems remain more common than neuter t-stems. They are regularly accented on the root and not uncommonly have vrddhi : mddhu ‘ honey ' (Gk. pid v t AS. medu), vdsu 1 property Ayu- ' life ' (RV. 1,89,9 an d 3>49> 2 * restored from evidence of metre) , jinu~ ‘ knee ' (Gk. yow, Lat. genu, Hitt, genu), diru ‘ wood ' (Gk. Sopu, Hitt, taru) sanu ‘ summit, top ' (also masc.), danu ‘ moisture 7 (also fem.), pdsu ‘ domestic animal ' (RV. 3. 53, 23, Goth, faihu , Lat. pecu ; elsewhere transferred to the masculine and accented on the suffix), imdsru 1 beard asm ‘tear' (Toch. A. akru-nt pi.), jatu 'gum' (AS. cwidu ), jdtru ' collar-bone * (also jatru- masc.), tdlu 1 palate trap a 4 tin ’, ydiu ' sexual embrace snayu ‘ sinew titail ‘ sieve The u ~ stem could serve throughout as the basis of inflection. This type of neuter inflection is seen in mddhu ' honey gen. sg. mddhvas and vdsu r wealth ' gen. sq. vdsvas. The same type appears in Gk. yowos, Sovpos, with the more ancient termina- tional accent in the gen. sg. which Sanskrit has abandoned for the fixed radical accent as in other kinds of neuters (gen. sg. ndmnas , vdcasas, etc.). The old terminational accent is seen in gen. sg. pa&vds but it has come to be associated with an altered type of stem. The old IE neuter pdku (Lat. pecu , etc.) is pre- served in one instance in the Rgveda (see above) as pdsu , show- ing that the change is of recent origin. The change of gender in the usual fiasu- is due to natural gender. The change of accent which has also occurred is due to the old terminational accent of pa£vds gen. sg. f etc. Since the normal neuters have adopted the fixed radical accent, the only stems in which there is commonly alternation of accent from stem to suffix in declension are suffixally accented masculines with reduced grade of suffix in the weak cases : nom. sg. pita , dat. sg. pitre , nom. sg. Pusa f gen. sg. Pusnds , etc. On this analogy it is easy to see how a suffixally accented paivas (after the style of the old neuters) still preTHE FORMATION OF NOUNS 179 served at the time of the change of gender should bring about the suffixal accentuation of the new nom. sg. pasus. In the same way we may judge pitu- ‘ food gen. sg. pitvds, which from being an action noun and from having the neuter type in- flection, may also be regarded as a transferred neuter. The stem krdtu - ' intelligence old neuter for the same reasons, has on the other hand established radical accentuation throughout. Alternatively neuter w-stems operate in the weakest cases with an extended suffix -un-(ci. vdrinas gen. sg.) : gen. sg. mddkunas , vasunas , drunas ; Loc. sg. dyuni (cf. Gk. a ZFev, loc. without ending), sdnuni ; gen. loc. du. jdnunos (cf. Toch. A. kanwem * knees * with strong form of same suffix appearing in dual). Greek, when extending the suffix in these cases, has the extra ^-suffix as with other neuter w-stems : gen. sg. yovvaros, Souparo?. In addition the Veda has some forms from these stems in which the adjectival type of inflection is employed : dros, snos , mddhos. This is not surprising since the adjectival type of in- flection has become universal in i- and w-stems by the classical period, and the process was already far advanced in the earliest period, only a few of the old type being left. Beside the neuter i- and w-stems there is a fair number of action nouns in -i and -u which appear as masculines or feminines. The action nouns in -i are normally feminine : jalpi - ' mutter- ing rdmhi - * speed rdji- ( direction ’ (Lat. regio with addi- tional n-suffix ; cf. the same feature in connection with the {{-stems, above), ropi- ' pain dhrdji - ' impulse, force nabhi - ' navel 1 ; tvisi- 4 splendour ruci- ‘ light * ; with suffixal accent, n sani- r winning f , dji- ‘ contest krsi- ‘ ploughing citi- ‘ under- standing *, nrti- ‘ dancing bhuji- ‘ benefiting The few datives of {-stems classified as infinitives have likewise suffixal accent , dr say e ' to see yudhdye 4 to light etc. Masculines are rare: arci- * ray ', dhvani- * noise rayi- 1 wealth With the {-stems as with the {{-stems the old rules about accent and apophony have long been superseded. All possible types are represented (jalpi sani - , tvisi-, krsi-) and the varia- tions are of no grammatical significance. Obviously rdji-

• direction which has a connection in Lat. regio represents the

oldest type. In rayi-, gen. sg. rdyas (for rani-, rally as, earlier ram-, ranyds) we may recognise the same accent development as has taken place in pasu, pasvtis , The differences in apophony l80 THE FORMATION OF NOUNS between dhvani- 4 noise ' and dhuni- 4 roaring ’ are those that exist between action and agent nouns, but the accent position is reversed. Obviously very complicated changes, which cannot now be followed, have affected the ^’-sterns to produce this com- plete lack of system. This is associated with the fact that in this type of stem the earliest mass transfers from the neuter of action nouns took place. On the other hand the non-adjectival masculine and feminine -stems have as a rule accent and guna of root : masc. dsu- 4 life ’ (as- ' to be ' : Av. ayhu-), svdru- 4 stake, post mdnu- 4 man bdndhu- 1 relation ' ; fem. jdsa- 1 exhaustion dhdnu- 4 sandbank pdrsu- ‘ rib ', sdru- 4 dart hdnu ‘ jaw ' (Gk. yews, Toch. A, sanwem du.). The words amsu - ‘ filament, ray ' (Av. asu- 4 shoot '), renu- 1 dust ’ and sanku . - ' peg, stake ' have final accent, but at the same time guna of root. The accent of hdhu- 4 arm J disagrees with that of Gk. 7 rfjxvs ; the apophony shows the Greek to be original. The weak grade of the root' in tsu- masc. and fem. ' arrow r ' and sindhu - masc. and fem. ' river ' is ex- ceptional; they are probably of adjectival origin with the shift of accent seen frequently in nominalised adjectives. There is a small number of neuters in -ya to be classified as thematic extensions of i-stems : mddhya- 4 middle 1 (see above), kravya 0 4 raw flesh ", ndbhya- ‘ nave ajya- 4 clarified butter drdvya- ‘ substance rdjya- 4 kingdom ' (Ir. rige). The forma- tion is not uncommon in tatpurusa compounds of the type havirddya- 4 eating the oblation hotrvurya - ' choosing a sacri- ficial priest brahmabhuya - ' becoming Brahma In sakhyd- 4 friendship ' as opposed to sdkhi - the normal accentual relation is reversed because sakhyd - is a secondary neuter, just as in the case of hotrd- nt. : hotar - masc, (see above, p, 137). In hr day a- 4 heart ’ the thematic extension is added to the gunated suffix. The neuter suffix -ya originating as a simple extension of -i has developed independently and on a very w T ide scale in the formation of secondary neuters, either with vrddhi, as usually in the later language, saubkdgya - ' welfare etc., or in some cases without, duty a- 4 embassy etc. As with the other neuter suffixes adjectives and nouns of adjectival type could be made from neuter i and u stems by shifting the accent to the suffix. For instance we have, with the oldest type of apophony, the IE neuter pAu (Goth, filu 4 much ') and with accentuation of suffix and reduction of root THE FORMATION OF NOUNS l8l an adjectival stem pleu - which appears in Gk, ttAcc? (for

• 7 TeF-€s)> The system has undergone various modifications,

and several different types of declension have emerged as a result. (1) The first, and oldest type, is represented by sdkhi -, nom. sg. sdkhd 4 friend This is characterised by vrddhi of the nom. sg., and in declension it follows the general lines of the declen- sion of the agent nouns in r and n (nom. acc. dat. sg. sdkhd t sdkhayam , sdkhye ; svasd, svdsaram , svasre). This rare type is found also in apt aid RV. 8, 32, 16 (nd sdmo apratd pape 4 soma is not drunk which gives not a return '), and in Av. kavd nom. sg. (kavi- title of kings, Skt. kavi 4 wise man '). In Sanskrit this type has usually been replaced by that in which the nom. and acc, sg. terminate in -is and -im t but considerable traces remain to show that the formation was originally more widely used. The feminine derivatives Agndyt and Mandvi are based on the old vrddhied nominative singulars *Agnd(y) and *Maniv. An old vrddhied nom. sg. forms the first member of the com- pounds Agnavisnil (cf. mdtdpitarau) and kavasakhd - ' he who enjoys the friendship of the wise Above all the usual form of the loc. sg. of i- and u- stems ( agna t later agnau , sunau , etc.) can only be explained as a locative without ending equi- valent to the old vrddhied nom. sg, of adjectival stems. (2) The few w-stems which have retained vrddhi in the nom, sg. add the nominative s to this. The type is represented in Sanskrit nom. sg. dyaus ' sky, sky-god * (Gk. Zev? ; the mean- ing f sky-god ’ is the original one, and this accounts for the adjectival form of the word) and in gaits 4 cow It is somewhat more frequent in Iranian, examples being O. Pers. dahyaui f land ' and Av. hiOdus 4 companion ' and uzbdzauS 4 with up- lifted arms In Avestan a tendency to differentiate adjectival and nominal declension is seen in the contrast between uzbdzaus and bazuS 4 arm * in the nom. sg. and between darzgd-ar&aem and frddat-fHaom as opposed to arstim 4 spear ' and pasutn 4 domestic animal ' in the acc. sg. In Greek this type has become productive, and a distinction has been developed between agent nouns of the type /JacaAeu? 4 king ' yovcvs 4 parent ' (: Skt. janu-s - nt.) and the adjectives of the type The Vedic nom, sg. vis 4 bird ' is also of this type, but the guna instead of vrddhi is unusual and unexplained. The corre- sponding Latin word avis conforms to the usual type of f-stems. I 82 THE FORMATION OF NOUNS Gk. wov 'egg' *(<dwyom ) is a thematic extension of an old neuter *6wi ‘ egg ' (with vrddhi of root as frequently in i- and u - stems). Skt. ves is therefore a formation of adjectival type (' one connected with eggs, egg-bearer ') and the suffixal accent which is proper to adjectives has brought about the reduction and disappearance of the radical vowel. ( 3 } The usual adjectival type forms the nom. and acc. singular in -is, -im, -us, -urn. Adjectives in accented -u are very common and frequently have corresponding forms in other IE languages : trsu- 'thirsty' (Goth, paursus 'dry'), ripu- 4 treacherous, enemy J (Lith. lip us 1 sticky, slimy '), prthu - 4 broad ' (Av. pndQu- 4 id *, Gk. TrXarvs ' flat Lith. plains 4 broad '), raghu laghu- ‘ swift, light 1 (Gk. iXaxvs) , guru- 1 heavy 1 (Gk. j 8apvs Goth, kaurus), bahu - 4 much ' (Gk. naxv$ ' thick '), puru pulu - 4 many ' (Gk. ttoAuT ; the corresponding neuter in Goth, fiht), nru - 4 broad ' (Av. vaunt-, Gk. <vpv$), tan u- 4 thin ' (Gk. raw °) , dsn- ' swift ’ (Av. dsu Gk. wkvs ), svddu - ‘ sweet ' (Gk. rjSvs), amhu - 4 narrow ' (Goth, aggwus, 0. SI. azu-ku), rju- *' straight * (Av. vrdzu-), krdhu - 4 shortened, mutilated ' jdyu- 4 victorious ddru - 4 destroying nrtu- 4 dancer piyu- 4 spiteful valgu - 4 handsome vidhu- 4 solitary sayu- 4 lying sddhu - 4 good The accent in gurus nom. sg., etc., is what is expected for adjectives, but it is in complete contradiction to the weak grade of the suffix which appears in the nom. acc. sg. It follows clearly that such a form of adjective cannot have remained un- altered from the beginning. We have seen that an older type is represented in a few archaic examples and that evidence exists that this was once more widely spread. The forms in -us, - urn may therefore be regarded as substitutes for this older type, but they are very ancient substitutes because they occur not only in the languages represented in the examples quoted above, but also in Hittite : nom. sg. 4 good ", parkus 4 high ', etc. The strong form of suffix which should go with the accent, but which has been r eplaced in the nom. acc. sg. appears in the case of these stems in the dat. and gen. sg. Here the adjectival declension (agues, guros) is differentiated from what remains of the old neuter declension ( mddhvas ) in a way that accords with the position of the accent in the two types. The reduction of the termination of the gen. sg. to -s in accordance with the old THE FORMATION OF NOUNS 183 rules of apophony shows that this form is very ancient. The same guna is found in Goth, sunaus, Lith. silnaus and, without reduction of the termination, in Gk, r}8e os- and Hitt, assawas. The distinction between neuter declension and adjectival de- clension appears elsewhere : Hitt, genuwas as opposed to asHawas , etc. The extension of the gunated suffix to the dat, and gen, sg, distinguishes these stems from the adjectival n- and r-stems. The latter are not normally distinguished in form in these cases from the neuter stems (yaknds, uksnas), because the adjectival stems retain the terminational accent. But outside Sanskrit there is some evidence that even these stems shared to some extent the tendency, notably in the case of Av. pitarg gen. sg. which is parallel in every way to agnds and sunos. On the other hand we have seen that sdkhd , dat. sdkhye follows the type of the adjectival r~ and n-stems, showing that two alternative types of inflection existed for adjectival stems. One became pre- dominant in r- and w-stems, the other in i- and u-stems, but to a certain extent both are represented in each of the two classes of stem. The root in these adjectives has usually the weak form in accordance with the accentuation. In the case of su- 1 good ' as opposed to Hitt, assus, Gk. ivs the suffixal accent has resulted in the complete reduction of the radical vowel. Since the word has ceased to be used except in composition, there is no evidence as to its earlier inflection, but in Av. haosravah- the gunated suffix which is in accordance with the adjectival accent is repre- sented. The prefix ku- ' bad ' appears to be of the same nature. The usual association of this with the interrogative pronoun can hardly be justified, so we may take it to be a w-adjective, similar in form and function to su. There are a few instances where corresponding neuters exist beside adjectives in -u : ayu - ' length of life ' (formed with simple ^-suffix and vrddhi from the IE root ai- ‘ to apportion, give i.e. 1 one's allotted span ') : ayu- 1 alive, mortal ' ; Gk. ttwv 1 herd, flock ' : Skt. pdyu - ' protector The plural jatrdvas 1 cartilages of the collar bones ' differs in meaning from the radically accented neuter jatru- ' collar-bone ' and is therefore likely to be an adjectival form. Earlier it may be presumed that such doublets were more regularly met with, and the interaction of the two types will account for radical guna or vrddhi appear184 THE FORMATION OF NOUNS ing in the suffixally accented adjectives. It is clear that the strong form of the radical element in dyu- derives from the neuter dyu where it is in accordance with the rules of apophony, and it may be assumed that similar mutual influence of the two contrasting types accounts for the form of such adjectives as dsu- 4 swift etc. The Hitt. nt. assu is used as a noun in the sense of ‘ goods, property 1 (like Skt. vdsu), and as such it must be regarded as continuing the primary neuter on which the adjective was built (dsu : (e)seu~). From this source comes the guna in Hitt, asms, Gk. Ivs as opposed to the reduction of the root in Skt. su-. No doubt also the early prevalence of the form of the nom. acc. sg* in -us, -urn, as opposed to the forms with. strengthened suffix, was assisted by the coexistence of such neuters. Obviously an easy alternative for distinguishing the two types existed in the simple addition of the terminations -m to the neuter stem in these cases. As a general rule this was done in conjunction with the retention of the adjectival accent. On the other hand radical accent appears in dnu- ‘ fine, small mddhu - 4 sweet r and vdsu- 4 good The two latter stems occur more abundantly as neuter nouns, and their adaptation as adjectives with the minimum change of form appears to be comparatively late. Adjectives and nouns of adjectival origin terminating in the 2 -suffix, and inflecting after the same pattern as the w-adjec- tives above, are fairly numerous : agni- 4 fire 9 (Lat. ignis, etc,), dpi 4 friend '* (Gk. tJ™? 4 friendly, kind thematic), kapt- 4 monkey 1 (originally adj. of colour, cf, kapild-), kavi- ‘wise man ’ (older inflection in Avestan, cf. above), kiri- 4 singer kridi - 4 playing jdmi- 4 consanguineous, * closely related' (cf. Lat, gemi-nus 4 twin '), nodi- 4 roarer svart - 4 noisy etc, ; with radical accent Hi- ' seer kari- 4 singing hymns gfbhi - 4 containing ", muni- 4 sage suci- 4 shining, pure hdri- 4 green tUrvi- 4 overcoming ', plusi - 4 flea ' (cf. Alb. pl’eSt, Arm. lu t etc. ; J'plu 4 hop, jump '). As with the action nouns in -i there is complete absence of rule in the matter of accent and apophony. For this reason it is not always possible to decide to which class a noun originally belongs, e.g. asi- 4 sword ’ (Lat. ensis) , gin- 4 mountain ' (Av. Hlwc gairi -) t dhi- 4 snake ' (Gk. ex L s> mani- 4 jewel etc. The original system, however, has left its mark in declension (agnds as opposed to dvyas ; cf. Hitt. Sallaias : halkiai), in the same THE FORMATION OF NOUNS 185 way as in the ^-declension. As with the «-stems the adjectival declension has spread at the expense of the nominal, and to an even greater extent. As with the suffixes previously dealt with, thematic adjectives could be made on the basis of i- and w-stems, and this served as an alternative to the type with accented suffix. Thus Lat. socius ‘ ally ' bears the same relation to Skt. sdkhi nom, sg. sakhd ' friend ' (cf. the old neuter stem preserved as adverb, sdci), as Hitt, vestaras to Av. vdstar- or Gk. larpos to lanjp. The suffix -ya is very frequent and at an early period in Indo- European it developed widely as an independent §uffix, so that the connection with f-stems has for the most part ceased to exist. The way the suffix originated is made clear by such examples in Sanskrit as urmyci - ‘ undulating sfnya- * furnished with a sickle kavyd - ' wise dvya- ' coming from a sheep bhumyd- ' terrestrial ’, yonyd- ‘ forming a receptacle aryd-

• kind, devoted, pious ', which have corresponding f-s terns

(urmi-, sfni- t kavi- t dvi- t bhumi yoni- t art -). The suffix, originating in this way, became widespread at an early period producing adjectives meaning ‘ belonging to , connected with In the case of thematic stems the suffix is substituted for the thematic suffix. In the case of a word like ajrya - ‘ belonging to the plain ' (Gk. aypios ‘ wild ’) the derivative could have been formed on the old neuter r-stem (ager) from which djra- and Gk. dypos are themselves derived. It is in some such way that this type of substitution must have been evolved. In the case of the derivative -ya the accent in Sanskrit may appear either on the suffix or on the root : (a) ngrya agriyd- ‘ foremost divyd- ' heavenly ’ (cf. Gk. Scos), satyd - ' true grdmyd - ' of the village somyd - f relating to Soma rdjyd - ‘ regal ' (Lat. regius), ; (b) dngya - ' of the limbs gdvya- ' bovine ', nary a- 1 manly jdmbhya- ' an incisor tooth ' (Gk. yopbpios), ksamya - ‘ terrestrial ' (Gk. x^wlos), pitrya - 1 paternal ’ (Gk. TTarpios, Lat. patrius), surya - ' sun 7 (Gk. iJe'Aios, t^Aios), vdnya- 1 belonging to the forest etc. This suffix, associated with accented root, is commonly used to produce adjectives from verbal roots which function as gerundives : guhya 4 to be hidden tdya - 4 to be worshipped yodhya- 4 to be fought havya- * to be invoked jdyya-, jdya- ' to be conquered vacya - 4 to be spoken etc. Vrddhied derivatives appear abundantly from the earliest i86 THE FORMATION OK NOUNS period : ddityd - ‘ descendant of Aditi graivyd - ‘ relating to the neck prdjdpatya- ( relating to Prajapati ', p&ncajarvya- ‘ relat- ing to the five peoples etc. In the Sanskrit suffix -ya two suffixes have been confounded. These are distinguishable in the Veda by means of the metre which shows that there is a monosyllabic - ya appearing in such words as kavyd- ' wise J , dvya- ‘ belonging to a sheep vdnya-

• of the forest ', etc., and a disyllabic suffix -iya appearing in

ddmiya- ' belonging to the house rdthiya- * relating to a chariot jdniya- 4 relating to the people visiya - * belonging to the community udaniya- 1 watery etc. Of these -ya is the suffix discussed above, and it was formed on the basis of the suffix -i by the addition of the thematic vowel. On the other hand the suffix -iya represents earlier -in a and it was formed on the basis of the suffix These formations are therefore dealt with in the next section in connection with that suffix. A small number of thematic formations have the s-suffix in the guna grade. Such are : sasaya- ‘ abundant, frequent ' (cf. s a sty as- ‘ -more numerous ’ and sdsvant -} , susmaya - ' strength- ening gavayd - 4 Bos gavaeus sdnaya -, sandy a- ‘ old tdnaya -

• offspring, son kupaya - 4 seething ', ddsataya- * tenfold '

(: dasati- 'decade'); also a couple of double formations hit any dy a- ‘ golden gavydya- 4 bovine The same type of suffix is found in other IE languages : Gk. xpv&eos 4 golden ", Lat. aureus , igneus , etc. Though the adjectival suffix -va is not uncommon, it never received anything like the extension of - ya , and its connection with the M-suffix, or other derivatives from it, remains in most cases evident. Examples with final accentuation are rsvd- ' lofty ' (Av. ?r?sva - ' upright, exalted urvd- 4 enclosure ' (vf- 4 to surround, cover J , originally adj. ' enclosed '), hrasvd *

• short urdhvd- ' erect ' (Av. zrzSwa-, Lat. arduus), takvd-

' speedy ' (beside tdku-), yahva - * young, youngest ' (Av. yezivt fern. ; beside yahu- ' id '), rkvd- 4 singing hymns ' (beside fkvan fkvant 1 id ’), dhruvd- ‘ firm 1 (Av. drva 0. Pers. duruva- ' sound, healthy '), tnalva- 4 foolish iydva- 4 dark ' (Av. sydva- ; beside sydmd- with w-suffix), ranvd - ' joyful, enjoyable ' (be- side rdnvan-) , jtvd- f alive, living being J (O. SI. zivu, Lith. g$vas, Lat. vivus, Osc. tews nom. pi.), pakvd- ‘cooked, ripe" (adapted as past participle), sarva- n. of a deity (Av. saourva lit. * one armed with a sdru- t 4 dart '), sruvd- ‘ ladle sikvd 1 skilful ' (beTHE FORMATION OF NOUNS 187 side Hkvan- ' id "), kharvd- * mutilated ' (Av. kaurva- f id "). In these, the most numerous examples, the original' adjectival accent is preserved. A small number have radical accent : dsva- f horse ' (Lat. equus, Goth. atha°- f etc. ; cf. din- ' swift "), sdrva- 1 all * (Av. haurva Gk. oAos, ouAos-, etc.), ptirva - ‘ former * ( 0 . SI. pr&vu) , ptva- ' fat " (beside plvan - * id "), The stem kulva- ‘ bald ' (Lat. calvus) appears only in the cpd. atikulva-. The M-suffix appears with guna in arnavd - * waving ; flood, ocean ", keiavd- ‘ long-haired ’ (cf. Av. gaesav - * hair "), pdvirava- ' having a metallic share" (pdviru -) , vidhdvd * widow 1 (Lat. vidua , etc. ; cf. vidhu- * solitary '). In some cases -va has acquired the character of a secondary suffix : anjivd- ‘ slippery ", santivd- * beneficial ", sraddhivd- J credible rajtva- ( striped, streaked ; blue lotus ", saciva -

• companion, minister

Stems in -u are occasionally enlarged by the addition of the suffix -i : ghfsvi- 1 lively, joyful ' beside ghfsu - f id " ; other sy- stems are dhruvi- ' firm ", ddrvi - fem. * ladle jAgrvi - ' wakeful didhrvi - ‘ sustaining ", didtvi- f shining ", susisvi- * swelling or growing well (in the womb) ", cf. Hsu- ‘ child This addition appears also in other IE languages, notably in Latin, where all the old adjectival w-stems are supplanted by w-stems : levis ' light ‘ (Skt. laghu -), mollis ' soft ' (Skt. mrdu ), brevis, gravis, etc. Conversely -u is added to the suffix A producing the com- pound suffix yu : ydjyu * pious ", sundhyu- 1 pure ", sdhyu- ‘ strong ", manyu- masc. 'anger", mrtyu masc. "death" (Av. m$rddyu-) t ddsyu- ‘ barbarian ", bhujyu- ‘ wealthy ". In Sanskrit this has developed chiefly as a secondary suffix : duvasyu- 4 worshipping ", udanyu- ‘ irrigating J , adhvaryu - ' a kind of priest It has come to be specially connected with the de- nominatiye verbal base, devayu- ' pious * : devaydti ' he is pious ", etc., and it tends to acquire a desiderative meaning, vasuyu- 1 desiring wealth ", etc. The formation is very produc- tive in the Veda, but disappears almost completely in the later language. §17. The Suffix i as Union-Vowel It was noticed above that the suffix t acquired under certain circumstances the character of an augment or special insertion between root and suffix (kr-t-ya- ‘ to be done ", etc.). The suffix i also functions in this way in Sanskrit on an extensive scale. i88 THE FORMATION OF NOUNS The use of i as a connecting link between root and suffix or be- tween root and termination is particularly common in the verbal formation. The Indian grammarians call the i so used it, and according to their terminology the forms which take this i are called set (with it) and those that do not are called unit (without it). The two types of conjugation may be illustrated by the following examples : I. Present, root-class, isise, isidkve, Hire ; future, patisydti, bhavisyati, vardhisyati ; aorist, -is, arocisatn, djivisam ; desider- ative, jijtvisami ; perfect, i pi. oct. bubudhitnd , teninid, 3 pL mid. bubudhird, teniri ; pass. part, sankitd -, lajjita - ; gerund. patitvd , yacitvd ; infin. vdrdhUutn, ydcitum. II, Present, root-class, dtsi, vitsi , sese ; future, vaksyati , chetsyati ; aorist, -s, acchaitsam , ddrdksam ; desiderative, didrksati , vivitsati ; perfect 1 pi, act. cakrma , jagrbhmd , 3 pL mid. yuyujre, vividre ; pass. part, krtd-, drstd- ; gerund, chittva, krtvd ; infin. kdrtum, dr d$ turn. This i. is also found in nominal derivatives other than the participial and infinitival forms illustrated above. In the agent nouns in - tar it appears mostly in agreement with the forms of the infinitive and gerund: yacitdr vardhitar etc., as opposed to kartdr-, drastdr-, etc. Examples in the case of other suffixes are: prathitndn- * breadth khamtra - r shovel rocisnu - 4 shining vdrivas- 1 wide space dtithi- 1 guest ' (as opposed to Av. asti-) t dravitnu - r running ", etc. It is not possible to formulate any simple general rule govern- ing the presence or absence of this i in the verbal conjugation and elsewhere, but a general tendency is observed to use the set-forms where inconvenient consonant groups would result ( paptima , etc.). The use of -i- is more predominant in the later language than in the Veda, For instance the Vedic language has both -re and -ire in the 3 pi. mid. of the perfect, but the classical language knows only -ire ; in later Sanskrit the stems in -isya account for three-quarters of the futures, while in the earlier language the larger proportion (five-ninths) are still formed with simple - sya This continues a process which had been going on in pre-Vedic times. Old Iranian, close as it is to San- skrit, shows very few formations of this nature, which makes it clear that in the main the great extension of the use of -i- in the verbal conjugation is a special development of Indo-Aryan. Its adoption on such a large scale is clearly connected with THE FORMATION OF NOUNS 189 characteristic tendency of Indo- Aryan, observable from the very beginning, to get rid of consonant groups. To begin with the i in verbal and nominal derivation must, in such cases where it existed, have had etymological justifica- tion, that is to say it must have been the suffix i compounded with other suffixes in the usual way. A number of such suffixes containing i have already been enumerated, -is, -isa, - ira , etc., formed on the basis of the i-suffix in the same way as -us, -usa, - ura , etc., are formed on the basis of the w-suffix. The suffixes forming the verbal stem are the same as those of the corre- sponding nominal stems. For instance the stem of the - is aorist drocisam appears also in the noun rods - * light Since rods - ' light ' is an extension of the simpler z-stem in rod”, ruci-, the verbal stem also is originally built up from tMe i-stem. Similarly in the future the suffix syd is formed from the 5-stem with the addition of the denominative yd, (the stems, vaksyd-ti, etc., differ only in apophony from denominatives of the type namasydti). In the same way the future in -isya was to begin with based on the is-stem. As regards the form of the stem bhavisyd-, etc., are exactly parallel to the denominatives in - usyd ( urusyati , vanusydti), and in the Rgveda a number of such formations are clearly to be classed as denominatives. Such are avisydti, sanisyati , which have beside them nominal forms (avisya, avisyu sanisyu-) which are commonly associated with the denominative but never with the future. Those participles in -ita where the i may be considered to be original have incorporated an i-suffix which appears elsewhere in the inflection of the root. This is the case for instance in sitd 4 hound ' ( sd - 1 to bind ') where the suffix i is so frequently associated with the root that the simple root (aor, asdt) has be- come comparatively rare ; e.g. s-ydti 1 binds s-d-tave * to bind sisdya ‘ bound etc. The - ita of the participles from other roots in -d [sitd- ' sharp ditd- ( bound sthitd ‘ stood etc.) origin- ates in the same way. The participle of causative and tenth class verbs (gamita- : gamdyati , etc.) includes the suffix which is used in the formation of the present, and comparative evidence shows that this practice is old (cf. Goth, gatarhips : gatarhjan ; wasips: wasjan , etc.). The same connection is seen between participles in -ita and presents of the fourth class (kupitd : kupyati , cf. Lat, cupio , cupitus) or presents in athematic i (stanitd- : stanihi , cf. further stanayitnu-, tanyatu -, O. SI. stenjg , THE FORMATION OF NOUNS I90 etc., for the prevalence of the i-suffix in connection with this root). A similar incorporation of suffixal u is seen in ddbhuta- and dnatidbkuta- from the root dabh- (pres, dabh-n-6-ti). In the same way we may account for -itar beside -tar in the agent nouns: marditar - ‘ forgiver cf. mrddyati , mrdayaku-, mrdtkd- (Av. nmdzdika -), panitdr- 4 praiser cf. pandyati , pandydti , panista , aor., panayayya , pdnisti -) ; likewise vardhitdr-i vardhdya coditdr- : coddya -, etc. The process is illustrated by a similar development occasionally in connection with the u- suffix. The agent nouns tarutdr - ' conqueror dhanutar- ' run- ning swiftly ' and sdnutar- 4 winning ' incorporate the w -suffix which appears in the present tense ( tarute , tUrvati ; dhdnvati ; sanoti) and elsewhere {tdrusa- etc.). From these instances it is clear that the i in a fair number of verbal forms and nominal derivatives was etymologically justi- fied, and the analysis in such cases does not differ from that of any other forms containing compound suffixes. What Indo- Aryan has done is, on the basis of a modest number of such forms to extend the use of i in the verbal derivation on a vast scale. The analogical i which then comes to be so abundantly used, ceases to be subject to the usual analysis and acquires the character of a union vowel or euphonic augment. The i which appears in the alternative form of certain termin- ations (papt-imd beside cakr-md, etc.) was also in origin suffixal. There are some non-thematic presents in -i (svdpiti, svdsiti, dniti, jaksiti, etc.) with parallels elsewhere (Lat. capid , capis , capit, etc.). They are defective, and since they no longer form a complete present system, they have been attached to the root class, the i being treated as union vowel. There are also some scattered preterite forms lajayit, bddhithds , atarima, avadiran , asth-ithds, asthiran , etc.) which have become attached to the is- aorist. This type of conjugation is based on the t-stem, just as Vedic tarute is based on a w-stem , so that in origin the i here is not different from suffixal i elsewhere. But it came to have the character of an addition to the termination and this enabled this type of termination to be transferred to the perfect. The perfect stem is based only on the root, so the i in the perfect terminations can only be accounted for as a borrowing from the present and preterite forms of the above type where its presence is etymologically justified ( bubudhimd , dadimd ; bubudhird , dadhire, etc,, after atarima , iiire, etc.). THE FORMATION OF NOUNS 191 §18. The Suffixes d, t, u The suffix -a has two functions* On the one hand it forms the feminine of adjectives in -a ( bdla - masc. ' boy bald fem.

• girl '), and on the other hand it appears as an independent

suffix of derivation, producing action nouns, abstract nouns, and the like. In having the two functions, adjectival and non- adjectival, it resembles the suffixes previously discussed, but it differs from them in that its adjectival use is confined to the formation of feminines. This was the result of specialisation since the feminine gender was not present in the earliest Indo-European, and as a result of this development of adjectival -d, the action nouns too, which to begin with did not differ from the ordinary neuters {d, i.e. -#H, like - as , - ar , etc.), have acquired the feminine gender. The following are examples of action nouns and the like ter- minating in the suffix -a in Sanskrit : krida ' play day a * pity nindi f blame-, sarikd * doubt himsd '-injury ksamd ' pat- ience bhasd ' speech sevd ‘ service sprhd ‘ desire sdkhd ' branch siphd ' whip 7 (cf. Av. sif- ‘ to strike with a whip ’), did ' direction ', ukhti ' pot ulkd ‘ meteor vend ‘ longing ksapi ' night 7 , chdya f shadow 7 (cf. Gk. cr/ad), jar a ‘ old age 7 , dosi f evening ra$d ‘ moisture 7 (0. SI. rosa , Lith. rasa ' dew 7 ). There is no consistent rule about the accentuation of such words, but final accentuation is the commoner ; in the corre- sponding forms in Greek (yoinj, etc.) final accentuation is the rule. The old general rule about the distribution of accent has obviously long ceased to have any relevance to this class. The suffix -d is added to the other primary suffixes in the usual manner, producing a series of compound suffixes : -nd : tfsnd 1 thirst drna ' wool 1 (Av. varznd, Lith. vilna , etc^) ; -and : jar and ‘ old age arhand ' worth 7 , kapand ' cater- pillar 7 (cf. Gk. KafiTTTj ' id 7 ) ; -rd f surd ' strong liquor 7 , tdmisra 1 darkness 7 ; -sd, mantsd ' thought, wisdom * ; -vd, jihva ' tongue ' (Av. hizvd ), grtva * neck ' (O. SL griva ' mane f ), dmivd ' disease 7 , apvd f a kind of disease durva ' a kind of grass 7 ; -yd.jdya ' wife mdyi 1 magical, supernatural power ’, itya * course vidyd * knowledge ", kriya ' action 7 , samajyd ' assembly etc. ; -td, krsnatd ' blackness etc., etc. The suffix -T likewise is predominantly used for making feminines of adjectives, preferably from consonantal stems. In THE FORMATION OF NOUNS I92 addition there are a small number of primary action nouns and the like. Such are : (a) Dm-declension : idci ‘ power sdmi ' holy work *, tavisx ' strength dsandi ' stool (b) Fr&f-declension : nadt 1 river dehi- ' wall, embank- ment ", start- ' barren cow ' (Gk, crreipa), surmi ' tube ", srni - 1 sickle palalt- ‘ straw sphigt- * hip nandi- ‘ joy athari - ' flame oni- ' breast ksoni - ' flood khdrt- ‘ a measure *, cakri - * wheel tandri - ‘ sloth, lassitude tapani - ' heat ", laksmt- ' mark, auspicious mark, good fortune tari- * raft tantri - ' string The suffix -u is used for making feminines to stems in -u (alternatively -vl is used, or the unaltered stem functions as feminine), e.g. land- * thin ’, phalgu - ' reddish babhrd- 'brown', corresponding to masculines tanu-, phalgu-, babhru Tt also forms a small number of independent nouns, camd - 'dish', tanu- * body vadhd- ' bride kadru- r soma-vessel jatu- ' bat ', dhanu - ' sandbank nabhanu - ' well svasru - ' mother- in-law 1 (0. SI. svekry, La t. socrus ). Sanskrit -d in this suffix resolves itself ultimately in Indo- European into the thematic vowel -f H. Similarly % and u re- present -i - h and -u-h. In this form the parallelism of these three suffixes to the three s-suffixes -as, -is, -us, becomes clear. The simple suffix is -an, which like neuter -as appears with guna, and in the compound suffixes -in and -tin, old neuter i- and u- stems are extended by suffixal -h, in the same way as they are extended by $ in the compound suffixes -is and -us. Thus sdmi fern, is an extension of sdnu nt. (indecl.) (cf. the relation of ar el- and arcis-, etc.) and dhanu fem. ' sandbank ’ along with dhdnvan- nt. and dhanus- nt. are alternative extensions of dhanu- ' id ' (fem,, originally nt.) The two compound suffixes sometimes appear with guna (vatya 'whirlwind jihva ' tongue'), as is alternatively the case with other compound neuter suffixes, vdrivas, etc. The close similarity in function between the h- and s- suffixes in making primary action nouns is seen from the frequent instances in which the two are found side by side in words of the same meaning : tdnd, tanas ' offspring ', jar a, jar as- ' old age tanu , tdnus- ' body dhand-, dhanus, etc. Like the other primary neuter suffixes this -a has no specific meaning and stems in -d frequently occur side by side with root nouns, the extension adding nothing to the meaning: ksdp THE FORMATION OF NOUNS 193 ksapa 1 night tan-, tana ‘ offspring dis-, disci * direction etc. The action nouns in -d, %, u, are formed in precisely the same way as the neuter action nouns made with other suffixes, but differ from them in being feminine in gender. Since the feminine gender is a comparatively late development in Indo-European, it is to be expected that these stems were originally neuter. Direct evidence of this is provided by the neuter plurals of the type yuga. This -a is the same as the a-suffix, which was used (like some of the other neuter suffixes) as a collective, and eventually as a plural. In this use the suffix still retains the indifference to the distinction between nominative and accusa- tive which characterises neuter stems. The reason for the feminine gender of these action nouns is that these suffixes in their adjectival use became specialised as feminines, and the action nouns on account of similarity of form eventually fol- lowed suit. We have seen that the normal accent of neuter action nouns was on the root. Little trace of the general system remains in the formation of these stems. The accent of the a-s terns is variable, showing the same complete absence of rule which was observed in the i- stems. The nouns in i and u have a regular accent; on the final. This accent is in complete contradiction to the general rule, but the apophony, with w T eak suffix, and usually guna of root (tanti-, debt) shows that this is not origin- ally ; only radical accentuation will account for such forms, IE tenun, etc, A parallel shift of accent was observed in the neuter stems in -is, havis -, etc. There is a tendency with the other neuter suffixes for the suffix n either to replace or to be added to the other suffixes. There are some traces of that system here. Corresponding to kanyd * girl Av. kainyd , there is in Avestan a genitive singular 7 kainino which is related to it in the same way as Skt. Ursnds to > t Ur as. In Sanskrit this form of the stem appears in the Vedicgeni- / ^ tive plural kanindm. The sm - is normal in jhe genitive plu ral and j the agreement between Sanskrit and Germanic (OHG, gebono, ON. rundno), shows that its presence in d-stems goes back to Indo-European. It appears that this -n- is the heteroclitic n- suffix, which has been generalised in the genitive plural, but abandoned elsewhere, though Av. kainino shows that it could originally appear in other cases. — THE FORMATION OF NOUNS 194 Stems in ~d are used to form a number of adverbs, in the same way as is done with the neuter suffixes above. Such are ; sddd ' always purd f formerly dvitd ‘ doubly so mrsd ' falsely ’, saca ‘ with *, devdtd ' among the gods sasvdrta ' secretly In this class are the absolutives in -tva and -tvi (see above, P* I 7 1 )- In their adjectival function these suffixes are principally used to make the feminine stem of adjectives, etc, : navd ' new de vi ' goddess madhu ‘ sweet etc. This is the result of specialisation in the later Indo-European period. Originally, it must be assumed, adjectival -a, -i, -u were on a par with other suffixes used adjectivally, indifferent to gender, and having the usual relation to the corresponding action nouns. Traces of the more general use of a- stems as agent nouns survive in a number of languages which still have some masculine agent nouns in -a : Lat. scriba, agricola , nauta , Gk. vavrys, rroXlnqs, O. Sh sluga 1 servant *, etc. Stems of this kind have totally disappeared from Indo-Iranian, but there remain in the Vedic language a number of masculine stems in -i, of which the commonest is rathi- ' charioteer 7 as well as two rare and obscure masculines in -u ( praiu krkaddsti-). These are the remnants of an older system in which adject iv es_a nd agent n oun^ of. J^rpe could be made with these^ suffixes in the same way .aa.with other suffixes^ Some evidence that there originally existed a formal distinc- tion between the action nouns and adjectives of this class is pro- vided by the existence of tv/o types of declension of I-stems in the Vedic language. One type is fepresented in the declension of ratht nom, sg. rathis , gen. sg. rathiyas and the other in that of devi, nom. sg. devt, gen. sg. devyds. This distinction corre- sponds exactly to that between the two types of i- and ^-sterns (gen. sg. dvyas, mddhvas : agnds, bahos). In the case of the i- and w-s terns there was evidence enough to show that one type was the declension proper to the neuter action nouns and that the other was the special adjectival declension. It is likely therefore that the same is the cause of the different declensions of the f-s terns, The bulk of the feminines formed from adjectives and agent nouns [devt 1 goddess prthvi ‘ broad adati ' eating*, jagmusl ‘having gone ndviyasT ‘newer*, avitri ‘helper*, dhenumdti * possessing cows dmavatl ‘ strong *, samrdjni 1 sovereign *, rtavan ‘ pious * and dpatihgm * not slaying her THE FORMATION OF NOUNS 195 husband illustrate the various types) inflect in what according to this theory is the adjectival declension. The weak stem in the nom. acc. sg. is parallel to that of the i- and w -stems. The origin of the variation is not clear, but as with the i- and u- stems, it appears that a more original form of the adjectival stem is preserved in the gen. sg., etc. Stems of the vrkt type comprise both action nouns and agent nouns together with some miscellaneous feminines. The de- clension of this type ( rathi gen. sg. rathiyas) and its accentu- ation is closely analagous to that of the adjectives in -in (bait, balinas ) . In the case of the latter type it was observed that the generalisation of the weak stem (from gen. sg., etc., which were originally suffixally accented) was secondary, and that compara- tive evidence indicated original vrddhied nominatives of the usual type. The same type of generalisation of the weak stem is likely to have happened in the case of rathi , vrki, etc., and in Avestan some traces of an older type of declension are preserved. The Avestan word for tongue is declined as follows : nom. sg. hizva, acc. sg. hizvam , instr. sg. hizvd, gen. sg. hizvd, loc. sg. hizvd , instr. pi. hizubis, This is clearly the same type of declen- sion as is found in data } dat. sg. ddtre, sdkhd, dat. sg. sdkhye , uksan gen. sg. nksnds, etc., with weak form of suffix in the oblique cases. As remarked above there are two forms of adjec- tival declension, one with the weak cases having the same form as the neuters (ddtri, etc.), and another with strong form of suffix and weak form of the gen. sg. termination in these cases (Av. pitars, Skt. agues, etc.). Both types are found among t-stems [sdkhye : agndye) and among the f-stems they are repre- sented by vrki- (gen. sg. vrkiyas) and devi (gen. sg. devyds) re- spectively. The oldest type of inflection is that which appears in the vdju stem hizvd fu in Avestan. Leaving aside the hetero- clitic -n- the same type of inflection appears in Av. kainya , gen. sg. kainind and in Skt. kanyd, gen. pi. kanindm. Elsewhere the weak stem is generalised as in ball, balinas . Alternatively the strong form w^as generalised in which case there was a transfer to the d- declension : jihvd, jihvayds . The adjective mahd 1 great ' remains in Sanskrit the only non-feminine adjectival d- stem, and it is defective. Apart from compounds where it remains in use in classical Sanskrit, it appears only in the Veda in the acc. sg. masc. mahdnt. The gen. sg., etc., appear as tnahas, etc. The other IE languages show g T96 THE FORMATION OF NOUNS in this root (Gk. luiyas) and the h ( <gh) of Sanskrit is due to a combination and the h which originally belonged to the suffix. The original genitive was therefore meg-H-Ss with terminational accent and weakening of the suffix of the adjective (cf. uksnds , etc.). This is the only place outside the thematic and other ? derivatives to be mentioned below, where the a, i.e. an of the suffix appears in its weak form. Elsewhere the strong form is generalised in both action nouns and adjectives, between which no formal differences exist. In the neuter sg. of this adjective a stem mdhi with an extra suffix -misused (meg-H-i, cf. Hi tt.mekki-). The thematic vowel and other vocalic suffixes could be added to the suffix -an (d) and the latter, being unaccented was weakened to h. This h remains in Sanskrit in the form of the aspiration of a preceding occlusive. Thus caturtha- ‘ fourth 7 ^ may be explained as *caturtd ( <°an) 4 fourness, group of four ' + adjectival -a, i.e. 'one connected with four, fourness', Similarly ratha- 4 chariot ’ is formed by the addition of the thematic suffix to *rotati> Lat. rota ‘ wheel Originally an adjective ' wheeled 1 it has had the accent thrown back on to the root, in common with many other nominalised adjectives [vfka- * wolf etc.). The compound suffix iha out of -t-w-a is not uncommon : driha - nt. ‘ object, aim varutha - nt. 4 protec- tion ' ; yajdtha- 4 worshipping vaksdtha- m. 4 growth sapdtha - m. 'curse', sacatha- m. 'companionship', sravdtha- m. ' flowing ucatha- nt. ' utterance viddtha - nt. ' worship ' ; with final accent, gdthd- m. ' song bhrthd- ‘ offering ' ; (neut.) ukthd - ' utterance tirihd- ' ford ’, yuthd- 4 herd rikthd- 4 inheri- tance Most of these forms seem to belong to the class of themat- ically extended action nouns formed in exactly the same way as ndksatra pdtatra -, vddhatra-, etc., above. Those finally accented are presumably adjectival in origin, e.g. rikthd - ' that which is left The th in the suffixes - tki (dtitki- 4 guest etc.) and -thu (vepdthu- ' quivering etc.) is in the same way a combination of the suffixes t and h ; cf. -tri, - tru . Other examples of aspirates concerning this suffix are : sdkhi - ' friend ' from sac - ' to associate (simple aK>d suffix in sacd ' with } ) ; makha- ' happy, exulting ’ from a *makan to be compared with Gk. pVap (old neuter adapted as adjective) with variant r-suffix ; nakhd- c nail where ultimately both k and h are suffixal since other languages have a variant g (Lat. unguis , etc.), satykha- 4 shell Gk. Koyxos (adjectival accent. THE FORMATION OF NOUNS 197 original meaning ' curved, coiled saphd- 4 hoof ' ( 4 striker ^ cf. Gk. K07TT0J and SL kopyto), sindhu- 4 river ' for sind-n-u-, of. syand- 4 to flow 1 ; sddhis- 4 seat, abode 1 for sad-a-is- : sad - * to sit * ; aghd - 4 wicked 1 perhaps from an *dgan side by side with agas } Gk. ayos 4 sin § 19 . Other Suffixes Of the remaining suffixes the most important is the suffix -ha. In other IE languages the element k may appear as a non- thematic suffix : Lat. senex 4 old man Gk. ixelpa^ * lad, lass This state of affairs has disappeared in Sanskrit, which has corresponding to these forms the thematic stems sanakd- 4 old ’ and maty aka- 4 young man The suffix rarely appears in primary formations in Sanskrit ; such cases are iuska- 4 dry J (Av. huska-) t sldka - 4 call ; fame ; verse ' (: sru- 4 to hear ’) and dtka- 4 garment ' (Av. ahka -). Normally it is added after vocalic suffixes, in particular after the thematic vowel. The latter type is rare in other IE languages but in Sanskrit it is exceedingly frequent. Examples are durakd - 4 distant vamraka - 4 ant arbhakd - 4 small kumdraka - 4 little boy putrakd - 4 little son It is often simply an extension which adds nothing to the mean- ing, but also it has in some cases a diminutive sense seen in the last two examples. More rarely it is used to form adjectives from nouns : dntaka - 4 making an end rupaka - 4 having an assumed form '. Other vocalic suffixes are likewise extended : aviki 4 sheep 1 (O. SI. ovica) i mrdlkd - nt. 4 grace ", dhdnuka 4 cow ghdtuka - 4 killing jatukd 4 bat It is more rarely added to consonantal suffixes : amyaskd- 4 thinner masttska- 4 brain ", viksinatka- 4 destroying The feminine of the com* bination -ika is extended beyond its original field ( avika , etc.) and it comes to function as the feminine to masculines in -aka : kumdraka - 4 little boy kumdrika 4 little girl etc. Suffixal k is followed by i and therefore palatalised in mdrfci - 4 ray of light J and svitici- 4 bright It is followed by -u in a few formations like pfddku - 4 water-snake * and yuviku- 4 be- longing to you two The corresponding suffix in the cetttara-languages is k to which in Sanskrit s is the most frequent corresponding sound. But in the case of this suffix such s appears only rarely, e.g. yuvasd - 4 young * (Lat. iuvencus), r omasa- 4 hairy # , babhruid - 4 brown kapisa- 4 tawmy 198 THE FORMATION OF NOUNS The voiced guttural, palatalised to j appears in the following athematic formations : dhrsdj - ' bold sandj- 4 old bhisaj- 4 physician ' (cf, Av. °bis t bisaz-) srdj - 4 garland ’ (cf . pratisara- and Lat. sero), trsnaj- 4 thirsty asvapnaj- 4 not sleeping 1 ; (with weak form of suffix) usij- 4 a kind of priest ' (Av. usig-) y vanij- 4 merchant bhurij- 4 shears sphtj- 4 hip ’ (: cf. sphyd - 1 flat ladle 1 ). In the nominal forms trsnaj - and asvapnaj - we have the same sufiixaL combination as is used in the formation of the seventh class of verbs. A thematic -ga appears rarely, e.g. in sfnga- nt. f horn ’ (cf, %**** Lat. cornu , etc.), vdmsaga- 4 bull ' (of uncertain etymology), n WV c patag a > patanga- 4 bird Sufhxal d appears in a small number of athematic formations : drsdd - £ stone * (cf. Gk. Set pds), darad - ‘ cliff, ravine ' (themat. darada- 4 Dard ’), sardd- 4 autumn ' (cf. Av. sardd- 4 year '), bhasdd - * posterior, rump vandd- 4 desire kakud- 4 summit ' *p (cf. kaku-bh - with different suffix). It appears compounded with n in sadandi- 4 permanent 1 (cf. the -nd- of the Lat. gerundive). The suffix ghjh is likewise extremely rare. It occurs in dirgha 4 long ’ (cf. Gk, SoAt^off, Hitt, dalugas ; note that it is preceded by three different suffixes in the three languages, Hitt. u } Gk. i, Skt. h, i.e. dl-n-gho-). The root appears without this suffix in O. SI. dlina 4 length Other cases are vardhd- 4 boar ’ (; Av. vardza- ; the root in Lat. verres, etc.), sardgh- 4 bee and com- pounded with r it appears in sfghrd- 4 swift ’ (: itbham adv. f quickly ’ with different suffix). Sufhxal dh appears in a number of combinations. The suffix -dhra appears in vdrdhra- 4 thong 1 (: varatrd 4 id ') ; it is common in certain other IE languages. The combination - dhya - appears in infinitives like gdmadhyai 4 to go', bhdradhyai 4 to bear', sahadhyai 4 to overcome * and the like (some thirty-five in- stances). A suffix -pa appears in a few rare instances like yUpa - 4 sacri- ficial post ' (yu- 4 to attach, join cf. yutkd- 4 herd ' for u) and stupa- ‘ top-knot ’ (cf. °stu and stuka 4 id '). Non-thematic -bh appears in kakubh- 1 summit * (cf. kaku-d- above), and thematic -bha in a small number of nouns, mostly animal names : vrsabhd- and rsabhd - f bull gardabhd- 4 don- key rdsabhd - 1 id salabhd - f locust sthulabhd- 4 big The specialisation of this suffix in animal names is known also THE FORMATION OF NOUNS I99 in other IE languages : Gk. eXapos ' stag O. SI. golgbi 4 dove etc. This completes the list of Sanskrit and IE suffixes. As will be seen all IE consonants were capable of being used as elements in the suffixal system. §20. Vrddhi in Derivation In its use of vrddhi Sanskrit has developed a system of deriva- tion which was totally unknown in the parent IE language. It is used in connection with a number of suffixes, which may also function without being associated with vrddhi, and as such have already been treated. Its use is a speciality of the secondary as opposed to the primary derivation. In the earlier language alternative forms without vrddhi occur. Thus the suffix -a (normally accented) which makes adjectives on the basis of the old neuter suffixes, may be associated with vrddhi, e.g. tnanusa - 1 connected with men ; man ' beside mdnusa - 4 man and vapusd - beside vdpusa- 4 beautiful ' from vapus-. This type of formation may be illustrated by a few examples classified according to the various suffixes. Suffix -a : angirasd - 4 descended from Angiras maruta- 4 re- lating to the Maruts tnanava - 4 human ; man jattra - 4 vic- torious ausijd- 4 connected with the priests called ustj-, mdrdtkd - nt. 4 mercy Suffix -ya : daivya- 4 divine saumyd - 4 relating to Soma vayavyd- 4 belonging to the wind # ; neuter abstracts, palitya- 4 greyness drtvijya - 4 priestly office Suffix -t : Agniveii- 4 a descendant of Agnivesa- Pauru - kutsi - 4 a descendant of Purukutsa This type is almost en- tirely confined to patronymics. Suffix -ay ana : patronymics , ' Kdnvdy ana-, Ddksayana etc. Suffix -tya : pdrvaiiya- 4 of the mountains This type is practically confined to the later language. Suffix -ka ; maniakd - 4 mine avasyaka- 4 necessary v&san- tika- 4 of the spring time dharmika - 4 religious etc. ; this type is mainly confined to the later language and few Vedic examples are quoted. Suffix -eya : arseyd- 4 descendant of a sage Janairuteyd - 4 son of Janasruti vast eya- 4 of the bladder 4 {vasti-) t pauru- seya- 4 relating to man 4 , etc. Derivational vrddhi with other suffixes is exceedingly rare ; 200 THE FORMATION OF NOUNS such cases are agnldhra- 4 belonging to the fire-kindler ' ( agnidh -) and dsvina- 4 a day's journey on horseback The exact details of the development of this use of vrddhi are somewhat obscure, but the material enables a number of observations about it to be made. The process began in the late Indo-Iranian period, developed rapidly in the pre-Vedic period of Indo-Aryan and continued to extend during the historical development of Sanskrit. The complete absence of any such formations in IE languages outside Indo-Iranian, makes it clear that it was a development confined to those languages, and the fact that in early Iranian examples of this vrddhi are exceed- ingly rare, shows that this type of formation was only in its beginning at the time of the separation of Indo-Aryan and Iranian. Iranian examples are O. Pers. mdrgava- 4 inhabitant of Margiana J , from margu- 4 Margiana Av. dhuiri - 4 belonging to Ahura- *, dhuirya- 4 son of a prince from ahura mdzdayasni- 1 belonging to the Mazdayasnian religion xsldvaenya- 4 de- scendant of Xstavi. The three suffixes -a, -i and -ya which occur in connection with vrddhi in these Iranian examples are the ones that most frequently occur in this connection in Sanskrit, and the suffix of the patronymic xsldvaenya is related to the -i dyana which makes patronymics in Sanskrit. The point of departure of this derivational vrddhi must be sought in the old IE phonetic vrddhi, which, as instanced from time to time above, is found sporadically in the radical syllable of nouns of primary derivation. Instances quoted are rajan - ' king bhdrman- 4 burden ', visas- 4 garment darn 4 wood sanu- 4 top grdhi - 4 seizure and the like. Examining some of the older cases of derivational vrddhi it can be seen that mdnavd- 4 man J , kdvyd- 4 one having the qualities of a kavi-, wise man dry a- 4 Aryan ', and nadyd - 1 born from a river ' and similar forms which from the point of view of Sanskrit show the vrddhi of' secondary derivation, are not essentially different in form from words like ddnava - * demon ', bhdvyd - 4 -which is to be ", vdrya- 4 to be chosen * and ddyd- 4 to be eaten ' which are either classified as primary formations^ or (in the case of ddnava etc.) are derived from primary formations with phonetic vrddhi. Likewise Skt. vdsard- 4 matutinal ; day J , would appear to have derivational vrddhi (cf. va$ar°), but elsewhere in Indo- European a long vowel is seen in the primary neuter (Gk. elap <*wesf, etc.). It may be assumed that primary formations THE FORMATION OF NOUNS 201 with vrddhi of the type diru-, nibht -, were originally more general, and that forms of the type *mdnu- existed in paradig- matic alternation with mdnu-, etc. ; and that when the guna grade, mdnu-, etc., was eventually generalised in the majority of the simple formations, there remained a class of thematic deri- vatives formed on the basis of obsolete vrddhied stems, i.e. mdnava- beside mdnu-. This nucleus showing the alternation guna in the primary and vrddhi in the secondary derivative, would then be the starting point of the system in which vrddhi came to be widely used in the formation of secondary deriva- tives. This means that derivational vrddhi must have appeared first in the case of the vowel d, and that the use of ai and au in these formations must be due to analogy, for the reason that ai and au do not occur in primary nominal derivation. There is evi- dence that this was the case, since in Iranian, which reflects this process in its early beginnings, d is the only vrddhied vowel which appears in this type of derivative. Forms such as Oraetaona- n. pr. (cf. Skt. traitdnd-) and haomanayha- ‘ kindness ’ (cf. Skt. saumanasd-) which have been quoted as parallels to the Sanskrit ai and au vrddhi, contain not vrddhi but guna. Since the words from which they are derived had by nature a guna as well as a weak grade ( trai-jtri - ' 3 ’ ; ( a)sau-/(a)su - 4 good ') there is no reason why the above formations should not be straight derivatives from this (like tretd ' triad '). It is in accordance with secondary origin of the ai and au vrddhi that cases of derivational vrddhi with these vowels are much rarer in the Rgveda than those containing d. On the whole the system has been fully built up by the time of the Rgveda. The main types are in existence, but examples do not occur with the same frequency as in the later language . There exist also a greater number of formations of the older type, in which the secondary suffixes are added without vrddhi, e.g. vdpusa- beside vdpusa- and visya- beside vaisya -. In the later period the popularity of the vrddhied forms rapidly in- creases, and it develops into one of the most characteristic features of the Sanskrit language. § 21 . Grammatical Gender In common with the other IE languages all Sanskrit nouns are classified according to the three genders. Masculine, Fern202 THE FORMATION OF NOUNS inine, and Neuter. This classification corresponds only partly to the natural order of things, that is to say in so far as the nouns or adjectives apply to human beings and to certain of the larger animals. For the rest of the language the choice of gender is arbitrary and without any logical foundation. In spite of this the system has proved remarkably tenacious in the majority of IE languages ; in the modern Indo-Aryan languages, for in- stance, where traces of the old IE grammatical system have been reduced to a minimum, the system of grammatical gender remains in operation. Languages such as English or Persian which have abolished the distinction remain a minority even now among the descendants of Indo-European. A study 'of the evidence provided by the comparison of the IE languages particularly of those which are recorded at an early period, enables some insight to be gained into the origin of this system. This is because grammatical gender was, at the period of Indo-European which can be reached by comparison, a comparatively recent innovation, and evidence enough can be gathered from the main existing languages, to understand the nature of its development. Two stages can be traced in this development. At the earliest stage there were two classes of nouns, on the one hand a 1 com- mon gender ' later differentiated into masculines and feminines, and on the other hand the ‘ neuters This state of affairs is faithfully reflected in Hittite, which is distinguished from all other IE languages by the absence of a special feminine gender. The next stage sees the development of the feminine, and it is only at this period that it is proper to speak of gender in the true sense. The existence of an earlier dual system is attested not only by Hittite, but also by abundant evidence gathered from the re- maining languages. MeiUet and others had adopted it on the basis of this latter evidence before anything much was known about Hittite, and the discovery of Hittite has gone further to confirm the theory. Attempts have been made to explain the dual system of Hittite as due to the loss of the feminine gender in that language, but no satisfactory evidence has been adduced for this. The fact is that the evidence of the other languages points unambiguously to the pre-existence of a dual system, and since such a system is to be found in Hittite, which in other re- spects preserves archaic features not known to the remaining THE FORMATION OF NOUNS 203 languages, there is no reason why the Hittite evidence should not be taken at its face value. The evidence from Sanskrit and the other languages is briefly that (i)The bulk of the masculine suffixes is also to be found in feminine nouns, and (2) that the specifically feminine suffixes a, % are used also in masculine de- rivatives. (i) The suffix Aar forms mainly agent nouns which are mas- culine. For the feminine the suffix l is added (ddtri) and a similar differentiation appears in other languages (Gk. ScmrZpa, Lat. ddtrlx ), On the other hand in the ancient group of nouns in Aar expressing family relationships the undifferentiated suffix is used for masculine ( pitdr -) and feminine (1 malar-, ydtar-) nouns. This conservative type preserves the older system which has been abandoned in the ordinary agent nouns in favour of a system in which masculine and feminine are distinguished. The suffix - sar which appears in svdsar- ‘ sister also in tisrds , cdtasras, Lat. uxor and thematised in Hitt, iihassarai ' lady ' functions adjectivally in the same way as -tar (both being opposed to the neuter -sar, Aar), but it tended at an early period to be specialised in feminine formations. The eventual adoption of il-suffix as the normal means of expressing the feminine •checked this development, and only these few survivals remain. The adjectival suffix - man is normally masculine (brahman-, etc.), but in bahuvrihi compounds it remains indifferent to gender according^ to the earlier system. Vedic examples of this are pnruiarma (Aditis), dyutddydmdnam (usdsam), sutdrmdnam (navam), acc. pi. sucijanmanas {usdsas), instr. pi. vdjabhar- mabhis ( utibhis ). The Atharva-veda first begins to show special feminine forms in compounds containing the word naman- ‘ name J : pancandmm ‘ having five names etc. The feminine suffix is added to the present participle in San- skrit and in Greek ( bhdranti , pipovaa), but in Latin the older undistinguished type is used for both masculine and feminine { ferens , jerentem ). The non-neuter (i.e. accented) suffix - as functions in both masculine and feminine nouns, e.g. apds- ' active ' masc. and fem. as opposed to dpas - neut, ' work ' ; likewise in bahuvrihi compounds, sumdnds nom. sg. masc. and fem. r well-disposed The same state of affairs appears in Greek, dXrjQri?, svjjLsvris, etc. The compound suffix -yas , functioning in a comparative THE FORMATION OF NOUNS 204 sense, adds the feminine -1 in Sanskrit ( bhuyast , etc.) but Latin preserves the undifferentiated usage ( maior , masc. and fem.). Non-neuter nouns in -i and -u are both masculine and fem- inine. The adjectives in i do not distinguish a masculine and feminine stem (sucis nom. sg. masc. and fern.) and those in -u optionally follow the same system {cdrus masc. and fem,). The latter may optionally form feminines in two ways ( bahvi 4 much ' tanti 4 thin ’), but the fact that this still remains optional shows that it is a comparatively recent innovation. The thematic suffix -a, accented and forming adjectives was originally in the same way indifferent to the distinction between masculine and feminine. This state of affairs has become alto- gether extinct in Sanskrit, but in addition to its being preserved in full force in Hittite, it has left considerable traces in Greek and Latin. It is preserved in Greek in compounds (poSo- SiIktvAos rjws, etc.) and in both Greek and Latin in a number of individual formations. A good illustration is provided by the word for daughter-in-law which appears with the thematic suffix in Greek and Latin (wo?, nurus) as opposed to the speci- fically feminine a-suffix which appears in Sanskrit [snusS] and Slavonic (Russ, snoxd). There is no doubt that the form pre- served in Greek and Latin is the more original, and that the form as it appears in Sanskrit and Slavonic is an innovation due to the growth of the system of grammatical gender ; IE snus~o - was formed at a time when the accented thematic vow r el was used simply to make adjectives on the basis of neuter stems in the way amply illustrated above (udra- : v&top, etc.) and was, as still in Hittite, indifferent to gender. The word is based on an obsolete neuter in -us, and etymologically this sn-u-s- is to be connected with sn-eu-bh - in Lat. nubo, etc. (2) Instances in the reverse direction are quotable from a variety of IE languages. In dealing with the suffix -d (-an) it was pointed out that it could appear with two functions, one originally neuter forming verbal abstracts, etc., and the other adjectival ; also that, since the usual variations in accent and apophony between the tw'o types were mainly eliminated in these stems, there is no formal difference between the two. The feminine gender developed with the specialisation of this suffix, in its adjectival function, as a feminine suffix, but there are still preserved a number of masculine adjectival formations with this suffix. Examples of such masculines are seen in Lat. scriba , THE FORMATION OF NOUNS 205 nauta, agricola , etc., O. SL sluga 'servant', vojevoda 'army commander ' which are in no way formally differentiated from feminines. Greek also has such masculine a-stems, but has differentiated them on its own by adding ~s in the nom. sg. (ttoit/ttJs*, etc.). In Sanskrit this type has become obsolete like that of the feminine o-stems. On the other hand there remain a number of masculines formed with the compound suffix -I (-i-H) which functions side by side with d in the formation of feminines. Skt, rathi - * charioteer ' is a survival from the time when adjectival -% was indifferent to gender, before it became specialised as a feminine suffix. In Italic and Celtic this adjec- tival -I, by an easy change of syntactical function, was adapted to form the genitive singular of o-stems [eqm stands to equus as rathi- to rdtha-). The existence of these common masc.-fem. formations so abundantly in Sanskrit and other IE languages, together with the twofold system of Hittite which shows no trace of a fem- inine gender, is capable of only one explanation. An older dual system has been replaced by a threefold classification into genders. The old system is preserved in its entirety in Hittite ; in Sanskrit and other languages it is still partly preserved, as the above examples show, but in the main it has been replaced by the threefold system. The process of this development cannot be followed in detail since it lies in the prehistory of the languages concerned. All that can be said is that at some period of later Indo-European the suffix d {-an) together with the compound suffixes l (-1-11) and u (-w-h) came to be specialised as feminine suffixes. This must have applied first to these suffixes in their adjectival use be- ginning possibly with a small nucleus of words'which happened to possess this suffix and were feminine by meaning (e.g. Skt. gnd , Gk. yvvrj). The suffixes so hsed are either an addition to the primary adjectival suffix {raj fit) or in the case of thematic stems a substitution for it {ndwosjndwd). The nature of the earlier dual system has been made suf- ficiently clear in dealing with the individual suffixes above. The words of f common gender ’ from which masculine and feminine nouns eventually derive are in origin adjectives or, what from the point of view of early Indo-European is the same thing, agent nouns. The fundamental division is the one repre- sented on the one hand by Gk. vBcop 1 water Hitt, arkuwar 206 THE FORMATION OF NOUNS ' prayer Skt. ydsas- ‘ fame ', brahma ' prayer ', *sthatar ‘stability', and on the other hand by Skt. udrd- ‘water- animal, otter fkvan- ' worshipping, worshipper yasds - ‘ famous brahman - ‘ priest and sthatdr- ' stander and in other examples copiously quoted above. It is therefore mis- leading to speak of an ‘ animate ’ and ' inanimate ' gender as if the twofold classification were in origin the expression of such a distinction. It is clear enough from the evidence that the origin of the system was primarily grammatical and not due to any psychological classification of objects in the external world. The so-called nouns of ‘ common gender ' or ' animate gender ' are in origin agent nouns, and they are predominantly ‘ ani- mate ’ (and in the main designative of human beings), because it is natural that the agent type of noun is most frequently applied to persons. It is not however exclusively so, and this may be illustrated by a number of Greek formations in -rryp, e.g. do/mjp 1 sword-belt apa mjp ‘ lamp-stand Kparrjp * mixing-bowl TptTrrrjp * pestle Y ^ukttjp ‘ yoke-strap etc. These represent an ancient type, better preserved in Greek than elsewhere, and show how in origin the adjective/agent-noun class of stem had nothing to do with the distinction between animate and in- animate. We have seen that these suffixally accented forma- tions are originally based on a class of neuters which are well represented by the Hittite formations in -tar. The latter are in the main verbal abstracts or nouns of a similar type. The adjectival type with suffixal accent means somebody or some- thing connected with the meaning of the primary neuter, and could originally apply to things as well as to persons. Because in practice such formations were most frequently applied to persons, the tendency was to eliminate their use as inanimates, so that in the case of nouns in -ter for instance such usage is rare outside this Greek type. Another type of archaism is preserved in the Vedic language. This is the occasional use of the masculine form of adjectives, in the case of consonantal stems, in agreement with neuter nouns. As examples of this we may quote vdcah . . . dvibdrhdh RV. 7. 8. 6, sdrdho . . . anarvanam 1. 37. i, visarmdnam krnuhi vittdm 5, 54, 9, sdrdho m&rutam . . . satydsravasam fbhvasam 5. 52. 8, tad rdstram ojasvi bhavati MS 4, p. 47,4. These reflect an early state of affairs when the formations with accented suffix and vrddhied nominative were purely adjectival, unconnected THE FORMATION OF NOUNS 207 with gender, and could therefore be used in agreement with any noun. With the growth of the gender system a new type of adjectival neuter was created, e.g. puru beside purus 1 much and traces like the above were eventually eliminated. The pro- cess is still to be seen in the course of development in Sanskrit in the case of the agent nouns in -tr ; the neuter formation kartf- f doer ’ (gen. sg. kartfnas) is unknown in the earliest texts and is a later analogical development. The older neuter forms of adjectives, though of early origin, were to begin with innova- tions of the same type. The foundation of the non-neuter class lies in the adjectival formations, but it was early augmented by transferences which introduced a growing number of action nouns. These have been classified separately in the above pages and are illustrated by such examples as bhdrd- masc. ' burden ', jigtsi ‘ desire of con- quest vac- fern. ' speech ’, ojmdn- masc. * strength bhiyds - fem. ' fear matt - fern. ‘ intelligence and idntu - masc. 1 thread The nature of this transference seems to have been mainly mechanical. Personification plays a certain part but this is strictly subsidiary. It is understandable that a stem like usds- should appear as feminine for this reason in view of the place of usds- in the Vedic pantheon ; or that omdn - 4 assist- ance ' and daman- 4 liberality ' which are invoked as divine attributes of the gods should be masculine rather than neuter. But no such consideration can apply to the majority of such nouns. For instance, while it is quite clear that vac- fem. 4 speech ' is in Sanskrit usage quite definitely personified as opposed to the neuter vdcas- f it cannot be said that it owes its feminine gender to this. On the other hand it is capable of being personified because for other reasons it has- acquired the feminine (derived from the originally common) gender. Stems terminating in occlusives in all IE languages take the nomi- native s and distinguish between nominative and accusative. In this they are distinguished from the mass of neuter action nouns and agree with the adjectival and agent noun type. It is clear also from the absence of cases to the contrary that this must have been the case from a very early period. There is no logical basis for this ; all that can be said is that there is a general rule that all stems of this type inflect in this way, that vdk($) f vicam is so inflected because it is a radical stem ending in an occlusive. The fact that it is inflected in this way, and 208 THE FORMATION OF NOUNS thereby acquires a non neuter, eventually feminine gender, enables it to be personified. In the same way we may judge the relationship between dpas, nom. pL fem. and udakam, udan - (Gk. v8cjp, etc.) ‘ water '* There is nothing about radical action nouns as such, from the point of view of their meaning which should cause them to become masculines or feminines ; only the mechanical development which caused them to inflect in the same way as the adjectival type of noun which formed the basis of the 1 animate 7 gender is responsible for their becoming such. In the same way other action noun stems eliding in occlusives early inflected in this way (Hitt, karlimmiyat- ' anger etc.). The neuter i-sterns were eliminated except for small remnants, and a similar tendency, though on a smaller scale is observable among the action nouns in -u. The thematic action nouns are extensions of root stems which were originally common gender, and this characteristic they retained ; when the common gender split into masculine and feminine they naturally became masculines because this is the masculine adjectival suffix. In the same way the action nouns in -d are feminine because this is the feminine adjectival suffix. An essential part is played in the development of the IE system of gender by the system prevailing in these languages by which an adjective must be inflected in the case, number and gender of the noun with which it is in agreement. This is one of the most characteristic features of Indo-European, as gram- matical congruence on this scale is hardly to be found elsewhere. Traces of an earlier system, in which the simple adjectival stem could function in attributive use, survive in nominal composi- tion, indicating that the full system was only gradually built up, but it is none the less of ancient origin. It is fully developed in Hittite and applies, there to gender in so far as the ‘ common gender ' and the ' neuter ' are distinguished, that is to say in the nominative and the accusative. With the growth of the feminine gender, which is the final stage in the development of the system, the system of congruence was correspondingly extended. §22* Nominal Composition The capacity to combine independent words into compound words is inherited by Sanskrit from Indo-European, and similar formations are found in other IE languages. Sanskrit differs THE FORMATION OF NOUNS 209 from the other IE languages in the enormous development which the system has undergone, which is unparalleled else- where. This development, however, is characteristic only of the classical-language, and in the Vedic language the use made of nominal composition is much more restricted. It is estimated that in the Rgveda the role it plays is not more important than in Homeric Greek. From the point of view of comparative philology it is mainly the Vedic language that has to be con- sidered. The unlimited development of nominal composition in the later classical literature is artificial and not based on spoken usage. The main features of a compound, though not invariably present, are (i) the appearance of the first member in its stem form, without the inflectional endings with which, except in the vocative, it is associated in independent use, and (2) the uniting of the two elements under one accent. The first feature is of great interest from the point of view of early Indo-European morphology, since it points to a time when the simple stem of a noun or adjective could appear in syntactical relation to other words of the sentence, without the case terminations which later became obligatory for the expression of such relationships. A compound comes into existence when two words appear so regularly and frequently together that they become to all intents and purposes a single expression, a process which is normally associated with the development of a specialised meaning. In the case of inflected groups this leads to compounds like Bfhaspdti- proper name of a divinity (‘ lord of prayer '). On the other hand a compound like vispdti - ' chief of a clan ' can only derive, as a type, from a state of affairs in which the relationship which is later expressed by the genitive case, could be expressed by the simple juxtaposition of two nouns in a certain order (vtk poti-s ), The compounds as a system are the fossilised remains of an earlier state of Indo-European which has long been supplanted by the consistently inflected type which appears in Sanskrit and the classical languages. Four main classes of compound were recognised by the Indian grammarians, Tatpurusa (with a special subdivision Kar- niadharaya) Bahuvrihi , Dvandva and Avyayibhdva, terms which will be defined below. Of these the last two are in the main specifically Indian developments ; the types inherited from Indo-European are those classified as tatpurusa and bahuvrihi. 210 THE FORMATION OF NOUNS For the purpose of this brief exposition the inherited types may be divided into two major classes, namely I, those which function as nouns and II. those which function as adjectives. These are followed by III. Dvandva and IV. Avyayibhava, The first class falls into two main divisions according to whether the first member is (a) an adjective or noun in apposi- tion with the second member or (b) a noun standing in such re- lationship to the second member as would normally be ex- pressed by a case termination. Conversely the adjectival type can conveniently be divided into two classes according to whether the final member is adjective or noun. Of the two major classes, nominal and adjectival, the former are rare in the early language, and this is the case elsewhere in Indo-European. On the other hand the various types of adjectival compound are abundantly represented, as elsewhere, particularly in Greek. We shad see that there is very good reason for this disparity and that it is of significance for understanding how the system evolved. I, A. Compounds in which the two members stand in apposi- tion to each other are named Karmadharaya by the Indian grammarians. The main class consists of an adjective followed by a noun. The type is rare in the Samhitas, but becomes more frequent in the later Vedic prose texts. Examples are candrd - mas- ' (bright) moon purndmdsa- ‘ full-moon J , ekavird - ' unique hero krsnaiakuni- ‘ raven mahdgrdmd- ‘ great host tnahavtrd- ' great hero mahddhana- 1 great wealth nllotpald-

• blue lotus ", rajatapdtra- * silver vessel daksindgm- ‘ southern

fire adharahanu- ' lower jaw ', trtiyasavand- * 3rd pressing navadavd- 1 land newly burnt for cultivation krsnasarpa - ' cobra Such compounds possess frequently specialised mean- ings, which would not automatically be expressed by the simple combination of the meanings of the adjective and the noun. The word krsnasakum literally * black bird ', means more speci- fically 4 raven ' ; ' black bird ' would be expressed by the un- compounded noun and adjective. Similarly mlotpald- means not merely ' a blue lotus ’, but a particular botanical species (Nymphaea cyanea). It is only in the later language that such compounds show a tendency to be used as simple equivalents of the combination adjective + noun, In a smaller class the first member is a noun in a relation of apposition to the second member. Such are purusamrgdTHE FORMATION OF NOUNS 211 4 male antelope ulukayatu- 'owl demon', vrsikapi 'man- ape r ajar si - 4 royal sage dhenustari 4 barren cow uksavehdt- 4 an impotent bull The karmadharaya is represented in other IE languages by such examples as Gk. aKponoXts, fieaoyaia, aypidfnreXos , Lat. angiportus, etc., and the second type can be compared with formations like Gk, tarpopLavn ? f physician-seer But just as in the earliest Sanskrit these formations are rare. This is natural in view of the origin of these compounds and their place in the grammatical structure of Indo-European in its various stages. They are the remains of a time when the adjective, when used attributively, took no inflections for gender, number and case. Such a state of affairs came early to be replaced in Indo-European by one in which the attributive adjective was inflected in agreement with its noun in all cases, genders and numbers, but there remained a few expressions which were so grown together in usage that they continued as relics of the old system. These could then serve as models for the creation of new examples of the same type. B. Tatpurusas with an ordinary substantive as their first member are in the earliest language somewhat more numerous than compounds of the karmadharaya type, but they are still distinctly rare in comparison with Bahuvrihis and other adjec- tival types. They are rarast in the earliest part of the Rgveda and become gradually more important in the succesive stages of Vedic literature. Examples are rdjaputra- 4 king's son ', mrtyubdndhu- 4 companion of death vispdti - 4 lord of the tribe drupadd - f post of wood hiranyarathd - ' car of gold devakilbisd- 4 offence against the gods indraseni 4 Indra's army camas&dhvaryu- 4 the priest connected with the cups J , drughand- 4 mace of wood deary aj ay & 4 teacher's wife purusardja - f king of men ', ajaloma - ' goat's hair asvavald- 4 hair (from the tail) of a horse ', udapdtrd - 4 bowl of water The relationship between the two members is most frequently that expressed by the genitive case, but being very general it can in various examples be transcribed by all cases, and the Indian grammarians have classified them on these lines : Dative, yupaddru f wood for a sacrificial post ' (yiipdya daru), ablative, caurabhaya - 'fear of thieves' ( caurebhyo bhayam) i locative grdmavdsa- 4 dwelling in a village *, accusative videiagamana f going abroad 212 THE FORMATION OF NOUNS Compounds of which the last member is a verbal action noun in -li form a special class. Such are dhdnasdti- ' winning of wealth devdhuti- 1 invocation of the gods ', sdmasuti - * pressing of soma ' and devdhiti - ‘ ordinance of the gods These have accent of the first member as opposed to the tatpurusas above and in this respect go with the adjective compounds whose final member is a participle in -ta (virdjdta-, etc.). They are also abundantly formed in the earliest language, a characteristic which is usually reserved for adjective compounds. A few instances where the final member in -ti has acquired a concrete sense are to be classed with the examples above, e.g. devaheti ‘ weapon of the gods ’. The corresponding type appears in other IE languages in such examples as Gk. ^TpoTrarajp 4 mother’s father ", narpaSeXpos 4 father’s brother ", oiKoSeoTroTrjs 1 master of a house ", Lat. muscerda * mouse dung ", Goth, pindangardi- ' king’s house ", O. SI. vodotoku 1 watercourse Besides these there exists in the Vedic language a new type of compound in which the first member retains its genitive ending, and, most frequently, its accent. These are commonest with pdti - as the last member : vdnaspdti- ‘ lord of the wood, tree ", gnaspati- 1 husband of a divine woman ", brhaspdli- 4 lord of de- votion ", etc. ; with one accent amhasaspati- f lord of distress ", name of an intercalary month. Other examples appear rarely : divodasa - 4 servant of heaven ", rdyasposa - ‘ increase of wealth ' ; later, gospada. - 4 cow’s footprint, small puddle dasydhputra- ‘ slave-girl"s son (term of abuse) Compare Gk. AiooKovpoi , etc. This is the type of compound an inflected language might be expected to form. Its emergence in the Vedic language is to be viewed in connection with the comparative rarity of the ordinary type. As in Greek, etc., these had come to play only a small part in the language, and were in comparison with other kinds of compound, unproductive. Later the reverse process sets in ; the frequency of the true tatpnrusa increases and the development of the new inflected type is checked. II. Compounds functioning as adjectives may be divided into two classes according as to whether the latter member is an adjective or noun. A. (i) (a) Compounds with verbal adjective as second member. In these the first member most frequently stands in the accusa- tive relationship to the verbal adjective which forms the THE FORMATION OF NOUNS 213 second member. They may be classified according to the various types of stem that appear in the second member. Root-stems : havirdd- ' eating the oblation aivavid- ' know- ing horses vrtrahdn - ‘ slaying enemies Roots in i, u, r may not appear as root nouns and add the augment A : dhanajit- ‘ conquering wealth somasut - ‘pressing soma', jyotiskft- ‘ making light Such compounds with root stems have some- times a passive meaning : manoyuj - ‘ yoked by the will hrdaydvtdh- 4 pierced to the heart The type is familiar in other IE languages^ cf . Gk. Lat. fidicen, artifex, etc. Thematic suffix : annddd- 4 food-eating J , goghna- * killing cows J , devavanda- 4 god-praising karmakara - * workman A newer type in which the first member takes the accusative (or occasionally some other) termination is common in connection with this suffix : dhananjaya- 4 conquering wealth purandard- 4 destroying cities talpesayd- 4 lying on a bed The thematic type is familiar in other IE languages : Av, hasidava - " betray- ing a friend Gk. OvpLopOopos, Spvro/xos, Lat. causidicus , mag- nificus , Russ, vodonos 4 water-carrier etc. Suffix -ana : kesavdrdhana- 4 cutting hair amitrad dm bh a na~ ‘ deceiving enemies/, devayajana- 4 worshipping the gods Suffix -in : ukthasamsin- 4 singing hymns vratacdrin- 4 performing a vow satyavadin - * speaking the truth J . Suffix -i : pathirdksi - ‘ protecting the road sahobhdri - ‘ bearing strength Suffixes -van and -man : somapdvan - r drinking soma *, baladdvun- * giving strength svaduksddman - 4 sharing out sweet things Cf. Gk. 7 roXv 8 €yfxojv. Other suffixes- raslradipsu - ‘injuring the kingdom', lokakrtnu - ' world making nrpdtf- 4 men-protecting Among formations of this kind there is a considerable class in which the form of the last member is modelled on the present stem taken by the root in question. Such are stems ending in : -ya (4th class) : punarmanyd - 4 again thinking of akrstapacyd- 4 ripening without ploughing asuryampasyd - ! not seeing the sun -ay a (10th class and causatives) : anilayd- ‘ not resting *, janamejaya - 4 rousing the people dharmadhdraya- 1 maintain- ing the law Cl. Av. naro vaepaya-. -nva : visvaminvd 4 stimulating all dhiyamjinvd- t danupinva 214 THE FORMATION OF NOUNS -na : duradabhnd kulampuna-, sadaprna -. Cf. Gk. jucr- dapvos , TroAuSa/LtvT]?. Infixed nasal : agnimindhd salyakrntd govinda -. Cf. Av. yimd'karznta -. Reduplicated formations : sardhanjaha-, manojighra -, adha-. Cf. Av. azro'daBa-. (b) Compounds with a past participle passive as second member differ from these in the syntactical relation of their members, and also in having their accent normally on the first member. For this reason they may be treated as a separate sub-class. They are a very productive type : hdstakrta- ' made by hand vtrdjdta - ' bom of a hero devatta - ‘ given by the gods prajapatisrsfa- ' created by P. ulkabhihata- ‘ struck by a thunder bolt ' ; indrotd- 4 helped by Indra This is an old Indo-European type which is also represented in related lan- guages : Av. ahura-Sdta-, Gk. IrnT^Xaros , etc. The type of compound instanced under (a) is characterised by the fact that the second member is very frequently, in the case of some classes almost invariably, a stem that cannot appear in independent use. Compounds like goghnd - may be formed at will but a simple ghna - does not exist. The same feature is shared by the related languages and goes back to the Indo-Euro- pean period. The origin of this type of compound goes back to an earlier phase of Indo-European with a different and simpler structure to that prevailing in the historical period and the period immediately preceding it. What in the historical period are compounds were to begin with constructions of a type which are familiar in languages with a less developed inflection than Indo-European. The relative participles known in Dravidian and certain other linguistic groups are instances of this type of construction. In Indo-European the growth of inflection led to the disuse of such simple constructions but this type of compound, based on them, continued to flourish. (2) Compounds having as their last member an ordinary adjective are comparatively few. Such are : tanusubhra - 4 shining in body yajnddhtra - 4 versed in the sacrifice sdmavipra - 4 skilled in Sama chants tildmisra- r mixed with sesamum ' ; with case termination of the first member, ntdderaghu - 4 quick in exhilaration vidmandpas - 4 working with wisdom THE FORMATION OF NOUNS 215 B. (i) Bahuvrihi-. The bahuvrihi or possessive compounds contain the same elements and in the same order as the kar- madharaya and tatpurusa compounds but differ in meaning in that the compound functions as an adjective qualifying some other concept. They also differ in accentuation from the karmadharaya and tatpurusa types, being characterised normally by the retention of the accent of the first member of the compound. The distinction between the two opposing types is illustrated by such cases as rtijaputra- ‘ having kings as sons * as opposed to rdjaputrd- ' son of a king * and stiryatejas- ' having the brightness of the sun ’ as opposed to siiryatejds - ' the sun's brightness The following will serve as typical examples of the bahuvrihi type : bahuvrihi - ' having much rice ' (after which the class is named), mayUraroman- ' having the plumes of a peacock indrasatru - ' whose foe is Indra ugrdbdhu - 4 having powerful arms ', dirghdstnasru- 1 long-bearded jivdputra- 4 having living sons iddhagni- 4 whose fire is kindled pra - yatadaksina- 4 who has presented sacrificial gifts ckinndpaksa - 4 whose wing is severed sucddratha- ' having a shining chariot pancdhguri- 4 five-fingered mddhujihva - 4 honey-tongued manigriva- 1 having a necklace on the neck pdtrahasta- ' having a vessel in the hand J , vdjrabdhu - 4 armed with the vajra khara - mukha- 4 donkey-faced In the Vedic language there are occas- ional examples with inflected first member : kratvdmagha- 4 constituting a reward gained by intelligence 4 , dsannisu- 1 having arrows in the mouth diviyoni - 4 whose origin is in heaven The type is widely distributed in the IE languages. Gk. Aev/coiAevo? 4 white-armed poSoBaKrvXos 4 rose-fingered Lat. magnanimus f great-souled capricornus 1 having the horns of a goat Goth, hrainya-hairts 4 pure-hearted O. SI. crlnovlasu 4 black-haired The bahuvrihi likewise originated in the earlier, less inflected period of Indo-European, and it remained after the system of declining adjective and noun in apposition was developed. That development was, as we have seen, unfavourable to the growth of a large class of karmadharaya compounds, since in the simple collocations of adjective and noun the inflected forms were used. On the other hand the bahuvrihi construc- tion could not be so simply transformed, since a substitute could only be found by clumsy periphrases. Consequently it H THE FORMATION OF NOUNS 2l6 survived in the more developed inflectional stage in the form of these compounds. Though the latter member of these compounds is always a noun, it does, in the case of consonantal stems always have an adjectival form, e.g, suyasas - ' of good fame From *(a)su yasas- * good fame ' (cf. Hitt. a£su~ ‘ good ') an adjective, nom. sg. suyasas is formed in the same way as yasas from simple ydsas , since the apophony indicates that the accent was origin- ally on the last syllable of the compound too. The same applies to the w-stems : nom. sg. putundmd ( having many names etc. Adjectival -d is frequently used in the same way as with simple nouns 1 anudrd - ' without water urunasd- ' having a broad nose trivatsa- 1 three years old sarvavedasd- ‘ (sacrifice) in which all property is given aw r ay Other adjectival suffixes are frequently appended, e.g. -ka : pvapitrka - ‘ whose father is alive punyalaksmlka- ‘ having auspicious marks ' ; -ya : hiranyakesya- f golden haired ", mddhuhastya - f having sweetness in the hand ' ; -in : mahdhastin - f having a large hand satagvin- 1 having a hundred cows (2) Adjectival compounds are formed on the basis of the combination preposition + noun. Corresponding to dty dmhas ‘ beyond distress * there exists the compound dtyamhas - ( one who is beyond the reach of distress Similarly dnuvrata - ' obedient abhidyu - ‘ directed to heaven upakaksd - f reach- ing to the shoulder urdhvdnabhas- ' being above the clouds parihastd ' something put round the hand, amulet These compounds frequently take the adjectival suffixes w r hich have been noted above in the case of bahuvrihis : djarasa - ‘reaching to old age apatki- ‘ being in the way paripanthin - r way- layer upatrnya- ‘ lurking in the grass {3) An archaic class, confined entirely to the Vedic language, is composed of a participial first member governing the second member. Examples : vidadvasu- ' winning wealth bharad - vaja - * carrying off prizes ’ , tardddvesa- ( overcoming hostility mandaydtsakha - * rejoicing friends The same type is estab- lished in Old Iranian : Av. vanat • pviana- ‘ wanning battles etc. Sporadically other verbal noun stems are used in the same w T ay : Trasddasyu - ' making enemies tremble raddvasu - ‘ opening up wealth ditivara - ' giving choice things Similar governing compounds are familiar in Greek ; pepiotKos ' carry- ing his house €K€ai 7 T€ 7 ros * dragging robes THE FORMATION OF NOUNS 217 Adjective compounds used as adverbs Compound adjectives may be used adverbially in the same way as simple adjectives, and as such normally appear in the accusative singular neuter. This is common with bahuvrihis and in the classical literature long conglomerations of this nature are frequently so used. It is also common with the com- pounds having a preposition or other indeclinable as first member, thus atimdtram adv. * excessively J from atimdtra- adj. ‘ excessive Such adverbial compounds are considered by the Hindu grammarians to be a separate class of compound and they are termed avyaytbhdva The reason for this lies in the fact that in the later language there is a considerable class of such adverbs without actual adjectives corresponding to them. This class is represented by such examples as upardjam ‘ near the king j upanadam ‘ near the river V pratyagni ‘ facing the fire J and pratinisam ' nightly A productive class of indeclinables is formed by those compounds, which have a relative adverb as prior member ; yathdkdmam ‘ according to wish ydvajjtvam ‘ as long as one lives ", etc. III. Dvandva Compounds . This type has nothing exactly corresponding to it in the re- lated languages and has developed mainly within the historical period of Sanskrit itself. The earliest type which is common to Sanskrit and Avestan consists of two duals, each retaining its own accents, which are juxtaposed in such a way that a + b is expressed by 22H-2&: Mitri-V arunau r M. and V.', dyavd- prthivi * heaven and earth usdsd-ndktd f dawn and night ' ; cf. Av. pasu viva ‘ beast and man gen. pasvd virayd. Sometimes elliptic duals may be used in place of this construction : Mitra 1 M. and V/, pitara r parents dyivd ‘ heaven and earth This represents the oldest state of affairs. Out of such con- structions, which are not in the proper sense of the term com- pounds, the Sanskrit system of dvandva compounds developed and some of the intermediate stages may be observed in €he early literature. Thus (1) the form of the nom. acc. dual is retained in the first member in cases other than nom* acc. : mitrd-vdrunayoh instead of mitrayor vdrunayoh ; (2) in a small number of instances in the RV, and in a somewhat larger number in the later Samhitas, the first member of such a 218 the formation of nouns combination loses its accent : indrdpusndh, somdrudrdyoh . The final stage appears when the first member appears in its simple stem form : indravayU (only example in RV.) vdyu- savitfbhydm, daksakratU, etc. Because the ending of the agent nouns, etc., in -tar in the nom. sg. coincides with that of the first member of the dual dv r andva, this form is chosen in dvandva compounds having such a stem as first member : pitaputrau ‘ father and son kotadhvaryu , etc. Plural dvandvas are exceedingly rare in the earliest language. Of the few examples indrdmarutah (voc.) and pitdputrah 1 father and sons J are modelled, as far as their first member is con- cerned, on the dual dvandvas. An example of the normal type, with simple- stem of first member, appears in a late hymn of the Rgveda : ajavdyah 4 sheep and goats In the later Vedic litera- ture such examples become more common : devamanusydh ' gods and men bhadrapdpah 4 good and bad In this period also dvandva compounds with more than two members begin to appear : prandpanoddnesu . Feminine nouns are not found employed in the oldest type of dvandva in the Veda, though such are known from Avestan : dpa urvaire 4 water and crops They appear in the later Vedic period in the fully developed type of compound : jdydpait 4 wife and husband A few neuter dvandvas of the fully developed type appear even in the Rgveda : satyanrU 4 truth and falsehood ahordtr&ni 4 days and nights There are also a few older types : idhmd - barhis - with two accents and istdpurta- with the first member modelled on the old type of masculine dvandva. In the plural dvandvas angapdrumsi r limbs and joints ' (two accents) and ukthasastrani 4 hymns and praises * the form of the first member may be interpreted as the old form of the nom. acc. pi. neut. At an early period there was created a type of neuter dvandva which functions as a singular collective stem : krtdkrtd- 4 what has been done and what has not been done trnodakd- 4 grass and water kasipupabarhand- 4 pillow and bolster In the examples both members are neuters. The same type also appears early in cases where one member is neuter, whether it be the last member as in kesasmasru- 4 hair and beard * and in klomahrdayd- 4 lungs and heart or the first member as in ahorat^d- 4 a day and a night J , sirogrivd- 4 head and neck ' and in yuga&amyd- ' yoke and the attaching pin Finally the stage THE FORMATION OF NOUNS 219 is reached (in the Brahmanas) when two non-neuters are com- bined to form a singular collective dvandva : osadhivanaspati- ' plants and trees candratarakd- ‘ moon and stars ustrakhara- 4 camels and asses Adjectival dvandvas are formed by the combination of two adjectives applying to the same noun, and such occur from the Rgveda onwards : nllalohita - ' dark blue and red tdmradhumrd- 4 dusky copper-coloured arundbabhru- 4 red brown krsna - Gabala- 4 speckled black sitosna- 4 lukewarm etc, There are parallels to these in other IE languages (e.g. Gk. XevKonvppos* XcvKopaios, yXvKVTHKpns) and it is likely that in contradistinc- tion to the nominal dvandvas the type is inherited from Indo- European. CHAPTER V THE DECLENSION OF NOUNS §i. Accent and Apophony The IE declensional system was characterised by a shift of accent from the stem in the strong cases (norm acc. sg. and du., nom. pi.) to the termination in the weak cases, that is to say in the majority of the oblique cases. This shift of accent entailed apophonic changes affecting stem and suffix. The system was already in decay in the late IE period, and tending to be replaced by a system of fixed accent. In Vedic the shift of accent is best maintained in monosyllabic stems, but considerable traces of it are found in the case of other types of stem, both radically accented neuters (ydkrt, yaknds) and suffixally accented masc,- fem. types. The three grades of apophony associated with this accent shift are clearly seen in the declension of vrtrahdn- : nom. sg. vrtraha, acc. sg. vrtrahdnam, gen. sg. vrtraghnds . It is seen also in the suffixally accented r- and w-stems of the type pit A, pitdram , pitri, uksa * bull uksdnam, uksnds. Elsewhere it has been modified and simplified in various ways. The vrddhi of the nom. sg. tends to be extended to the acc. sg. and nom. pi., e.g. pat * foot 9 , padam , pddas, as opposed to Gk. 7ro&a, noSes- The accent shift may remain while the vowel gradation is abandoned, e.g. dik 4 direction gen, sg. diids for what must originally have been *ddks : dikSs . Conversely the accent may be stabilised but the vowel gradation retained, e.g. pasuman ' possessing cows pasumdntani, pasumdtas . The system of accent shift is best preserved in radical con- sonantal stems. In these the accent regularly appears on the termination outside the strong cases. On the other hand the accompanying vowel gradation is only partially preserved. The three grades appear in the declension of ksdm- ‘ earth ' : nom. du. ks&md with vrddhi, loc, sg. ksdmi with guna, gen. sg. ksmds , jmds, gmds with zero grade of root. Elsewhere the zero grade is rare in alternating stems : cf. vrtraghnds already mentioned, 220 THE DECLENSION OF NOUNS 221 havyauha instr. sg. of havyav&h - ' carrying the oblation duras (once duras RV. 2. 2. 7.) acc. pL of dvaras f doors In other stems there is only alternation between vrddhi and guna, e.g. nom, sg. pAt ' foot gen. sg. padds , a pas nom. pi. f water acc. pi. apds, n&sa nom. acc. du. ' nostrils gen. loc nasos . The vrddhi grade is generalised in vtic- 4 speech ' (nom, sg. vak, gen. sg. vac as) as also in Lat. vox , vocis, in contradistinction to the guna grade in Gk. dp, dna. The guna grade is generalised in such nouns as ksdp - ' night instr. sg. ksapa , spas- 4 spy etc. Root nouns having i, u, or r as the radical vowel have generalised the weak grade in all cases : dik nom. sg. f direc- tion gen. sg. Jisds, instr. pL digbhis , similarly from re- e hymn J , fk, reds, rgbhis, and so on without exception. With these belong root nouns originally ending in h, namely radical stems in -F. and -u, e.g. dhf- ' thought ' ( dhis , dhtyam, dhiyds) and bhd- 4 earth ' [bhfts, bhuvam, bhuvds). Accent alternation has been abandoned as a general, rule in the case of those root stems which appear at the end of com- pounds, e.g. trivft- 4 threefold gen. sg. trivftas. The older alternating system is only preserved in vrlraghnds because the vowel of the root had been elided. In the alternating stem anadvah- gen, sg. anaduhas the apophony indicates that there was originally a shifting accent which has been replaced by a fixed accent. An exception to the general tendency is found in the various formations in -anc ( pratydne etc.) which are originally compounds of prepositions with the root of dksi ' eye Here the accent appears on the termination in the weakest cases (gen. sg. praticds) but it is shifted back in the middle cases with a corresponding difference of grades ( pratyagbhis ). The movable accent was originally characteristic of the neuter nouns formed with the various suffixes classified above. The tendency was from the late IE period for this to be given up and replaced by a fixed radical accent. Nevertheless there remain in Sanskrit, as also in Greek, various survivals from this system. The accent shift is usually preserved in the archaic neuters with alternating rjn stem : asrk, asnds 4 blood ydkrt, yaknds 4 liver sdkrt , saknds 1 dung ' ; likewise in the stems in ijn : dksi, aksnds 4 eye dsthi, asthnds ' bone dddhi, dadhnds 4 curd Similar terminational accent appears also in dsnds gen. sg. (dsya- ‘ mouth *), likewise in dosnds, yusnds, udnds and strsnds (gen. sg. to dos 4 forearm yds- 4 broth udakd - 4 water and 222 THE DECLENSION OF NOUNS sir as ' head '). The terminational accent in the oblique cases entails in some instances a reduction of the root, e.g. udnds as compared with Engl, water , etc. It is seen also in gen. sg. usrds 4 of the dawn 3 compared with the strong stem vasar° which appears in compounds. In two cases fixed accent has been applied to nouns of the old r n declension : dhar , dhnas ‘ day udhar, udhnas ‘ udder In the masc, and fem, r- and w-stems the accent shift to the termination is preserved in the weakest cases where the vowel of the suffix is lost : pita , pitrd, cf. Gk. ira r^p, tt cn-pos, murdha ‘ head gen. murdhnds, uksd 4 bull 3 , uksnds, cf. Gk. apr^v, Faprjv 4 lamb gen. apvos. In the middle cases where the n and r of the suffix were vocalised on the loss of the guna vowel the accent is retracted to the suffix : pitfbhis, murdhdbhis. In cer- tain cases the apophony indicates that the accent was originally of the alternating variety although it has become fixed : svd f dog gen. sunas (original accent in Gk. kvvqs), yuvi ‘ young man gen. yilnas, maghdvd, maghonas. With as-stems traces of this accent shift are exceedingly rare. The instr. sg. bhtsd ( bhiyds - 4 fear ') and the gen. sg. usds (for us-s-as) show both the terminational accent and the correspond- ing weak form of suffix and root. Otherwise these stems have been normalised. This accent was originally characteristic of the i- and w-stems, and traces remain notably in Greek : 019 sheep gen. olos, yow 4 knee * t yowos, Sopv ‘ spear *, Sovpos. Sanskrit in general has stabilised the accent ( dvyas , mddhvas) though occasionally the apophony indicates original accent of the termination, e.g. in drunas ( <drunds) gen. of ddru 4 wood The nom. sg. pdsu nt. and the gen. sg. paivds (masc. but originally neuter) repre- sent the original IE inflection, but they no longer belong to- gether, since the forms classed with pdiu have acquired a normal radical accent, and a masc. pasus has come into being, to which the gen. sg. pasvas is attached. Elsewhere termina- tional accent appears in a small number of suffixally accented nouns which take the gen, sg. termination -as : aris : ary as ; ray is : ray as ; pitus : pitvas. The accent shift remains in the case of those participles in -ant which are accented on the suffix*: nom. sg. ad an ( eating ' : instr. sg. adatd ; yunjdn * joining ' : yunjata ; sanvdn 1 press- ing 3 : sunvatd , etc. ; but this does not apply to the middle THE DECLENSION OF NOUNS 223 cases where the accent is on the suffix and not the termination. These are the main cases where shift of accent appears in de- clension in Sanskrit. Elsewhere, and these form the majority of stems, the accent has been stabilised either on the root or on the suffix. Fixed accent on the root becomes the normal accentua- tion of neuter nouns in Sanskrit : dhdnva ' bow dhdnvanas ; ndma ' name ndmnas ; brahma ‘ prayer brdhmanas ; dmhas ' distress dmhasas ; mddhu 4 honey b mddhvas , mddhunas ; vdri 4 water b virinas. Fixed radical accent is likewise the rule in those masculine and feminine nouns which are accented on the root : bhrqtd ‘ brother b instr bhratrd ; raja 4 king gen. r&jnas ; bhdvan 4 being b bhdvatas ; gdmdn 4 possessing cows gdmatas ; avis 4 sheep b dvyas ; pdtis 4 lord b pates , paly us (husband) ; krdtus 1 wisdom b kr divas ; sdtrus 4 enemy sdtros. Thematic stems, both those accented on the root and those accented on the suffix, have invariably fixed accent both in San- skrit and Greek. The same applies to verbal thematic stems. The evidence is that in Indo-European such stems were char- acterised by fixed accent from the beginning. Fixed suffixal accent in the case of non-thematic stems appears in a number of types. It is rare in Sanskrit in the r- and «-stems, though not uncommon in Greek (noi^v voipwos, and with vrddhi carried through, Sorrjp, So rypos). Such accent only appears in Sanskrit in those «-stems in which the suffixal vowel is not elided in the weak cases : brahma , brahmdnas. On the other hand this type of accentuation has assumed great importance in the case of the i- and w-stems, where it produced a special inflection of the suffixally accented type (adjective and agent noun), which was eventually applied to all masculine and feminine nouns however accented. Inflection of the type pita : pitrd is found only in the stem sdkhi- friend , norn. sg. sdkha t dat. sg. sdkhye , but there has been a secondary shift of accent to the root which must have originally been the same in the two cases. Elsewhere in the normal type ( agnis t agnds) there is fixed radical accent, and this type must be very ancient because accent and apophony are in agreement in the gen. sg., etc. The accent causes the retention of the guna grade of the suffix in the genitive, dative ( agndye ) and nom. pi. (agnayas), and the reduction of the gen. sg. termination to -s (agnd-s). Likewise in the case of w-stems the inflection of the type vdyus 4 wind gen. sg. vdyds arises from the fixation of 22^ THE DECLENSION OF NOUNS the accent on the suffix of suffixally accented agent nouns and adjectives. This type of declension eventually ousts the alter- native type, which was originally exclusively used with neuter nouns (mddhvas, pasvds originally neuter) and optionally in the agent noun-adjective type ( sdkhye with secondary radical accent). The decline of the neuter as opposed to the masc.-fem. types is largely responsible for this development. Fixed accent on the suffix is to be found also in masc. and fem. stems in -as : nom. raksds, gen. raksdsas and in the stems in -mant and -vant in so far as they are accented on the suffix : pasumdn , pasumdtas. In the latter case the apophony indicates an original shifting accent. It does so also in the perfect parti- ciples whose fixed accent must be of secondary origin : cakrvdn , cakrusas. The accent of certain infinitival forms calls for mention since it differs from any of the types listed so far. This appears in certain dative infinitives which are accented on the suffix. This is most commonly found in infinitives formed from 5-stems, e.g. reuse 1 to praise cardse 1 to move spurdhdse ‘ to strive bhojdse * to enjojr Some examples are also found from man- and vans terns : vidmdne 1 to know 1 ddvdne 1 to give turvdne ' to overcome This accent cannot be original since suffixal accent is pioper to the adjective and agent-noun type, whereas the neuter action nouns, to which these infinitives belong, are accented on the root. It is also hardly possible that this type of accentuation should have supplanted a radical accentuation, since that has become the normal type, and the reverse would be expected. The accent normal to neuter nouns does some- times occur in these infinitives, rarely in those with suffix -as (dyase 1 to go dhdyase ‘ to cherish more preponderantly elsewhere (ddmane ' to give dhtirvane ' to injure Since it is unlikely that this normal type would be supplanted, the in- finitival accent on the suffix must be explained as a substitu- tion for older terminational accent (*rcasd, etc.). The cause of this change is not altogether clear but it may be associated with the tendency observed elsewhere to avoid final accentuation in forms of more than two syllables : cf. trivftas as opposed to rcds t and aksdbhis as opposed to aksnd. It may also be due partly to influence of locative infinitives where the regular accent was on the suffix. A few' action nouns in -as have also acquired suffixal accent, THE DECLENSION OF NOTJNS 225 notably bhiyas - 4 fear instr. bhiydsd beside older bhtsd (p. 159). When the neuter nouns had substituted fixed radical accent for the shift of the accent to the suffix in the oblique cases, the only type where it normally took place were masculine stems in which the vowel of the suffix was elided in these cases : uksa t uksnds . On this analogy the few remaining neuter stems which retained oblique cases with terminational accent received suffixal accent in nom. acc, (acc, bhiydsam corresponding to bhtsd like uksdnam to uksnd ; later bhiydsd is created by stabilis- ation of the new accent). In this, noun the feminine gender results from its changed accent. Suffixal accent has become normal in the majority of the neuter nouns in -is : arcis- ' flame gen. sg. arcisas, etc., as opposed to the rarer type jyotis - ' light '. The anomalous nature of this accent is clear from the weak grade of the syllable on which it is placed, and also from a comparison with the related -as and -us stems. The same type of accent appears in the I-stems of the vrkt- type and in ii-stems (originally -in and -wh stems). The system here is more complicated inasmuch as these classes contain both action-noun types (dehi 4 rampart tanU 4 body ') and agent-noun/adjective types (vrkt- 4 she- wolf agrii- ‘ maid '). The accent of the former type is exactly parallel to that of arcis-, etc. An exact parallel to the latter type is found in the adjectives in -in : bait , balinas. In both these adjectival types the suffixal accent is regular, but its weak grade is to be explained out of forms in which the accent was originally on the suffix (*vrkiyds, * balinas). The weak grade associated with the latter forms has been generalised, but also the suffixal accent of the nom. acc. where originally the strong grade of the suffix must have prevailed. The action nouns of the {vrkt) t- and w-stems have fallen together with the adjective/agent-noun type in accent as in other respects. The same kind of development seems to have taken place in a number of originally neuter i- and u- stems. This is clearest in the case of the stem paid - 4 domestic animal A neuter paiu is preserved in one instance and comparative evidence shows that this form with its radical accent is original (cf. Lat. pecu , Goth. faihu , 0 . Pruss. pecku : IE piku). The old form of the gen, sg. to this, with its terminational accent is preserved in Sanskrit ( pasvas ), but by the analogy of the masc, stems mentioned above this form is the cause of the creation of a new. masculine, 226 THE DECLENSION OF NOUNS nom. sg. pasus . The same seems to have occurred with pitus 1 food gen. sg. pitvds , since this by its meaning is an action noun, and among i-stems with rayts : rdyds ‘ wealth ’ (for

• rdms : ranyds). Possibly also some masc. w-stems which are

not of the adjective/agent-noun type arose in this way, e.g. murdha t gen. sg. murdhnds ‘ head §2. Heteroclitic Declension The mutual relation of the r- and n-stems has been dealt with at some length in the section dealing with the formation of nouns, and may be briefly summarised here. The neuter r- stems that remain in Sanskrit are normally not declined out- side the nom. acc. sg., w-stems being used in the remaining cases : dharjdhnas f day yakrtjyaknds ‘ liver etc. This type of inflection is found elsewhere in Indo-European, but always, outside Hittite, as an archaic survival, and not as a productive formation. In Hittite, on the other hand, this type of alterna- tion is exceedingly common, and appears regularly in the in- flection of neuters in rjn t and in the compound suffixes -mar, war , sar , tar/n. It was therefore at an earlier period of Indo- European much commoner than later, and its decline is due partly to the decline of the old neuter types in general, and partly to the extension of the w-stem to the nom. acc. sg. This system arose too early for it to be possible now to say how pre- cisely it came into being. It does not however appear that the neuter, r-stems were from the beginning incapable of inflection, since such examples occur in all languages (Skt. svar j stir as ; vasar 0 jusrds ; Gk. eapjeapos Hitt, kururjkururas , etc.), and there is no reason to believe that this type is not ancient. Nor can it be said that the w-suffix is in origin either a case termina- tion or a formative making an oblique case. It is a suffix in its own right, on a par with the others, and it appears like them in the nom. acc. sg. in many ancient examples (e.g. Skt. n&ma ‘ name ', Lat. notnen, Hitt. Idman, etc.). It is therefore dif- ficult to say how exactly these two stems so often combined to form a single paradigm, but this took place in the early period of Indo-European, and though the system was beginning to become obsolete in the final stages of the parent language, it persists as an archaic survival in several of the existing lan- guages. THE DECLENSION OF NOUNS 227 In the same way the few neuters in -x substitute an n-stem outside the nom. acc. sg. : dsthi : asthnds ' bone etc. Besides these the instr. pi. naktdbhis may correspond to an old neuter ndkti ' night ' which has been replaced by a feminine (nom. sg. ndktis) , In the case of vdri ‘water' the n-suffix is added to nistead of being substituted for the i- suffix (gen. sg. vdri^as). This process appears commonly in neuter w-stems : ddru drunas , mddhn ‘ honey mddhunas, etc., and its antiquity is guaranteed by similar formations elsewhere : Gk. Sopu, Soparo? (*dorwntos, with the additional /-suffix characteristic of Greek). In the Vedic language this is only one means of inflecting the neuters (the alternatives are mddhvas and madhos , the latter an innova- tion borrowed from the masc.), but it becomes the general rule in the classical language. This ^-extension in the oblique cases is found occasionally with other suffixes : e.g. the os-stem sir as * head gen. sg. strsnds , and the yd~ stem kanyd , gen. kaninatn , Av. also gen. sg. kainind. The use of this n is much extended in certain cases, e.g. instr. sg. of masc. i- and w-stems (agmnd, vdyund) ; in the gen. pi. it has been intro- duced in the case of all vocalic stems : devdnam, agnlnam, pitfnam , etc. Certain defective neuter n- and m-stems appear in the Veda mainly as instrumentals, and have become attached to the corresponding man- stems : bhund , mahnd, prend, prathind, mahind } varind (bhumdn- 1 abundance etc., prathimdn - ‘ width etc.). It has been noted above that it is unnecessary in these cases to assume a change of -mn- to -n-. An alternative us Ivan similar to the rjn alternation appears in the Vedic declension of dhdnus- ‘ bow The ws-stem appears in the nom. sg., while elsewhere the stem dhdnvan- is used. It is probable that the two stems pdrus - and pdrvan - ' joint ' were originally distributed in the same way. This combination of n- and s-stems is found also in the corresponding masculines. The voc. sg. of certain ^aw-stems in the Veda is in -vas : rtdvas, sva-ydvas , vibhavas , mdtavisvas from ftdvan- ‘ righteous etc. There are also doublets like fbhvan-, fbhvas - ; sikvan- } sikvas - (both meaning 1 skilled ') in which the two suffixes alternate without any apparent rule. This voc. -vas appears also, and more regularly, in the stems in -vant, which is a /-extension of the i>a«-suffix, and in the parallel stems in - mant : rsivas, gndvas, patnlvas, tuvismas, bhdnumas , sucismas. In the case of 228 THE DECLENSION OF NOUNS these stems the 5-forms are more extensively used in Iranian, since Avestan has nominatives of this kind attached to steins in -if ant : amavd ( amavant -). On analysis the Sanskrit forms of the nom, sg T are derived from this ( ndmasvan , pasuman , etc.). The nominative formed from the vant- and mant - stems would have appeared as °van and °tnan {like °an in the ant- stems). The nominatives in -van, -man are derived from *-vdns, -mans (-vdms, mams) which have replaced *-vas, *-mas by analogical extension of nasalisation to the nom. sg. This phenomenon is found elsewhere in Sanskrit in 5-stems, and since it does not appear in Iranian, it is to be taken as a special Indian development. The alternation of nasalised forms in the strong cases with forms without nasal in the weak cases in such classes as the present participles (< adan . addntam , adatA , etc.), which is due to the change of the sonant nasal to a, led to the extension of n to the strong cases of other classes where the nasal does not originally belong. This is found notably in the comparatives in y as {srdydn ‘ better srdydmsam, srdyasas) and in the perfect participles in -vas ( vidvdn , vid- vAmsam , vidusas). It is found also, in the declension of pums- 1 man ' : pumAn, pumAmsam, pumsds . This is a masc. -as- stem, but one which in contradistinction to the normalised type (raksas, raksdsam , raksdsas) has preserved some archaic features. These are (i) the weakening of the radical vowel as a result of the accentuation of the suffix, (2) the old terminational accent of the oblique cases as in pitrd, uksnds, etc., and {3) the con- sequent weakening of the suffix in these cases. In addition the inflection is complicated by the introduction of the nasal into the nom. acc, sg. (replacing * pumas, *pumasam ). 1 There is one other example of this nasalisation among the masc. as-stems, namely svdvdn , nom. sg. of svdvas - f helpful The introduction of -n- into the heteroclitic nom. sg. of the vant- and mant - stems follows this general principle, and it was further facilitated by the existence of -n- regularly in the acc. sg. which was formed with the vant- stem. The distribution of the two stems corresponds to that of the neuter us and van in dhdnus-jdhdnvan above except that in the masculine the acc. sg. 1 The masc. pumas - would correspond to a neuter *petimos- ‘ pubes The Lat. words pubes, puber, have different suffixes. Since the final root here is likely to be that which appears also in Lat, pu-d-or, b and m may also be vary- ing sulbxal elements, alternatively b in Latin may be for m before r in puber as in hibemus , tuber (; tumdr ), THE DECLENSION OF NOUNS 229 has a form different from the nom. sg. and this form follows the analogy of the majority of cases. The yaw^-suffix is built on the van-suffix and though no nom. sg. in -vas is recorded for the latter type of stem, the vocatives in -vas are an indication that the nom. sg. may once have been so formed. In some stems the suffixes van - and vant- are combined heteroclitically. The stem maghdvan - (nom. sg. maghdvd, gen. sg. maghdnas uses the vant - stem before terminations beginning with a consonant (instr. pi. niaghdvadbhis). An instr. sg. fkvatd appears beside the usual stem fkvan - ' praising The stems dr van- and qrvant- ' steed * are interchangeable. The stem yuvan-, yun- makes its neut. sg. yuvat, and this extended stem is the basis of the fem. yuvatv The perfect participle is formed mainly with the stem in -■ yarns- jus , but before the terminations beginning * with a con- sonant, there appears a stem in vat ( vidvddbhis instr, pi.). This suffix reappears in Greek, where it forms the normal basis of the declension IS tbs', elSoro s) and it is attested also in Gothic [weitwod- ( witness '). The comparative evidence shows it to be different from the vat [wnt) which is the weak form of the vant- suffix, since it has no nasal. The word for ' path, way ' declines with a variety of stems. The strong form in the Rgveda appears as nom. sg. pdnihds , acc. sg. pdnthdm , nom. ph pdnthas, to which correspond Aw pantA, pantam. In the weak cases the stem appears as path- (instr. sg. patha, etc.). The relation of these two stems is one of apophonv : strong form of suffix a h (> a), weak form h. The weak form of the suffix, H, aspirates the preceding t, and this aspiration is then extended to the nom. sg., etc. The same development occurs in the case of the strong and weak stems mahd mah- 1 great ' (megeu 2 -lmegn 2 ). In the middie cases of path - an i- stem is used, which occurs elsewhere (O. SI. pgii, O. Pruss. pintis) : inscr. ph pathibhis , etc. (on the other hand Av. has padzbis without -?'-). In the same way mahdjmah - has a supple- mentary i-stem, in this case in the nept. sg. (mahi ; in Av. also in. the instr. ph mazibiS). After the Rgveda there appears another strong stem of path-, an n- stem (acc. sg. jpdnthdham , nom. ph panthdnas). This is also ancient since the same formation appears in Avestan : pantdmm , pantdno. The same kind of inflection is laid down by the grammarians for rbhuks- ‘ n. of a divine being ' and math- 4 chuming-stick 230 THE DECLENSION OF NOUNS §3. The Case-terminations Nominative Singular , Masc. and Fem. The nominative singular of masculine and feminine nouns is formed in three ways, (1) by vrddhi of the suffix, (2) by the termination s, (3) by the simple stem uncharacterised in any way. The basis of the first method has been dealt with at length in the section dealing with the formation of nouns. There it was seen that adjectives and agent nouns were formed on the basis of the various types of neuter stem by the transference of the accent to the suffix, and that for phonetic reasons which are not now clear, this led to the vrddhi of the suffix in the nom. sg. Assoc- iated with this vrddhi there is a tendency for the final semi- vowel of a suffix to be elided : pita , brahmi . The vrddhiejd type of nom. sg. appears regularly in the case of masc. r- t n- and s- stems (i data , cf. Gk, Sorrjp ; brahmi , cf. Gk, ; raksas , sumands , cf. Gk. pevSrjs, tvfievrjs), rarely in the. case of i-s terns (sdkha). A similar vrddhied nom. sg, originally existed in the case of w-stems of this type, but in all cases where such vrddhi is preserved it has secondarily acquired the addition of -$ : Skt. dyaus , Av. uzbdzaus, Gk. fiaaiAev? (for etc. The termination -s appears in Sanskrit, which in this respect is in close agreement with the related languages, in the mas- culine a-stems (devas, Lat. deus) in masc. and fem. i- and sy- stems, both of the action-noun {malts, krdtus, Gk. /teeny, 7777^9, etc.) and adjectival type {iucis, purus ; Gk. tSpty, 7 toAuV, etc.), in t-stems of the vrki - type and w-stems, in consonantal, stems (action-noun or agent), including the monosyllabic stems originally ending in -11 (dhts, bhis). It is clear that there is no common principle uniting these various formations, and distinguishing them from those classes in which the nom. sg. masc. and fem. is formed differently. It is also clear that the range of s-nominatives has extended at the expense of other types. This has already been observed in the case of the vrddhied nominatives in -ans. It is also clear that the feminines in -i of the vrki class, and those in -u ( tanUs ) have acquired their -s from the radical stems in -% and -u which else- where are declined like them. The feminines in ~a and those in -I of the devi class preserve the uncharacterised nominative which was original to the stems in -h. Furthermore it is doubtful whether the 5-nominative was originally attached to the u and THE DECLENSION OF NOUNS 231 tf-stems, though it must have extended to them at a very early period. The reason for this is that the action nouns of these classes would be* expected, according to the general analogy, to have been originally neuters, and this is borne out by the exist- ence of survivals of the type ; on the other hand in the case of the agent-noun/adjective types there is evidence of a vrddhied treatment parallel to that of the r- y n- and 5 -stems, and though this is rare in the historical period, it can be shown to have been much commoner earlier. The probability is that the nominative s was originally proper to the adjectival thematic stems, since they are definitely a class apart. Its extension to the i- and u- stems is not difficult to understand, since like the thematic stems they are vocalic stems. In this process the morphological distinction and agent-noun /adjective appears to have counted very little, and the 5 of the nominative (and with it the distinc- tion between norm and acc.) became early attached even to the action nouns of these classes. In this way the majority of such nouns acquired the masc, or fern, (earlier, common) gender. Historically they do not take s in the nom. sg. because they have gender, but rather they have acquired gender as a result of taking -s. In the same way sterns in occlusives were from an early period characterised by s in the nom. sg., and this regard- less of whether they were action or agent nouns. In mono- syllabic stems the s was also added after consonants other than occlusives (nasals, Gk. Krais ; H, Skt. dhis, bhUs). The feminines in -d and those in -1 of the devi class have no special sign for the nom. sg. To this extent they agree with the neuters. The action nouns ending in these suffixes originally were neuters, and in the case of this suffix the adjectival type, which was specialised as a fern, formation was always less clearly distinguished from the action noun type than was the case with other suffixes. Accusative Singular, Masc. and Fern. The accusative singular masc. and fem. shows no such variation. The termination appears as -m after vocalic stems ( dsvam , Lat. equum ; agnim , Lat. ignem, etc.) and - am after consonantal stems ( padarn , rdjanam, pitdram , etc.). In the latter case Greek has a out of m ( 77 -oSa, irarapa) and, this is the form which would be expected phonetically, but Indo-Iranian substitutes the fuller form which has the advantage of greater clarity. In some languages (Gk., Celt., Germ.) this final -m changes to -n, as it also does in 232 THE DECLENSION OF NOUNS the nom. acc. sg. neut. (Gk. AuVov, WSov), and it is certainly wrong to assume, as is sometimes done, that the latter is the more original form. Nom -Acc. Sg. Neuter. Apart from a-stems neuter nouns have no endings in these cases : fidhar, Gk. ofidap ; mddhu , Gk. p.£9v ; nama, Lat. nomen, Hitt. ldman } etc.). In thematic neuters the termination in both cases is -w. It has been suggested above that the -m of the old neuters of this class was originally the suffix -m {yugdm : yugmd etc.) and as result of these forms coinciding with the acc. sg. of thematic adjectival stems, a neuter thematic type was developed. Instrumental Singular , The instrumental singular shows no united formation in Indo-European. Forms corresponding to the -d which is the regular ending in Indo-Iranian, appear only in certain languages, and there only in certain classes of stem. In addition there appear the endings -bhi (Gk. 0 co</h, Arm. mardov) and -mi (O. SI. vlttkomi , Lith. sunumi). The former element is that which appears in the instr. pi. in Sanskrit (-bhi-s). In Greek it is used indifferently either as singular or plural, and further in a wide sense, covering instr. loc. and abl. Hittite has a different formation of its own (-et) which is not to be reconciled with any of the other forms in Indo-European. It appears that the instrumental with its various forms is a comparatively new case, and consequently has no common form covering the whole of Indo-European. The Sanskrit form is normally i.e. -an : padd , pitrd, rdjna , etc. But it may also appear in the zero grade, -h, notably in the case of feminine f-stems : citti (Av. iisti ) , uti t justi etc. In Avestan this form is also attested for w-stems : mainyu , xratii, etc. It must further be assumed for thematic stems (vfk&y Av. whrkd , etc.), since the acute accent which appears elsewhere (Lith. vilku , geru-ju , and cf. the Gk. adv. iTTiax^pdt which is interpreted as an old instrumental) speaks against contraction (IE therefore wk w o- H, not ujk^o-oujeu). The quality of the long vowel that developed from this varied between -6 (Lith. vilku , OHG. wolfu with u<o) and -l (Goth. kamme-h } he , Skt. (adverbs) paicd, uccd with palatalisation indicating This implies an original IE metaphony enjou^ Dative Singular. The termination is - e , Av. -e, oi, Indo-Ir. -ai : padi, pitrd t iune, mdnase ; Av. bzrdzaite , vise, pidre , paiQvit-fia. zrdzzjyoi, etc. The IE ending -ei is preserved in THE DECLENSION OF NOUNS 233 Gscan {paler ei t regaturei, leginei) and Phrygian (FavaKrei). Elsewhere phonetic developments have obscured it (Lat. main, O. SI. materi , etc.). There has been some dispute about the original form of the dative ending, since alternatively the Greek infinitives in -at have been compared (Sd/imit, SoFevcu, Sovvai, etc.). But the existence in Greek of certain traces of the dative in -ex (dtFct^tAo?), shows that, whatever the explana- tion of these infinitives, they should be discounted in settling the form of the IE dative. Ablative Singular . A special form for the ablative singular, which elsewhere has the same form as the gen. sg., is found only in the declension of thematic stems : -at (-ad) in vfkdt, etc., Lat. lupo(d), This represents the IE state of affairs. In certain languages, notably in Italic and the later Avestan, this form is extended to other classes (Qsc. touiad ‘ civitate Lat. magis- trate, Av. adrat , garoit , etc.). In Slavonic this termination serves both for the ablative and the genitive of the thematic de- clension (O. SI. vluka). It is not possible to determine whether the final consonant was originally d or t. The vowel was b in ordinary nominal declension alternating with e in adverbial forms (Lat . facillumed). In the latter type the termination was accented (Skt. pascdf , sanat ) . The vowel was of circumflex quality (Lith. to] Gk. (dial.) rtbSe ' from here ’) indicating con- traction (-dd<-o-od), and this is reflected by occasional disyl- labic scansion in the Veda. Genitive- A blative Singular . The termination, which outside the thematic class combines the functions of the ablative and the genitive, is -as representing IE -es and -os. The difference between the two depended on accentuation, -es occurring in connection with the original terminational accent, - os in those cases (Gk. owfiaros, etc.) where the accent had become fixed on the root. This distinction is nowhere preserved, since in the various languages one or other form is generalised, e.g. -os in Gk. (awfiaros, ttoSos*) and -es (> is) in Latin (corporis, pedis , but O. Lat. regus, etc.). In addition there exists a reduced termina- tion -s which appears in conjunction with adjective and agent noun types with accented suffix. This appears in Sanskrit in connection with i- and w-stems (agnds, siinds ), in Avestan also in some r-stetns ( pitars ). In the i- and w-stems the form has spread from the adjectival typ^ to which it properly belongs, to the majority of action nouns (maids, dtos ). Only a few 234 THE DECLENSION of nouns examples remain in the Veda of the alternative type (dvyas, mddhvas ) . Locative Singular. Three types of locative singular are found in Sanskrit, illustrated by the alternative forms of locative of the word for ' eye ’ : aksan , aksdni, aksni . Their chronology appears to be in this order. The type aksni is the latest. According to the grammarians the locative of n-stems may be in -ani or -ni (r&jani, rdjni ; sakthdni, sakthni ), but in the lan- guage of the Rgveda the latter type does not appear, and is therefore clearly an innovation. It is due to an analogical tendency to put the loc. sg. on the same footing as the other oblique cases by accenting the termination and weakening the suffix. In 'many of the consonantal stems this tendency had already become general in the pre-Vedic period ( ddati, bhdga- vati f vidusi , etc.), but the older type with accent and guna of suffix is preserved in the tfw-stems, in r-stems [svdsari, pitdri), to which certain monosyllabic stems can be added : ksami, dydvi (beside divt). The oldest form, the locative without ending, appears in n- stems ( dhan , murdhdn, sirsdn ; cf. Gk. allv * always and in- finitives like So/xcf, etc.), and in the vrddhied forms of the t- and westerns. It also appears sporadically elsewhere, e.g. in parut ‘ last year ' as opposed to Gk. Trdpvai, irdpvri, a compound whose last member {-ut) is the weak form of the wet that appears in Hitt, wett Gk. Feros ‘ year In Avestan there appears a locative without ending from a root noun man - 1 mind ' in the phrase mm ca daidydi 4 and to put in the mind, remember The locative in -i is based on the older locative without end- ing, to which a suffix or particle has been added. This pro- duces a clearer form which tends to oust the earlier form with- out ending, but the process is not yet complete by the Vedic period. To a large extent this form of locative preserves the accent and guna of the suffix which characterised the form without ending, and it is thus sharply differentiated from the genitive and dative singular with their accented termination. At the same time analogy has tended to adapt the loc. sg. to their type, in some cases in the prehistoric period ( adati , etc.) and in other cases during the history of Sanskrit itself ( rdjfii , etc.). The suffixal accent of the old locatives without ending is parallel to that which has been observed to occur in adverbs based on neuter stems (prdtdr, etc.). THE DECLENSION OF NOUNS 235 Vocative Singular. The vocative singular consists of the simple uninflected stem, and it is therefore a survival from the time when the inflection of the noun had not been built up to the degree which appears later. In this respect it points to the same conclusion as the system of nominal composition, indicat- ing an early period of IE in which the bare stem could function as a word. In thematic stems the vocative is formed simply by dropping the -s which characterises the nominative : vrka , Gk, A ukc, Lat. hipe . The various languages agree in having the e- grade of the suffix in this case instead of the usual o (Au^oy, etc.). Stems which take vrddhi in the nom. sg. substitute guna in the vocative (svan, pitar as opposed to svA, pita , etc.), and this characteristic is found also in related languages (Gk. kvov , 77 glt€/}). This guna appears also in i- and w-stems ( agne , suno) since the formation of adjectival i- and w-stems was originally parallel to that of the adjectival r- and w-stems. The feminines in -t and -u substitute the short vowels i and u and this is pro- bably to be regarded as the regular development of -iw and -wh when followed by a pause. Compare the similar development of -ah to d in Greek vocatives like vvfuf>a, av^wra. The vocative of the d-stem is anomalous (bale, voc. of bald ‘ girl ') and is per- haps due to the addition of an enclitic particle i (-e for -<zh-i). The vocative is unaccented in Sanskrit, except when it appears at the beginning of a sentence or pa da, and in these conditions it has a special accent of its own, namely on the first syllable, regardless of the normal position of the accent in the word. There are traces of this latter type of accentuation elsewhere (Gk. 7 TaT€p } d&eA <f>€, as opposed to irar-qp, aSeApos in the nom. sg.), but nowhere to the extent that is found in Sanskrit. The system cannot be very ancient, otherwise there would not be regular guna of the suffix, which is due to the suffixal accentua- tion which characterises these types normally. Nominative Plural , Masc , and Fem . The Sanskrit - as (pAdas ‘ feet ') corresponds to IE -es which is preserved elsewhere (Gk. 7 rdS€f). The termination always appears in the full grade though it is never accented. It is associated with the strong stem in stems of varying grades, and this may be either guna {pitar as, uksdnas , agnayas) or vrddhi ( datdras , rajdnas). As in the accusative singular the latter type is due to extension from the nominative singular. An 5 appears in most of the plural cases, e.g. acc. -ns, instr. 236 THE DECLENSION OF NOUNS -this, dat. abl. - hhyas , loc. -su. It is possible,, but not certain that this 5 is identical with that of the nom. pL The IE plural system is complicated by two unusual features. On the one hand if this s is the sign of the plural it is distinguished from the type found in most linguistic families by being added after instead of appearing before the case terminations. In the second place the terminations of the plural are for the most part different from those that appear in the singular and this appears to be quite an unusual phenomenon. A further problem is pre- sented by Hittite. In this language the nominative plural has a form of its own, and likewise the accusative {hum antes, humandus f all ') but for the gen. dat. there appears normally a form identb cal with the gen. sg., and otherwise the inflection is undeveloped. It is uncertain to what extent this is due to Hittite innovation, but it may be an indication that the plural inflection in IE is a later development than the singular. Accusative Plural, Masc. and Fern. The ending in IE was - ns after vocalic stems, -ns after consonantal stems. This is preserved in Gothic and certain Greek dialects, notably Cretan (Gk. eA €v0epov$, vlvvs, Goth, wulfans , gastins, bropruns). In Sanskrit the accusative plural of masc. vocalic stems {-an, -in, -un, -fn) preserves this -s in sandhi (-dms, etc., before t-). In the Veda its effect is seen also before a vowel (-dm, imr ), The long vowel in Sanskrit is not original but arises in thematic stems from the analogy of the nom. pi. (-as with long vowel whence acc. - dns for - ans ). From this declension the long vowel has spread to the stems in -i, -u and -r. The ending -ns after con- sonantal stems becomes regularly -as in Sanskrit as in Greek (padds, TToSas). The acc. pi. is a weak case in Sanskrit, that is to say the termination may be accented and the stem appears in its weak form. This is in contradiction with the fact that the termination itself appears in the weak grade and it is therefore in all probability an innovation. If IE -ns in this case is de- rived from -ms the form can have arisen by the addition of the plural sign -s to the acc. sg. The feminine vocalic stems show no trace of n in Sanskru (-as, - is , -us, fs). This absence of n is shown to be IE in the case of stems in -d by the agreement of Indo-Iranian (Skt. kanyds , Av. urvard) and Germanic (Goth, gibos). Non-distinction of nominative and accusative, which characterises neuters was originally characteristic of -d (~#h) stems when these had not THE DECLENSION OF NOUNS 237 become differentiated from the other neuters. It is preserved in the a-formation which serves to provide the plural of neuter thematic stems (nom. acc. pi. yugd). The normal feminines have acquired the plural -s of the non-neuter classes but they still retain in the plural the absence of distinction between nom. and acc. From the i-stems the type spreads to the fem. u- and r-stems which did not originally in IE have a form distinct from that of the corresponding masculines (Gk. (dial.) ofivs ' sheep ). Nominative- Accusative Plural Neuter . (1) The neuter plural appears still in the Vedic language in some cases undifferen- tiated from the singular : e.g. in udhar divyini ' divine udders vtsvdni vdsu ' all goods ydjand pur it 1 many leagues sdm aranta pdrva ‘ the joints came together This is a survival from an early stage when the inflectional system was less de- veloped. (2) There exists in Iranian, beside this type, a series of neuter plurals characterised by vrddhi of the suffix : Av. ayarz ‘ days va 6 d ‘ words ndman 1 names etc. This type is ancient since examples are also quotable from Hittite, e.g. widdr, pi. of watar 4 water In Greek on the other hand such vrddhied neuter forms appear merely as singulars : vScjp, T€Kfiojp, etc. These may be old plural forms utilised as singulars after the type had died out as a plural formation. Sanskrit has in the main replaced this type by that which is extended by the suffix -1 (nAmdni 4 names ') but the Vedic language still re- tains it (beside the alternative form) in the case of neuter n- stems (bhumd ' beings aha 1 days sirs a 4 heads ') in which the n of the suffix is elided as elsewhere in connection with vrddhi ( rdja , etc.). (3) The neuter plurals which are made by suffixing i to these vrddhied forms appear also in Avestan (nampni ‘ names sdx v Jnt * teachings *, vardidhi 1 energies *) as an alternative to the plurals with simple vrddhi. A neuter plural suffix -i is found in Hittite (kururi pi. of kurur nt. ' hostility '), which testifies to its antiquity in Indo-European as a method of forming the neuter plural. The i is apparently identical with the suffix -i which appears in the formation of neuter nouns. Other IE languages have mainly the suffix a or a which originates from the thematic stems (Gk. ovopara like Ivyd, etc.). In Vedic the i-form of the plural has been much extended in comparison with the Indo-Iranian state of affairs which can be deduced from the comparison of Avestan. The 238 THE DECLENSION OF NOUNS formations with simple vrddhi have disappeared in the case of most types of stem. In addition the nasal of the n- and -wi- st e ms (nitndni, ghrt'atidnti) has been analogically introduced into other types of'stem, e.g. mdnamsi ‘ minds' for munas-i , simi- larly havinisi ‘ ablations cdksumsi f eyes etc. A non- nasalised form remains only in the case of caivdri * four The process is continued further in the post-Rgvedic period by the creation of nasalised i-plurals for consonantal root-stems, e.g. °sanki from °sak- ‘ able °bundhi from °budh - ‘ understand- ing In addition, on the analogy of the neuter n-stems like ndmani there is created a new type of neuter plural for a-, i - and n-stems : bhuvandni f worlds suclni ‘ bright vdsttni ' riches In the Vedic language these forms occur in competi- tion with the older forms { bhuvand , suet, vdsu), but in the later language they are exclusively used. Furthermore, on this analogy the later language creates a neut. pi. -Ifni for stems in - ir . (4) The inherited neuter plural of thematic stems is in -d (yugd, Gk. £uya, Lat. iuga, Goth, juka, 0 . SI. iga). This -d is identical with the suffix -a which in the historical period forms feminines. As already pointed out this d ( <a) was not origin- ally distinct from the usual type of neuter suffix. These plurals were originally singular neuter collectives, and in Greek they retain this character to the extent that they are still construed with a singular verb (Vd £o>a rpc^t). In this function the suffix -a retains its primitive characteristic of being indifferent to the distinction between nominative and accusative. The variation in the IE languages between a and d appears to be due to dif- ferent sandhi developments of IE -an (-d before vowels or a pause, otherwise -d) and it is paralleled by a similar fluctuation in the case of feminines in -d. The neuter stems in -1 and -w also* make plurals by lengthening the vowel of the stem, and if these forms are not simply made on the analogy of the thematic neuters, they can be analysed -t-H and -u-u with the weak form of the suffix added to the stem. In the Vedic language they exist side by side with forms undifferentiated from the singular and with the innovating type -Ini, - uni which later becomes the rule. h Inmental Plural. The ending of the instrumental plural v ,*t$ (Av; -bis) contains an element -bhi- which according to the evidence of Greek (Oeorf) i, dycAij^i, t<j>t t va.v<f)i, €pef}€o<f>L, etc.) was at an earlier period of IE of much vaguer and wider appliTHE DECLENSION OF NOUNS 239 cation, being used both in the singular and the plural, and covering the meanings of instrumental, locative and ablative. In Indo- Iranian, as opposed to Greek and Armenian ( gailov , pi. gailovh 1 : gail ‘ wolf ') this formative appears only in the plural, the instrumental singular being formed in quite a different manner. The final -s may be interpreted as the -5 of the plural added to this element, or possibly in view of such adverbial forms like Gk. A iKpufals 'crosswise* and Av. mazibis 1 greatly * may be merely some adverbial suffix (cf. afxfts, etc.), which in view of the regular occurrence of -s in the plural led to its being understood as such. As elsewhere Balto- Slavonic and Germanic have -m- in this case (Lith. summits, etc.) which it is not possible phonetically to relate to the -bh- of the other languages. Dative- Ablative Plural. Anomalously the ablative which in the singular has mainly the same form as the genitive, has in the plural a form identical with that of the dative. The ending is -bhyas, Av. byd. The western IE languages have a form similar to this going back to original -bhos (Lat. -bus, Osc.fs, ss, Venet. -bos, Gallic -/3o). It is possible but not certain that this - bhos has developed out ‘of -bhyos through the sporadic loss of post-consonantal -y- t easily understood in a weakly stressed termination. The analysis of the form is indicated by the com- parison of the datives of the personal pronouns. Beside the usual forms tubhyam , asmdbhyam the Vedic language preserves also a form without -m, whose antiquity is attested by Iranian (Av. maibyd). The -bhyas of the dat.-abl. plural can be inter- preted as this - bhya followed by the -s which characterises the plural. In this way the case would originally be a dative, and its use also as ablative can naturally be explained by the fact that the -as which comes at the end of the termination is similar in form to the -as of the gen.-abl. sg. Genitive Plural . The termination of the genitive plural is distinguished from the majority of the plural cases by the absence of $ (w r ith the exception of the pronominal forms tesdni , tdsam). The termination is -dm which is frequently scanned as disyllabic in the Veda, and this in conjunction with the circumflex accent in Gk. -Gjv, points to an original contrac- tion of - o-om . This can only have come about in thematic stems, and it must be assumed that the original termination -am has elsewhere been replaced by the long contracted -dm 240 THE DECLENSION OF NOUNS which arose in this class. The shorter termination -om has been generalized in Slavonic (>u f mdteru, imenu). t and also probably in Latin ( hominum ), where it appears unnecessary to assume that -um has developed out of an earlier form with a long vowel. A variant £-grade of this formative appears in Gothic only {wulfe, suniwe). The Sanskrit vocalic stems are characterised by an -n- inserted before the termination, and the Avestan de- clension agrees with this system with the exception of the stems in -r. The only agreement elsewhere is in Germanic, in the tf-stems (OHG. gebono 4 of the gifts ’) and it is likely that the inserted - n - began in this class and from there spread to the other vocalic classes. Locative Plural. The Sanskrit termination -su (patsu) appears also in Iranian, Slavonic (-chuCsu) and dialectically in Lithuanian. In Greek on the other hand the termination is -<n {ttogoI, etc.). This variation indicates that the termination is analysable into two elements, on the one hand s+w and on the other hand s+i. The s can be identified as the plural s which appears in other cases, to which the further elements Land u are added in the two types. The -i of Greek is apparently to be identified with the -i of the locative singular, and the -u of the other languages in an alternative suffix performing the same function. The case would thus originally be formed by the addition of plural -$ to the endingless form of the loc. sg. (in thematic stems to the loc. sg, in -oi), and the addition of i and u is secondary, just as is the addition of -i in the loc. sg, Nom . Voc. Acc . Dual. This case was made by various form- atives in IE, according to the type of stem. The ending - au , -a of Sanskrit was originally, from the evidence of the related lan- guages, confined to the thematic stems (Gk. Xvkoj, Lith. vilku , O. SI. vluka) from which in Sanskrit it has been extended to other types of stem (pddau, pitdrau , etc,). In these latter classes Greek and Lithuanian have an ending - e [fi^Tepc, auguse ). It has been suggested that this termination, elsewhere replaced by - au in Sanskrit, is preserved in the dual dvandva matara- pitarau 4 parents ’ which the grammarians quote as a northern form. The termination is regularly - au in classical Sanskrit, but in the Vedic language it varies between -au and d. As a general rule -au is used before vowels, becoming ~av, elsewhere a. Some such variation must go back to the IE period, and it is the latter form which has been generalised in the related languages. THE DECLENSION OF NOUNS 24I The feminines in -a and the neuters take the termination -t. This agreement is another sign of the close relations existing be- tween the feminine a-stems and the neuters. Examples from neuter consonantal stems are : vricast, cdksusi, n&mni, brhatl . In the thematic neuters and the feminines in -a this -l combines w r ith the vowel of the stem to form the diphthong -e ; yugi, sfnge ; dsve , sene. This - 1 is found also in Old Slavonic, though rarely ( imeni , telesi) , and Slavonic shows the same diphthong in 0 - and a-stems (sele, ryce) a form which then spreads to con- sonantal stems (imene ' two names '). The 1 - and w-stems make the dual by lengthening the vowel of the stem : ftati, sunii, and with them are to be classed the feminines in -l of the devt type in the Vedic declension (du. devt, later devyau). This type is ancient, appearing also in Avestan (gairi, mainyn ), Slavonic (O. SI. pgti, syny) and Lithuanian (nakti, s&nu ). Instr . Dat . Abl. Dual . The termination that serves for all the three cases is -bhydm, and it contains the same element - bhi - that occurs in the dat. -abl. and instr, pi. A corresponding - byam appears only once in Avestan ( brvatbyqm from brvat- ‘ brow '), Elsewhere it has -by a and O. Pers. has -biyd, which makes it clear that -m is an element secondarily added, as else- where (cf, tubhya, tubhyam, etc.). The Balto-Slavonic languages have, as in the plural, - m - instead of -bh- here (O, SI. ocima, etc.). The termination is ordinarily added to the normal stem, but in the earlier language sometimes to the form that serves as nom. acc. du., e.g. akstbhydm , cf. aksi ' the two eyes ’ (cf. O. SI. ocima ; oci ‘ the two eyes '). This has become the normal form in the case of thematic stems : vfkdbhydm , cf. vfkd(u) t etc. Genitive-Locative Dual. The termination common to these two cases is -os : padds f pitros , etc., which is added to the weak form of the stem. Avestan on the other hand has two separate terminations, -6 (zastayo) for the locative and -d ( ndirikayd ) for the genitive. The ending -d is derived from - au and is equi- valent to the Sanskrit ending minus the final -s. The genitive -d ( <-as) is peculiar to Avestan. Slavonic has a termination -u which could represent either -ou (Av. ~d) or -ous (Skt. -os). Lithuanian which keeps this inflection only in some adverbial forms has both -au and - aus : dvejau y dvejaus * in twos, as a pair *, cf. Skt. dvdyos. The -ay- which in Sanskrit appears before the termination in a and a-stems has spread from the declension of the pronouns and the numeral * two ' ( tdyos t dvdyos ). It re242 THE DECLENSION OF NOUNS mains confined to these cases in Slavonic : toju, dvoju, but vluku, igu, rgku. §4, The Declension Classes The complication of the Sanskrit declension consists not so much in the system of terminations thus briefly described, as in the combination of these with the various types of stem, and the alternation of the stem itself in respect of accent and apophony. The classification of the stem types and the de- clensions based on them falls naturally into five main divisions : (1) consonantal stems, (2) stems in -r, (3) stems in i, u, (4) stems in d, t, u , (5) stems in -a (thematic stems). After the practice of the grammar of the classical languages, and also for reasons of convenience, the descriptive grammars normally deal with these classes in the reverse order to that given here. Since however the normal scheme of terminations as described above appears most clearly in the consonantal declension, and since the thematic declension is the most aberrant from this, having adopted a variety of special declensional forms from the pronouns, it is more convenient from the point of view of comparative gram- mar to proceed in this order. §5. Consonantal Stems The consonantal stems consist of the root stems (_ pad etc.) and derivative stems in -n, -nt t - s , etc. The latter fall into tw r o classes, neuters and masculine-feminines. The particulars of their formation, and the mutual relation of the tw r o classes have already been dealt with. The inflection of the neuters and non- neuters is distinguished only in the nom. and acc. In this re- spect the consonantal stems differ from the stems in i and u (mddhvas : stinds), and also from the practice of certain other languages with consonantal stems (Gk. ov8a to?, dyd/xajos*, neut. : (frepovros , tt ot/xeVo?, masc.). The declension of these stems calls for little extra comment. The normal endings are added with little modification throughout this declension. The special development of Sanskrit phonetics cause some complica- tion (e.g, vis- * settlement ' : nom. vit } acc. visam , instr. pi. vidbhis , loc. pi. (vedic) viksu) but this aspect of the problem belongs more properly to phonetics than to morphology. For the rest the complications that occur in this class have already been described under the headings of (1) Accent and Apophony THE DECLENSION OF NOUNS 243 (pat : padds, etc.) and (2) Heteroclitic Declension (tidhar, udhnas, etc.). The strong tendency of Sanskrit to nasalise the stem in the strong cases has also been noted (vidvan, vidvAm- saw, vidusas ) . It spreads by analogy from those cases where it is historically justified ( bhdvan , bhdvantam, bhdvatas , etc.) and it is paralleled by a similar development in the neuter plural. § 6 . Stems in y Sing. Xom. pitA, acc. pitdram , ddtaram , instr. pitra , dat. pitre , gen.-abl. pitur, loc. pitdri , voc, pitar. Du. N.A.V. pit dr an, datdrau, I.D. Ab. pitfbhydm , G.L. pitros ; PI. N. pitaras , daturas, Acc. pitfn , mdtfs , I. pitfbhis, D. Ab. pitfbhyas , G. pitfnAm , L. pitfsu. The fact that the stems in r are classed in Sanskrit as vocalic stems rather than consonant stems is due to certain develop- ments of Sanskrit which have tended to enhance their vocalic character. This appears particularly in the acc. and gen. plur., forms which are Sanskrit innovations. On the analogy of the consonantal stems the acc. plur, would have been pitrds , but this is replaced by a new form in -fn, based on the analogy of -an, -in, - iin . By this process Sanskrit creates a new long vowel f which has no phonetic basis among the inherited IE sounds. The old type of gen. pi. appears in Av. dugddram , etc. In San- skrit it is preserved occasionally in the Veda, e.g. ndram (: Osc. nerum ), gen. pi. of ndr- * man and once svdsrdtn. Else- where it has been replaced by the innovation -fndm, created by the same type of analogy on the pattern of -dndtn, ~indm, - unam . In the vrddhied nom. sing, the r is elided in the same way as the -n of n - stems (piti : raja) . This elision appears also in Iranian (Av. mala, etc.), Baltic (Lith. moil, sesuo) and Slavonic (O. SI. mati). In other languages the -r of the stem is preserved (Gk. fxrjTTjp, etc.). The acc. sg. has guna of the suffix in most of the names of family relationship (mat dr am, duhitdram, etc.), but in svasr- ' sister and in the agent nouns in - tr vrddhi appears which has been introduced from the analogy of the nom. sg. The same distinction appears in the nom. acc. du. and nom. pi. In the weakest cases the old type of inflection, with transference of the accent to the termination is normally pre- served, Elsewhere in IE this type is found in the conservative names of relationship (Gk. Trarpos* narpi, Lat. patris , etc.), beTHE DECLENSION OF NOUNS 244 side which there is found an alternative type with guna of suffix (Gk. ixrjT€pos beside p^rpos, cf . 7roLpevos t Osc. pater et f O. SI. materi , ef, agnaye ). Indo-Iranian alone preserves the primitive type in the case of the agent nouns in 4r. Elsewhere this has been replaced by innovating forms with guna or vrddhi of suffix in these cases (Gk. horijpos, Scoropos, Eat, datoris, etc.). In Sanskrit guna of the stem appears in the de- clension of nar- ‘ man ' (D. nare, G. ndras) as opposed to the older type of inflection seen in Greek (dvSpos, dvSpi). In this respect the Sanskrit /'-stems differ markedly from the adjectival i- and w-stems which keep the guna and accent of the suffix in the dat. and gen. sg. (agnaye, agnes). The difference between the two classes becomes less when the nature of the gen. sg. of /'-stems is examined. To agree with the form of the dative this would normally have been in -as with accented termination, and such forms are in fact found in Iranian (Av. brddro , dadro) as well as in other IE languages (Gk. rra rpos, etc.). The form which actually occurs (- ur , -us t -nh) goes back on the evidence of Iranian (Av. n?r?s) to -rs (Spiffs). Such a form with weak grade of both suffix and termination cannot be original and it must therefore be regarded as an innovation which has replaced something else. There is no way by which it could have developed from *pitrds if that had been the only form, and its origin is therefore to be sought in yet another type of gen. sg. which Iranian preserves : nars, zaoiars, sastars. This type, with which we may compare Lith. motcrs is of exactly the same for- mation as the gen. sg. of adjectival i- and w-stems (agnd-s) . It has arisen by the same process, i.e. by the extension of the accent and guna of suffix proper to adjectival stems to the gen. sg. and since it involves a reduction of the termination it must be ancient. In Sanskrit and partly in Avestan the - ars has been replaced by -rs (> Skt, -ur). The reason for this is that elsewhere in the weak and middle cases the suffix appears in its weak form ( pitrd , pitfbhis, pitfsu ), and this grade has been analogically extended to the genitive singular. No forms of the loc. sg. without ending are preserved, though such presumably existed at one time. This case always retains the guna of the suffix which is proper to it, in contradiction to other stems ( rijni , etc.) and the practice of other IE languages in nouns of this class (Gk. narpL , etc.). Of the old neuter nouns in -r such few as remain inflect THE DECLENSION OF NOUNS 245 heteroclitically, and they have lost connection with the masc.- iem. r- (r-) stems. On the other hand Sanskrit creates a new adjectival type of neuter in -r which has no prototype in IE. This differs from the masc.-fem. in the nom.-acc. as usual (( dhdtf , dhdtrni, dhdtfni), and also in the weak cases by inserting after the style of the neuters in i and u (dhdtrnd, etc.). It is not of frequent use. § 7 . Stems in i and u Sing. N. agnts, stinus ; vari t mddhu, Acc. agnim, sunum , I. agnind , sunilnd ; gatyd , dhenva, D. agndye , siindve ; pdtye , paive ; v arine, mddhune ; gdtyai , dhenvat ; Ab. G. agnds , sunos ; dvyas, mddhvas ; v Arinas, mddhunas ; gdtyds, dhenvas, L. agnau, sunau ; sdno, sanavi t sanuni ; gatydm, dhenvAm ; V. dgne, sfino. Du. N.A. agni t sunA, I. D. Ab. agnibhydm , sunubhydm, G. L. agnyds, sunvos ; varinos, ntddkunos , Pl. N. agndyas , sunavas ; ^ryas ; suct(ni), purU(ni). Ac. agnin , sunun ; dhentis ; aryds, pasvds , I. agnibhis , sunubhis, D. Ab. agnibhyas, sunubhyas, G. agnindm, sununam , L, agnisu, sunusu. The most ancient and fundamental division of these stems is between the neuters on the one hand and the masculine- feminines on the other. The latter two classes were originally identical in declension and the distinction between them which is observed in Sanskrit is a secondary development. On the other hand the distinction between the neuter and masc.-fem. types [mddhvas : agnds) which is caused by variation of accent goes back to an ancient period of Indo-European. The inflection of the neuters was effected by the addition of the normal endings, which in the weak cases originally bore the accent (Gk. Sovpos , yov vos), and in this respect it did not in principle differ from that of the consonantal stems. This type of declension was not originally confined to the neuters (see below, sdkhye t etc.), but the special type developed by the masc.- fem. class as a result of their suffixa! accentuation ( agndye , agnSs) was foreign to them. In Sanskrit the accent has become fixed on the root throughout the declension, mddhvas gen. re- placing *madhvds, as in other neuter stems. A few traces remain. The IE declension of the neuter w-stem meaning * domestic animal ' was of the type p6ku t pekwds . Corresponding to these forms Sanskrit has a neuter pdiu (once in RV.) and a gen. sg. THE DECLENSION OF NOUNS 246 pasvds which directly continue the old type, but pasvds lias be- come the gen, s g. of an analogically created nom. paius T while corresponding to pasu there has been created a dat. sg. pdive by the usual levelling process. As a result of the stabilising of the accent on the root in the normal type of neuters, the only type remaining in which the accent normally changed from the stem to the termination in declension consisted of those sufhxally accented masculines and feminines in which the vowel of the suffix was elided in the weak cases (murdhd : murdhnds, etc.). The result was that in certain cases, where an old neuter noun had preserved the terminational accent in the gen. sg,, a new sufhxally accented stem was created on this analogy, and with this change of accent was associated a change of gender : paius masc. for pain neut. Of the same type is pitus : pitvds * food The number of stems inflecting in this way in the Vedic language is very small, and in addition to neuters it includes some masc. and fern, nouns : avis, dvyas ' sheep krdtus , kr divas ‘ intelligence cf. Av. xratus, xraOwo . These may be regarded as transferred neuters. Such a development is easily understandable in the case of dvi- on account of its animate nature. The action nouns in -z and -it were originally, in accord- ance with their meaning, of the neuter type, but in general they have been transferred to the masculine and feminine classes. In doing this they have normally adopted the adjectival type of declension (tnatis : mates , etc.) but the neuter type has re- mained in a few cases as an indication of their originally neuter class. The masculines of this type use sometimes special forms of the nom. acc. plur. (nom. Av. pasvas ‘ cattle Mdzvb ‘ fingers ’, acc. Skt. pasvds) but also those of the normal type [paidvas, fiastin). The Vedic declension of the stem rayt-.rdy- is of this type (n. ray is : g. ray as, etc.). It represents earlier *ram-s ; ranyds. It is a transferred neuter of the dvi- type, and the termina- tional accent of the gen. sg. has effected a change of accent from root to suffix in the nom. sg., just as has happened in the case of paid-. Besides this there exists a root stem rd- corresponding to Latin res. In the classical declension this is combined with the ray- form of the z-stem. Similar is the declension of nails, navds ‘ ship ’ (Gk. vavs, vrjcr) for *nanu-s f *nanvds . In the only place where the nom. sg. occurs in the Rgveda it is proTHE DECLENSION OF NOUNS 247 nounced as a dissyllable : naiis ( navus ) and in this respect it can be compared directly with rayis, The long diphthong is a later contraction of these vowels ia hiatus. The acc. sg. is an innovation like the r&yam which supplants rayttn. The stem an- in which two different words have been con- fused (an- ' pious ', *ali- 4 alien, hostile ') inflects according to this type, although adjectival in sense. Since there are found elsewhere in IE examples of old neuters being adapted to this use without the usual modification of the stem, its character- istic inflection (gen. sg. aryas } etc.) might be explained by its being an old neuter of this type transferred to the masculine. If it were originally an adjective it would have to be assumed that it inflected originally like sdkhi (I. sakhyd, D. sdkhye ) and that from these weak forms the type of inflection characteristic of this class has spread to the nom. pi. (ary as like pasvds), On the whole the absence of any trace of the strong form of the suffix makes the first explanation more probable. The same doubt exists in the case of pdti-. In the meaning 4 lord ' this word follows the normal inflection (type agni -), but in the meaning f husband ' it forms cases after this style (dat. pdtye). The accent and the - n - of the derivative pdtni ' wife ' might be held to indicate an old alternating neuter ; on the other hand the nom. pi. is always normal and weak cases of this type are shown also by the adjectival sdkhi-. The stem jani- 1 woman ' has a gen. sg. jdnyus with a termination - us which appears also in sdkhyus, pdtyus and which is clearly borrowed from the nouns of relationship ( pitas , etc.). In all three cases the normal gen. ending -as has been replaced. Avestan has janyois, a compromise form replacing *janyas. This type of genitive inflection indicates that the stem is an old neuter transferred to the feminine. There existed an alternative way of inflecting the neuters of this class in IE, by the employment of the heteroclitic w-suffix In the few neuter i-stems that remain this n replaces the i- suffix ( dksi , aksnas) , so that these stems are in the main re- moved from this declension. On the other hand the neuter stems in -u add this -w- to the stem before the vocalic weak ter- minations : dat. sg. madhune, gen.-abl. mddhunas , sammas, drunas , loc. mddhuni , vastuni , nom. acc. du. janum ‘ knees gen. loc. du. janunos. Similar forms in other languages show that this practice is ancient (Gk. gen. sg. yovaros for *gonwnto$, 1 248 THE DECLENSION OF NOUNS Toch. du. kanwem ' knees '). It is probable that originally a corresponding extension -r could be added in the nom. acc. sg, whence the two types of suffix -ura/una- in derivative adjectival stems. In the early language these endings are, with very rare exceptions, used only with neuter nouns. In adjectives the neuter is not normally distinguished in form outside the non.- acc. In the classical language the n-forms are the rule for neuter substantives, but optional in the case of adjectives. The -n- has already in the earliest language spread to the instrumental singular of the masculines ( sununa ). Examples of this kind of inflection in neuter i-stems are ex- ceedingly rare (aksini ' eyes ') and the only common neuter that inflects in this way, vdri ' water gen. sg. vdrinas , does not appear in the earliest language. This is to be expected in view of the rarity of such stems, the existence of an alternative type of heteroclitic declension (dsthi/asthnds) , and the fact that the adjectival i-stems do not, any more than the corresponding 11- stems, distinguish the neuter in these cases in the early lan- guage (gen. sg. nt. bhUres). Nevertheless it may be assumed to be old from the existence of the alternating suffixes - ira-jina - and the fact that this -n- has already in the earliest language spread to the instr. sg. of the masculines. In the Veda the neuter nouns in -it may as a third alternative inflect according to the normal masculine type: gen. sg. mddhos, dros, snos, etc. This is an innovation which is elim- inated in the classical grammar. The neuter type of declension shows three types of loc. sg. in the w-stems which differ from the normal type of the masculine stems ( sunau ). (1) A locative without ending appears in sdno , vdsto . This type appears also in Iranian (Av. pdtdto, O. Pers. Babirauv , gddav-d), and it corresponds to similar formations from w-stems ( aksdn ), except that the characteristic accentua- tion of the loc. sg. has been eliminated. In sanavi this forma- tion is extended by the addition of locatival -i as has happened in aksdni , etc. In the Vedic language this type has been ex- tended to a small number of masculines ( dnavi , ddsyavi , etc,). (3) The locative may be made on the basis of the stem extended by -n- : sinuni , vastuni. This becomes the regular inflection in the classical language. The common masc.-fem. type, consisting originally of adjec- tives and agent nouns, but at an early period enriched by THE DECLENSION OF NOUNS 249 wholesale transfers of action-nouns from the neuter, was dif- ferentiated in IE from the. corresponding neuters by its sufffxal accent : pttu nt. subst, 4 much 1 (Goth.//«), pleu- adj. f much, many ' (Gk. nXees). On the basis of this adjectival stem there could be formed a vrddhied nom. sg,, uncharacterised by the termination -s, on the same pattern as in the n-, r- and s- stems : sdkha 4 friend 1 (stem, sdkki-) like pitd, r&ja. In Sanskrit this word, in which the accent may be presumed to have been secondarily transferred to the root, remains the only example of this type of formation from an i- stem. As a general rule the forms -is, -im and -us, -um are substituted in the nominative and accusative, forms which go back to an early period in Indo- European. The vrddhi which appears in the acc. sg, and nom. pi. ( sdkhdyam , sdkhdyas) is as elsewhere [dataram, etc.) an ex- tension of the form of the nom. sg. Guna was originally proper to these cases. Some forms of the acc. sg. with guna are pre- served in Avestan ( kavaem , frddat-fsaom) . Sanskrit has no such forms but it preserves the regular guna in the nom. pi. of the ordinary declension (agndyas, sundvas). The accented and gunated suffix could also appear in the dat. and gen. sg. (i agnaye , agnds) and this type has become the normal one in Sanskrit for masculines and feminines. Alternatively, on the analogy of pitri , etc., the accent, could appear on the termina- tion in stems of adjectival type, so* that in these cases their declension is not distinguished from that of the neuters. Apart from the shift of accent this type is preserved in the dat. sakhye ; and also in pdtye if this word is of adjectival origin. The Avestan declension of haxd 4 friend * corresponds in general to that, of sdkha (nom. s. haxd, acc. haxdim , dat. hase, etc.) thus establishing it as Indo-Iranian, In the gen. sg. this stem has been influenced by the names of relationship of the /--declension (sdkhyus after pitus), The old endingless locative has been re- placed by one in which the -y- of the dat. sg., etc., has been introduced ( sdkhyau ). A similar form appears in the case of pdti - 4 husband ' (pdtyau). Although the vrddhied nom. sg. which appears in sdkha is isolated in Sanskrit, signs are not wanting that it was originally more common in the i- and w-stems of the adjectival and agent- noun type. In Avestan the stem kavi » which has been nor- malised in Sanskrit still inflects in this way : nom. sg. kavd. In the acc. sg. this word has the original guna (kavaim, i.e., THE DECLENSION OF NOUNS 250 kavayam , cf, pi tar am), as opposed to the vrddhi in Skt, sdkhd- yam t Av. haxdim . In the gen. sg. in Avestan the gunated suffix normally appears, but once apparently it is inflected on the analogy of the heteroclitic neuters ( kzvino t cf. varinas). A vrddhied nom. sg. in -dus appears in Iranian in the case of some w-stems : Av. hiOdus 4 associate ', uzbazdus * with arms aloft' (and uyra°) ; 0 . Pers, dahyduL These have developed from the asigmatic vrddhied nom. sg. by the secondary addition of the nom. sg. -s. The adjectival nature of this form of termina- tion emerges clearly from the juxtaposition of uzbdzaui and bdzus ' arm In the acc, sg, we may have the old guna grade preserved (Av, dainhaom) or vrddhi extended from the nom. sg. ( 0 . Pers. dahydum, Av. nasdum ‘ spirit of the corpse '). The contrast between *the acc. sg. pasutn 4 domestic animal ' and frdiat- fsaom ‘ increasing cattle ' illustrates the adjectival nature of this kind of inflection. A similar variation appears between arstim ‘ spear ' and darzyd * arstaem ' having a long spear In the gen, sg. these stems have either the old type of inflection undifferentiated from the neuter {uzbdzvd) t the normal type with guna (daitjJizus), or by later innovation forms with vrddhi (nasdvo). In Greek the adjectives and agent nouns have separated into two types in the case of w-stems. The adjectives have, as in Sanskrit, adopted the endings -us, - urn (-vv) in the nom. acc. sg. 7 toAu£ : purus), at the same time preserving the original suffixal accent. On the other hand the agent-noun type [/ 3 a- cnAetfe, yovevs, povevs, etc.), have developed on the basis of the old vrddhied nom. sg. to which -5 has been secondarily added as in Iranian (-eu? for - rjvs ). The vrddhi is carried through the declension as in the case of certain other types of stem (f 3 acnr}(F)os, cf. Sorrjpos). To return to Sanskrit there is possibly one example of a formation parallel to sdkhd : aprata in RV. viii, 32, 16, nd sdmo _ aprata pape * Soma is not drunk without recompense ' ( prati cf,^ Lat. pretium). This is usually interpreted as loc. sg., but as non-adjectival compounds with a- are against the normal usage of the Vedic language, it is probably better taken as a nom, sg. interpreting the compound as a bahuvrihi The inflectional type of which Avestan preserves traces in examples like uzbazdus is preserved in Skt. dyaus * sky This is conventionally classed as a diphthongal stem, but, ts elseTHE DECLENSION OF NOUNS 251 where this classification is unsound. On the one hand the normal i - and w-s terns are themselves partly diphthongal {agnds, suno's ), and on the other hand this word is in part of its inflec- tion not diphthongal (gen. sg. di-v-as , cf. madh-v-as ). It is an adjectival u/eu stem dy-eu- with accent and guna or vrddhi of suffix according to the general rule. The addition of -s in the nom. sg. is secondary but of IE date (dyaiis : Ztvs). In the oblique cases of the singular there appears most commonly the undifferentiated type with accented termination (divas : Gk. At os) but also the special adjectival type with accent and guna of suffix (dyos : Av. dyaos). With this must be classed gaus 1 cow ' (dat. gave , gen. gds) whose accent and declension show it to have this adjectival suffix [g-6- f i.e. g w n z -eu, cf. Gk. fiouKoj). It inflects only according to the adjectival type and goes further than other ujo - stems in introducing the guna into the cases of the plural (instr, gobhis , etc.). In the acc. sg. the forms dyam, gdm appear to be from *dyaum , *gaum, with vrddhi from the analogy of the nom. sg. and elision of the final element of the diphthong before -m. The stem vi- 4 bird ' is an adjectival formation based on an old IE neuter *awi ‘ egg ' (whence with thematic extension Gk. <Lov) and the accentuation of the suffix has resulted in the total elisionof the radical vowel (as opposed to Lat. avis ) . In the RV. it has a nom. sg. vts with guna and nom. $. There is no parallel to this formation. An old nominative agnd(i) is traceable from the derivative Agnctyf ' wife of Agni and it appears also in the compound Agndvisnu 4 Agni and Visnu ' (cf. mdtdpitarqu ) . Similarly Manavi ’ Manu’s wife ’ is based on an old nominative *Manau. More important than this the vrddhied nominative of the adjec- tival type is preserved intact in the form that appears in the locative singular : agna(u), sunau. These forms are a special adjectival type of the locative without ending, being identical in every respect with the forms that originally served as the nom, sg. In that function they have normally been replaced by the termination -is, -us, but in their locatival function they have been retained. The terminations -is, -im, -us, -uni are therefore innovations in IE, and this accounts for the association of the accent with the weak grade of the suffix (purus). Nevertheless they are of considerable antiquity, and they have become the normal type THE DECLENSION OF NOUNS 252 in the various languages (Hitt, sallis ' great a$£u$ 9 good Gk. ttoXvs much Lith. lipus ' sticky Goth kaurus ' heavy etc.). Their origin is to be sought in the fact that in the mas- culine and feminine 1- and -w-stems two classes have coalesced. To the adjectival class consisting of adjectives proper (puru-, etc.) and nouns of adjectival type (. siinu etc.) there has been added a large class of action nouns (gdti-, situ etc.) which were transferred at an early period from the neuter class. This in- volved the adding of the case terminations -s, -m in the nom. acc. sg. on the analogy of the thematic stems, thus producing a type ending in -is, -im, -us, -um. In the amalgamation of the two classes in a common masc.-fem. declension, this type of formation in the nom. -acc. is generalised, but in the dat. and gen. sg. the formation proper to the adjectival type is general- ised {gates after agnds, etc.). The type of inflection prevalent in i- and n-stems in Sanskrit appears also in other IE languages : cf. gen. sg. Goth, anstais, sunaus , Lith. naktes , siinaus , dat. sg. O. SI. synovi. In Hittite and Greek the forms of the gen. sg. with non-reduced termina- tion (Hitt, sallaias , as taw as, Gk. o<f>eo$, iJSeo?) may be regarded as innovations replacing this type. In the locative singular the - au of the w-stems (sunau) has been introduced into the x-stems {agnail), but the Vedic lan- guage also has agni (he. agnd(i)). The locative without ending appears also in other IE languages (Goth, anstai , sunau, 0 . SL synu, etc.), but it is not possible to determine whether a long or short diphthong is represented in these cases. The special feminine terminations in the dat., gen.-abl. and loc. sg. {gdtydi, -as, -dm ; dhenvai, -as, -dm) are adapted from the t-declension. They are still rare in the RV. but become very common in the later pre-classical literature. The grammarians allow optionally in the case of feminine nouns either these ter- minations gdtyai, etc.) or the common masc.-fem. terminations (gataye, etc.). §8. Stems in d, %, u - d : Sg. N. sdnd, A. sdndm, I. sdnayd, D. sdndyai, G. Ab. sdndyds , L. sdndydm , V. sine ; I)u. N.A.V. sine, I.D. Abl. sdndbhydm , G.L. sdnayos, PI, N.A.V. sinds , I. sdndbhis , D. Ab, sdnabhyas, G. sdndndm, L. sdnasu. -f ; (A) Sg. N. vrkfs , A. vrkydm, I. vrkyd, D. vrkyd , Ab. G. THE DECLENSION OF NOUNS 253 vrkyds, Loc. gauri ; Du, N.A.V. vrkyd(u), I.D. Ab. vrkibhydm, G.L. vrkyds ; PI. N.A.V. vrkyds, I. vrkibhis t D, Ab. vrkibhyas, G. vrkindm, L. vrkisu. (B) Sg. M. devi, A. devhn, I. devyti, D, devyai, Ab. G. devyds, L. devyim, V. devi ; Du. N.A. hfetyaw], I.D. Ab. devibhydm , G.L. devyos ; PI. N, [^evyas], A. I. devibhis , D. Ab. devtbhyas, G. devindm , L. devisu . - u : Sg. N. tanus, A. tanvdm , I. tanvd, D. Ab. G. L. ianvt, V. /anw ; D. N.A.V. tanvdu, I.D. Ab. tanUbhydm, G.L. tanvds ; PI, N.A. tanvas, I. tanubhis , D. Ab, tamibhyas, G. tanU- ndm, L. tantisu. The nouns of these classes have in common an IE suffix -H (-ah, -t-H, -w-h> a, L w). With this suffix action nouns could be formed which originally did not differ from the usual neuter types (cf. the neuters in - as , -is, -us), and also adjectives which came eventually to be specialised as feminines. The dual type of inflection which appears in the f-stems is traceable to the dis- tinction of these two types, since wffiereas the vrki type inflects according to a system which may appear in both neuter and adjectival types, the devi declension contains inflections (devyd-s, etc.) of a specifically adjectival nature, in which the strong form of the suffix is due to the original adjectival accent. The stems of the vrki type consist of both action nouns (dehi ‘ rampart J ) and nouns of adjectival type, masculine and feminine (rathi- f charioteer vrki ' she-wolf ’). In the former the accent has been shifted to the suffix, where it remains throughout the declension, in the same way as has happened in the stems in -is (havis, havisas, similarly -in, -i'hos). In the adjectival class the strong form of the stem (*vrkyd), whose original existence is to be assumed on the basis of the accentua- tion and of the general system {pitd, etc.), has been replaced by the weak form, a process for which a parallel is to be found in the m-stems [bait, balinas). As a result of these processes, and becahse of the acquisition of feminine gender by action nouns with suffix -H, the two classes become completely fused to- gether in declension. The same two types are found among the i?-stem$ ((1) tanU -

• body ', (2) svasrti- ‘ mother-in-law ') and their fusion has pro-

ceeded in the same way. Since the -s of the nom. sg. was originally not characteristic of the H-stems. from which it remains absent elsewhere, its THE DECLENSION OF NOUNS 254 existence in these two classes must be regarded as an importa- tion from the root stems in i and u which are declined in the same way ( dhis , dhiyds , bhils, bhuvds) . For the rest the declen- sion is of the normal consonantal type calling for little comment. The stem and ending are pronounced as separate syllables in the Veda (tanuvam, etc.) although written according to the later system (tanvdm, etc.). The normal type of loc. sg. appears in catnvi , tanvl, etc., the endingless variety in camu, etc. The few locatives in -t of the f-s terns could either be the result of con- traction (<*iyi <*iui) or be locatives without ending. In the gen. pi. -n- has been introduced on the general analogy of the vocalic stems. The devi type is the one that normally appears in the feminine of non-thematic and some thematic stems (rdjnl, datri, prtkvf, kalydnl). It is thus predominantly an adjectival suffix, and although the accent of stems of this type has become variable in Sanskrit, the suffixal accent which frequently appears maybe considered to be the more original type. The strong forms of the suffix, which are to be explained by this adjectival accent, appear in the dat., abl.-gen. and loc. sg. In the nom.-acc. the weak form of the suffix appears, so that there exists an alterna- tion here parallel to that which appears in the i- and w-stems. Theoretical considerations indicate that the weak form of the nom.-acc. sg. (and of the nom. pi. following suit) are innova- tions, just as the similar formations in the adjectival i- and u- stems, and the related languages provide some evidence that this is so. This is clearest in the case of the acc. sg. which for phonological reasons cannot be original, since these stems were originally consonantal and *-inam could only produce -iyam (-yam). In Balto-Slavonic and Germanic the strong form appears in the acc. sg. (Lith. ncfusia, O. SL nesusa, Goth, bandja, etc.) and there is no reason to believe that these forms are in- novations.. On the other hand these languages have the weak stem in the nom. sg. (Goth .frijondi, Lith. vezanii, O. SI. vezasti) a fact which indicates that the weak form was earliest established in the nom. sg. Greek on the other hand has *-ya in the nom. sg, (ttotvmx, pepovoa, fila), and this cannot be phonetically equated with the l of the other languages since IE -in- develops into t in Gk. as elsewhere. The final -a here as elsewhere (wfipa, etc.) represents IE -an and the short vowel, as opposed to the long a elsewhere, arises from the pre- vocalic sandhi of this combinaTHE DECLENSION OF NOUNS 255 tion. It is clear that the accent of fiia is not original since it rests on a weakened syllable and it follows that the accent of the oblique cases (fuas) must have originally prevailed in the nom. sg. also (*sMidn). The distinction between the two types of declension of the f-stems, which is strictly observed in the Vedic language, is not retained in later Sanskrit. The devt inflection is preserved at the expense of the vrkt inflection, but it adopts the inflections of the latter class in the nom.-acc. du. and nom. ph : devyau , devyds. Among the u-s terns inflection of a type parallel to that of the devt stems is exceedingly rare in the RV. In the later language it becomes the normal type ; vadhvat, vadhvds , vadhvdm, the development running parallel to that of the t- stems. The fusion of the termination and suffix in certain cases {devyai, devyds) produces the special feminine terminations which are later applied to the feminine i- and u-stems. The loc. sg. has a special termination -m. This -m is absent in Iranian (O. Pers. haranvatiya ), which shows that the loc. was originally without termination. The secondary addition of -m is paralleled elsewhere : tubhya : tubhyam ; instr.-dat.-abL du. -bhyam : Av. -by a. The stems in -d have been influenced in declension by the stems in -t“. The cases of the singular from the dative onward are formed by the addition of -yeti, -yds, -yam which have been abstracted from the devt declension. This peculiarity is shared by Iranian : Av. dat. sg. daenaydi , etc. The older IE endings are preserved elsewhere : gen. sg. Gk. ^O. Lat. vids, Goth, gib os, etc. §9. Stems in - a Sg. N. devds , A. detain , I. devena , D. dev ay a, Ab. devat , G. devdsya, L. deve, V. diva ; Du. N.A.V. devau , LD. Abl, devabhydm , G.L. devdyos ; PL N. devds , A, devdn , I. devais, D. Ab. devibhyas , G. devandm , L. devisu. Neuter N. A. yugdm, Du .yngi, PI .yngdni. The stems in -a are the most numerous type in the language (45 per cent of all nominal stems in the Rgveda ) . They are char- acterised by the absence of any shift of accent in declension, and this seems always to have been the case. Thematic stems are either masculine or neuter, and these differ in declension only 256 THE DECLENSION OF NOUNS in the nominative and accusative. This declension contains some special features which may be briefly summarised. It has been considerably influenced by the pronominal declension. In the instr. sg. the termination -ena has been taken from that source. The older termination -a still exists in the Vedic lan- guage, though it is a good deal less common than -ena. In Avestan only the ending -a appears. The Indo-Iranian termin- ation of the dat. sg. was ~di , a contraction of the stem vowel and the termination -at, and this was inherited from Indo-European (-0 + ei) : Av. haomdi, Gk. lttttqj, etc. To this a postposition a might optionally be added as in Av. ahurdi a and in Skt. this element has become permanently attached, producing the extended termination -dya. The ablative sg. which is distin- guished from the gen. sg. in this declension alone, is inherited directly from IE (O, Lat. - dd , etc.). There exists no common IE form of the gen. sg. In Balto- Slavonic the old abl. sg. functions also as gen. sg. and in view of the identity of the two cases elsewhere this could be ancient. Italic and Celtic have an ending -f, which is the adjectival suffix -t substituted for the thematic suffix. This t appears in Sanskrit in constructions of the type saml-kr * to make even In Hitt, the gen. sg. of the thematic class is equivalent to the nom. sg. There exists in the Veda a small number of compounds like rathaspdti which possibly contain such a form of gen. sg. A form corresponding to the Sanskrit termination appears in Greek and Armenian (Gk. <no, Arm. - oy ). A similar formation, but without the -y- appears elsewhere : Goth, wulfis (<°eso) i O. SI. ceso ‘ whose These terminations may be presumed to have originated in the pronominal declension, as has happened elsewhere. The elements -so and -syo which are thus added to the stem appear to be demonstrative pronouns of that form. The loc. sg. is analysable into the stem vowel and the normal termination -i : cf. Gk. olkoi, etc. The nom, pi. -as (a + as) appears in a similar form elsewhere (Goth, wulfos , Osc. nuv - lanus ), but in its place an ending -oi, derived from the pronouns is also frequent (Gk. Xvkoi , Lat, lupt, Lith. vilkai, Toch. B. yakwi). The ending -as is in the Vedic language sometimes pleonastically extended to - dsas , a feature which is also ob- served in Iranian (Av. -dyhd). This innovation is again abol- ished in classical Sanskrit, but it lives on in some early Pali forms {pandiidse , etc.). The acc. pi. has acquired its long vowel THE DECLENSION OF NOUNS 257 from the nom. pL (original form in Gk, (dial.) Au/cov?, etc.). The gen. pi. has acquired its -n- from other classes, the innova- tion being common to Indo-Aryan and Iranian (Av. masyanam, 0. Pers. bagandm ). The original termination is preserved only in the phrase dev An jdnma 1 the race of gods ' ; cf. Av. staoram, Gk. decov, Lat. deum, etc. Two forms of the instr. pi. appear in the Vedic language, in ais and in - ebhis . In Iranian Avestan has -dis, 0. Persian -aibiL Elsewhere there exist only forms corresponding to -ais ; Gk. AyVotff, Lat. lapis, Osc. nuvlanuls, Lith. vilkais. This form of the case has ’the appearance of being a pluralisation of the form that appears in the singular as dative. If so it must go back to a time when the cases were less differentiated than they became later. The ending -ebhis is an Indo-Iranian innovation after the pronominal declension. The innovation is later eliminated in classical Sanskrit, but it is the basis of the Middle Indo-Aryan forms of this case (: Pali - ehi , Pkt. -ehi(m)). The -e- which appears here is from the form of the pronominal stem which appears in most of the cases of the plural (nom. pi. ti, etc.). It also appears in the dat.-abl. and loc. pi. of this de- clension {-ebhyas, -esu ) . CHAPTER VI NUMERALS, PRONOUNS, INDECLIN ABLES §i. Numerals The Sanskrit numerals from i-iooo are inherited from Indo- European. They are constructed on the decimal system, the numerals i-io being the foundation of the whole series. The numerals from i-io are adjectives, as also 11-19 which are com- pounded with ddsa ' ten The higher numbers are properly collective nouns, though a tendency to treat them as adjectives appears as the language develops. The numbers from 1-4 are fully inflected in the three genders ; those from 6-10 are defective and appear originally to have been uninflected. 1. Two roots appear as the basis of the numeral 1 in Indo- European, of which the one ioi-) appears to have had the mean- ing of ' alone the other {sent-) that of ‘ together In Sanskrit dka- f one ' is formed from the first root with the suffix - ka , and it is declined according to the pronominal declension. The only form exactly corresponding is the Aryan aika- which is found in the Hittite documents. Elsewhere different suffixes appear ; Av. aeva Gk. oto? f alone 1 ; Lat. unus, O. Ir, din, Goth. ains. The root sent - provides this numeral in Greek (efe, ^ua • *sems,

• smia) and Tocharian (A. sas, B. seme). This root appears in

Sanskrit in its reduced form (sm-) in sakft 1 once ' (: Av. hakdrdt), and in compounds of the type samanas- 1 of one mind, the same mind The adjectives samd samiind- f same ' are derived from it. 2. The numeral dvau (: Gk. SJo>, Lat. duo, O. SI. dva , etc.) is like the normal dual of an tf-stem : N.A. nt. dvd (; O. SI. dvi), I.D. Ab, dvdbhyam, G.L. dvdyos (: O. SI. dvoju) . The uninflected thematic stem appears in the Gk. variant 5 vo. There is an alternative stem formed by means of the suffix -i, which appears in compounds {dvi pad- * two-footed Gk. Sittovs, Lat. bipes) and in various derivatives (dvittya- 1 second etc.). Beside the numeral proper there is a stem ubhd- ‘ both ' which inflects in the same way. The exact nature of its relation to Gk. 258 NUMERALS, PRONOUNS, IND ECLIN ABLES 259 Lat. amho, Q. SI. oba, Lith. abu t Goth. bai } Engl, both , etc., is not altogether clear. 3. The stem tri- contains a suffix -i which is absent in the ordinal trtiya It is inflected like a normal adjective in -i (nom. pi. trdyas : Gk. Tf>eZ$ } etc.), except in the gen. pi. where trdydnam has replaced an earlier tray dm (: Av. dray am, with guna from the nom. pi. as opposed to Gk. rpitiv, etc.). It pre- serves the alternation of accent in declension : instr. pi. tribhis as opposed to vtbhis ‘ with birds 4. This numeral is formed on the basis of a root k w et- which seems originally to have meant something like 4 angle ’ (cf. Lat. triquetrus 4 triangular '), whence 4 square ' and from that 4 four In the masc. and neut. (catvdras, catvari, Lat. quattuor, etc.) the stem is formed by means of the suffix -i jar, with adjectival accent and vrddhi in the nominative. In the other cases (acc. caturas , etc.) the suffix has the weak form according to the general rule. A neuter noun *cdtvar, or its IE prototype, is pre- supposed by the thematic extension catvara- 4 square, cross- roads Elsewhere the simple r-suffix may appear (Gk. Dor. TeVopcy, Lat. quater), or the elements of the suffix may be reversed (Av. caOruA. The feminine of these two numerals (nom. acc. tisrds , cdtasras : Av. tisro, tatarjrb) is made by means of the suffix - sar , which elsewhere (see p. 141) shows a tendency to become a specifically feminine suffix. In tisrds the -r- of tri- has dis- appeared through dissimilation. The common form of the nom. acc. is in origin accusative. The original forms of the nom. to be assumed for IE, *tisores , *k w etesores t are continued in Celtic : O. Ir. teoir, cetheoir. The numerals from 5-10 have a less developed system of in- flection than the preceding ones. With the exception of the dual ending of astau 4 eight they do not inflect in the nom. acc. In the Vedic language they may appear uninflected also in the other cases : pane a ksitisu 4 in the five tribes sapid hotrbhih 4 with seven priests etc. In Greek and Latin the correspond- ing words are uninflected, also in Germanic when used attribu- tively. This is the oldest state of affairs. The beginnings of inflection may be put in the late IE period, the uninflected type surviving by the side of the new inflected type down to Vedic times. 5. In pahea (: Gk. TreWe, Lat. quinque , etc.), representing IE 2&0 numerals, pronouns, indecunables peyk w e, we find an uninflected thematic stem. As such are to be assumed, from the evidence of compounds, etc., to have existed freely at an early stage of IE, this is an interesting archaism. The root peyk xc - is in all probability that which appears in Engl. finger {fiperjk^rc-) and fist ( <*pyk^sti- t cf. O. SL p$sti ' id '). The derivative panktt- (: O. Sl. p$tt 1 five ') means from the Rgveda on, not only ' group of five but more generally ' group, series 6. Skt. sat ( <*sats <sats with assimilation of initial) repre- sents IE *seks : cf. Lat. sex , Goth, saihs , etc. The forms of this word in Indo-European are rather complicated, since there also exist forms beginning with sw- (Welsh chwech ), ks- (Gk. £d<7Tpi£ ‘ in rows of six O. Sl. sesti ‘ six '), ksw- (Av. xsvas) and simple w- (Arm. vec , O. Pruss. uschts 1 sixth ’). The original initial con- sonant group has been simplified variously in the different lan- guages. Middle Indo-Aryan cha goes back to an original differ- ing from the Sanskrit form, and beginning with ks-. 7. The final - a of saptd (: Gk. i-nra) represents IE -rji, as is clear not only from Lat. septem but also from the ordinal saptamd The agreement between Greek and Sanskrit shows that the final accentuation existed already in IE (septrfi), but it cannot be original since it appears on a reduced syllable. 8. In astau (: Gk. oktw, Lat. odd, Goth, ahtau, etc.) there appears the termination of the dual. The meaning of the stem okto- of which this is the dual may be inferred from a related i-stem, a£ti-, which is found in Avestan. This is a measure of length meaning ‘ the width of four fingers from which it may be inferred that the dual *okt6(u) meant originally ' two groups of four fingers In classical Sanskrit there exists beside this a form asta with short a from the surrounding numerals. 9. The - a of ndva f nine ' (: Gk. ir-vea) may go back to -m (cf. navamd-, Lat. novem) or possibly -« (cf. Lat, nonus 1 ninth '). The word has been considered to be related to IE niwos ‘ new which in view of the formation of the preceding numeral is not without plausibility. 10. The primary series closes with ddsa * ten ' (: Gk. 8cKa t Lat decent , Goth, ialhun, etc.) representing IE dikrjx, about the etymology of which no likely suggestions exist. 11-19. These are dvandva compounds of 1-9 with 10 : dvidasa * 12 trayodaia 1 13 *, cdturdaia * 14 sddate * 16 etc. ; cf. Gk. SoiSeKa, Lat. duodecim •, etc. The long vowel of NUMERALS, PRONOUNS, INDECLINABLES 261 dkddasa ' 11 is from dvtidasa. The numeral 19 may alterna- tively be expressed ekonavimiati * 20 less I The tens from 20-90 are feminine substantives and as such decline properly in the singular, with the gen. pi. of the things enumerated ; navatim navydnam ' 90 navigable streams *, etc. But they may also be construed either (1) agreeing in case, but not in number, with the noun enumerated, vimsatyd harihhih

• with 20 bay horses b or even (2) adjectivally, agreeing also in

number with it : pancdsadbhir banaih * with 50 arrows The intervening numerals are constructed like those from 11-19 : trdyastrimsat- ' 33 ', cdtuhsasti ‘ 64 etc. 20-50 : vimsati- ; cf. Av. vrsaiti, Gk. *likogl, Lat. viginti, etc. trimsdt : cf. A. Orisas , acc. sg. drisatem, Lat. trigintd, etc. catvdrimiat ; cf. Av. caOwarzsatdtn (acc. sg.), Gk. rerrapd- kovtcl, Lat. quadrdginta, etc. pancasdt- : cf. Av. pancdsat-, Gk, TrevTrjKovra, Lat. quin- qudginta, etc. The element -sat- which appears in these four numerals is out of kmt -, which further stands for dknit-, a reduced form of the numeral 10, with the ^-suffix that appears in Skt. dasdt - ' de- cade The reduced form belongs properly to the w r eak cases, since Av. Orisas 1 30 and visas 1 20 show that the strong stem {-iant-) was originally used in the nom. sg. This stem was originally neuter and could be inflected as such along with the preceding numeral (Gk. rerrapa- kovtcl 1 4 tens etc.). In San- skrit the neuter pi. inflection appears in the first member in catvdri-m-sdt and pancd-sdf-. Its absence from the second member is explained by the fact that this inflection could be optionally omitted (cf. visvdni vdsu, etc., p. 237). The f-stem vimsati- was originally a dual *vi salt ‘ 2 tens The vi which appears here in the sense of f 2 ' is the normal word in Tocharian in place of the usual dvau , etc. (A. wu t we (fem.), B. wi). It may be identified with the prefix vi- ' apart, separate The presence of the nasal at the end of the first member is unexplained, and outside Indo-Aryan it appears only in Oss. insdi * 20 60-90. sasti, saptati-, atitt- , navati-. These are formed in a manner quite different from the preceding. They are abstract or collective nouns formed by means of the suffix - ti, meaning primarily ‘ hexad ", etc. The primary meaning is retained in the similarly formed pankti- ‘ group of 5 and O. SI. Se$tl % which corresponds in form to Skt. sasM-, means simply 1 6 The 262 NUMERALS, PRONOUNS, INDECLIN ARLES specialisation of these forms as names of the tens is common to Indo-Aryan and Iranian, cf. Av. xsvasti haptditi -, astdili - t navaiti -. The for m of the numeral 80 in Sanskrit is an inde- pendent formation from the root which is not paralleled else- where but which is obviously ancient. It has been replaced in Iranian”By a n o rm al i se 3 “ form based on aHd. 100. The original form of satdm ‘(Av. satdm, Gk, e-Karov, Lat. centum , etc.) w T as *kmtdm, as is clear from Lith simtas. In view of this, and in view of the fact that the -sat- in trimsat , etc., means ' ten the IE original is derived from dkmtom , a neuter collective noun meaning ' a decad (of tens) 1000. sahdsra- (: Av. hazayra-) is connected with Gk. XeiXiot, ^eAAtot (*^€<jAtot) r and the initial element sa- is identi- fied with IE sm- ‘ one, together These two numerals are neuter substantives, but the same variations of construction are found as with the tens. The numerals above a thousand are purely Indian : ayuta - To,ooo', laksd- ‘ 100,000 prayuta- 1 1,000,000 koti- ' 10,000,000 etc. The series is carried to great lengths, particularly among the Jains and Buddhists ( asamkhyeya - = io 140 ), but among the higher numbers there is little agreement in the names between the various texts. Ordinals : 1 prathamd- (Av. fratzma-), 2 dvitiya - (: Av. daibitya bitya-, 0 . Pers. duvitiya -), 3 irtiya - (Av. Or tty a-, Lat. tertius, etc.), 4 (a) caturtha- (Gk. rerapros, Lith. ketvirtas , etc.), (b) turfy a-, turya- (Av. tuirya-), 5 (a) pakthd- RV. 10, 61, 1 (Av. puxha-, OHG. funfto from pyk w t[h)o- : Gk. Trepirros, Lith. penktas, etc. , from peyk w to-) , (b) pancathd -, Kafh. (: Gall.pimpetos, 0. Welsh pimphet), (c) normally pahcama- (Pahl. patijum , Oss. pdnjdm), 6 sasthd- (Gk. <ektos k Lat. sextus ), 7 (a) saptatha-, RV. (Av. haptada-), (b) normally saptamd- (Pers. haftum , Gk. efl&opos, Lat. Septimus , etc.), 8 astamd- (Av. asthma-), gnavama- (Av. naoma- } O. Pers. navama-), 10 dasamd (Av. dasdma , Lat. decimus , etc.). The simplest type of ordinal is made by adding the accented thematic vowel to the numeral in its usual adjectival function : saptam-d- } dasam-d . From these ordinals the formation is further extended to cases where it did not originally belong : astamd as opposed to the more original Gk. oySoo?, Lat. octdvus ; navama- as opposed to Lat. nonus . The suffix -thd, 1. e. -t-n-d- must have originally arisen through the addition of NUMERALS, PRONOUNS, INDECLINABLES 263 the accented thematic vowel to a collective or abstract deriva- tive in -id, i.e. -tan (*caturta + d> caturtha-) . There is no evidence of an h outside Indo-Iranian : Gk. Trlpirros, etc., are formed on the basis of the simple t- stem. The first ordinal differs in the various languages (Gk. irpwTos, Lat. primus , Lith. ptrmas , etc.), but they agree in deriving it from the same root meaning ' in front and not from the corresponding cardinal. The ordinals from 11-19 are thematic formations with accentuation of suffix : dvddasa etc., cf. Av. dvddasa etc. The suffix -tama which serves also as a superla- tive suffix is used to form ordinals from the tens, 100 and 1000 : trimsattama- (Av. Orisastema -, Lat. trtce(n)simus) ‘ 30th r ), §astitama - ' 60th satatama- ‘ 100th sahasratama- ‘ 1000th ' {Av. hazarjrd * t?ma -). Alternatively there exists for 20-50 the type vimsd-, trimsa-, catvdrimsa pancasa-, to which nothing corresponds outside Indo- Aryan. They are formed analogically on the pattern of ekddasd etc. Of adjectival derivatives other than ordinals dvayd- (Gk, Soto?, O. SI. dvojt) and trayd- (O. SI. troji), meaning ' of two (three) sorts, parts 1 are inherited. Sanskrit has created by means of the suffix -taya a series ikataya cdtustaya etc., used in the same sense. Old adverbial derivatives are dvis (Av. biS t Gk. Sts Lat. bis) and iris (Av. dri§ } Gk. rpls) ' twice, three times It is not certain whether catuh ' four times ' is the simple stem catur used adverbially, or whether it is from

• catur-s, with the addition of this adverbial -s. Av. caOruS

1 four times ' has such an -s (though the elements of the suffix are arranged in different order), but it is not necessary to assume its original presence in Lat. quater. Elsewhere this meaning is expressed by the use of kftvas ' times ' : pdnca kftvah f 5 times, etc/. The root of this word appears in the compound sakft ' once and in Lith. kartas, O, SI. kratu. Other adverbial formations are made by means of the suffixes -dhd (tridhd 1 in 3 parts *, etc.) and -sas (satasds * in hundreds '^etc. §2. Pronouns Personal Pronouns 1. Sg. N. aham } A. mam , ma, I. mdyd t D. mdhyam , me , Ab. mat , G. mdma t me, Loc. mayi. Du. N.A.V. avdm , I.D. Ab. avabhyam , G,L. dvdyos , and A.D.G. nau. 264 NUMERALS, PRONOUNS, INDECLIN ABLES PL N. vaydm, A. asmdn, nas, L asmabhis, D. asmabhyam , nas, Ab. a smut, G. asm&katn, nas , L. as mas u. 2. Sg. N. tvdm, A. tvdm, tvd, I. tvdyd , D. tubhyam , te, Ab. toa/, G, /aua, te, L. /u&yi. Du. N.A.V. yuvim , I.D, Ab. yuvdbhydm, G.L. ynwayos, and A.D.G. uatw, PL N. yuydm , A. ywsmaw, vas, I. yusmdbhis, D. yusmdbhyam , vtfs, Ab, yusmdt, G. yusmdkam, vas, L. yusmdsu. The inflection of the pronouns differs in many respects from that of the nouns, and this difference is most marked in the personal pronouns. (1) These pronouns show no difference of gender which is in accordance with the earliest IE practice, and contrasts with the practice of other language families (e.g. Semitic) in which gender is distinguished. Only an isolated yusmas acc. pi, fern, is quotable from the Vedic texts. (2) The distinctions of number are expressed by the use of different stems, which contain different radical elements. (3) The same distinction appears in the first person between the nom. sg. and the other cases. (4) The terminations of the plural are partly identical with those of the singular, and this was much more marked in the prehistoric period. (5) The individual terminations differ widely from those of the noun. The nom. sg. ahdm (Av. azzm) contains a suffix -am which is elsewhere prevalent in the declension of the personal pronouns. The form is found also in Slavonic (Q. SI. azu with u<-om). Elsewhere there appear forms without ending (Lith. es , as) and forms terminating in -b (Gk. eycb, Lat. ego). The latter form was originally egoii, and the aspiration in Sanskrit shows that it was to this form that the -dm has been added (egon +o'm> egu-om). The -am of tvdm (Av. twzm, turn) is not found outside Indo* Iranian and is therefore of more recent origin. The other lan- guages have tu which also survives in Iranian, and possibly also in the Vedic particle tu (cf. RV. 8, 13, 14 d tu gahi, prd tu drava ). In the other cases of the first person there appears a stem beginning with m-, before which on the evidence of Greek (ifjLe acc. sg.) and Hittite (ammuk acc. dat. sg.) a vowel has been lost in most of the IE languages. On the other hand the pro- noun of the second person does not differ radically in these cases. The base tl which appears elsewhere (O. SI. Skt. tvdm, etc.) may have arisen out of *twe by sporadic loss of -w- after initial t-. The final -m in the acc. sg. forms mam, tvdm NUMERALS, PRONOUNS, INDECLIN ABLES 265 (: Av. mam, 8wqm) is found outside Indo-Iranian only in Slavonic (O. SI. mg, t$). It is an innovation replacing the older forms which are preserved only as enclitics : md, tvd (: Av, md t dwd). The relation of these forms with long vowel to the forms with short vowel like Gk. €fxd, fit, al, ac is not clear. The enclitic forms me, te (: O, Pers. fnaiy, taiy , Av. moi, me, toi , te, Gk. fioi, trot, Lith. mi, ti) are forms of stem and contain no case ending. As such they are more ancient than the accented forms which have evolved a full case system. It is also an ancient feature that their use is much wider than that of a normal case form. They are used regularly in the sense of both genitive and dative, and occasionally even more widely. Originally there must have been corresponding accented forms with similar wide use. The growth of a full system of in- flection for the accented personal pronouns has abolished these, but the older undeveloped system is preserved in the enclitics. The lateness of the fully inflected case forms is shown by the fact that a number of them have no exact correspondences in other IE languages. Such is the case with the instr. sg. mdyd . This is formed on the basis of the stem form preserved in the enclitic me . On the other hand the original form of the instr. sg. of the second person was tvA, as is shown by the agreement of this rare Vedic form with Av. 6wq. It is replaced by tvdyd formed on the pattern of mdyd . The same form of stem is the basis of the loc. sg. mdyi. The original loc. sg. of the second person is tvd which appears in the RV, It is replaced from the AV. on by the analogical tvdyi. In the absence of Iranian evid- ence it is not possible to say anything about the earlier history of this case. The forms of the abl. sg. mat, tvat (: Av. mat , Swat, cf. O. Lat. mid, ted) are formed with the same element that appears in the declension of thematic stems. A form mdmat , influenced by the gen. sg. appears in the RV., and later the extended forms matids, tvattds, formed with the ablatival suffix - tas , come to be frequently used. The forms of the dat. sg. are mdhyam, tubhyam, but tubhya is attested in the RV., and both mahya and tubhya are frequently required by the metre. These are the earliest forms and the -m is a secondary addition of Sanskrit, as elsewhere. The -w is absent in Iranian : Av. maibya, tnaibyd, taibyd, taibyd. The final element d found in Indo-Iranian is 266 NUMERALS, PRONOUNS, INDECLIN ABLES absent elsewhere (Lat. mihi, tibi, Umbr. mehe, tefe, O. SI. tebS, O. Pruss. tebbei ), and it is therefore to be regarded as a post- position which has become attached to the original case form. The h ( <gh) in Skt. mdhyam) is shown to be original as opposed to the -b- in the Avestan form, because it appears also in Italic. The -u- of tubhyam is peculiar to Sanskrit ; all the other lan- guages have forms derived from original te-. The gen. sg. tava (: Av. tava) representing IE *tdwo is an uninflected thematic adjectival stem, and therefore an isolated survival of archaic IE usage. Elsewhere this stem is found inflected as a full adjective : Gk. red?, Lat. tuus . On this analogy mama may be explained as a substitute for *dma f the initial m- being intro- duced from the rest of the paradigm. This *ama would corre- spond to the stem of Gk. ifios * my ' in the same way as tava to reos. Such an IE form of gen. sg. is represented in Armenian im . The gen. sg. in Iranian (Av. mana, 0. Pers. mana) which corresponds exactly to that of Slavonic (0, SI. mene) is a dif- ferent formation, containing an w-suffix which in Germanic is attached to the diphthongal base (Goth, meina ). The nom. pi. vaydm (: Av. vaem) contains the same additional element -am that appears in the nom. sg. It points to an original IE wei which appears in Gothic with the secondary addition of the plural -s [wets, cf. also Hitt. well), This -s appears also in the nom. pi. of the second person in Av. ytil, Goth, jus, Lith. jus. A second form yuhm exists in Avestan with the addition of the common pronominal termination -am. In Sanskrit yuydm the -am is added to the stem without -s, with the intervention of a -y- taken from vaydm. The cases of the plural from the accusative on are made from the bases asmd- and yusmd - to w T hich corespond exactly Gk. (Aeol.) a/i/xe and v/x/xe. The as- of the first person is for ns - ( = Goth, uns) which is explained as the weak grade of the form of the pronoun which is used as an enclitic, nas . It is possible that in yusmd - the initial y has been secondarily introduced from the nom. pL, and that an original *us- was in the same way the weak form of vas. The stem extension resembles that in the masc. sg. of the demonstrative pronouns [tdsmat, etc.). These bases originally d:ook the inflections of the singular, and the introduction of the plural termination is of comparatively recent origin. The old state of affairs is still preserved in the dative and ablative : asmabhyam (: Av. ahmaibyd ), yusmdNUMERALS, PRONOUNS, INDECLIN ABLES 267 bhyatn (: Av. yuimaibyd ) ; asmdt (: Av. ahmat ), yusmdt (: Av. yusmat, xsmal ) . For the accusative the simple uninflected base was originally used as in the Greek forms quoted above and in Avestan dhmd, ahma. The introduction of the termination -an is an innovation of Sanskrit. The original (singular) ending of the instr. pi. is preserved in Av. xsmd, and also in Vedic in a few compounds, like yusma-datta - ‘ given by you The long -d- of asmabhis, yusmdbkis is derived from this form of instrumental, and from here it has been introduced into the loc. pi. : asmasu , yusmdsu . The forms of the gen. pi , asm akam, yusmakam (: Av. ahmaksm-, O. Pers. amdxam ; Av. yusmakzm) are based on the adjectival stems asm aka- and yusmaka- which still function as such in the RV. The original form, which still occurs occasionally in the Vedic language, was the untnflected stem as in tava, and the -m as elsewhere is a later addition. An earlier form of the locative, asmS, yusmi appears in the Vedic texts, with the peculiarity that it can be used also as dative and genitive. This archaic characteristic of combining the meanings of several cases indicates that the forms are ancient, though nothing exactly parallel is quotable elsewhere. Only one form of enclitic appears in the plural, nas , vas , which serves as acc. dat. gen. In Avestan the corresponding no, V9, vo, serve only as dat. -gen. while for the acc. there appear forms with a long vowel, nd, va. These correspond in form to Lat. nos , vos and 0. SI. ny, vy. In classical Sanskrit three case-forms are distinguished in the dual of the personal pronouns, as elsewhere. In the Vedic language more numerous distinctions are found. The nomina- tive dvdm, yuvdm are distinguished from the accusative avdm , yuvdm . The ablative forms dvdt and yuvdt appear. A separate instrumental is attested by compounds of the type yuva-datta-

• given by you two The uninflected adjectival stem yuvaku is

found functioning as gen. du. In Avestan there exists a gen. du. yavakdm more closely parallel to the forms of the plural. The forms of the dative and locative, yuvdbhydm , yuvds are replaced by more regularised forms in the classical language : yuvd- bhydm , yuvdyos. The nom. du. yuvdm is formed from the same radical element as the nom, pi. The unextended form is seen in Lith. jit-du 4 you two For the first person a nom. du. vdm is attested once 268 NUMERALS, PRONOUNS, I ND ECLIN ABLES in the RV. (6. 55. 1). This corresponds to Av. vd , the -nt being obviously a secondary addition of Sanskrit. Av. vd corresponds exactly to O. SI. ve, and from these forms Goth, wi-t and Lith. ve-du differ in having a short vowel. All these forms contain the same radical element as the nom. pi. A form corresponding to dvarn is found only in Av. wdvd (acc.), and there is nothing similar in the rest of IE. The most plausible explanation of this formation peculiar to Indo-Iranian is that a dual d of the pro- nominal stem a- has been prefixed to the original vd, va (IE we, we). The enclitic forms of the dual, whose usage corresponds to that of the plural enclitics, are nau, vdm. Forms corresponding to nau appear in Av. nd (gen.), 0 . SI. na (acc.) and Gk. vat (nom. acc.). It is clear, particularly from Greek, that this formation was not originally confined to enclitic use. A form of the second person without -m appears once in the RV. (4. 41. 2), and a comparison with Av. vd (acc.) shows this form to be original. In O. SI. the corresponding form va is an accented form used as both nom. and acc. There existed in IE a reflexive pronoun which inflected after the fashion of the personal pronouns (Lat. se , sibi , etc.). The initial varied between sv- and s- in the same way as that of the second personal pronoun. There are some remnants of this in- flection in Avestan (dat. sg. hvdvdya, i.e. *hvawya), but it does not remain in Sanskrit. The stem sva- ' self ' is used in com- pounds (sva-yuj- * yoking oneself etc.), in the adverbial svatas 'from oneself', and in certain derivatives (, svatvd -, etc.). Apart from these cases the stern sva - is a possessive adjective corresponding to Lat. suus , etc. There is also an indeclinable svaydm ' self which is formed by adding the usual pronominal increment - am to a base *svai- [sve-, cf. me, te). As an enclitic this base appears with initial $- in Av. hoi, he, se, 0 . Pers. saiy, functioningTn the same way as me, te. It has been thought that this enclitic pronoun is represented in Pkt. se, but the latter is more likely to be of secondary origin. The Vedic enclitic sim (acc.) appears to be radically related to this group, though differing widely in formation. In Iranian there are correspond- ing forms of the dual (Av, hi) and plural (Av. his, 0 . Pers. sis). The old possessive adjectives based on the first and second personal pronouns (cf. Lat, mens, tuus, etc.) have been lost in Sanskrit, although they are preserved in Iranian (Av. ma- ‘ my NUMERALS, PRONOUNS, INDECLIN ABLES 269 8wa- 4 thy *). It is pointed out above how the stem of such adjectives is adapted as a gen. sg. In their place there are some fresh creations of Sanskrit, e g. mdmaka tdvakd- with vrddhi, based on the gen. sg. ; mdmakina etc. ; madly a , tvadlya asmadlya -, yusmadiya- , based on the compositional stem ; matka - ‘ mine etc. Substitution of some honorific term for the singular of the second personal pronoun is a characteristic shared by Sanskrit with many languages. To some extent the plural serves this purpose, as in English, etc. (yuyam me guravah 4 you are my teacher '), but the normal substitute is bhavdn (stem bhavant-) 4 your honour ' with the third person singular of the verb. The word is an irregular contraction of bhdgavdn 1 the fortunate, blessed one which is itself used in this way. The vocative bhos ( <bhagavas) which is still further contracted, preserves the old Vedic form of the vocative. Substitutes for the old, reflexive pronoun are provided by the nouns tanu- 4 body ' and atmdn- 4 soul The former is so used in the Vedic language (stir a updkd tanvdm dddhdnah 4 placing himself near the sun ') and the usage is paralleled in Avestan. It disappears in this usage in the classical language, but derivatives of it appear in later dialects of the North-West (N.W. Prakrit tanuvaka-, tanuvaga- ‘ one's own Torwali tanu 4 id etc.). The classical alternative atmdn- appears also in the RV., and supersedes the other word from the early prose onwards. Demonstrative , Interrogative and Relative Pronouns Sg. N. m. sd-s, f. sd, n f tdd t A. m. tarn, f. tam, n. tad , I. m. n. tena f f. tdyd ; D. m. n. tdsmai , f. tdsyai, Ab. m. n. tdsmdt , f. tdsyds, G. m. n. tasya, f. tdsyas , L. m. n. tasmin , f. tdsydm. Du. N.A. m. tau s f. n. td, I.D. Ab. m. f. n. tabhydm , G.L. m. f. n. tayos. PI. N. m. U, f. tas } n. ia t tint , A. m. tan, f. tas, n. ta , tdni , I. m. n. tais, Ubhis , f. tdbhis, D. Ab, m. n. Ubhyas, f. ttibhyas, G. m. n. Us dm, f. tasdm , L. m. n. tesu, f. tasu. Those demonstrative, interrogative and relative pronouns whose stem ends in the thematic vowel inflect according to the above pattern. The masculine and neuter pronouns inflect partly in agreement with the nominal a-stems and the feminines partly in agreement with the a -stems. In addition they have 270 NUMERALS, PRONOUNS, IN DECLIN ABIES forms of inflection which are not shared by the nominal stems. These are as follows : The nom. sg. in the pronouns sa, esa and sya appear without final -s when followed by a word beginning with a consonant : sa dadarsa ' he saw ", but so ’bravtt ' he said purusa esah ‘ this man '. Forms without the nominative -5 appear also in the corresponding Gk. 6, Goth. sa. The nom. acc. sg. nt. ends in djt : tat (: Av, tat, Gk. to <*tod, Lat. is-tud ) } etdt (: Av. aetat), tydt, ydt (: Av. yat, Gk. ot-tl), kdt RV. (: Av. kat, Lat. quod), tvat , enat. The Sanskrit sandhi does not allow any decision as to whether the original consonant is -d or -t, but it is clear that the final consonant was originally -d both from the evidence of other languages (Lat. quod, Goth. pat-a) and from forms in Sanskrit where a further suffix is added to this stem : tadd , iddm, etc. The instr. sg. masc. nt. is in the classical language identical in noun and pronoun. In the Vedic language the noun has also the termination -a, which is not used in the pronoun with the exception of end (classical enena) and the adverbial and 'thus It is clear that -ena is the termination proper to- the pronoun and that this has been transferred to the noun. The form is based on the diphthongal stem ( ke- t etc.) which elsewhere is confined to plural use. The -n- appears to be of the same nature as the -n- which appears in the instr. sg. of masc. and neut. i- and u- stems. As opposed to classical -ena, the Vedic language has boti -ena and -end, due to different developments in sandhi of final -an (>-a before vowel). No forms corresponding exactly to these are found * outside Indo-Aryan. In Iranian there are some forms with the intrusive -n~, but they are formed on the- ordinary thematic, not on the diphthongal stem : Av, kana , 0. Pers. tyand, avand . The instr. sg. fern. ( tdya , etc.) is likewise based on the diphthongal stem, without the -n-. It has likewise been adopted by the nominal declension ( sdnayd ). In the dat. abl. loc. sg. the stem of the pronoun is enlarged by an element -sm(a)-. This element is fairly widespread in IE : Umbr. esmei pusme, Goth .imma, pamma (-mm- <-sm-), 0. Pruss. stesmu, etc. It is not altogether clear whether the simple -m- which appears in O. SI. tomu, etc., is a development out of this or stands for an originally variant form. If -sm- only is original it could perhaps be connected with the root of samd-, so that tasmai , for instance, would mean originally ‘ to that same NUMERALS, PRONOUNS, I N D ECLI N ABLE S 2J1 The dat. sg. preserves the old termination of the thematic stems, which in the noun has been replaced by the extended -dya. The -in which forms the termination of the loc. sg. appears nowhere else. In Iranian there are forms with simple -i Av. ahmi , kahmi , cahmi , yahmi. These forms are clearly more original, and the -n of Sanskrit must be regarded as a secondary addition, whatever its origin. There is also absence of nasalisa- tion in some of the middle I A forms : Pa. tamhi, Pkt. tamsi , as opposed to Pkt. tassim. These may be connected directly with, the Iranian forms, and they point to a dialect variant *tdsmi, etc., in Old Indo-Aryan, existing beside tasmin , etc., adopted by the literary language. In the dat. gen. abl. loc. sg. fern, an element -sy- appears before the termination. This is found also in Iranian (Av. ahydi, aiyhdi, aiyhd, etc.) And in^Old Prussian (dat. sg. stessiei , gen. sg. stessias ). In Germanic there appears in these cases simple -s- (Goth. dat. sg. pizai , gen. sg. pizos ). The most plausible explan- ation of these forms is that they are based on the gen. sg. tdsya, etc,, which were originally common to both genders, the fem- inine being eventually discriminated by the addition of the termination -as. , From this starting the rest of the cases could easily be formed on the analogy of the noun. The terminations of the dual are the same as those of the noun. In the gen. -loc. this is due to the transference of the pro- nominal forms to the noun, as can be seen from the opposition of two types of formation in Slavonic (vluku, toju). A few forms with simple -os appear in the Vedic language ( avds , enos ) . The nom. pi. masc. is formed by the diphthongal stem : id, kd, etc. Similarly in .other IE languages : Gk. to/, Goth. pai> Lith. tie , O. SI. ti f etc. Since in O. Lat. quoi (gen. quoins , dat. quoiei) we find such a form of stem used in the singular, and since in Sanskrit it appears in certain cases outside the plural (instr. sg. td-n-a , fern, tay-d , gen. loc. du. tdy-os) } it must be assumed that this form of stem was not originally exclusively plural, and that it gradually became specialised as such. The form of stem that appears in t]je hpttu pi. forms the basis of the other cases (i iebhis , etc.) with the': exception of. the acc. pi. which is in all probability borrowed from, the nominal declension. In the RV. the only forms of the instr. -pi. that occur are of the type tdbhiSy and the nominal declension shows a tendency to borrow 272 NUMERALS, PRONOUNS, INDECLIN ABLES this type. Later the pronominal form of the case is not only excluded from the noun, but also in the pronoun it is replaced by the nominal form (tats). The only exception is the stem a- which preserves the old form of instrumental in the later lan- guage (< ebhts : Av. aeibis). In the gen. pi. both in the masc.-neut. and in the feminine an -s- appears between the stem and the termination. The same -5- appears elsewhere : masc. Av. aesam, O. Pers. avaisam , O. SI. techu, 0. Pruss. stetson ; fern. Av. dyham, Gk. raaiv, Lat, istdrum , Osc. eizazun-c. Allowing for the fact that in Germanic (AS. bar a) and Slavonic {techu) the masculine forms have re- placed the feminine in this case, the -s- forms are clearly more widespread in the feminine than in the masculine (e.g. Gk. rao>v but ru)v). This may well indicate that the formation is more original in the feminine, and if so the -s- would be the -s of the nom. pi. ids, to which the gen. pi. termination is added. Apart from the gen. pi. the inflection of the feminine does not differ from that of the nouns in the plural. The thematic pronominal stems that appear in Sanskrit are sdjid- ‘ that ; he, she , it esdjetd - ' this sydjtyd - ' that ", ena- ' him, her, it dma- ‘ this ", avd- ‘ that ", tva- f one, one , . . the other ", kd- ‘ who, which ? ", yd - ‘ who (relative) ", sama - ' any, every ", simd - ' self ", ndma- * a certain ", a-, ana - and imd - "this". The stems sa- and td- combine together to make one paradigm, and they are divided so that sa~ appears in the nom. sg. masc. and fem., and td- elsewhere. In this respect Sanskrit is in agreement with Iranian (Av. hd, ho. hd, tat , etc.), Greek (o, fj , to, etc.) and Germanic (Goth, sa, so, pata, etc.). This con- tinues the IE state of affairs, and wherp t- appears in these two cases (Lith. ids, ia , 0. SI. tu , to) it is an innovation. The absence of the 5 -termination in the nom. sg. masc. was also an original characteristic of this pronoun. In Sanskrit the pronoun has this termination isah] when it appears at the end of a sentence, and in sandhi before vowels it is treated as if it were sds {sa aha, so f dya ). Greek also has a form os * he ", which appears pre- dominantly at the end of a clause, which suggests that this variant form of the nom. sg. goes back to the IE period. For the rest of the paradigm td- follows the regular form of the de- clension of these pronouns, with the possible exception of RV. sdstnin , loc. sg. Since however the meaning of this in some NUMERALS, PRONOUNS, INDECLINABLES 273 contexts appears to be ' the same, one it should perhaps be connected with the IE root sem- 4 one rather than with this pronoun. The stems esd-jetd- and sydjtyd- alternate in the same way as sajia The former of these is compounded e-, which is the ay- oi the ay dm- pronoun, and the above stem sa/ta-. The combina- tion appears also in Avestan (nom. sg. masc. aefa, aeso, fem. aesa, nt. aetat) but not elsewhere in IE. The pronoun sydjtyd - appears mainly in the RV. There are a small number of occurrences in the later preclassical literature, and even in early Pali texts (Jat. 'tyamhi, tydsu), but it is not used in the classical language, although it is recognised by the grammarians. Out- side Indo-Aryan the only corresponding forms are 0. Pers, hyajlya used as a relative pronoun in place of Skt. Av. yd-. The interrogative pronoun kd- 4 who ? * is declined regularly according to the pronominal declension with the exception of the nom. acc. sg. nt. Mm, beside which the RV. has also kdi (: Av. kat, Lat. quod , Goth, ha) formed regularly from the a- stem. In Indo-European there existed both an a-stem (Skt. kd-s, Av. kd, Lith. kds, Goth, has, etc.) and an r-stem (Gk. tL$, Lat. quis, Hitt. kwis). Indo-Iranian had likewise both forms of stem, and this state of affairs is continued in Iranian : Av. nom. sg. (is, acc. sg. Urn nom. pi. cayo , etc. The tendency in Sanskrit has been to eliminate this form of the pronoun. An isolated interrogative Ms is quoted once from the RV., elsewhere this form only occurs in the combinations nd-kis and md-kis 4 no one 4 (; Av. nai-cis, md-cis). A particle klm of the same forma- tion as Im, sim , which occurs in combinations like nd-klm , mi-ktm (cf. Av. nae-Sim , tnd-cim) also belongs here. The only ki- form which maintains itself in the regular paradigm is nom. acc. sg. nt. Mm. This form does not correspond to that found in other IE languages, which has the normal pronominal -d of the neuter (Hitt, kwit, Lat. quid Av. lit, etc.). This form is preserved in Sanskrit in the enclitic particle cit (kdscit ‘ any- body '), from which the existence of an interrogative *cid may be inferred at an earlier stage of the language. How the final -m should be explained is not quite clear, but the existence of the Avestan particle am would seem to indicate that it is at least of Indo-Iranian date, and further connections with the Latin adverbial termination in inter -im t olim, and of Skt. kimcit with Arm. in-c 4 something ' have been suggested. The restora274 NUMERALS, PRONOUNS, INDECLIN ARLES tion, by analogy, of the k- before the palatalising vowel is a common feature of Sanskrit in contradistinction to early Iran- ian. It has also taken place in certain of the thematic stems which originally contained the vowel e t e.g. gen. sg. kdsya , Av. cahyd t Gk, reo, O. SL ceso. There is a variety of adjectival and adverbial derivatives from this pronoun based on the three stems ka- } hi- and ku-. ka- : katard- 4 which of two ' (: Av. katdra Gk. rrorepas, Goth, hapar ), katamd - ' which of many > (; Av. katdma-), kdti 4 how many ' (: Av. caiti, Lat. quot), kathd , kathdm ' how ' (: Av. kaOd) kadd 4 when ’ (: Av. kadd, kaha ), kdrhi 4 when ki- ; kivant- 4 how much ' (: Av. cvant-) ktyant - * id ku - : ku in kUcit 4 everywhere kuv-it particle of interroga- tion (; Av. kil 4 where ’), kvd * where kuha £ id 1 (: Av. kudd s O. SI. kude), kutra 1 id kutas * whence J . The relative pronoun yd- is paralleled in Iranian (Av. ya-) Greek (os) and Phrygian (ios), and in various derivatives else- where. Its declension is of the normal pronominal type (yds, yd, ydd , etc.) and calls for no further comment. Among the various derivatives from it mention may be made of yatard - ' which of two ' (: Av. yatdra -), ydti ' how many yavant - ‘ how big J (: Av. yavant-), ydthd ' how r (: Av. yaOa), ydtra 4 where ' (: Av. yaOrd) , yadd 4 when ' (: Av. yadd) , yddi 1 if ' (Av. yeSi, yeiSi, 0. Pers. yadiy) . The enclitic pronoun ena- 4 him, her, it ' is used only in the acc. of all numbers, in the instrumental singular, and in the gen, loc. du. In other cases unaccented forms of the ay dm pronoun are used with the same syntactical function (asya, asmai , etc.). No pronoun which can be compared with this is found in Old Iranian, nor in the rest of IE, but possibly M, Pers. en, Mod. Pers. in go back to the same source. The pronominal stem avd- is nearly extinct already in the Vedic language, being confined to three occurrences in the gen. loc. du. (avos). In Iranian on the other hand it remained in common use (Av. 0. Pers. ava-) and a corresponding pronoun is found also in Slavonic (0. SI. ovii ). The unaccented pronoun tva- 1 a certain one, many a one ' (when repeated 4 one . . . another ’) occurs in the Veda but is absent from the later language. Outside Indo-Aryan it is pos- sible to compare Av. Owat, nt. sg. used adverbially (as also is Skt. tvat). Avestan has also a pronoun hvo ' he which suggests that NUMERALS, PRONOUNS, IN DECLIN ABLES 275 there was originally an alternating stem *svajtva- after the fashion of syajtya The pronoun dma- 4 this one ' appears only in one ritual phrase. Elsewhere the stem appears only in the 0. Pers. adverb amata ' from there and possibly in the Vedic adverb amd 1 at home ' . The Vedic pronoun n&ma- ‘ a certain one, many a one ' is used in much the same sense as the pronoun tva- t and may be combined with it in phrases like ndma u tva aha ‘ one or the other said It is the stem of Skt. nema- 4 half Av. naema- ‘ half, side ’ used adjectivally, 4 he, they on the one side The unaccented sama- ‘ any, every * is likewise confined to the earliest language. It corresponds to Av. ham a- ' every Goth, sums 4 a certain one ', and it is ultimately derived from the IE root sent- * one An adverb samaha f somewhere, some- how ’ is derived from it. The pronoun simd- 1 oneself which is also confined to the Vedic language, has nothing exactly corresponding to it outside Sanskrit. In structure it can be compared to the pronominal stem imd-, simd- having the same relation to sun, for instance, as imd- to im. The stems a-, and- and imd form part of the ay dm declension, and are treated below. This form of declension is followed by a variety of adjectival stems, consisting partly of pronominal derivatives, and partly of certain other adjectives. The full pronominal declension with nom. acc. sg. nt. in -at is taken by any a- 4 other * (anydt, cf. Gk. aXXo, Lat. aliud ) and such pronominal derivatives as katard kaiamd-, yatard yatarna-, and itara- ( other Pro- nominal inflection, but with nom. acc. sg. nt. in -am is taken by such stems as vtsva- ‘ all ' (nom. pi. masc. visve, gen. pi. masc. neut. visvesam , etc. : Av. vispe , vispaesam) t sdrva- 1 all ' (Indo- Aryan development on the analogy of visva-) , ika- 4 one 4 (cf. Av. aevahmdt , etc., from aeva- 4 one ; ). Other adjectives which show declension of this type are certain comparatives and superlatives such as ddhara-, adhamd-, dniara- t antama-, dpara - t apamd- and the like, and a few other adjectives such as pUrva- 4 prior, east ddksina 4 right, south ' and ubhdya- ‘ of both kinds In this latter class there is considerable fluctuation of usage ; in some cases pronominal inflection is optional, and in others pronominal inflection only occurs in specific senses. 276 NUMERALS, PRONOUNS, INDECLIN ABLES The Pronouns ay dm and asau Paradigms : ( 1 ) Sg. N. m, aydm, f. iydm, n, iddm, A. m, imam , f. imam , nt. iddm, I. m. n. andna , f, andya , D. m. n, asmai, f. asyaf, Ab. m. n. asmdt , f. asyds, G. m, n. asya, f, asyas, L. m. n. asmtn, i. asydm. Du. N.A. m. imau, f. n. ira^, I.D, Ab. dbhyam, GL. andyos . PL N. m. f. n. imdni , A. m. iman, f. irais, n. imdni, I. m. n. f. afcAfs, D. Ab. m. n, ebhyds , f. dbhyds, G. m. n. f. as dm, L. m. n. f. asw. ( 2 ) Sg. N. m. f. asaw, n. arfas, A. m. amum , f. amdm, n. atfas, l. m. n. amiina, f. arawya, D. in. n, amusmai, f. amusyai , Ab. m. n. amusmdt , i. amusyds, G. m. n. f. ara«sya$, L. m. n. amusmin , f. amusydm . Du. N.A. ara«, I.D. Ab, amUbhydm, GL. amwyas. PL N, m. ami, f. amtis, n. aniuni , A. m. amun, f. amtis, n. amUni, I. m. n. amtbhis, f. amUbhis, D. Ab. m.n. amtbhyas amti- bhyas , G. m. n. amisdm , f. amUsdm , L. m. n. amisu,t. amusu . These two pronouns may be classed together as being in many ways aberrant from the normal pronominal declension. The ayam pronoun is distinguished by the number of different stems that combine to form the paradigm. The fundamental stems are ay-(e-)ji and a-. The nom. sg. masc. ayara (: Av. a£w) is made by the addition of the common pronominal -dm (cf. ahdm, etc.) to the gunated form of the first stem. The nom. sg. fern, iydm {Av. im for *iy?m f 0. Pers. iyam both masc, and fern.) is a similar extension of i- <i-H. In the same way the nom. acc, sg. nt. is an extension of id (: Lat. id) which remains in use as a particle. Latin has the same extension in idem J the same with a specialised sense that is absent in Sanskrit. The same extension applied to an acc. sg. masc. *im (Gk. tv ■ avrov, Lat. im ‘ eum ’) has produced imam (: Av. mum, 0. Pers. imam), from which a new stem imd- is abstracted and extended to the acc. sg. fern, imdm (: Av. imam , 0. Pers. imam), nom. acc. du. masc. imau (: Av. ima), fern. nt. ime, nom. pi. masc. imd (: Av. ime, 0. Pers. imaiy), acc. pi. masc. imdn (: Av. ima), nom. acc. pi. fern, imds (Av. imd(s), O. Pers. imd) nt. imd(ni) (Av. imd). The stem is occasionally extended to other cases in the Vedic (but never in the classical) language : imdsya, imasmai ; similarly in Iranian (Av. nom. acc. sg. nt. imat, NUMERALS, PRONOUNS, INDECLINABLES 277 O. Pers. gen. pi. imaiSam), Middle Indo-Aryan (Pa. imassa , etc.), and in Buddhist * and other incorrect forms of Sanskrit (imesu, etc.). The stem a - appears in the D.G. Ab. L. sg., in accordance with Avestan [ahmdi, ahmat, ahydjahe , ahtni, fem. ahydi, etc.) with the normal corresponding forms in the instr. pi. etc. (ebhis : Av, aeibis ; fem. dbhts, : Ay. dbis, etc.). In the Vedic language the instr. sg. appears as end, fem. ayd, and the gen. loc. du. as ayos. These are (allowing for the alternation -a Id in kina, etc., as opposed to end) the normal thematic endings, and the stem therefore is a-. On the other hand in the classical language these are replaced by andna, andyd, andyos . In Avestan there is an instr. sg. and, formed like kana, etc., from the stem a-, and corresponding to it there is a Vedic adverb ana 4 therefore It seems that the stem ana-, on which the above Sanskrit forms and the Av. instr. pi. andis are made, originated in this form and. On the other hand there is in Slavonic a pronoun onu which can be compared. It is possible that the Slavonic pro- noun has arisen by the generalisation of a stem which arose in the same way as Indo-Iranian ana-. There is a difference of accentuation between this pronoun and the corresponding forms of the normal pronominal declen- sion ( asyd as opposed to tdsya, etc.), which is apparently due to a generalisation of the final accentuation of aydm * As an anaphoric pronoun ( asntai 4 to him etc.) the cases of the ci- stern are unaccented. A variety of adverbial forms are made, on the basis of the pronominal stems a- and i - : dir a 1 there 1 (; Av. adrd)^ dtas 4 from there ida (: Av. iSa), iddnlm 1 now ihd 4 here ’ (Av. ida, t8a, O. Pers. ida), itthdm 4 thus ', etc. The only part of the asau pronoun for which anything corre- sponding can be found in another language is the nom. sg. asau. Corresponding to this Iranian has Av. hdu, O. Pers. hauv, but in the other cases it uses the stem ava-, which has become almost extinct in Sanskrit. The most likely explanation of

• sau is that it consists of the pronouns, sa, sd and a particle -aw

indicating distance. In the same way the acc. anium may be explained as replacing am-u, with a variant grade of the same particle. The am- would originally be the acc, sg. of the pro- nominal stem a-, the specific sense of the pronoun (‘ that over there ') being provided by the added element -w. Once this is 278 NUMERALS, PRONOUNS, I N D ECLI N A BLE S interpreted as a stem, and the acc. sg. termination added it naturally forms the basis of a fully inflected pronoun declined on the analogy of the other pronominal stems. If it had been an older stem with suffix -u, inflection after the style of the u- stems would be expected. The fact that in. contradistinction to all other u-s terns it inflects according to the thematic type (in its pronominal variety) indicates that it is a late creation peculiar to Sanskrit which has arisen in some such way as described above. The Sanskrit nom. sg. masc. fern, has an initial a- which is absent from Iranian haa. It is clear that this has arisen from the analogy of the amu- stem, and that Iranian hdu , represent- ing Indo-Ir. *sdu is the more original form. The nom. acc. sg. nt. adds is a form for which no analogy appears elsewhere. The most likely interpretation of this form is that it was really ado misinterpreted as being for adds in those sandhi contexts where -as becomes -0. In support of this explanation one instance of ado before initial p- can be cited from the RV. The nom. pi. amt and the remaining cases based on this stem are not easy to explain. It is suggested, but without any degree of certainty that an analogical diphthongal form amui was created corre- sponding to the diphthongal plural stems of the thematic pro- nouns (U <tai <toi, etc.) and that since this combination was new and unfamiliar (inherited u + i becoming vi~), it was un- stable and became changed to -1. §3. Indeclinables Some of the oldest types of adverb have been mentioned already in connection with the formation of nominal stems. It was observed that adverbs of the type pratdr r early ' are formed by means of suffixes that were originally used in the formation of neuter stems. Adverbs of this type have most commonly accent on the suffix which also appears in the endingless locatives, with which they are identical in formation. There is also a type with radical accent (anti, etc.) having the form of an unaltered neuter stem. The following is a list of adverbs of these types arranged according to suffix : Neuter stems without suffix : yugapad ‘ simultaneously anusdk ‘ in order -or ; avdr 4 down # , punar 4 again - tar : pratdr ' early sanutdr 4 away, apart antdr f inside ’ ; -tur : sanitur ‘ away, NUMERALS, PRONOUNS, INDECLINABLES 279 apart ' ; -var : sasvdr 4 secretly ' ; -ur : mtihur 4 suddenly 9 ; -i : s ami 4 half ' (usually in compounds, cf. Gk. rjpu-, Lat. semi-), pari 4 around ' ; -ti : a'ft/x 4 near 4 , prati 4 opposite, to- wards J ; -it : miihn 4 falsely maksu 4 immediately 4 (cf. Lat, mox), muku 4 suddenly 4 (cf. Av. m$rdzu°- 4 short Gk. f3paxvs) t anusthu 4 at once ’ ; -as : mithds 4 falsely Ay as 4 yesterday ' (: Gk. Lat. Am with additional suffix), si‘<is 4 tomor- row ' (as opposed to r-stem in Av. sur- 'morning'), avds 4 down adhas ‘ id prdyas 1 generally sadyds , sadivas 4 to- day, immediately liras 4 across paras 4 beyond pur as 4 be- fore 4 ; -is : avis 4 openly AaAxs 4 outside 4 ; -us : anyedyus 4 on the following day ', prddus 4 forth to view * ; -at : dr aval 4 quickly drahydt 4 stoutly J (RV. once), isat 4 a little 1 ; -it : pradaksinil 4 so as to keep something to the right ', cikitvit ; -ad : smi/, sumdt 4 with 1 (: Av. mat) ; -A : jyok 4 lor a long time 4 (cf. Lat. did). A number of adverbs functioning as verbal prefixes have the form of uninflected thematic stems, namely dva 4 down ’, apa 4 away from 4 (; Gk. ano), upa 4 up to, near 4 (: Gk. vno 4 under 4 ), and prd 4 fprth 4 (Gk. npo, etc.). Since the thematic suffix was originally Used for the formation of adjectives, words of this type may be regarded as fossilized adjectival stems without in- flection which have acquired the function of prepositions and verbal prefixes. The above adverbs are formed on the same lines as the corre- sponding nominal stems. Other adverbs contain specifically adverbial suffixes. The more important of these are as follows : -tas : This suffix makes adverbs which have in general an ablatival sense : itds 4 from here 4 , tdtas 4 from there anydlas 4 from another place 4 , daksinatds 4 from the right, on the right hrttds 4 from the heart etc. In some, cases such forms function like forms of the ablative case : sarvato bhaydi 4 from all fear 4 , kutas cid desad dgatya 4 coming from some district or other It has already been pointed out that this formation may be explained as deriving from the gen.-abl. of old ^-sterns. When simple ^-formations became rare forms of the type daksinat-ds were reinterpreted as daksina-tds, etc., and a new adverbial suffix created. Corresponding formations appear in other IE languages : Av. x'ato 4 of oneself 4 , aiwito 4 around O. Pers. haca paruviyata 4 from of old amata 4 from there 4 , Gk. £kt6$ 4 outside Lat. funditus, caelitus, etc. 280 numerals, pronouns, indeclin ables -tat : This formation which arises from a contamination of the above with the ablative termination -at appears in such adverbs as udaktdt 4 from above praktat ( from in front adhdstdt * (from) below purdstdt 1 from or in the front 1 and the like. There are no parallels in other languages. -trd : With this suffix are formed adverbs with a locatival sense from noun stems (in the Veda only) and from pro- nominal stems : devatrd 4 among the gods ', purusatrA 4 among men sayutrd 4 on a couch daksinatra 1 on the right side atrfi ‘ here tatra 1 there kutra 4 where etc. The forms (com- pare those in - tas ) are occasionally used as substitutes for the locative case : hdsta d daksinatrd 4 in the right hand Similar formations in Iranian are Av . vayhaQra 4 at the place of dwell- ing ' , idra 4 here V adrd 1 there ', kudra ‘ where etc. As already indicated these adverbs are based on the extinct class of neuters in -tar ( *sayutar 4 couch *vasatar 4 dwelling-place '), of which they are instrumentals, with the locatival sense which instrumental forms always have when used adverbially. The fluctuation between a and d is due to variant treatment in sandhi of final -a<an, of which examples are noted elsewhere. -thd, - tham : The suffix Aha forms adverbs of manner ; rtutha 4 regularly pratndtha 4 as of old visvatha 4 in every way anyathd 4 otherwise tathd 4 so kathd 4 how itthd 4 thus etc. More rarely -tham appears in the same sense : kathdm 4 how ’, itthdm 4 thus A similar formative appears only in Iranian : Av, ka6a } ku8a 4 how avada 4 thus hamaBa 1 in the same way ', etc. -dd, -ddntm, - di : The suffix -da in kadd 4 when tad a 4 then ida 4 now J , sarvada ' always 1 contains clearly as its first ele- ment the -d of the neuter pronouns (tad, etc.). Similar forma- tions appear in Iranian (Av. kaSa, ta8a, etc.), and Lithuanian (kadd f tadd ; vis add 4 always '). These forms may be extended by the addition of an element -mm of obscure origin : iddntm, taddnim ; vibadamm ' always The same pronominal -d appears in the -di of yddi 4 if ' (: 0. Pers. yadiy, Av. yeiSi) , -dhd : This suffix meaning ' in so many parts ' appears in such words as tridhd 4 triply caturdhd 4 four fold katidha 4 in how many parts /, bahudhd 4 in many ways vibddhd 4 in every way bahirdhd 4 outside mitradha 4 in a friendly manner The -dhd is not in origin suffixal, but the root dhd forming a compound with the previous member (cf. tridhitu * consisting NUMERALS, PRONOUNS, INDECLINABLES 281 of 3 parts ' beside Iridhd ), but it has come to function exactly like a suffix. -dha, -ha : A suffix -dha appears occasionally, e.g. Vedic sadha- 4 with ’ (in certain compounds; usually it is weakened to -ha : sahd ' with ’). The same suffix with the same weakening appears in ihd 4 here ’ (Pa. idha, Av. 18a) f kuha 4 where ' (Av. kudu, O. SI. kude ), visvdha 4 always ' (Q. SL viside ‘ everywhere ') and samaha 4 in some way or other -s' as : This suffix makes distributive adverbs from numerals and other words : dvisds 4 in twos, two by two satasds ‘ by hundreds sahasrasds 4 in thousands ', sreniids 4 in rows devaids 4 to each of the gods *, etc. ; cf. Av. navaso ' in nines The corresponding suffix in Greek is -/ca?, in eVas- # by oneself ’ and avSpaKas ' man by man The ultimate analysis of this element is not certain. -rhi : karhi 4 when ? tdrhi 1 then etdrhi 4 now yarhi 4 where (rel.) amurhi 4 there The first element of this double suffix appears independently in various IE languages to make adverbial derivatives from pronominal stems : Lat. cur 4 why Goth, par 4 there Lith. kuf 4 where visur 4 everywhere etc. The second element is best explained as a weakening of -dhi (cf. -ha above) and this -dhi may be compared with the which appears in Greek in such words as 7 toOi 4 where r68t 4 there Adverbially Used Case Forms. It has been pointed out that a large number of the adverbs dealt with above are, in their ultimate analysis, case forms of nominal stems, e.g. the simple neuter stem which elsewhere functions as nom. acc. sg. nt. (jatu), endingless locative (prdtdr)., instrumental (iayutri). In principle such adverbs are formed in the same way as those below, but the stems on which they are based no longer exist apart from the adverbs concerned. The following list consists of adverbs formed from stems which are used also as substan- tives or adjectives. The most common case form used in the making of adjectives is the nom. -acc. sg. nt. By this means adverbs are formed, occasionally from nouns, copiously from adjectives, including all compounds which are adjectives. Typical examples are puru 4 abundantly uru 4 widely mdhi 4 greatly bhiiyas ' more, again rdhas 4 secretly nyak 4 downwards ', nima 4 by name sukham 4 happily bdlavat 4 strongly ', dhrsnu 4 boldly 282 NUMERALS, PRONOUNS, INDECLINABLES satyam ‘ truly ’, nltyam * continually dram 1 for a long time sddaram * respectfully ndndratham ' on different chariots ’ , praddna-purvam * accompanied by a gift f . Those compounds which have a preposition as their first member are classed by the grammarians as avyaylbhdva . Such ar epratyagni ' towards the fire anusvadhdm * by one's own will pratidosdm 1 towards evening and the like. Adverbs meaning : like — ' are formed by means of the suffix - vat in its nom. acc. sg, nt. form, the accent being on the suffix (as opposed to bdlavat , etc,) : manusvdt 1 like Manu pur an aval * as of old ', etc. The acc. sg. is occasionally so used in the case of masculine and feminine nouns : kamam ' at will, if you will ndktam * by night vasam 1 freely, as one wills The feminine accusative functions adverbially in certain cases where the suffixes - tar a and -tama are added to adverbs : uccaistardm * higher sanaistaram ' more gradually etc. These suffixes are allowed by the grammarians to be added even to finite verbs — sTdate- tardm, etc., thoqgh no trace of such usage appears in the earlier language. From the adverbial accusative there develops a special form of gerund in -am : abhikrdmam juhoti ' approach^ ing (the fire) he offers viparyasam avaguhati ' he buries it up- side down These formations are common in the Brdhmanas , but rare earlier. In the later language only a repeated variety is used : madkukardndm kvanitdni srdvam-srdvam pariba- bhrama ' constantly hearing the humming of bees he wandered .about The adverbial use of the instrumental may be illustrated by such examples as sdhasd * suddenly anjasa ‘ suddenly distyd * fortunately aiesena ‘ completely daksinena ' to the south 9 , sanakais * slowly uccais * on high meats ‘ below A number of adverbial instrumentals in - d have a locative rather than an instrumental sense : diva * by day ', dosd ‘ in the evening etc. Shift of accent sometimes characterises the form as an adverb : so in diva, madhya 1 in the middle daksind * to the right'. In the Vedic language there occur adverbs in - ay A such as naktayd ' by night rtaya ' in the right way sumnayd * piously svapnaya 1 in a dream Avestan has similar formations, angrayd ' evilly as ay a 1 rightly etc. These have the appearance of instrumental singulars of fem- inine i-stems, but no such a-stems occur. Possibly they have NUMERALS, PRONOUNS, INDECLIN ABLES 283 developed out of the locative singular with postposition d of the type O. Pers. dastaya 4 in the hand On the other hand a similar formation appears (also confined to the Vedic language) in connection with stems in -u : raghuyd 4 quickly ", dhrsnuyd 4 boldly etc. ; cf. Av. cisaya 4 quickly A satisfactory explan- ation to account for both types is difficult to find. Examples of other cases used adverbially are : dat. (rare) apardya 4 for the future cirdya f for long ahndya 4 presently abl. pascal 4 behind saksdt 4 evidently sandt 4 from of old ’ (with changed accent), adhardt 4 below ’ ; gen. (rare) aktos 4 by night vasios 4 by day ' ; loc. diird 4 afar ’, rdhasi 4 secretly ’, sthdne 4 suitably sapadi 4 immediately Miscellaneous Adverbs and Particles. In addition to the adverbs classified above, mention may be made of the follow- ing. The particle evd 4 only ’ is possibly to be identified with the stem of that form of the word for one which appears in Iranian and Greek (Av. aeva-, Gk. ofo?), and evdm 4 so ' is probably the neut. sg. of the same word. Of the particles of comparison iva 4 like ' appears to be based on the pronominal base i- and its formation may be compared with that of the stem ava-. In the Vedic language a particle nd f homophonous with the negative particle is used in this sense : guard na trsitdh piba 4 drink like a thirsty bison etc. A comparable form elsewhere appears only as an enclitic particle in certain combinations : Av. yadand, ciBmd, Lat. quidne . This enclitic may be seen also in Skt. cana ( kdscana 4 anybody '). The negative na is a common IE base : cf. Av. na-, O. SI. ne , Lith. ne, Lat. ne-. Goth, ni, etc. As the first member of com- pounds it appears in a weakened form, a- ( <n-), before vowels an- ; similarly Av. O. Pers. a-, an-, Gk. a-, av-, Lat. in-, Ir. an-, Goth, an-. The adverbs nu, nti, nunam 4 now 1 belong to a family well represented in Indo-European : Av, nu, niirzm, Mod. Pers. nun, Oss. nur, O. SI. nyne, Lith. nu, Gk. vv , vvv, vvv, Lat. nun-c , Engl, now, etc. The radical element is that which appears in the adjective ndva- 4 new Formed directly on the adverb Sanskrit has nutana-, nUtna- 4 new, belonging to the present time From a comparison of Greek it appears that the form nundm is to be analysed nun-am, and that it contains the same strengthening affix which was frequently met with in the pro- nominal formations. The njr alternation between Skt. nundm 284 NUMERALS, PRONOUNS, INDECLIN ABLES and Av. nurdtn is of the same origin as that of the nominal formations (cf. Gk. veapo? : vcavtas, etc.). The adverb nand 1 variously ' (the Veda has also an extended form nananam) is of unknown derivation. The conjunction ca ‘ and ' (Av. 6a, Gk. tc, Lat. que, Goth. -h)' is enclitic, as in all the languages. The same applies to vd ' or *, cf. Lat. ve. A non-enclitic in the former sense is utd * and, also ' (Av. O, Pers. uta ‘ and ’). Of miscellaneous particles of asseveration, etc., mention may be made of anga ‘verily' hdnta (expressive of incitement), kila * forsooth ’, khdlu ' indeed tu ' but hi 1 for ' (Av. zi)> gha , ha ' indeed ' (the latter a weakened form of the former ; cf. O. SI, zt), vat 4 verily viva * id ' (with two accents that have not been explained), «, aha , sma, bhala , The particle sma used in conjunction with a present tense gives it an imperfect value. In the later language particles are less frequent than in the earlier, and those that remain tend to lose their significance^ and serve in poetry simply as devices for filling out the metre. A few interjections may be merely listed : d, hd, ahaha , he, ayi f aye , aho, bat , bata , dhik . Some noun and adjective forms have acquired this function ; re, are (voc. of ari- * enemy '), bhos (for bhavas , voc. of bhavant - f your honour ’), kastam ‘ woe is me ! svasti 4 hail ! susthu, sadhu ‘ good, excellent ! etc. Prepositions and Postpositions In contradistinction to other IE languages Sanskrit has not a developed series of prepositions. Furthermore those adverbial formations which are used to define more closely the case- relationship are normally placed after the noun used in this case, and not before it as in other IE languages. In comparison with the Vedic language later Sanskrit is noticeably poorer in words of this type, so that the distinction between it and the usual type of IE language is partly due to regression. On the other hand the system as it appears in the Vedic language, with freer order and looser connection of such words with the nouns they govern, is clearly more primitive than that found in Greek, Latin, etc., and is closer to the IE beginnings of the develop- ment of the prepositional system. Of the words so used in the Veda the most important class, as elsewhere, consists of those words which are also used as verbal prefixes (see below). The majority of these can be so NUMERALS, PRONOUNS, I N D EC L I N A B L E S 285 used, but ud , ni, para, pra, ava and vi are exceptions. Their use, mainly postpositional, may be illustrated by a few examples : dti : yd devo marly dn dti ' the god who is beyond mortals ddhi : prthivydm ddhy osadhih ‘ the plants upon the earth dpi : yd apdm dpi vratd 1 who are in the domain of the waters’. dnu : mama cittam dnu cittdbhir dta 1 follow after my mind with your minds abhi : ydh ■ pradiso abhi stiryo vicdsie ' what quarters the sun looks abroad to a : mdrtyesu a 1 among mortals up a : amur yd up a surye 1 those who are near the sun pari : jdto hundvatas pari ' born from the Himalaya prdti : dbodhy agnih prdty dyaiim usdsam 1 Agni has been awakened to meet the approaching dawn sam : te sumatibhih sdm pdtnibhir nd vfsano nashnahi t may we be united with thy favours as males with their spouses The use of the last one is rare, and in its place the radically related sahd commonly appears as a postposition with the in- strumental in the earlier and later language. Of the other words listed above the only ones so used in the later language are dnu , prati and a. The first two are used as postpositions, the latter as a preposition with the ablative meaning ‘ up to d samudrat ‘ up to the ocean In addition a variety of adverbs, both the old inherited type and the newer adverbially used case-forms, are used to define more closely the relation expressed by a case affix or in con- junction with it to express a relation which cannot be expressed by a case-termination alone. Such are ; With accusative : liras ' through antdr , antard ‘ between avarena ' below parena r beyond uttarena f to the north of daksinena c to the south of nikasd ‘ near ’ ; e.g. yd 'varena* dityam, yd parenddityam ' those who are below the sun, those who are beyond the sun daksinena vddim ‘ to the south of the altar ', nikasd Yamunam ' close to the Jumna ; . , With instrumental : Mainly words meaning * with e.g. saha (above), sdkam t sardhdm , samdm, samdya , sardtham, but also vind 1 without ’ which follows the analogy of the words of contrasting meaning. The dative is the only case which is not used in conjunction with words of prepositional character. Nevertheless, as noted 286 NUMERALS, PRONOUNS, IN DECLINABLE S above, the form of the dative of a-stems (-aya) can only be ex- plained by the incorporation of what was originally an inde- pendent postposition. With ablative : bahts 1 outside of ' pur as 4 in front of 4 , avds , adhds 4 below 4 , pura 4 before paras 4 beyond vina 4 without 4 (also instrumental), arvdk 4 this side of pascAt 4 behind urdhvdm 1 above Hi 4 without 4 , etc. With genitive : mostly case-forms of nouns or adjectives which take this case by virtue of retaining their nominal char- acter. Such are agre 4 in front of abhydse, satntpe 4 near arthe, krte 4 for the sake of madhye 4 in the midst of Words of more purely adverbial character used with the genitive are upari 4 above pardstat 4 beyond purastdt 4 before etc. With locative : antdr, antara 4 inside sued 4 with Verbal Prefixes . A widespread feature of Indo-European is the compounding of verbs with prepositional prefixes. It is normally the same words which appear in use as the common prepositions which are compounded with the verbs. In San- skrit, it has been noted, the system of prepositions (or post- positions) used in conjunction with nouns is much less developed than in the related languages. On the other hand the use of the same class of words as verbal prefixes is as fully developed in Sanskrit as in the other IE languages. The common prefixes so used are as follows : dti 4 across, beyond 4 (Av. aiti, O. Pers. atiy ; Gk. cti : also, still 4 , Lat. et ‘ and 4 which are used differently), ddhi ‘ above, on, on to 4 (Av. aidt, aiSi, 0. Pers. adiy) dnu 4 after, along, towards 1 (Av. anu, O. Pers. anuv ; Gk. dvd with variant suffix), antdr 4 with- in 4 (Av. antara, O. Pers. antar, Lat. inter), apa 4 away, from 4 (Av. O. Pers. apa, Gk. diro, Lat. ab ; Hitt, appa 4 back, be- hind 4 ), dpi c unto, upon 4 (Av. aipi , 0. Pers. apiy, Gk. k-nl ; in Sanskrit used rarely in this way but mostly as a conjunction 4 also cf. Gk. ert, Lat. et above), abht 4 to, towards, against 4 (Av. aiwi, O. Pers. ably, O. SI. obu t obi , Lat. ob), dva 4 down, off 4 (Av. O. Pers. ava, O. Pruss. au- t O. SI. u- t Lat. au-) f A 4 to, up to. at ' (Av. O. Pers. d ), ud 4 up, forth, out 4 (Av. us -, uz- f O. Pers. ud Ir. ud -, od- t Goth, ut 4 out 4 ), upa 4 to, toward, near 4 (Av. upa, O. Pers, upd, Gk. vtto, Goth, uf), nt 4 down 4 (Av. ni- t 0. Pers. niy-), nts 4 out, forth 4 (Av. mi-), para 4 forth, away 4 (0. Pers. para), pari 4 around 4 (Av. pairi , O. Pers. pariy , Gk. nepl), prd 4 forward, forth 4 (Av. O. Pers. fra - O. SI. pro, NUMERALS, PRONOUNS, INDECLINABLE S 287 Lith. pra , Gk. irpo, Lat. pro-) ; prdti r against, back, in return 1 (Gk. rrpoTi, TTport, npos, O. SL protivu , etc.), vi ‘ apart, asunder, away * (Av, 0. Pers. vi- ; cf. Toch. wi ' two etc., above, p. 260 ), sam ‘ together, with ' (Av. 0. Pers. ham-, 0. SL sg , su, Lith. sa sw). These are the regular and normal prefixes. In addition there are a few of more restricted application. In the Veda dccha 1 to, towards ' is fairly common, but it dies out later. Others occur, in the Veda and later, only in connection with a restricted number of roots : avis ‘ forth to sight, in view ’ (with bhu , as and kr), prddus r id ' (with the same roots), liras "through, across, out of sight ' (with kr, dhd , bhu) pur as ‘ in front ’ (with kr, dhd , i and a few others. More than one prefix can be combined with a verb (as in Greek, etc.). Combinations of two are common, of three, not unusual, but more than three are very rarely found. There are no particular rules as to the order in which they may appear, but the prefix d is practically never separated from the verb. All these prefixes were to begin with independent adverbs. In the language of the Veda they partly retain this character and it is only in the later language that they become insepar- ably combined with the verbal stem. A similar difference is to be observed between Homeric and later Greek, which makes it clear that the development of the full system of verbal com- position is largely a parallel development in the various lan- guages. In the Veda, a prefix most frequently stands immediately before the verb (a gamut ' may he come ') but it may be separ- ated from it by another word (d tvd visantu ' may they enter thee f ) and it may even follow the verb (Indro ga avrnod dpa 1 Indra disclosed the cows '). Whatever its position, in a prin- cipal clause the preposition is regularly accented, and the verb, according to the general rule, is unaccented. When two pre- fixes are used both are accented normally in the RV. [upa prd ydhi ‘ come forth here ') a fact which emphasises their status as independent words. But besides this there is a system, showing the transition to a closer form of union, according to which the second only of two prefixes is accented when they immediately precede the verb : dthdstani vipdretana 1 then scatter ye away to your home In subordinate clauses the process of composi- tion has preceded further, the preposition generally appearing 288 NUMERALS, PRONOUNS, IN DECLIN ABLES compounded, and since the verb in these cases is accented, the prefix is without accent : e.g. ydd . . . nisidathah 4 when ye two sit down Even here, however, it may appear separate from the verb and accented (vi yd ntame rdjasi ' who measured out the two spaces ’), while occasionally it is treated as a separate word and accented even when it immediately precedes the verb : yd ahutim pari vddd ndmobhih ' who fully knows the offering with devotion In the preclassical prose texts the prefix is still to some extent separated from the verb, but on a much more limited scale. By the classical period its independence is totally lost, and except for the few that continue to function as postpositions, the verbal prefixes have ceased to exist as independent words. In combination with the nominal derivatives of verbal root the verbal prefixes appear fully compounded from the begin- ning : adhivdsd- 1 garment apaciti- 1 retribution ' (Gk. dwucHs), abhidruh- 4 treacherous avapana - r drinking place ' ; uddyana- 4 rising (of the sun) upasruti f overhearing nidki- 4 deposit, treasure nirdyana- 4 going out 1 ; prabkangin - ' crushing etc. In such cases the general rule is that the prefix loses its accent in favour of the second member of the compound but in some cases it is regularly accented, namely (i) in com- bination with the past passive participle, par eta- 1 gone forth ', antarhita- 1 concealed dvapanna - 1 fallen down etc. ; (2) with the verbal action nouns in 4 i, apaciti , etc. In both these cases there is agreement in the matter of accent between Sanskrit and Greek (aTro/JA^ro?, dvafiATjcns, etc.) ; (3) with the infinitival forms based on the /w-suffix : sdmhartum 4 to collect dpi - dhdtave 1 to cover up ' ; dvagantos ( of descending CHAPTER VII . THE VERB §i. The Verbal Root The roots of the Sanskrit language as enumerated by the Hindu grammarians comprise a list of some two thousand. Something like half of these are not attested in actual use, and since it is unlikely that many of them will ever turn up they may for all practical purposes be neglected. Of the rest a considerable number may be dismissed as being either reduplications (( didht -), stem forms (urnu-) } denominatives (arth- } etc.) or in some other respect not primitive. When allowance is made for these there remain somewhat over eight hundred roots, which form the basis not only of the verbal system, but also the larger part of the inherited nominal stems of the language. Chiefly owing to its antiquity the Sanskrit language is more readily analysable, and its roots more easily separable from accretionary elements than is the case with any other IE lan- guage. This is because the suffixes with which the present and aorist stems are formed, are normally kept out of the other forms of the finite verb, and from nominal derivatives : sunoti ' presses out ' : perf. susdva , susuma , fut. sosyati , part. pass. sutd-. Nevertheless even Sanskrit is not wanting in cases where suffixes whose primary function is the formation of the present stem, have become permanently attached to the root, and con- sequently appear throughout the conjugation of the verb. For instance from the present -prcchati * asks ', formed with the IE suffix -she- an extended root prch-jprach - is made which appears in the perf. paprdccha and elsewhere. Similarly Latin has posed , poposci ( <prk$hd). The simpler form of the root appears in Lat. precem (acc. sg.), procus and in Skt. prasna- ' question The root ksnu - ■ to sharpen ’ contains the ww-suffix which else- where forms the fifth present class. From a number of presents in which this suffix enlarged by the thematic vowel appears, extended root forms ending in -nv develop : pinv- * to fatten pres, pinvati , perf. 2 du. pipinvdthus (beside simple root pi- 289 THE VERB 29O in pdyate, etc.). Similar extended roots are inv- 7 to send ’ (: 1 -) and jinv- 7 to quicken 7 (: ji-). From present stems in -va a number of roots ending in -v are created : jiv- 7 to live 7 , pres. jivaii (: simple root in gdya- 1 livelihood, belongings '), dhurv - 7 to injure 7 (: dhvr-), tiirv- 7 to overcome 7 {: //-) and bharv- 7 to chew 7 for which no simpler form exists. Accretions of this type are of comparatively recent origin, and it is quite clear how they have arisen from particular stem forms that are current in the verbal conjugation. Besides them there exists another class of extended roots, of much more ancient date, containing accretions whose functions it is for the most part no longer possible to discern. These elements are fairly easily recognisable, either through the coexistence of a simpler form of the root, or by the existence of synonymous roots which differ only in the final element. They are identical with the individual suffixes which have been enumerated in treating of the formation of nouns, and may conveniently be enumerated in the same order ; -arjr : dhar- (dhr-) 7 to hold 7 (i.e. dh-ar cf. dha-), svar - 7 to sound ' (: cf. svan- 7 id 7 with alternating -n). -an : ksan- 7 to wound 7 (: sas - 7 to cut 7 : Gk. KreLvay for

• kstenyo has compound suffix -ten- alternating with -ter- in

KTtpts ' veKpol), svan- 7 to sound J (see above), khan- 7 to dig ' (i.e. kn-an cf. khd- without w- suffix, likewise dkhu- 'mole 7 and akhard- masc. 7 hole 7 with alternating r). -as/s : tras- 7 to fear Gk. rpdoj (: Lat. tremo), bhyas- 1 to fear 7 (: bhl- ‘ id ’), gras - f to devour ' (gf- 7 to swallow 7 ), dhvas - 1 to scatter 1 (dhu- 7 to shake 7 ), srus- ' to hear 7 , O. SI. slysati, Toch. klyos etc. (: normally sru- 7 id Gk. kAu-, etc.), aks-, naks- 7 to attain 7 (: as-, nas - 7 id 7 ), nks- 7 to sprinkle ' (: Gk. vypos ‘ wet etc.), niks- 7 to pierce 7 (: O. SI. vu-nizQ "id 7 ), bhaks- 7 to partake of, eat 7 (: bhaj- 7 to divide, share 7 ), miks- 7 to mix 7 (: misrd- 7 mixed 7 ), mrks- 7 to rub 7 (: mrj- 7 wipe Gk. 6p,opyvvp.L t etc.), raks- 7 to protect 7 , Gk. ahefa 7 to ward off 7 (: Gk. dAaA/ceu' 7 id 7 , AS. ealgian 7 to protect f ) ; vaks- 7 to in- crease 7 , Gk. de£cn (: Lat. auged 7 id etc.,) has- 7 to go forth 7 (: ha- 7 id 7 ). In contrast with most of the extensions the suffix s plays a considerable part in the conjugation of the verb, and this no doubt accounts for the comparative frequency of such forms. -am 1 dram - 7 to run Gk. eS pap,ov (: drd Gk. StSpa<j/coi ; dru- ' id '), gam- 7 to go Goth, qiman, etc. (: gd- 7 id Gk, t/Jd ; THE VERB 29I Gk. jSguVoj, Lat. venio from g*-en-), ksam- * to endure 4 , by meta- thesis for *zgham-, cf. Pasta zyamdl 4 id 1 (: sagh - 4 to be equal to, endure ’), bhram - ‘ to revolve, wander ' (: bkur- 4 to be in uneasy motion -i: jri-, jrdyati , 'to move y ray a sand- 'extending": jar- 4 to move Ir, zar - 4 id 4 ; ksi-, ksdyati 4 to rule ", as opposed to ksatrd - * sovereignty ' and Gk. Kraopiai ‘ possess 4 ; sri-, srdyati 4 to lean Gk. kXIvqs, etc. (:Lith. atsikalti 1 to lean against Russ, klon 4 inclination etc.). Alternative forms of root appear in the conjugation of svi-, su- ' to swell si (' :5m-), sa- 1 to bind and si-, sd- 4 to sharpen It has not in these cases become completely incorporated, but it shows a tendency in the direction. -u : sru- 1 to hear 4 : the unextended form of the root appears in the Skt. present, sr-no-ti ; elsewhere the ujo, which appears as part of the suffixal complex in this form, is per- manently attached to the root. Other examples are dru- 4 to run 4 (: dram-, drd- above) and sru- ‘ to flow ’ (sat- 4 to move, flow cf. sarit- 4 river '). -«h/h : gd - 4 to go 4 (: gam- above), yd - 4 to go 4 (: i- 4 id '), psd- 1 to devour Gk. iftw cu 4 chew 1 (: bhas - 4 to devour ), drd- 4 to run ' (: dram-, etc.), mnd- 4 to note 4 (: man - 4 to think '), tr a- 4 to rescue 4 (: if- 4 to cross 4 ), pyd- 4 to swell (udder) 4 (: pi- 4 id '), pi - [piy-) * to abuse 1 (i.e. pi- H- : pis - in pisuna - 4 malicious, tale-bearing Gk. m*rpos, etc.). A series of roots belonging to the ninth class contain this enlargement. The simple form of the root appears in the present tense, where the #h/h suffix is separated from it by the intervening n-suffix with which it is combined : e.g. prnati 4 fills 4 for pl-n-du-ti as opposed to purna- 4 full 4 for ppu-no-. Similar cases are yu- 4 to be swift 4 (jundti, jiitd-), pu- 4 to purify 4 ( punati , putd -), stf- 4 to strew 4 ( strndti , stirnd-) and so on. The enlargement tends to be introduced into the present tense, so that from mi - 4 to damage 4 for instance there appear both mindti and mindti. In some cases the root appears only with the enlargement in the present tense, e.g. bhrinati, Av. brinmii f cut * and krindti 4 buys though in the latter case the metre of the Rgveda in- dicates a pronunciation krindti. -t : krt- 4 to cut 4 (: Gk, Kelpaj ) , cit- 4 to perceive ' {ci- 4 id ') ; in combination with i, mrit- ' to fall in pieces 4 (mf- 4 to crush 4 , mrd- 4 id 4 ), ivit- 4 to be bright' (: iuc- 4 to gleam 4 , iubhTHE VERB 292 ' to be bright '), with u , dyui - ' to shine r ( 1 it - in d'ma- ' day etc.), -th (i.e. t-H-) : prath- ' to extend' (; Hitt, palhis 'wide', Lat. planus, etc.), vyath- ' to be unsteady ' (vij- 4 to tremble vip- 4 id '), inalh- 1 to pierce ' (: simple root in iisnd - ' organ of generation cf, Gk. Ktvrl co ' pierce ' with guna of root and simple ^-suffix) ; similarly irath- 4 to loosen grath - ' to tie mith- ' to alternate ' (mi- 1 to exchange '), -d ; ksad- * to divide ' (: &as- 4 to cut cf. ksan- above), chid- 1 to cut Lat. scindo, etc. (: cf. chd-, chi - ' id '), rud- 4 to weep Lat. rudo, AS. reotan, etc. (: ru- 4 to cry O. SI. rjuvg, etc.), mrd- 4 to rub, crush ' (: mf - 4 to crush '), pid - 1 to press '( <*pizd-

pis 4 to crush ') ; in combination with n, syand - ' to flow '

(: sic - ' to pour Toch. stk-, etc.), krand Hand - f to cry out ', cf. Gk. /ccAaSos* 1 cry, shout 1 (: Gk. *aAetu ‘ call etc.). -dh : mrdh - ' to neglect, be careless ', cf. Gk. paAdaKos 4 soft etc, (: Gk. apaAos 4 soft etc.), edh- 4 to prosper cf, Av. azdya- 4 thriving, fatness Gk. £a8$ 4 good ' (: IE es- in Gk. ivs 4 good Hitt. asSu -, Skt. su-), sprdh- ' to contend Av, spared-, cf. Goth, spaurds 4 race-course ' (: spr- 4 to win ', sprh - * to be eager Gk. airepxofJLcu* etc.) ; combined with suffix, ksudh- ( to be hungry ' (: Hitt, hast-, Toch A, hast, B. kest, with guna of root and dental suffix whose exact nature cannot be speci- fied). -p : dip - ' to shine ’(: cf. di-, dyut- above), nilup- 4 (sun) to set ' (: mruc mine- 4 id ’), rip-, lip - ' to smear ' (: li- * to cleave to', Lat. lino, etc.), rup J to break Lat. rumpo (: ru- 4 to break Lat. ruo ; ruj- 4 to break '), vip- 4 to tremble ' (: cf. vyath-, vij- above), svap- 4 to sleep AS. swefan (: Gk. etJSaj 4 id ' <seu-d). - bh : £ubh- 4 to be bright ' (: sue- 4 to gleam etc., above) stubh - 4 to praise ' (: stu- ' id '). -c : mine - 4 to set ' (: mlup -), ydc- 4 to ask ' (: Av. yds- with different enlargement), rue- 1 to shine ' (ruiant 4 bright '), sic - 4 to pour ' (: syand- above). -j : tarj- 4 to threaten ' (cf. tras-, etc., above, Lat. terreo), yuj- ' to join ' (: yu- 4 id '), ruj- 4 to break vij- 4 to tremble ' {rup, vip-). -h : sprh - r to be eager ' (: sprdh-, etc., above), druh- 1 to injure'* (: dhru- Vid '). The identity of these elements with the suffixes enumerated THE VERB 293 in dealing with the formation of nouns is evident, and it is in accordance with the fact that nominal and verbal stems are formed fundamentally on the same principles. The enlargements of the verbal roots are simply incorporated suffixes, and do not require, as is sometimes considered, a separate morphological classification. All the IE consonants and semivowels can appear in this function, just as they can in the formation of nouns. In the latter case those so used with any frequency are, as has been seen, necessarily limited in number. In the case of the enlargements the distribution is more even, and with the ex- ception of s and H, which had a considerable part to play in the IE conjugation, the common nominal suffixes do not appear correspondingly frequently as enlargements. The suffixes n and r, for instance, which are of very great importance in nominal stem formation, appear only rarely as enlargements. This would suggest that, on the whole, the enlargements of the verbal root reflect a comparatively early stage of IE stem- formation, that is to say a period when the emphasis on a com- paratively few suffixes out of the large number available had not developed to the extent with which we are familiar later. It is also clear that the growing clarity of the distinction between verb and noun, which evolved in the later pre-history of Indo- European, tended to prevent the incorporation of such suffixes (e.g. r and n) which were felt as predominantly nominal. In their guna grade such roots may appear in two forms, on the one hand that which appears in cet- * perceive sec- ‘ pour rod- ‘ weep etc., and on the other hand that which is seen in tras - ' fear ksad- ' divide sro- ‘ hear ', etc. That is to say, either the root or the enlargement may have the guna, but, in accordance with the principles of IE apophony, it is not pos- sible that both should have it. The difference between the two types of guna form is without any significance as far as the meaning of the roots is concerned, or their conjugation. It is only in connection with nominal stems that this kind of dis- tinction is significant. There it provides the distinction between neuter action nouns and adjectives or agent-nouns. Since these roots were originally stems, and in the early period, when noun and verb were less clearly distinguished, as much nominal as verbal, it is reasonable to assume that the distinction in form between the two types of extended root, was originally the same as that which is fundamental in the formation of nouns. THE VERB 294 That is to say a root form *tres- (Skt. trdsati) would originally be a nominal stem meaning * fearing, one who fears and the alternative form *tirs- (Lat, terreo) would be a stem meaning

• fear

In these cases when the ultimate root has been reached by analysis it is seen never to consist of more than two consonants and the guna vowel {ter-, etc.), or if the root begins with a vowel, of this vowel and a following consonant ( ed-, £s-). The number of roots which can be so reduced is sufficiently great to justify the extension of this principle to other roots of three consonants, even where shorter or variant forms are not pre- served. There is little reason to doubt that the third consonant of all roots which have it is to be regarded as an incorporated suffix. §2. System of the Finite Verb Voice . The Sanskrit verb contains two voices, Active and Middle, which are distinguished by means of two sets of per- sonal terminations throughout the conjugation. The difference in meaning between the two is expressed by the names given to them by the Sanskrit grammarians, parasmai padam ' a word for another 1 and dtmane padam ' a word for oneself The middle is used when the subject is in some way or other specially implicated in the result of the action ; when this is not so the active is used. For instance katam karoti ' he makes a mat ' would be used of the workman employed in the trade of mat- making who makes a mat for another, while the middle katam kurute would be used of one who makes a mat for his own use. The same distinction is seen between pacati ‘ (the cook) cooks ' and pacate ' he cooks (a meal for himself) ' and between yajati ' (the priest) sacrifices (on behalf of another) ' and yajate ' (the householder) sacrifices (on behalf of himself) Again the special sense of the middle is seen in those cases where the direct object of the verb is a member of one's own body : nakhdni nikrntate ' he cuts his nails ', dato dhavate ‘ he cleans his teeth In another class of roots there appears a distinction of a different nature, that between transitive (active) and intransitive (middle) : drmhati ' makes firm drmhate * becomes firm vardhati ‘ increases, makes bigger vardhate * increases (intr.), becomes bigger ', vahati ' (chariot) carries (man) ', vahate ' (man) rides (in chariot) From this the development is not far to the THE VERB 295 distinction between active and passive, and the use of the middle to express a passive sense becomes common in the perfect and the future, which possess no other means of expressing the passive. One sense that the middle does not normally express is that of a direct reflexive, which is expressed by means of the accusative dtmdnam ' self Not all verbs are capable of appear- ing in both voices. Some are used only in the active, e.g. ad- ' to eat as- ‘ to be ksudh- ' to be hungry 7 bhuj- ‘ to bend sarp - 9 to creep ', etc, ; others only in the middle, e.g., as- ‘ to sit hsam - * to endure labh- 4 to receive ’, vas- 4 to wear (clothes) sac- f to accompany Occasionally a different voice appears in different tenses of the same verb, the most usual alternation being that of active perfect and middle present : vartate : vavarta. The distinction between active and middle is inherited from Indo-European (cf, the equation sdcate , Gk. e^rat, Lat. sequitur ), and outside Indo-Iranian the language which shows the greatest similarity to Sanskrit in form and usage is Greek. In later Indo-Aryan the distinction dies out, and this is re- flected in the Epic and other less correct forms of Sanskrit. Tense . The Sanskrit verb has four tense stems : Present, Future, Aorist and Perfect. The present stem forms the basis of a preterite, the so-called Imperfect, in addition to the present tense. In the same way there is formed a preterite of the future which functions as a conditional. In the Vedic language a form of preterite is formed on the basis of the perfect stem. These pluperfect forms are rare even in the earlier language, and dis- appear later. The aorist stem forms only a preterite. The clearest division to be found in this somewhat compli- cated system is that between the perfect on the one hand and the other three systems on the other. The perfect is distin- guished from the other tenses not only in stem-formation, but also in the fact that it possesses a special series of personal endings. Between the perfect and the rest of the conjugation we have clearly the most ancient and fundamental division in the Indo-European system. On the other hand when we examine the future and the aorist in their relation to the present system it is clear that they are in origin only special modifica- tions of the same type of formation. The future for instance is only one subdivision of the class of present stems in ya in which are included the verbs of the foprth class and the various types 296 the verb of denominative. The close relation between the aorist and present systems is seen by the fact that certain types of aorist stem are identical in form with certain types of present stem. This is so with the root aorist (dkar, etc.) which is formed like the imperfect of the root class ( dhan , etc.), and the a-aorist [druhat, etc.) which resembles the imperfect of the sixth class ( dtudat , etc.). The aorist or imperfect character of these two types of formation is determined not by the form itself but by the existence or non-existence of a present from the same stem. In other cases there is evidence for the one time existence of presents from those forms of stem which in Sanskrit are used exclusively as aorists, Thus corresponding to the reduplicated aorist ajijanat, Avestan has a present zlzanznti "they give birth Even in the case of the s -aorist, which is most clearly marked off from the present system, the existence of presents like Av. ndisnvi ' I insult ' demonstrates that such formations were not always exclusively aoristie. The relation of the present-imperfect on the one hand, and the aorist on the other, can be discussed only in view of the meaning of the three tenses. In Sanskrit this is not at all com- plicated. The present indicates simply present time, and the imperfect past time in contradistinction to this, no more and no less : hdnti ‘ he slays dhan ‘ he slew etc. There exists no trace of an ' imperfect ’ sense in the Sanskrit tense of that name, and such a sense, if it is needed, is expressed by the present tense with the addition of the particle sma. The aorist in contra- distinction to the imperfect expresses a special kind of past time, inasmuch as it is used for describing an action which has just recently been completed : ud asau sUryo agdt 4 yonder sun has risen etc. This clear distinction of meaning between the aorist and present stem is found only in the case of the indicative, in these two kinds of preterite. There exist also various moods — in- junctive, subjunctive, imperative and optative — and also par- ticiples, active and middle, which may be formed alternatively from the present or aorist stems. But in all these latter types of formation no serious distinction of meaning can be found in the Vedic language between those formed from the present and those formed from the aorist stem, e.g. karat subj. ' he will do ' does not differ in any demonstrable sense Troir krndvat 1 id * from the present stem. THE VERB 297 The absence of distinction of meaning in all these types of formation between the present and aorist stem, in contradis- tinction to the clear distinction between the two types of pre- terite, points to the conclusion that it was specifically in these preterite forms that the aorist developed as a special gram- matical category. It appears that originally Indo-European distinguished in the indicative simply between present and pre- terite, the forms of which could be made from a wide variety of stems. This state of affairs is continued in Hittite, which shows no sign of ever having had a terS2 corresponding to the aorist of other languages. The next stage of development is the evolu- tion of a double set of preterite forms, one with a correspond- ing present (imperfect) and one detached from the present tense (aorist) and having a special sense. This stage is represented in Indo-Iranian. In Greek the distinction between the present and aorist systems is carried further, and applied to moods, participles and infinitives derived from the two stems. The two stems in all formations express different modes of action, namely punctual (aorist) and durative (present). Consequently the preterite of the present acquires an ' imperfect ’ sense which is absent from it in Hittite and in the corresponding formations in Sanskrit. The perfect is independent in formation from the present/ aorist system, and is also characterised by the possession of a special series of personal endings. It appears to be one of the more ancient IE verbal formations, and to bear some relation to the conjugation of the Hittite verbs in -hi. In that language there are two conjugations of verbs, one making the 1st person singular in -mi (like Skt. asmi, etc.) and the other in -hi. The relation between the two is not at all that which exists between the present and perfect in other IE languages, but the endings of the -hi conjugation are comparable in some ways to the per- fect endings of Sanskrit, Greek, etc., so that while the detailed relation of the two formations remains obscure, there is general agreement that some definite connection exists between them. The fundamental meaning of the perfect, as it emerges from a comparison of Sanskrit and Greek, and is confirmed by the evidence of the other IE languages, is that of state as opposed to process which is expressed by the present : e.g. bibhdya * he is afraid ’ as opposed to bhayate * he becomes afraid ' ; ciketa 4 he is aware of, knows ’ : ceiati * he becomes aware of, notices ' ; THE VERB 298 tasthau ' stands J (permanently) : tisthati ‘ takes his stand etc. Closely related is the meaning of continuous action seen in such examples as : nd srdmyanti nd vi muncanti ite vdyo nd paptuh * They do not become weary or stop, they fly (continu- ally) like birds The perfect is thus in origin a special kind of present tense, not a preterite form, and in such cases it is normally to be translated by the English present. Its develop- ment to a preterite takes place in two stages, both of which are represented in the Vedic language. The first stage is the de- velopment of a sense which is rendered in translation by the English perfect. Since a state is normally the result of a pre- ceding process, it was natural that the perfect should be used to express the fact that such an action had already taken place. As examples of this use we may quote : ydt stm Aga£ cakrtna tdt su mrlatu * whatever sin we have committed, let him forgive that and ydthd jaghdntha dhrsata purd cid evA jahi sdirum asmakam indr a ' as thou hast boldly slain (enemies) in the past, so slay our foe now, O Indra The difference in meaning be- tween this use of the perfect and the aorist remains clear, be- cause the aorist is confined to those actions which have taken place in the immediate past, while the perfect indicates comple- tion of the action regardless of the precise time. The final step takes place when the preterital sense acquired by the perfect in contexts like these becomes the predominant sense, with the result that the perfect becomes a tense of narrative u r ith a meaning that does not differ materially from that of the imper- fect. The last usage has already become quite common in the Rgveda : ahan dhim dnv apds tatarda, ‘ (Indra) slew the dragon, he penetrated to the waters The same usage has developed widely in other sections of Indo-European, notably in Italic, Celtic and Germanic. The pluperfect, the augmented preterite of the perfect, is rare even in the Rgveda , and it quickly dies out. It appears to have no specific meaning proper to itself, being used as a rule in sense of the imperfect (narrative), occasionally in that of the aorist. Mood. Five moods are enumerated in Sanskrit grammar, the Injunctive, the Imperative, the Subjunctive, the Optative and the Precative. In the older language modal forms may be made from all three types of stem, present, aorist and perfect, without any apparent difference of meaning. In the classical language THE VERB 299 injunctive forms are confined to the aorist stem, imperative and optative forms to the present stem, while the subjunctive, except for such forms of it as are incorporated in the impera- tive, dies out. The precative is in the later language connected with the root aorist in the active and with the is-aorist in the middle. Earlier, while confined to the aorist it appears assoc- iated with a greater number of forms. Injunctive . The so-called injunctive is not strictly speaking a separate morphological category at all. In form the injunc- tives are unaugmented aorists and imperfects : dhat, vrnak t etc., but forms of this nature may also be used as simple pre- terites of the indicative, in the same way as the augmented forms. Comparative study of the languages makes it clear that the augment was originally an optional prefix in the formation of these preterites, and that the unaugmented forms are the most ancient. The fact that these forms can also be used in an 4 injunctive 1 sense, that is to say as futures, imperatives and in the expression of a wish, takes us back to an extraordinarily primitive state of the language when owing to the undeveloped nature of the verbal system one form had perforce to serve in many meanings. The three main meanings of the injunctive may be briefly illustrated by a few examples : (1) as future : ko no mahyi dditaye punar dot 4 who will give us back to the great Aditi Indrasya nu viryhii pra vocam 4 I will now proclaim the manly deeds of Indra * ; (2) as imperative : gdrbham A dhdh * deposit the embryo pari tvesdsya durmatir maht gat ' let the great malevolence of the impetuous one avoid us Where special forms of the imperative do not exist, in the 2nd person plural, etc., the injunctive remains the only way of expressing the imperative, and such forms are incorporated into the impera- tive system. The injunctive is used exclusively with ma to express prohibitions : ma gdh 4 do not go *, etc. This construc- tion remains in the classical language where it is the only type of injunctive (except for those forms incorporated in the im- perative) which continues to exist ; (3) in the expression of a wish : agnim hinvamu no dkiyas tena jesma dhdnam-dhanam 4 let our prayers urge Agni ; may we continually win wealth through him Imperative. The imperative possesses distinct forms only in a certain number of persons and numbers, namely in the 2. 3, THE VERB 300 singular and 3. plural. Of these the forms of the 3. sing, and plur, have originally evolved from injunctive forms by the addition to such forms of a particle -u : vahatu from vahat + u. The forms of the 2. 3. dual and 2, plural are unaltered injunc- tive forms. The forms of the first person in the later language are subjunctives which have been incorporated in the impera- tive system ; they do not belong to this system in the earlier language. The imperative expresses commands just as the in- junctive may do, but it is not used in the other senses which the injunctive has. It is also used more commonly in this sense than the injunctive. Subjunctive. Morphologically the subjunctive arises by the evolution of a particular kind of injunctive. It is natural there- fore that its sphere of meaning should correspond in general to that of the injunctive. Like the injunctive the subjunctive may be used (1) simply in a future sense : prd nu voed suidstt vdm 1 I will now praise you two at the libations ", uvisa usa ucchac ca nu 1 Dawn has shone forth (in the past) and she will shine forth now ' ; (2) equivalent to an imperative ; a vdm vahantu . . . divah y pibdtho asmd madhuni f let your horses bring you here and drink ye draughts of mead with us ' ; (3) in the expression of a wish : pari no helo vdrunasya vrjydh t urum na indrah krnavad u lokdm 1 may the wrath of Varuna avoid us, may Indra procure for us a wide space While the sphere of the subjunctive coincides with that of the injunctive, there is a difference of emphasis inasmuch as the future meaning is much more prominent in the case of the subjunctive. The subjunc- tive is, in fact, the normal means of expressing the future in the Vedic language. There are also certain important syntactical differences between the use of the two moods. The injunctive for instance is alone used to express prohibitions in connection with mi. Another important difference is that the injunctive is rarely used in subordinate clauses (relative, conditional, etc.). On the other hand the use of the subjunctive is very widely developed in this connection, and even more distinctively so in the related languages, whence the name of the mood. Optative . The optative differs from the moods so far de- scribed in that it is formed on the basis of a special stem formed by the suffix yd/%. Its original meaning appears to have been the expression of a wish (from which its name is derived) and this meaning is well preserved in Sanskrit : vaydm syama pdtayo THE VERB 301 rayvndm ' may we be lords of riches From this there arises a potential meaning (the mood is sometimes so called) which from the testimony of the various languages was already well estab- lished in the Indo-European period : ydd agne sydm ahatn tvam , tv dm vd ghd sya aham, syus te satyd ihasisah, ‘ if, Agni, I were you or you were me, then your prayers would come true A usage widely developed in Sanskrit is that of the prescriptive optative, which appears largely in law books and similar texts : sdmvatsarikam dptais ca rdstrdd aharayed halim , ‘ he should have the annual tax collected from the kingdom by suitable officials Precative. The precative is formed on the basis of the optative stem by the addition of s to the optative suffix, pro- ducing the combination yds /is. Its use is in all cases confined to the expression of a wish : bhdgo me agne sakhyd nd mrdhydh f may my good fortune, O Agni, not relax in (thy) friendship yd no dvdsti ddharah sds padista * may he who hates us fall down §3. The Verbal Stem The foregoing analysis of the root shows how from the earliest period the verb could be built optionally on the root itself, or the root already provided with suffixes. These suffixes are in all cases identifiable with the corresponding suffixes which appear in the formation of nouns. In these formations of the oldest type the suffixes are completely incorporated and new, fuller roots are created. Besides these suffixes there exists a series used only to form the present stem but excluded from the other verbal formations. These are likewise identifiable with corresponding nominal suffixes, e.g. dhrsndju of the fifth class, from dhrs - ' to be bold with the adjectival stem dhrsnu- 1 bold The general structure of nominal and verbal stems runs closely parallel. Both may be based either on the simple root or on the root provided with suffix. The suffixes may be simple or compound and the compound suffixes arise always by the- addition of one suffix to another. The formations are divided into non-thematic and thematic classes, the latter in both cases progressively increasing in importance. Verbal stems may be accented on the root or the suffix, e.g. in the case of non-thematic stems vdmiti , jdsma as opposed to &rndti> prn&ti , and in the case of thematic stems, bhdvati as opposed to tuddti . The fact that such accentual difference corresponds to no difference of meaning in the* verbal system, but clearly does THE VERB 302 so in the nominal system, may be held to indicate that these steins are primarily nominal in origin. In the classical language the present stem of a verb is nor- mally made according to one only of the ten different types. In the Vedic language greater latitude is observable. While in general the distribution of the roots among the ten present classes corresponds to that of the later language, a large number of roots is found which form their present tense according to two, three, or even more different types. Such cases are illus- trated by krs- ‘ to plough I karsati, VI krsdti ; jf- ' to grow old I jdrati , IV jiryati ; dd- ‘ to divide ', II dati, IV dydti ; dhu- ‘ to shake V dhunoti, VI dhuvdti ; pf- ‘ to fill III pipdrli , IX prndti ; bhi- * to fear I bhayate, III bibkdti ; rdh- ‘ to prosper IV fdhyaii , V rdhnoti , VII subj. rnddhat ; if- 1 to cross, overcome I tarati , III ptc. titrai -, VI tirdti, VIII tarute. To a large extent this variation of stem is not associated with any difference of meaning, hut sometimes the difference be- tween transitive and intransitive is connected with the use of alternative stems ; jdvate ‘ hastens, is quick jundti 1 speeds, urges on tdpati ' heats tdpyati 1 becomes hot ", pdcati f cooks, ripens (trans .) pdcyate 1 becomes ripe In particular an intran- sitive sense tends to be associated with stems of the fourth class. It is clear from the comparative study of the IE languages that this variability of stem formation was even greater in the prehistoric period. Avestan shows in some respects even greater variety than the language of the Veda. Furthermore the discrepancies in stem formation between the various IE languages (Skt. rindkti : Gk. Xelrno, etc., etc.) shows that in the earlier IE period the greatest freedom prevailed in the formation of present stems. In all this variety of stem formations no inherent difference of meaning was attached to the various types, just as in Sanskrit the ten present classes are equivalent in function. At the same time there arose early in Indo-Euro- pean a tendency for certain of these formations to acquire a special meaning and function. The suffix -sco- for instance acquired an inchoative meaning which is represented in a variety of languages. In Hittite formations in -nu- (corresponding to the Skt. fifth class) acquired a special function as causatives, a development which is not shown by the other languages. In Sanskrit those present formations which acquired a special meaning became the foundation of what is called secondary THE VERB 303 conjugation, namely causatives with stems in - aya t desidera- tives with reduplicated stems in sa , intensives with strong re- duplication and radical stem or stem in ya. These were origin- ally conjugated in the present system only and it is a special development of Sanskrit which allows them to be inflected in other parts of the verbal system. With these must be classified the passive which is a special development of the fourth class and which arises from a tendency of stems of that class to specialise as intransitives. §4. Accent and Apophony of Verbal Stems By a rule peculiar to Sanskrit, the like of which is not trace- able in other IE languages, the verb is unaccented in an inde- pendent clause, except at the beginning of such a clause and under certain special conditions ; it retains its accent in de- pendent clauses. When accented the verbal stem has an un- changeable accent in the case of thematic formations, which in this respect agree entirely with the nominal thematic formations. In the non-thematic formations the accent varies between stem and personal ending, and this variation corresponds to a varia- tion between the guna (occasionally vrddhi) grade of the stem and the zero grade. The general rule is that in the indicative the stem has the accent and the guna grade in the three persons of the singular active, and that in the dual and plural of the active and in the whole of the middle the accent is on the termin- ation and the stem appears in its weak form : dvisti f hates ', 3 pi, dvisanti , 3 sg, mid. dvistd, yundkti ' joins 3 pL yunjdnti, 3 sg. mid. yunktd . 'Exceptions to the rule (e.g, in the s-aorist) are comparatively rare. This old IE system appears also in other languages (e.g. Gk. etfju t/xev), though nowhere so clearly and consistently as in Sanskrit. §5. Augment and Reduplication In addition to a large variety of suffixation Indo-European made use of two types of prefixation in the formation of tense stems, Augment and Reduplication. The augment (IE e-, Skt, a -) is prefixed to the various pre- terites (imperfect, aorist, pluperfect, conditional) to indicate past time. It is found in Indo-Iranian (Skt. abharai ), Greek (e<£ep€), Armenian (eber), and Phrygian (eSa*? 'constructed'), but it is absent in the rest of Jndo-European. It is thus an THE VERB 304 important feature in connection with the dialectal divisions of Indo-European, since it is clearly of late origin, and has estab- lished itself over only part of the IE linguistic area, among dialects which for other reasons also may be held to have been contiguous. Even where it established itself it existed origin- ally only as an optional formation, augmented and unaugmented forms being optionally used. The unaugmented forms were of course alone used in an injunctive sense, but they could be also used as preterites just like the augmented forms. The co- existence of augmented and unaugmented preterites is a char- acteristic both of the earliest Greek and the earliest Sanskrit. It is only in the later stage of both languages that the augment ceases to be optional and becomes obligatory. In Iranian the augment is regularly used in Old Persian, but only rarely in Avestan, where the unaugmented type of preterite has mainly prevailed. In the early stage of Middle Indo-Aryan, which still preserves an old preterite made up of imperfect and aorist forms, the old Vedic freedom of usage is maintained, but the unaugmented instead of the augmented forms become the most common. The augment seems in origin to have been a separate word, namely a particle i meaning ' there, then J which came to be compounded with the verb. It invariably bears the acute accent whenever the verb is accented. When the verb is com- pounded with a preposition, it always appears between the preposition and the verb: samabharat, etc., and likewise in Greek. An irregular sandhi appears when it is combined with a root beginning with i, u or r [aicchat ‘ wished aurnot ' covered ' ardhnot r thrived ' from icchdti, urnoti, rdhndti, with vrddhi in- stead of the expected guna, and this indicates that up to a late period it was pronounced as a separate syllable with hiatus (. aicchat , etc.). On the other hand its coalescence with initial a (IE e , a, 0) appears to be ancient, judging by parallels between Greek and Sanskrit (Skt. as ' was Gk. Dor. 77s, Skt. ijat ' drove ", Gk. Dor. 3ye). Before roots beginning with v t y, n and r the augment may appear as long d in the Vedic language (dvrnak, ayunak , etc.). The reason for this is not very clear but a parallel phenomenon in the case of initial v- is found in Greek (Horn. Att, 7jSei). Reduplication consists normally of the repetition of the initial consonant of a root with a vowel which may or may not THE VERB 305 be the same as the radical vowel. It appears in one class of present (the third class), in the reduplicated aorist, in the per- fect, in the desiderative and in the intensive. The main varieties which will be detailed below under the separate formations are as follows : (1) Reduplication with the vowel a (IE e) : dadhiti 1 places tatdna ' stretched cf. Gk. ycyove, Lat. pepigl , etc. (2) Reduplication with long d ; jdgarti , jdgara 4 is awake cf. Gk. Sr}Sexo.raL 1 they welcome etc, (3) Reduplication with the vowel i when that is not the vowel of the root ; Hsthati 4 stands didrksate 4 desires to see cf. Gk. tcrr T)fu, ylyvofiai, Lat. sisto, etc. (4) Similar reduplication with long i : djljanat t 1 gave birth to cf. Av. zizanmti. (5) Reduplication with weak form of vowel of roots in diphthongs : juhoti 4 sacrifices bibhdda * split cf. Lat. pupugi, scicidl . (6) Intensive reduplication with guna vowel of such rpots and similar reduplication with repeated final r t n, etc. ; nenikU 4 washes dMiste ‘ points out vdrvrtaii ‘ they turn (continu- ally) ' ; cf. Av. naenizaiti, daeddist, etc. This involves the com- plete repetition of roots containing only two consonants: nondva 4 roars mightily janghanti ' smites violently (7) Such intensive reduplication with 1 suffixed to the re- duplicating syllable : bhdribharti 3 sg., bhdribhrali 3 pi. This type alternates with one in which the i is suffixed after the root : johavlti 1 calls loudly etc. (8) Initial a may be reduplicated producing d {dsa 4 was ') or some more complicated process may appear. In the case of roots beginning with two consonants the first is reduplicated with the exception of the combination 5 y occlu- sive : susrdva 4 heard cf. Av. susruma, Gk. ice/oWe, sisriye 4 rested on J , cf. Gk. /ceVAirat, sastndra 4 remembered sasnau ' bathed etc. When the root begins with the group 5 + occlu- sive, the occlusive is repeated in Sanskrit : tisthaii 4 stands caskdnda 4 sprung J , pasparia 1 touched On the other hand, s (> h ). appears in the reduplicating syllable in such cases both in Iranian and Greek. The same type of reduplication appears also in Lat. sisto and in Celtic (Ir. - sescaind from skinnim ‘ jump out cf. Skt. skand -). Yet other varieties of reduplication in the case of these groups appear (a) in Lat. perfects of the type THE VERB 306 steti, scicidt, spopondl , which preserve the consonant group in the reduplicative syllable but simplify it in the root, (b) in Gothic where the full group appears in both root and reduplica- tion : skaiskaip [skaidan 1 cut ’). In view of this variety it is unlikely that there was any universally consistent usage in the IE period. Certain sound changes have affected the reduplicated forms in Sanskrit. By the rule which prevents a succession of two aspirated consonants, a non-aspirate is substituted in reduplica- tion for an aspirate ; dddhdti , etc. In the case of A<IE gh the non-aspirated form appears as; : jahara. Since a in a redupli- cating syllable was originally e, the old velar series is palatalised before it ; cak&ra , jagdma, jaghana. The same treatment appears regularly when the vowel of the reduplicating syllable is i ( ciktrsali ) and it is applied analogically in the case of u : cukopa, §6. Personal Terminations Sanskrit, like the parent Indo-European, has two sets of personal terminations, one for the active and one for the middle voice. These two sets embrace further subdivisions which appear in different parts of the verbal conjugation. In the present-aorist system the so-called primary endings appear in the present and future, while a different series, the secondary endings, appear in the imperfect, aorist and optative. The sub- junctive has optionally either. The perfect endings, 'where they differ from the above, do so more fundamentally than the primary and secondary' endings differ from each other. The imperative has special endings only in the 2 singular and in the 3 singular and plural. The subjunctive has a separate termina- tion in the 1 singular, which is old, and some special middle endings w r hich are an Indian innovation. The primary, secondary and perfect endings are listed in the following table : 1 s. Active d. A. Primary ph s. Middle d. pi. I mi vas mas e vahe make 2 si ihas tha se athe dhve 3 ti tas anti te die ante , ate 1 For table of imperative and subjunctive endings see under the respective sections. THE VERB 307 B. Secondary I am, m va ma i, a vahi mahi 2 s tarn ta thus dthdm dhvam 3 t tam an, ur ta dtdm ant a, at a ran C. Perfect 1 a va ma e vahe make 2 tha athus a se dike dhve 3 a atiis ur e ate re Active Terminations: 1 Sing . P. Skt. as mi ' I am ', Gk. eip.1, Lith. esmi, Hitt. esmi ; emi 1 I go Gk. elpu, dddami f I give Gk. &{&a>pu, etc. This ending was originally confined to the non-thematic classes, and a different ending -0 (-on) appeared in the thematic classes : Gk. pipco, Lat . Jero, Goth, baira . Some such forms are preserved in Iranian (Av. spasyd : Lat. specio), but usually in Iranian, and always in Sanskrit, mi from the non-thematic verbs is added to the older form : Skt. bhardmi , Av. bar ami. S. dbharam, Gk. epepov, dgdm , Gk. eftrjv, syam 1 sim O. Lat. stem. Non-thematic verbs have the fuller ending -am, dsam ‘ I was *, O. Pers. a ham, as opposed to Gk. a<m in Horn. rja. A similar difference between the two languages was observed in the case of the acc. sg. of non-thematic stems. Pf. Skt. veda, Gk. ol8a; daddy sa, Gk. hihoptca. Sbj. The Vedic language has either a or dni, brava ' I will speak bhardni 4 I will bear '. In the classical language, where these formations have entered the imperative system, only the fuller ending dni is used. The ending -d is identical with the -6 which in other languages appears in the present indicative of thematic verbs as well as in the subjunctive : Gk. ayto, pipw, etc. The extension -ni appears only in Indo-Iranian and is of uncertain origin. 2 Sing. P. e'si 4 you go Lith. eisi, bharasi 4 you bear Av. barahi ; cf. O. Russ, veil si ‘ you command Hitt, iyaii 1 you do S, dsthds, Gk. lords', dbharas, Gk. epcpe?, bhares Gk. pipois, Goth, bairais. Pf. vettha, Gk. otaOa, Goth, waist , daddtha, Av. dadddd. Impv. ihi 1 go ’ ( <*idht), Av. idt, Gk. Wi t jahi 'slay' (han-) t Av. jaiSi, viddht 4 know Gk. ta9 1 . In thematic verbs the simple stem serves as the second person singular of the imperative : bhara , Gk. pipe, Goth, bait ; Lat. lege, etc. THE VERB 308 3 Sing. P, dsti, Lith. esti , Gk. e<m, Hitt, eszi, (zi<ii) ; 'slays *, Av. jainti , Hitt, kuenzi ; hhdrati, Av. baraiti, O. SI. (beside herein), S. abhor at , sya/, cf. Av. 6ara/, Gk. (with loss of -/) efape, fa'poL, Lat. erat , sf/ (O. hat. sfetf), etc. Pf. daddrsa, Gk. hihopKe. Impv. Hitt, ebu Met him go hh ara tit, O. Pers. baratu. An alternative ending -tfoi appears in Greek and Latin (eoraj, Forms correspond- ing to this are found in Sanskrit, e.g. vittat ( - Gk. Iotuj) but they are used for both the second and third persons, and for all numbers. A different kind of inflection in the primary endings of the 2, 3 sing, appears in Gk. thematic verbs : 2 a yet?, 3 ayei. A comparison with Lith. vedi ' take : you take 1 shows that the 5 of the 2 sing, is a later addition, and that the two persons were originally identical. They contain no personal terminations, only an appended i indicating present time. Such an i, un- associated with a personal ending, is found in Hittite verbs of the -hi class : aki 1 dies ddi 1 takes Sanskrit has innovated here by applying the endings of the ^^--conjugation, just as in the 1 sing., but in this case the innovation is more widely shared by other IE languages, Lat. agis f agii, Goth, bairis , bairip, etc. 1 Plur. P, (a) imds ' w r e go bhdrdmas ' we bear cf. Gk. Dor. tpLts, Lat. tnius, ferimus, O. SL damn, neseniu , etc., (b) smdsi 1 we are O. Pers. aniahi, bhdrdmasi , Av. bardmahi , cf. O. Ir. ammi ' sumus bermai, bermi 'we bear S. and Pf. dbhardma, sydma , Av. hydmd , vidmd ' we know Goth. witum. Forms with long vowel which appear in the Vedic lan- guage, particularly in the perfect {vidmd, etc.) appear to be ancient, and not merely metrical lengthening, on account of the occurrence of similar formations in other languages : Lith. sukome-s (reflexive), Goth, batraima opt. (out of °me or O mo). The variations between IE-ra^s and -mos were due to dif- ferences of accentuation, like the similar phenomenon in the genitive singular : originally *imes but *bheromos . The alterna- tive ending -masi appears in the Vedic language beside mas , but it is disused in the classical language. In Iranian the correspond- ing mahi has come to be exclusively used as the primary ending. In Greek there is an alternative termination -p,ev, used both as a primary and secondary ending. It was customary to regard the final-v of this form as ephelcystic, and to equate the Gk. ending with the Skt. secondary ending, but it is now clear from THE VERB 309 Ilittite that this is not so. The Hittite terminations are : P. went, meni (with i appended as in Skt. masi) , S. wen, men . The variety with w- is related to the dual endings of other IE languages ; the forms with m- appear after u ( arnummeni ‘ we bring etc.) and the secondary form -men corresponds exactly with the Greek ending. In Indodranian this form of ending is traceable in Khotanese: hamdmane ‘ we may become ' (sub], mid.). 2 Plur . P. (a) bhdratha, cf. Av. xsayaOa, (b) vadathana 4 you speak S. (a) dbharata (impf.), bhdrata (impv.), cf. Gk. <f>ip€T€, Lat. ferte (impv.), Goth, bairip , O. SI. berete, (b) djahatana ‘ you abandoned hantana ‘ slay ' (impv.), cf. Hitt, kuenten pret, and impv. (kuen- ‘ slay ’). Pf. cakrd (hr- 4 to do J ), vidd. The primary endings with aspiration (- tha <tne) do not appear out- side Indo-Iranian. The other languages have normally one form which serves as both primary and secondary ending, and this corresponds to the secondary ending of Indo-Iranian. Hittite has evolved a distinction between primary and secondary ending here in quite a different way (P. teni, S. ten). The longer forms were analysed as tha-na and ta-na, the na being regarded as an appended particle, and the whole form as a Sanskrit innovation. In view of Hitt, -ten we should analyse rather -tan-a of which tan corresponds exactly to Hitt, -ten, and the a is simply a thematic enlargement such as is found elsewhere in the formation of words. The perfect form is properly without termination and nothing like it is found outside Indo-Iranian. 3 Plur. P. bhdranti, Gk. Dor. <f>epovn, sdnti „ Gk. Dor. ivrl, Lat. sunt, Goth, sind, O. SI. sgti (beside sgtu), Hitt. aSanzi ; ghndnti * they slay Hitt, kunanzi ; dddati ' they give cf. Gk. (Delph.) KadeoraKart, (Horn.) AeAoy^acrt. S. (a) abharan , Av. barm, Gk. e^pov ; (b) ddadur (impf.), ddhur (aor.), syur (opt.). Cf. Av. adard ‘ they made hydra 1 they might be jamyaras 4 they might come Hitt, wekir * they wished ekuer 4 they drank Pf. dsur 4 they were cf. Av. dykara, cikitur 1 they are aware of cf. Av. dikoitaras. Impv. bhdrantu, cf. Hitt, iyandu. The full form of the ending varied between -onti and -enti according to accentuation, like the variation between -mes and -mos in the 1 plur. Also due to accent is the weakening to -pti> ati which appears in Sanskrit and Greek. A corresponding w r eak form of the secondary ending, -at<nt, appears in Iranian : Av. dadat , jlgarazat . In such cases Sanskrit always has the 3io THE VERB alternative ending -ur. The secondary ending was originally - ant which has been reduced to - an by the normal cause of phonetic development in Sanskrit as in most of the related languages (Gk. -ov<ovr t etc.). The alternative secondary ending -ur appears in the imperfect of the reduplicating class (ddadur), in the imperfect of root stems ending in -d (dyur : ytiti ‘ goes in non-thematic aorist stems and in the perfect. Outside Indo-Iranian r-endings of the 3 plur. appear in Hittite, Latin [dixere] and Tocharian ( kdtkar ' they arose tnrasar ' they forgot '). In Iranian this r may be enlarged by an additional element s (cikoiterds, etc,). The form of the Sanskrit ending shows that it also originally contained this enlargement, since -ur has developed out of -rs here in the same way as it has done in pitur <*pitfs. Of the various types of r-formation which Iranian preserves, one has been generalised in Sanskrit at the expense of the rest. i Dual. P. bhdmvas , cf. Goth, bauds (<*bherdwes), svds ‘ we two are The corresponding ending in Iranian is an extended -vahi (cf. rnasi beside mas) : Av. usvahi 1 we two wish ' ; a form -vast after the style of mast does not appear in Skt. S. and Pf. ; impf. dbharava, opt. bhdreva , pf. vidvd, cf. A v.jvdva ( we two lived Lith. pret. sitkova , -vd-s, O. SI. pres jesve, vezeve, Aor. vezovi, Goth. opt. bindaiwa . As in the i plur. only Indo-Iranian distinguishes between primary and secondary terminations. The variation between the long and short vowel (Skt. va : SI. ve) was noticed also in the plural. The fir&t person of the dual is found only in Indo-Iranian, Balto-Slavonic and Germanic. Hittite contains a termination which is related to these forms, P. weni, S. wen, but it is used as a plural side by side with the rarer -mem /men. The most satisfactory explana- tion of this is that there existed originally in IE parallel forma- tions beginning with w or m which were optionally used as I plur. ending. Hittite has preserved this state of affairs but restricted the use of the men- termination to stems in it. The above-mentioned languages have kept both types of ending but specialised as duals the ^-variety. About the remaining IE languages it is not possible to say anything owing to lack of evidence. 2, 3 Dual . P. 2 bhdrathas , 3 bhdratas , cf. A v. yuidyado, baratd - Goth, bairats , In Avestan no 2 dual is found ; the terminations to and -83 are used promiscuously as endings, of the 3 dual, the THE VERB 31 * two forms having been confused. The Gothic form is used only for the 2 dual. S. 2 dbharatam , Gk. tyeperov, 3 dbharatam , Gk. ifaptrrjv. Pf. 2 cakrdthur , 3 cakrdtur , cf. Av. yaetaiar 9. With the primary terminations of the 2, 3 dual it is possible to compare the -its ( <-tes) of Lat. legitis , etc., which is used as 2 plur. As in the first person the dual endings seem to have arisen by the specialisation in dual usage of a variant form of the plural ending. It is also noteworthy that the variation between t and th serves quite a different purpose in the two numbers, to distinguish 2 from 3 in the dual, and to distinguish primary from secondary in the plural. Secondaiy adaptation in both cases would account for this difference. The 3 dual is not of independent origin but merely a variant in form of the 2 dual. The same form may appear for both as in the Gk. primary ending (cftcpeTov). In Balto-Slavonic there appears an ending in -id which in Lithuanian appears only in the second person (sukata, sukato-s) in Slavonic in both (O. SI. 2, 3 vezeta). It corresponds to the secondary ending of the third person in Sanskrit {-tam : Gk. -ttjv, Dor. -rdv) to which -m is a secondary addition. The perfect endings are an Indo-Iranian creation with -ur (Ir. ar) introduced from the 3 plur. Middle Endings 1 Sing. P. bruve 1 I speak Av. -mruye, ydje ‘ I worship Av. yaze. A different formation with -m- as in the active of mi-verbs appears in Greek : <f>£pop.ai, etc. S. non-thematic, dduhi f I milked akri 4 I have done Av. aoji ( aog - 4 to say ’), nwnghi— Skt. (a)mamsi (man- 4 to think', s-aor.). Thematic stems have -e as in the primary system, dbhave> etc. Greek has a quite different ending : ifapofjurjv, Dor. °fidv. Pf. susruvd, Av. susruye with same ending as primary system. Sbj. mdmsai, Av. mdnghai, ydjai, Av. yazdi. Opt. tanmyd , Av. tanuya. 2 Sing. P. bharase, cf, Av. pdtgsahe ‘ you ask Gk. Repeat <f>epr) t Goth, bairaza ; datse ' you give Gk. SiSoa-ai. S. dkrtkas, adhatthds , etc. In contradistinction to Sanskrit, Iranian and Greek both agree in having forms representing IE -so : Av. -zayayha, Gk. €<f>£peo, £p£pov. These bear the same relation to the primary ending -sai as the 3 sing, secondary -to to primary -tai. Pf. same as primary, ririksi , dadki$e l cf. Gk. AeAtu/rcu, SeSoccu. Impv. bhdrasva, Av. barayuha. 3 Sing. P. bhdrate, Gk. (fxpcrai , Goth, bairada f dste * sits THE VERB 312 Gk, tforau. In the Vedic language there occur some forms with- out the the termination being identical with that of the first person as in the perfect : duhi t idye, srnvd, etc. In view of the Hittite middle formations of the third person of the type esa , kisa (cf. aduha below), this type must be regarded as ancient and not as an importation from the perfect system. S. ab karat a, Gk. €<f>€pero, opt. bhdreta, Gk. pepoiro. The 4a which appears in Hittite as primary ending aria ( stands up kitta ‘ lies 1 beside artari, kittari) is equivalent to the Sanskrit secondary ending. Rare forms without t corresponding to the presents duhd, etc., are aisa and aduha (is- * to rule duh- ' to milk ') r cf. Hitt, primary esa, kisa , etc. Pf. dadhd cakrd, etc. On the other hand Gk. has 4- as in the present, SeSorcw, etc. Impv. bhdratdm , dhattam , cf. Av. v Wdzyatam. This form and the corresponding plural antdmjatdm are without parallel outside Indo-Iranian. A form without 4-, corresponding to the indica- tive endings P. -e, S, -a, appears occasionally in the Veda : duhdm . 1 Plur . P. Pf. ydjdmahe , Av. yazamaide, brumahe ' we Say Av. mrumaide , cakrmdhe , etc. S. dbharamahi, cf. Av. varzmaidi ( var - 1 to choose '), etc. Sanskrit -h- is from -dh-, as in the im- perative -hi, etc., the original Indo-Iranian terminations being

• -madhai and - madhi . Closely related to these forms but differ-

ing in the matter of the final vowel is Gk. -pLtda <*medha, which serves as both primary and secondary ending : </>cpo- tytpofieBa. Greek has also a fuller ending -peaftz, with which Hitt, -wasta is to be compared, there being the same alternation of m and w as appears in the active (men/wen). 2 Plur. P. Pf. bhdradhve , dadidhvi , cf. Av. mzrmgdduye (marzk- 1 to destroy ’). S. dbharadhvam , cf. Av. ddrayahwzm, etc. The Greek ending -<j8e (primary and secondary) may be related presuming it is out of -zdhwe, and more closely the dual ending -udov ( <zdhwom ) the dual use of which is, as elsewhere, a secondary adaptation. The Hittite termination, primary dutna (pahhasduma * you protect '), S. dumat is more obviously con- nected, -dum- being the weak grade corresponding to - dhvam . 3 Plur . P. (a) bharanie , Gk. fitpovraa, asate ‘ they sit ', Gk. Horn. Tjarax (<*esytai) ; (b) Ure ‘ they lie Av. soire , saere, duhrd , sunviri , with the same endings as the perfect. A combin- ation of a and b appears in i&rate, duhrate. S. (a) abharanta , Gk. €<f>€pQVTo, dsata f they sat Gk. Horn, i^aro ; (b) Three THE VERB 313 varieties of f -ending appear: (1) - ra : dduhra , (2) -ran: dduhran , dseran, opt. dstran, hhdreran , (3) -ram : asasrgram > cf, Av. vaozirsm (vaz- ' to carry ’). The types (2) and (3) are clearly enlargements of type (1) which must be regarded as most original. By combination of (a) and (b) arise the termina- tions -ranta and - rata : dvavrtranta ( vrt - ' to turn '), bharerata . Outside the optatives in -frara, the r-endings are compara- tively rare and archaic formations in the Vedic language beside the normal ^-formations. They are almost completely dis- carded in the classical language. It will be observed that the r is identical with the r which appears in the third plural of the active, and it is the elements added to it which characterise the terminations as middle (- e in duhrd , etc., -a in aduhra). Pf. Ex- clusively r-en dings : duduhri , cakriri , etc., cf. Av. idxrare (variant -arai which does not occur in Skt., cf. Av. -ard active which is likewise missing in Skt.). 1 Dual. P. Pf. bhdravahe , cakrvdhe, S. dbhardvahi . Avestan has only - vaidi : dvaidt ( dav- 3 i.e. du-vaidi). The Hitt, plural ending - wasta has formally the same relationship to Indo- Iranian - vadhi as Gk. -fitoda to - madhi . 2, 3 Dual . P. Pf. thematic 2 bhdrethe , cdrethe , cf. Av. 3 dual cardtOe , 3 bhdrete- cf. Av. visaete ; non-thematic 2 bruv&the , mamndthe, 3 bruvdte , mamndte . S. thematic 2 dbharetham , 3 dbharetam , cf. Av. jasaetzm (with short vowel of termination as opposed to long vowel in Sanskrit) ; non-thematic 2 airw- vdthdm (sru- ' to hear *), 3 airuvdtdm , Av. asrvdtzm. The variation between Iranian -tom and Skt. -tom in the third person corresponds to a similar difference in the active (Av. jasatzm, Skt. dgacchatam). No corresponding endings appear outside Indo-Iranian. The Greek terminations, P. 2, 3 -cr#ov, S. 2 -crflov, 3 -oOrjv are connected with the second plural of the middle. On the other hand the Indo-Iranian forms are con- nected with the corresponding active terminations of the dual. The same variation between th and t appears between the two persons in the primary endings, and in the middle this is intro- duced also into the secondary endings. The final -e of the primary endings is the same -e characterising the middle which appears throughout the primary system. The influence of this -e of the middle is responsible for the substitution of e for the a of thematic stems (abhavetdm as opposed to active dbhavaiam , Av, fasaitvm beside jasatzm). Non-thematic verbs are distinTHE VERB 314 guished by an a of the middle ending, but examples like Av. dazde 3 du. pf. and Skt. ciMthe indicate that this is an innova- tion. It can only have come from formations of roots in d like dad&the , daddte where the d is originally part of the root as in 2 sg. daddtha (beside daditkd) and 2 pi. dddhdtana (beside dhattana ) . §7. Structure and Origin of the Terminationai. System A comparison of the primary and secondary endings shows that from the historical point of view they are incorrectly named. It is the ' secondary ' endings that are primary, and vice versa. The relation of the series -m, -s, -t, - an(t ) with the primary -mi, -si, -ti, - anti can only be explained by the assump- tion that in the latter series a particle -i indicating present time has been secondarily added. In the same way in the imperative endings -tu, -antu, a particle -« is added to the same basic ter- minations. This is made clear, among other things, by the fact that these same elements -i and -11 may appear by themselves in formations that have no personal termination, e.g. in Gk. <f>ep€i (fcpe + i), Hitt, sakki ' knows ' and in Hittite imperatives of the -hi class : aku , ddu ( ak - f to die dd- 1 to take '), This addition of -i to -t, etc., implies an earlier period when secondary terminations alone existed ; the ‘ primary ' system, and there- fore the present tense is formed on the basis of the * secondary ' system of the preterites. The unaugmented preterite and the ' injunctive ' form the primary basis of the IE present-aorist system. It does not seem that the distinction between primary and secondary terminations was fully worked out in the IE period. For instance in the 1 plur. and in the 2 plur. Greek makes no distinction (P. S. and this indifference is shared by other languages (O. SI. nesemu, nesomu, Goth, bindam , witum, -budum). The distinction appears in Hittite and Indo-Iranian, but it is effected by quite different means. In Hittite -went, -meni beside -wen, -men is clearly a private innovation modelled on the three persons of the present and the 3 plur. In Indo- Iranian the distinction is effected by the choice of two different forms of the suffix (1 masjma , similarly du. vasjva ) and there is no evidence to show that this variation was connected with the distinction between secondary and primary in the IE period. THE VERB 315 Possibly -ma was originally the perfect ending in view of the frequency of final -a in forms of that tense {veda, vdttha, vdda, 2 pi. vidd) . The innovation Skt : -masi t Av. - mahi , Av. -vahi resembles the Hittite development, but it is quite independent, and expresses no such distinction (P. mas and mast as opposed to Hitt. P, -meni, - went , S. men/wen). Hittite treats the 2 pi. in the same way ( tenjteni ) and here again Sanskrit expresses the difference in quite a different way (4ajtha). In all other lan- guages the distinction does not exist. Apart from lack of sup- port from other languages, the fact that the distinction be- tween t and th is used in the related dual endings for a totally different purpose (2 -thus, 3 -tas) makes it altogether unlikely that the difference between the two forms of suffix was from the beginning connected with the distinction between primary and secondary ending. In the middle the priority of the secondary endings is no less clear. The primary endings all terminate in -e, and the com- parative evidence shows that the extension of this -e to all the primary persons of the middle is a special Indo-Xranian de- velopment. In Greek, which is closest to Indo-Iranian in its verbal inflection, a corresponding -at is found only in the three persons of the singular and in the 3 plural. Elsewhere {-odov, -peda, ^odt) the endings are not characterised by this element, the same ending functioning as both primary and secondary. It is certain that in this respect Greek represents the more original state of affairs, and that in Sanskrit -make and -dhve (and the same applies to the dual endings) are new formations on the basis of mahi and dhvam which originally functioned indifferently as primary and secondary endings. Comparison between Sanskrit and Hittite confirms this, Hittite differs from Sanskrit more than Greek does, and it shares with Italic, Celtic and Tocharian an element r in the middle endings which Greek and Sanskrit agree in ignoring. Nevertheless there exist forms in the two languages which can be directly compared, and these are invariably secondary endings in Sanskrit : Hitt. 3 sg. pr. arta : cf. Skt. dkrta ; 3 pi. aranta : Skt. aranta ; 3 sg. esa : cf. Skt. dduha ; 2 pi. pahhaMuma : cf. Skt, dbharadhvam. The Hittite terminations all belong to the present (primary) system, as opposed to the identical forms in Sanskrit which belong to the secondary system. The formation of the middle endings, as is clear from a survey THE VERB 3i6 of the Sanskrit forms, is remarkably heterogeneous, and its study is made more complicated by the existence in Indo- European of two distinct types, one (Hittite, Italic, Celtic, Tocharian) which makes extensive use of an element r in its formation, and another (Indo-Iranian, Greek, Germanic) which ignores this element. At the same time there is a nucleus of forms, as instanced above, which bridge the gap between the two types. Among the oldest forms we can observe several types. (1) In the 3rd person a thematic variation of the same suffix serves as the middle ending : dkar(t) ; dkrta ; dbharan{t) : dbharanta ; duhiir : aduhra . (2) In the 1st person dual and plural the middle ending is formed by the addition of particles (Skt. -hi<*dhi y Gk. Oa<*dha) to a form of the active ending. The Greek alternation -fiedaj-fieoda, which contains as its first element two variant forms of the active ending (cf. Skt, -mas and -ma), makes this quite clear. Hitt, -waita, with the -w- that elsewhere appears in the dual, but in Hitt, in the plural, is naturally to be explained in the same way as Gk. -fie uQa. )3) In the 2nd plur. an ending which is quite different from the active ending is used. The 2nd sg. -thus is likewise quite dif- ferent from the active but it is clearly connected with the perfect active ending -tha. The final s may be explained as a secondary addition, since -s characterises the 2nd person else- where. The relation between *-thd and -tha is apparently the same as that between mu and ma (IE me t me) of the 1st person plural. The simplest form of the 1st person middle termination in Hittite is -ha (zahhiyahha) which occurs rarely beside the more usual - hart , hahari , This ending is represented in Sanskrit in the 1st person middle of the optative ( bhdveya ). Elsewhere there is a secondary ending -i ( dduhi , dkri) for which there are no parallels outside Indo-Iranian (Av. aojt, nipnght). It is clear from the agreements between Sanskrit and Hittite that the oldest nucleus of middle endings is common IE pro- perty. Further developments based on this show remarkable divergence, since Hittite shares with Italic, Celtic and Tocharian an element r which is not known to Sanskrit and Greek. In Hitt, this r, which appears with the addition of the primary -i of the present, is optional, e.g. 3 sg.' aria and ariari, 3 pi. aranta and arantari . It is clear that it must have been in the same way optional in Indo-European, and that in the further course THE VERB 317 of development it became established as a necessary element in Italo-Celtic and Tocharian, and on the other hand went out of use in that dialectal area of Indo-European from which Indo- Iranian and Greek derive. The primary middle endings of Sanskrit arise in the first place, as in the active, from the addition of -i to the secondary endings : bharaia 4- i> bhdrate . Corresponding to the -e t Greek has -at. This vocalism is most simply explained as due to the ending of the 1st person (-H a + i>ai) from which it spread by analogy to the other persons. In Sanskrit this final - e appears in all the primary forms, but its presence in the dual and in the first and second persons of the plural is due entirely to analogy, and these are to be regarded as the latest parts of the system. The active endings of the perfect are in the singular identical with the oldest forms of the middle endings : (1) -H a>a, vida, cf. bhaveya ; (2) -tha, cf. -thas, Hitt. - ta ; (3) - a (IE - e ), veda , cf. aduha, Hitt. esa. That this is no accident is clear from the frequent cases in which active perfects with intransitive sense appear by the side of middle forms in the other tenses, e.g. Skt. variate : vavdrta, Gk. ylyvo^iat : ydyova, etc. Originally, it seems, the perfect had no distinction of the two voices, and both in form and sense it was closer to the middle than to the active. The development of separate middle forms may be regarded as a late Indo-European feature. In Sanskrit these perfect termina- tions are identical with those of the present, and these, as already observed, are later formations than the corresponding secondary endings. As to the nature of the personal endings it is quite clear that they have nothing to do with the corresponding personal pro- nouns. The theory that these endings are of the nature of suffixed pronouns has often enough been put forward in the course of IE studies, but concrete evidence in the form of de- tailed comparisons is lacking. It is possible to find an -w- in the ending of the 1 sg., and a t in the 3 sg., which are letters that occur in the corresponding pronouns (acc. sg. md : ta-}, but beyond this there is practically nothing. Since no theory can be based on the comparison of one or two single letters, the attempt at explaining the personal endings as suffixed pronouns has to be abandoned. When this is done, and the terminations are analysed in such detail as the comparative evidence will permit, it becomes clear that the elements of which the system 318 the verb is constructed are in the main identical with the suffixes of derivation which are met with in the formation of nouns. This is clearly so in the purely thematic formations which have no ending in the proper sense : 2 sg. impv. dja, 3 sg. impf. mid. dduha, 2 pi. pf. vidd. Such formations are in no way to be dis- tinguished from ordinary thematic nominal stems. The same correspondence is generally seen between the ending of the 3 pi. (Skt. -an(t)lanti, IE -ent'-ont) and the suffix which forms present participles (Skt. bhdrant Lat. ferens). In addition it has been pointed out that the relation of the two kinds of ending in the 3 pi., the above and that consisting of or containing an element r, is reminiscent of the alternation found in nominal stems between r and njnt : Gk. v 8 wp, vBaros, etc. Elsewhere too there are features about the verbal terminations which recall those of the nominal suffixes. There appears to exist the same relation between the terminations of the active and middle of the 3 sg. ((d)kar(t), (d)krta) as is found in the nominal suffixes in krt- : krtd The suffix of the 2 pL contains also a -ta which may be equated with the corresponding nominal suffix ( bharata , cf. the nominal stem bharata-). In Indo-Iranian alone there exists a variant ending -tha which functions as primary ending. It is probably no coincidence that Indo-Iranian is also the only branch which shows a suffix -tha beside -ta in the formation of nouns ( yajdtha - ‘ worship ' : yajata- 1 to be worshipped '). The behaviour of the suffix of the 1 pi. is in several ways reminiscent of the corresponding nominal suffix. In the first place the coexistence of two forms, one beginning with w and one with m, which is seen in Hittite, is matched by a similar duality in the infinitival forms containing the same elements : tiyawar, tiyawanzi ; tarnummar , tarnummanzi. In Sanskrit the suffixes -vant and - mant are found in the same way side by side with similar function. Another similarity between the verbal and nominal forms is seen in the variation of the latter part of the suffix : IE wen/wes , menjmes . This is paralleled by the variations in nominal declension, e.g. in the vocatives rtdvas, patnlvas , tuvismas from the stems rtdvan -, pdtnxvant- and tuvismant Correspondences of this kind make clear the original nature of a considerable section of the verbal terminations. By some process of adaptation, the course of which it is not now possible to follow, certain nominal formations became associated with THE VERB 319 particular persons and numbers, and at least a fair proportion of the existing personal terminations came into being in this way. §8. The Ten Present Classes The roots of the Sanskrit language are arranged by the Hindu grammarians in ten classes, according to the way in which they form the present system, and named after a verb taken as typical of its class. The order in which these classes are placed corresponds to no discoverable grammatical principle, and for convenience of exposition it needs to be rearranged. The verbs are divided into two major types, (a) non-thematic (classes 2 , 3, 5, 7, 8, 9) and (b) thematic (classes 1, 4, 6, 10). A. Non-thematic Presents Root Class (Second or ad- class) Paradigm : (1 dvis - 4 to hate *). Present, Active, S. 1 dvdsmi, 2 dvdksi, 3 dvdsti , D, 1 dvisvds, 2 dvisthas , 3 dvistds, P. 1 dvismds , 2 dvisthd , 3 dvisdnti. Middle, S. 1 dvise, 2 dviksd , 3 dvistd, D. 1 dvisvdhe, 2 dvisdthe, 3 dvisite , P. 1 dvistndhe , 2 dviddhvd , 3 dvisate . Imperfect, Active, S. 1 ddvesam, 2 ddvet , 3 ddvei, D. ddvisva , 2 advistam, 3 ddvistdm , P. 1 ddvisma , 2 ddvista, 3 ddviscm . Middle, S. 1 ddvisi , 2 2 ddvista, D. 1 ddvisvahi, 2 ddvisdthdni , 3 ddvisdtdm , P. 1 ddvismahi , 2 ddvi ddhvam , 3 ddvisata. Inflection of this type in Sanskrit is made from nearly 130 roots. In most other IE languages it has largely died out, its place being taken by thematic formations. Consequently direct comparisons with forms of other languages are confined to a few common roots : dsti ‘ is Gk. eari, Lat. est, etc. ; dmi ' I go J , imds 4 w T e go Gk. eqa, tfxtv, Lith. eimi , etc. ; dtti, 4 eats Lat. est, Russ, jest’ ; dste 4 sits Gk. fjorcu ; sdte 4 lies ', Gk. KeZrai, Hitt, kitta , kittari. Hittite is the only language beside Sanskrit in which this type of formation is well preserved, and here further parallels are available : hanti 4 he slays ghndnii 5 they slay Hitt, kuenzi, kunanzi ; vasii ‘ he wishes ' (va 4 -) Hitt, wekzi (Gk. only ptc. Ikwv) ; sdsti ' sleeps Hitt. seszi. Beside the regular endings given above there exists a variant THE VERB 320 type : active, 3 pi. impf. caksur, duhur ; middle, 3 sg. pres, tie, citd, duhe, bruve, idye, vidi, 3 pi. duhri, sire ; duhrate , Urate ; impf. 3 sg. atia, aduha , 3 pi. aduhra ; aduhran, dseran ; dierata ; impv. 3 sg. duhdm , viddm, iaydm, 3 pi. duhrdm ; duhratdm, ieratdm. These forms (for which see above, § 6) are confined to the Vedic language with the exception of the root si- ‘ to lie ' which preserves such inflection in the classical language (3 pi. Urate). This series is important because it shows that there were originally two types of conjugation in the case of root stems, corresponding to the Hittite -mi and -hi conjugations. Sanskrit has generalised the mi- type in the active, but in the middle the Vedic language preserves these traces of the old dual system. With certain exceptions the normal system of accent and apophony prevails in this class, that is to say the root has accent and guna in the three persons of the active, while else- where it appears in its weak form and the accent is on the termination : hdnti : ghndnti ; vdsmi : uimasi ; dsmi : smds, etc. Roots in -u followed by endings beginning with a con- sonant, take vrddhi instead of guna in the strong forms : stauti , ' praises yauti f joins ’ ; also certain others, e.g. mdrsti ‘ rubs ' ; 3 pi. mrjdnti . A number of roots retain accent and guna throughout for reasons which are not clear : e.g. Ute Mies vdste ‘ wears clothes Certain roots with long vowels where this applies, e.g. isle ' sits J , iste ' rules ' have perhaps been adapted from the perfect system (as- originally perfect stem of as- ‘ to be J ). In the Vedic language the strong form of the root is optional in the 2 pi. : pres, nethd , impv. stota, impf. dbrainta. The weak form of the 3 pi. mid. termination {dvisdte as opposed to dvisdnti ) indicates original final accent which is preserved ocassionally in the Veda : duhatd , rihatd , The root sds- has the weak termination also in the active [sdsati 3 pi.) which accords with its radical accent. The conjugation of this class is complicated by changes due to internal sandhi. As this is a matter of phonology rather than morphology, a few examples will suffice : duh- ' to milk doh f si> dkoksi , doh 4- ti> dogdhi ; lih- * to lick leh + ti> ledki; sds- 4 dhi> iadhi. Analogy is responsible for the 3 sg. impf. as at (instead of *aias <*aidst) and in the same way for diet . Different formations are occasionally substituted where the operation of phonetic laws would leave a form too short or THE VERB 321 obscure : 3 sg. impf. adat with thematic vowel from ad- 4 to eat 1 islt with -1- beside Vedic as from as- 1 to be Some of the roots of this class contain enlargements, e.g. trd- 4 to save, protect sd-s- £ to command v-as- 4 to clothe which means that originally they did not belong to the root class. Such roots tend to be irregular in the matter of accent and apophony. Some reduplicated formations have come to be classed here, e.g. jaks- (1) 4 to laugh ’ ( has -), {2) r to eat ’ (ghas-) which retains some features of reduplicated inflection (3 pi. act. jdksati ), and nims- 4 to touch closely, kiss J (was-) ; likewise certain intensive formations which are treated as roots by the grammarians : jdgarti 4 is awake daridrdti 4 runs about, is poor dxdeii 4 shines . Here are classified certain roots making a stem by means of the suffix t, namely, in the classical language, rud- 4 to weep svap- 4 to sleep an - 4 to breathe ', svas - 4 to breathe 1 and jaks- 4 to eat : 3 sg. pres, roditi , svapiti , etc. Further examples are found in the Vedic language : vdmiti 4 vomits janisva 4 be born vdsisva 4 wear snathihi 4 smash stanihi 1 roar and the M.Bh. has iocimi. This formation corresponds to the Latin 3rd conjugation verbs of the type capio (capis, capit . , . cap - iunt). Like the other non -thematic classes it shows changes in accent and apophony (1 pi. ruditnds), but it had originally nothing to do with the root class, being an independent forma- tion. But it is a formation which from the earliest period of the language is on the way to obsolescence. Most of the forms quoted are isolated and not parts of complete paradigms. Even in the most .stable group which the classical language preserves, the t-suffix is absent before endings beginning with a vowel (3 pi. Yuddnti ) and in the 2 and 3 sg. impf. it is replaced either by long t (dntt), or by a thematic formation (anat). The type has ceased to form a full separate class, and by interpreting the suffix as the union vowel i (it), and attaching it to the root class, the grammarians were able to account for most of its characteristics. A suffix t appears in the conjugation of brii- 4 to speak but only in the strong forms before terminations beginning with a consonant ( brdviti , abravit ; dbravam , bruvdnti). In the corre- sponding Avestan verb it does not appear at all : 3 sg. mraoiti , impf. mraot . The suffix has importance elsewhere in the forma- tion of verbal stems, namely in Latin (audit e) and Slavonic THE VERB 322 (O. SI. supitu * sleeps 1 mluvitu 4 matters '). Like the short i above it is obsolescent in Sanskrit, and in addition to bru- only a few roots are found to take it in the Vedic language : dmtti (am- 4 to injure ’), tavili (tu- 4 to be strong '), samxsva (saw- * to labour ’). The Reduplicating Class (3rd or bu- class) Paradigm : (hu- 1 to sacrifice '). Present, Active, S. 1 juhomi, 2 juhosi, 3 juhoti, D. 1 juhuvds, 2 juhuthds , 3 juhutds , P. 1 juhumds , 2 juhuthd , 3 juhvali . Middle, S. 1 juhve , 2 juhusd, 3 juhuti , D. 1 juhuvahe, 2 /w&- vdthe, 3 juhvate , P. 1 juhumdhe , 2 juhudhv 6 t 3 juhvate , Imperfect, Active, S. 1 djuhavam , 2 ajuhos, 3 ajuhot , D. 1 a/w- A&wz, 2 djuhutam , 3 djuhutdm , P. 1 ajuhuma , 2 djuhuta , 3 havur. Middle, S. x djuhvi, 2 djuhuthds, 3 djuhuta, D. x ajuhuvahi , 2 djuhvdthdm , 3 djuhvatdm , P. 1 djuhumahi , 2 ajuhudhvam , 3 djuhvata . Forms according to this class are made from some 50 roots all told, but only from 16 in the classical language. The forma- tion is well represented also in Greek : m(/x]7rAT7/xi, ' I, we fill ', Skt. piparmi, piprmds ; elairippavax ‘ to introduce cf. Skt. bibharmi, bibhrmds ; Suku/n, Skt, dadami ; rl&Tjpu, Skt. dddhdmi ; tarTji/Lu, Skt. tisthdmi (transferred to thematic class). Elsewhere it has become comparatively rare. The vowel of reduplication corresponds to the radical vowel in the case of roots containing the vowels z and w : aMz ' ob- serves ' , jihreii 4 is ashamed vivesti ' is active bibheti 4 fears ninikta 2 pL impv. f wash ' ; yayoti 4 separates ' . In other roots it is sometimes z and sometimes a. (a) jigharti 4 sprinkles piparti 4 fills ', bibhdrti 4 bears jigati 4 goes ', mimdti 4 bellows sisdti 1 sharpens sisakti c cleaves to * (saj-). (b) ddddti 4 gives dddhati r places jahdti 1 leaves bdbh - asti 4 eats vavartti 4 turns sdsasti 4 sleeps saicati 1 they accompany This is in contradistinction to Greek which has almost exclusively -i- in the reduplicating syllable. The accent of verbs of this class is somewhat unstable. It may appear on the root in the strong forms ( juhoti , etc.) which is in accordance with the guna of the root, or in the case of THE VERB 323 certain verbs on the reduplication (dadhdti, etc.). The latter type is most prevalent in the Vedic language, appearing often when the later language has radical accent ( bibharti : bibhdrti ). Greek has also accent of the reduplication (SIScj/m, etc.), but the apophony indicates that radical accent must be original in the three persons of the singular active. On the other hand accent of the reduplication appears to be ancient in the 3 pi. active, where both root and ending appear in weakened form ; dddati , sascati .. In the weak forms the normal accent nation of the terminations is found, with the exception that the accent is thrown back on to the reduplicating syllable when the ter- mination begins with a vowel (juhve, bibhre, etc.). This de- velops in the post- Vedic period (V. juhvd) from the analogy of the 3rd person plural. The weakening of the radical vowel results in samprasarana in the case of vyac- [viviktds) and hvar - ( juhurthds ), and in loss of syllable in sac- and bhas- (3 pi. sascati , bapsati ). Roots in long a are treated variously. In the commonest, del- and dhd- t the root is fully reduced and the vowel elided : dadvds , dadmds ; dadhvds, dadkmds, etc. In the case of other roots this type is superseded by one in which the vowel -i- t or more usually -2- is inserted between the reduced form of the root and the termina- tion. The short appears in some forms from the root hd- 1 to leave ' : jah-i-mas, jahihi, etc. This reduplicated formation may be compared to the type svapiti of the root class. Normally however the vowel is long : sis-i-hi (sd- ' to sharpen '), mimtte [md- 1 to measure ’), rarithas (rd- ' to bestow '), etc. The pre- valence of the long vowel is due to rhythmical reasons, and the suffixal l balances the d of the root in such a way that the two have acquired the appearance of being the strong and weak forms of the root. The nu - and u- Classes (Fifth and Eighth, su- and tan- classes) Present, Active, S. 1 sunonti , 2 sunosi, 3 sunoti „ D. 1 sunuvds, etc. ... P. 3 sunvdnti, Middle, S. 1 sunvd, 2 sunusd . . . . P. 3 sunvdte . Imperfect, Active, S. 1 dsunavam, 2 dsunos , 3 dsunot, D. 1 dsunuva . ... P. 3 asunvan , Middle, S. 1 dsunvi , 2 dsunuthds, 3 asunuta , D. x dsunuvahi. ... P. 3 dsunvata. About 50 roots make presents according to this class. Typical examples are : rnoti ' rises ' (cf. Gk. opvvfii ). strnoti ‘ strews * (cf. Gk. OTopvvpu), ksinoti ' destroys ' (cf, Gk. <j> 6 tvio, THE VERB 3^4 <f>divv 0 {jj) minoti 4 harms, lessens ' (cf. Lat. minuo), dhunoti 4 shakes ’ (cf. Gk. diva* 4 rage ’), trpndti 4 is satisfied rdhnoii 4 thrives apnoti 4 reaches, obtains 4 , asnoti 4 obtains etc. This suffix, which is compounded of n and u t appears also in the formation of nouns, frequently from the same roots, e.g. dhrsnu - 4 bold ' beside dhrsndti 4 is bold In a series of roots the alternative suffix nd (ninth class) appears side by side with nojnu : vrndti : vrndti ; strndti : strndti ; ksinoti : ksinati . A simple suffix «, without the n, often appears in related formations : rnoti, cf. Gk. opovw ; strndti , cf. Goth, stranjan ; vrndti 1 covers cf. vdrutra - ; dhrsndti 4 is bold cf Gk. Bpauvs ; dabhndti 4 injures, deceives J , cf. ddbhuta-* (that cannot be hurt, divine) wonderful ' ; jinoti 4 enlivens cf. jivd- 4 alive ' ; sadhndti 4 accomplishes cf. sadhu- 4 straight, good T he same relation exists between ksurd- 4 razor ’ and ksyduti f sharpens which the complete incorporation of the suffix has caused to be transferred to the root class. In sru- 4 to hear 4 (parte, srutd-, Gk, kAutoss etc.) this u has been incorporated in the root every- where except in the present tense (srndti : IE * kl-n-eu~ti) . In Iranian, by later substitution it appears even here (Av. sitru- naoiti). Verbal classes corresponding to this appear in Greek (opvvpu etc.) and Hittite, where the formation has developed a special causative sense {arnummi, etc.), but in neither case is the old apophony seen in Sanskrit preserved intact. Elsewhere forma- tions of this type have been replaced by thematic formations (Ir. ro cluinethar 4 hears etc.). Thematic formations based on this suffix are found : pinvati 4 fattens J (cf. pinute, Av. pinaoiti) , invati 4 drives, attacks ' (cf. inoti), hinvati 4 impels ' (cf. hi noti) pnvati 4 enlivens 4 (cf. jinoti). In accent and apophony this class conforms to the normal type, with the usual Vedic irregularities, e.g, strong form of suffix in 2 pL (dkrnota(na)) and final accentuation of 3 pi, middle ( krnvatd , vrnvate, etc.). Concerning the form of the endings it may be noted that the u of the suffix may optionally be omitted in the 1 du. and pi. (sunvds, suntnas ; this starts of course in the 1 du.), and that before vocalic terminations - nuv - appears in- stead of -nv- after roots ending in a consonant ( saknuvanti 4 they are able '). The terminations -d of the 3 sg. middle [srnvd 1 is heard sunvd 4 is pressed J ) and ~rd of the 3 pi. ( srnvird , sunvird THE VERB 325 etc.) appear sporadically in the Veda, the latter being always associated with the union vowel -i-. This union vowel is found also in Vedic srnvisd 2 sg. middle. Under the eighth class are classified certain roots formed by the simple suffix oju instead of nojnu. These consist of a number of roots terminating in n : tan - ‘ to stretch 3 sg. tanoti , similarly sanoti ‘ wins 1 (cf. Gk. avvyu ' achieve '), vanoti ‘ wins manuld * thinks ksanoti ' wounds ’, and the root kr- ‘ to do 1 : 3 sg. karoti * There is some uncertainty in interpreting the forms from the roots in -n, since it is possible to argue that here the suffix is really nojnu before which a stands for n appear- ing in the reduced form of the root (tn-neu-ti). On the whole it seems preferable to adopt the simpler theory, and evidence for it may be seen in the existence of Vedic tarute which appears to be exactly parallel to manutd. The other important root classed here, namely kr - ‘to do * [karoti, kuriitd) also presents a problem because both the Vedic language and Iranian agree in inflecting it as a raw-verb (V. krnoti, krnute , Av. kzrdnaoiti , 0 . Pers. akunavam <akrnavam ) . This might suggest that the classical forms are ‘ prakritisms but this is rendered implausible by the fact that a phonetic development of rn to or fur is wholly anomalous in Middle Indo- Aryan,(cf. tfna- * grass f >tdna-, tina t etc., and Pkt. kunai from the Vedic form of the verb). The formation must therefore be regarded as a genuine and ancient dialect variant formed, like tarute , with the simple suffix u t which also appears in the Vedic noun karuna- nt. * action The weak form of the root presents some complication since normally either complete reduction {kr-) or restoration of guna as in tarute might be expected. Since suffixal H which produces the combination -ur- in the weak grade elsewhere [guru-, etc.) is here out of the question, the only theory that remains to explain the form of the w T eak grade is the assumption that the root originally began with a labio-velar. The form kur- may then be classed with those sur- vivals where this labial element is found to function as a vowel (Gk. ywri, Hitt, kunanzi , etc., see p. 74). As to the etymology it seems that the IE root k w el- diverged in Indo-Iranian, con- sequent on the second palatalisation, to produce two roots, on the one hand an intransitive car - ‘ to move, go and on the other hand a transitive kr - ' to do, make The u of the suffix of this verb is always omitted in those THE VERB 326 cases where such omission is optional in the nw-verbs (kurtnds, etc.). It is also omitted in the active of the optative : kuryam as opposed to kurmyd. These may be radical formations incor- porated in this conjugation. The nd - Class (Ninth or kri- class) Present, Active, S. 1 krindtni, 2 krindsi, 3 krinati, D. 1 krini- vds, 2 krintthds, 3 krinitds , P. 1 krinimds, 2 kriniikd, 3 knnanti. Middle, P. 1 krind, 2 knnisd , 3 krtnitd, D. 1 krinivahe , 2 krin- ithe, 3 krinate, P. 1 krinimdhe , 2 krinidhvd, 3 krinate. Imperfect, Active, S. 1 akrindm, 2 dkrinds , 3 dkrlndt , D. 1 dkriniva, 2 dknmtam, 3 dkrinitam , P. 1 dkrinima, 2 dknmta, 3 akrinan. Middle, S. 1 dkrini , 2 d krintthds, 3 dkrinita, D. 1 dkrinivahi , 2 dkrindtham , 3 akrinatdm, P. 1 dkrinimahi , 2 dkrinidhvam , 3 dkrinata. Some fifty roots all told make presents according to this class. Typical examples are: krinati 'buys' (cf. Ir. crenaid ), ZxftiZi ‘ sticks, adheres to 1 (cf. Ir. lenaid ( id '), srnati 1 smashes ’ (cf. Ir. ■ chrinat 3 pi. ' collapse ’), jinati ' inflicts loss mrndti 4 crushes prndti ' fills etc. The suffix is compounded of w and a (-#h-), and these ele- ments often appear separately in related formations. The n which appears in isndti appears in connection with other ele- ments in isanat and isanydti. The 4-suffix without w appears in a number of parallel formations in : grbhdydti , rnkthayati , skabhaydti beside grbhndti, etc. The 4 is often partially incor- porated in the roots, e.g, in ‘ to inflict loss ' and prd- ' to fill ' beside the presents jinati , prndti ; cf. the same process, though morecomplete, in sru-jsro beside irnoti. With the addi- tion of the weak form of the suffix roots in z, w and r become roots in i, u, and f. This is how they are normally given and how they normally appear outside the present system : pritd- 4 pleased pitta- ' purified ' (: pundti), piirnd - 4 full ' (: prndti ). In the tense there are two kinds of treatment : (a) the unex- tended form of the root appears before -nd-, e.g. jinati , pundti , mrndti ; (b) the extended form of the root is introduced even into this formation, e.g. prinati, bhnndti , a process which is found only in the case of roots in i. The root vli- * to press down ’ makes both types and Pali has kindti ' buys 1 (corre- sponding to the Celtic form) as opposed to Skt. krinati . THE VERB 327 The alternation between the strong form of the suffix -na- with -nt- in the weak cases is not found outside Sanskrit. In Greek there is alternation between long and short vowel, bdfAvrjfju, Bafjivdfxev. In Avestan complete loss of a in the weak grade, which is the regular Indo-Iranian phonetic develop- ment, is found : 3 sg. mid. vvrznte, °ster 9 nte. In Sanskrit this type is replaced by one containing an extra suffix in a way similar to that observed in the third class, e.g. vrnitd having the same relation to Av, v*r?nte as Skt. mimltd, etc., to datU, This l is only introduced before the consonantal terminations ; before vowels the simple reduced form of the suffix is found : jdnate , cf. A. zdnaite. Nasal-infixing Class (Seventh, rudh - class) Present, Active, S. 1 yundjmi , 2 yundksi, 3 yundkti , D. 1 yunjvds , 2 yunkthds , 3 yunktds, P. 1 yuiijmds, 2 yunkthd , 3 yun - jdnti. Middle, S. 1 yunjd, 2 yunksd, etc. . . . P. 2 yungdhvd, 3 yufijdie , Imperfect, Active, S. 1 dyunajam, 2 dyunak , 3 dyunak, D. 1 dyunjva , etc. ... P. 3 dyunjan. Middle, S. 1 dyunji , 2 dyunkthds , etc. . , . P. 3 dyunjata . About thirty roots inflect in this manner. Common examples are ; nndkti ‘ leaves ' (Lat. linquo ), cAtrf- ' cut chinatti (Lat. scindo ), ' split bhinatti (Lat. findo), pis- * crush pinasti (Lat. pinso), anj- ‘ anoint andkti (Lat. unguo ), bhuj-, bhundkti ' benefits bhunktd ‘ enjoys 1 (Lat. fungor ). In most languages outside Indo-Iranian the strong forms have been replaced by the weak forms, and the whole type transferred to the thematic class. The same tendency is seen in Skt. vinddti finds as opposed to Av. vinasti , and in the later history of Indo-Aryan such forms completely replace the older type (Pa. yunjati , etc,). In Hittite the corresponding class has a double nasal infix and no apophony, e.g, harnink- ‘ to destroy ' ; both these features are special developments of Hittite, out of the regular IE type which Sanskrit preserves. This type appears superficially to be different from the two preceding types, the present stem being made by infixation in- stead of suffixation. Ultimately they are not really different, since in most of the roots of this class, the final consonant can be interpreted as an extension, that is to say originally a suffix which in course of time has become incorporated in the root. THE VERB 328 Beside the root yuj- ' to join ' for instance there is also a simpler root yu- (yduti) with the same meaning. Evidence of the same kind is available in the case of a number of roots : e.g. chid- ' to cut cf. chydti 1 cuts child- ‘ cut ' ; tic- ‘ to leave rindkti , cf. rinati 1 lets flow ' ; ltd - ' to pierce trndtti , cf. tfna- ' grass Engl, thorn ; krt- ' to spin *, krnatti, cf. Lat. colus ‘ distaff In such cases the analysis of the forms of this class is the same as of those of the other two nasal classes : 5 kl-n-dw-ti ( srndti ), 9 pl-n-dn-li (prnati), 7 yu-n-dg-ti (yunakti). At the same time not all forms can be explained in this way, e.g. andkti ‘ anoints Once the infixing class was established through the incorporation of the second suffix, it attracted a certain number of other roots which did not belong to the original nucleus. B, Thematic Presents Radically Accented Class (First or bhu - class) Present, Active, S. 1 bhdvdmi , 2 bhdvasi , 3 bhdvati , D. 1 bhd - vdvas, 2 bhdvathas , 3 bhdvatas , P. 1 bhavdmas, 2 bhdvatha , 3 bhdvanti. Middle, S. 1 bhave, 2 bhdvase , 3 bhavate , D. 1 bhdvdvahe , 2 bhavethe , 3 bhdvete, P. 1 bhavdmahe, 2 bhavadhve , 3 bhavante. Imperfect, Active, S. 1 dbhavam, 2 dbhavas , 3 dbhavat, D. 1 dbhavava , 2 dbhavatam , 3 dbhavat dm, P. 1 dbhavdma , 2 dbhavata, 3 dbhavan. Middle, S. 1 dbhave , 2 dbhavathds , 3 dbhavata, D. 1 dbhavavaki, 2 dbhnvethdni, 3 dbhavetdm, P. 1 dbhavamahi, 2 dbhavadhvam , 3 dbhavanta. This is the commonest of all the present classes in Sanskrit, being formed by nearly half of the verbal roots in the language. The predominance of thematic formations is paralleled in the nominal stems, and it recurs in other IE languages. Direct equations between Sanskrit and other languages, attesting IE forms, are commoner in this class than anywhere. Examples are : pldvate, prdvate ‘ floats Gk. 7 ri(F)a) ; srdvati * flows Gk. ptei ; svdnati 1 sounds O. Lat. sonil ; stanati ‘ roars Gk. crrtvei ; bhdrati 1 bears Gk. pepw, La t. fero, Goth, bairip, Ir. berid, O. SI. beretu ; cdrati 1 goes J , Gk. WA ofiai, Lat. colo ; bodhati 1 understands Gk. ntvdo^cu ; josate f enjoys Gk. ytvopLCLL * taste ’, Goth, kiusip ‘ tests, chooses 1 ; rodhati, rohati "grows 1 , Goth, liudi p ; osati ‘bums', Gk. evco, Lat. uro ] THE VERB 329 variate * turns Lat, vertitur ; pdrdate ' breaks wind Gk. TzepSerat ; sdrpati ' creeps Gk, epTret, Lat. serpit ; ydsati 4 seethes Gk. £e'a> ; vdhati 1 carries Gk. Pamph. Pegu*, Lat, vehit, O. SI. vezetu ; vdsati ' dvrells Goth, wisip ' is 1 ; ndsate ' resorts to Gk. viofxat ‘ return Goth, ganisip ' is saved 1 ; mdjjati ‘ plunges Lat: mergit ; trdsati ' trembles Gk. rpcw ; pdtati ' flies Gk. ttcto/acu ; sthdgati ' covers Gk. crrtyei, Lat. tegit ; sdcate 4 associates with Gk. Ittctcu ' follows Lat. sequitur 4 id ' ; ddhati ' burns ’, Lith. degu ; pdcati ' cooks Lat. coquit , O. SI. pecetu ; tdksati ‘ (carpenter) joins, constructs ; hews Lat. texit ' weaves ' ; hdvate ' calls Av. zavaiti, O. SI. zovetu ; djati 4 drives Gk. ayet, Lat. agit, Ir. ad * aig ; dnati 'breathes* (beside dniti), Goth, uzanip 'breathes out, expires The majority of roots conform to the normal type, the stem consisting of the accented and gunated root followed by the thematic vow T el. In a small number of verbs vrddhi instead of guna is found: badhate 4 repels ', bhrajate 'shines', dhdvati ‘runs' (Gk. ddu) with g^ma grade), krimati 'strides* (beside middle kramate), dcdmati 4 stfps *. The class is augmented by a number of varied thematic formations with accent on the root or first syllable wdiich did not originally belong here : e,g. (1) a form with infixed nasal, nmdati ' blames * (cf. nid - ‘ contempt, insult ’, Gk. ovciSos-) ; formations of this type are commoner in the sixth class ; (2) forms with a suffix - va : pvati 4 lives tUrvati ' overcomes etc. ; in such cases it can be seen from the apophony that the initial accent is not original ; (3) stems con- taining the IE inchoative suffix -skejsko- (>ccha~) with second- ary radical accent : gdcchati ' goes ' (cf. Gk. /Lumc, impv.) yacchati ' holds ' ; (4) reduplicated thematic formations : tisthati 4 stands' (sthd-), cf. Lat sistit , pibaii ‘ drinks' (pd-) f Ir. ibid, jigkrati 4 smells * (ghrd-) ; in sldati 4 sits Lat. sldo (IE sizd - from sed-) the normal phonetic development would have given d in Sanskrit, but d appears here through the influence of other parts of the conjugation. Suffi-xally Accented Class (Sixth or tad- class) This and the following two classes are conjugated in exactly the same way as the preceding. The sixth class is fairly com- mon, something like 150 roots being conjugated in this way. Typical examples : rujdti ' breaks visdti * enters tuddti THE VERB 330 ‘ pushes disdti 1 points out mrsati ‘ strokes sprsati 4 touches suvdti ‘ drives kirdti ‘ scatters srjdti 4 lets go In contrast to Sanskrit this type is rare in most of the IE lan- guages, because the preceding class has become normal for thematic verbs. In Greek this type of stem is found usually only in aoristic use, where its contrasts in apophony with the normal thematic presents, pevyco : zpvyov. In such presents of this type as can be found the accent has been transferred to the root (yXvp tu), A fairly common sub-class here is formed by roots taking an infixed nasal : sincdti r sprinkles ', muncdti 1 releases vinddti 4 finds krntdti ‘ cuts lumpdti 4 breaks *, limpdti 4 smears Some are obvious transfers from the seventh class, e.g. unddti , yunjdti beside undtti , yundkti, and it is possible that the whole series arose in this way. The accented suffix -cchd- (IE -ske-) appears in icchati

• wishes ucchdti 4 shines rcchdti 4 goes J (roots is-, vas- t r).

In prcchdti 1 asks ’ (Lat. poscit) it has been incorporated into the root (pf. paprdccha) , but the unextended root appears occasion- ally in nominal derivatives : prasna- ‘ question Y a- Class (Fourth or div- class) The fourth class contains some 130 roots which form their present by means of the suffix ya : e.g. kupyati * is angry krudhyati 1 id tusyati ‘ is pleased yudhyati ‘ fights ’, vidhyati 1 pierces * (vyadh-), divyati r plays hfsyati 4 rejoices tdpyate

• is hot pdiyati 1 sees ndhyati 1 ties This suffix is also used

in the formation of denominative verbs, and the form of the passive differs from the middle of this class only in having suffixal accent. The formation is found in Hittite {wemiezzi 1 finds ", zahhiezi r fights ') and Greek (jLuuVeTa* ' is mad cf. Skt. mdnyate ' thinks fialvaj, etc.). In Latin on the other hand non-thematic stems in -i appear instead of this type ( cupio , cupit as opposed to Skt. kupyati). In Sanskrit the existence of a present stdnyati ‘ thunders ' (O. SI. stenjg) beside the non-thematic stanihi illustrates how the formation may arise by the thematic enlargement of an i-stem in the same way as in nominal stems ( °kravi° : kravya etc.). The apophony of the majority of forms indicates original suffixal accent, such as is found in the passive. On the other hand there are certain forms with vrddhied root, e.g. THE VERB 331 mddyati * becomes intoxicated iramyati 4 becomes tired which must have had radical accent from the beginning, which makes it appear that there were originally two types combined in this class. Certain roots in a which belong here, e.g. gd- 4 to sing ' (gdyati) , gla- 4 to be weary ' ( gldyati ), trd- 4 to save * ( trdyate ) and dhya - 4 to think ' (dhydyati), are by the grammarians attached to the first class by the wholly unnecessary assumption of roots of the form gdi etc. Since the roots are certainly to be set up as gd- t etc., these presents must be attached to this class, in which their form and accent are regular. There are a number of roots in d which lose this vowel before the accented suffix -yd- : da- ' to bind ' ( d-yaii ), chd- 4 to cut J (chydti), sd- 4 to sharpen ' ( sydti ) and sd- 4 to bind ' (sydti ; cf, Hitt, ishiya- 4 bind '). They retain what must have been, as noted above, the original accent of this class, because the com- plete reduction of the radical syllable made impossible any shift of accent to the root.' Th$ Tenth Class (cur- class) The suffix is -dya-. This has normally been specialised in the formation of causative verbs, but it is not exclusively used for this purpose, and a nucleus of forms remain which belong to the primary rather than the secondary conjugation. In the lan- guage of the Veda there is a fairly clear distinction between presents in dya which do not have strengthening of the root (guna or vrddhi) in which a causative sense is usually absent, and those in which it is so strengthened which are normally causative. Examples of the former are : citdya - 4 notice, ob- serve isdya- 4 flourish J , turdya- 4 hasten, speed ', dyutdya- 4 shine rucdya- 4 id ’, subhaya- 4 be splendid J , mrddya- 4 par- don sprhdya- 4 desire patdya - 4 fly about Here belong originally such presents as hvdyati 4 calls svdyati 4 swells ' and dh-ayati * sucks which the grammarians have attached to the first class by the assumption of the root forms svi-, hve dhe In the classical language the verbs classified here are more miscellaneous and they include a variety of formations of de- nominative or causative origin : e.g. kdmayate 4 desires f , corayati 4 steals chddayati 4 covers avalokayati ' looks upon dusdyati 4 spoils bhusdyati 4 adorns taddyati 4 beats etc. 332 THE VERB §9. The Future The stem of the future is formed by means of the suffix -syd-, or, with connecting vowel -i~, -isya- added to the gunated root, and it is inflected in precisely the same way as the thematic presents : ddsydti 4 he will give dhoksydti 4 he will milk bhavisydti 4 he will be karisydti 1 he will do etc, There are no simple rules by which the distribution of the two forms can be stated. In Iranian there exists a corresponding formation in ~$ya- (Av. vaxsyd 4 I will say '), but none corresponding to - isya The same formation appears in Lithuanian ; duosiu 4 I will give On the other hand the Greek future (Scl^oj, etc.) appears to be based on a simple so-stem. In the early period of the Sanskrit language the future tense is comparatively rare, as it is in the A vesta, the sense of the future being most usually expressed by the subjunctive, but it rapidly becomes more common. The future is a specialised type of present stem and it belongs with the various denominative formations in -yd-. It is capable of forming a preterite in the same way as the present stems proper. This formation functions as a conditional : yad evam ndvaksyo murdhd te vyapatisyat 4 If you had not spoken in this way, your head would have fallen off Only one example of the conditional is found in the Rgveda and it is never very com- mon during any period of the language. In addition to the ordinary future Sanskrit created a second or periphrastic future based on the agent nouns in -tar. In the third person the nom. sg., du. and pi. of such nouns functions as the second future without any addition : karta, kartdrau, kartdras . In the first and second persons forms of the verb ‘ to be J are added to the nom. sg. of agent noun, in the dual and plural as well as in the singular : kartdsmi, kartdsi ; kartdsvas ; kartdsmas: The function of the second future is to express the future in connection with some specified time : $vd vrastd 4 it will rain tomorrow etc. This type of future first begins to appear in the Brahmana period, and its use continues later, though it is never anything like as common as the first future. There was created also a corresponding middle, which, how r ever, is exceedingly rare, since it appears to have been current in the living language for only a very limited period. The special middle forms, which exist only for the first and second persons, THE VERB 333 are as follows : S, I kartdhe, 2 kartase, D. i kartasvahe , 2 karta- sdthe, P. 1 kartasmahe , 2 kartadhve. §10. The Aorist The aorist is formed by seven different types of stem which fall clearly into two classes, non-sigmatic and sigmatic. Of the non-sigmatic types the root aorist (ddhdt) and the a-aor. (druhat) do not differ in their formation from the imperfects of the corresponding present classes (ay at, dtudat ), It has been pointed out that the two types of preterite, imperfect and aorist, have arisen by specialisation out of a system with un- differentiated preterite, and the continued existence of forms common to the two remains as an indication of this. The difference in function depends on whether a corresponding present exists or not. The reduplicating aorist is less closely connected with corresponding present stems and it has under- gone special developments of its own, but it cannot be separated in origin from the reduplicating type of present. On the other hand the various aorist s- stems are formations which are con- fined to this use, with the rarest exceptions (Av. naismi is a present from an s-aor. stem.) The Root- Aorist The root-aorist is abundantly represented in the Vedic lan- guage. The apophony differs from that which is regular in the imperfect in that guna of the root appears normally in all the persons of the active with the exception of the third person plural. The normal weak form of the root appears in the middle. Typical forms are; S. 1 dsravam, agant, dkar am, 2 dgds , dsres, akar (for akars), 3 dsrot, dsthdt, dkar (for dkart, cf. A v.cor 9 t), D. 2 agdiarn , dbhutam 3 dkartdm, adhatam, P. 1 dkarma , dddma, dhema, 2 dkarta , dgcita , ahetana , 3 (a) akran, dksan (, ghas -), agman, (b) adhur, dsthur , dkramur. Middle : S. 1 dkri, dyuji , 2 akrthds , agathas , ayukthds t 3 dkrta, dmata (man-), dyukta, D, 1 ganvahi (gam-), P. 1 dgan- mahi , dmanmahi (with strong form of root), ayujmahi, dhumahi , 2 acidhvam, ayugdhvam, 3 (a) akrata , dgmata , (b) ddrsran, abudhran , (c) ddrsram , abudhram . The root aorist of the root bhu- (which is conjugated only in the active) is anomalous in having the weak grade throughout : dbhuvam, abhus, dbhut, . , . dbhuta , abhuvan. TH E V ER B 333 are as follows: S. I Artike, 2 Rrtise, D. I artisuage2 karth. sethe, P. I artishe, 2 Hartfdhye. 8 I 0. TH EA O RIST The aorist is formed by seven different types of stem which fall clearly into two classes, non-sigmatic and sigmatic. Of the non-sigmatic types the root aorist (id.dit) and the a-aor. (#vadhut) do not differ in their formation from the imperfects of the corresponding present classes (cytit!dat). It has been pointed out that the two types of preterite, imperfect and aorist, have arisen by specialisation out of a system with un differentiated preterite, and the continued existence of forms common to the two remains as an indication of this. The difference in function depends on whether a corresponding present exists or not. The reduplicating aOrist is less. closely connected with corresponding present stems and it has under gone special developments of its own, but it cannot be separated in origin from the reduplicating type of present. On the other hand the various aOrist s-stems are formations which are con. fined to this use, with the rarest exceptions (Av. Is is a present from an s-aor. stem .) The Root-Horist The root-aorist is abundantly represented in the Vedic lan guage. The apophony differs from that which is regular in the imperfect in that guna of the root appears normally in all the persons of the active with the exception of the third person plural. The normal weak form of the root appears in the middle. Typical forms are : S. I structgan, Rara21, 2 gis, ystes, gar (for Rus, 3 isrot, asthat, ihrer (for dikart, cf. Avtral), D.2 agalant, ¢bhita 3 tBartion, adhaitant, P. I Racront, ddiai/%, 2 keerta, right, afeta120, 3 (a) diar wron. asa has-), gal, (b) adhystlur, krch. Middle S. T \, ifyj, 2 kthas, aguthis, ayukthas, ata, inuata (nual-), iyukta, D. I gauagha (goz), P. I digan Mahi, juarchi (with strong form of root), ayujhi, ihnahi 2 acidhuum, aywgdhul3 (a) tructa, gatt, (b) addsrc/ , abodyal(c) dysra, budra. The root aorist of the root b- (which is conjugated only in the active) is anomalous in having the weak grade throughout bhain, abhis, abhnt, ... bit, ¢bhuvan THE VERB 335 sequent ambiguities were largely responsible for the develop- ment. All that remains of the root aorist in the classical lan- guage is the active inflection of certain roots in d ( dddt , etc.) and of bhu - (dbhut). In addition there are certain forms of the 2 and 3 middle of the root aorist (where the root terminates in a short vowel according to the grammarians) which in the later history of the language became heteroclitically attached to the s-aorist, e.g. we find 2 sg. dvrthds and 3 sg, dvrta forming part of the same paradigm with 1 sg. dvr si and 3 pi. dvrsata, etc. In the same way the middle forms quoted above from the roots da-, dhd-, sthd - with connecting vowel -i- are combined with -is- aorist forms to produce a series like sg. 1 a-sth-is-i, 2 d-sth-i-thds , 3 dsthita , The a- Aorist The form and conjugation of the a-aorist agrees with the imperfect accented a-class : dsicam, dsicas , dsicai , etc., cf. dtudam , dtudas , dtudat, etc. The stems agree in apophony, and also in accent, in the comparatively rare instances where the unaugmented aorist forms bear the accent : S. 1 ruhdm, 2 vidds , 3 dhrsdt, viddt, P. 3 dhvasdn, viddn , trpdn ; regularly in the participle trpdnt- dhrsant-, sucdnt etc., middle, sucdmdna-, etc. Correspondences with other languages attest the IE date of a number of such formations : dvidat, Gk. TSoy, inf. IStiv, Arm. egit ; drsan, Gk. eS pattov dricat, cf. Gk. eXnrov, Arm. elik 1 ; budhanta, cf, Gk. ihrvdovTo. At the same time there is some- times disagreement between languages in the assignment of a form to imperfect or aorist : Skt, ddasat impL ; Gk, ISa/coy aor. ; Skt. dbhujat impf. : Gk. epvye aor. The suffixal accent is retained in Greek in participles and infinitives (A lttljv, Xtrrelv), In addition to the regular type there are a number of a-aorists in Sanskrit whose form agrees rather with imperfects of the first class rather than of the sixth class since they have guna of root ; e.g. dsakam, dsanam, dsaram , akaras, dgamat, atanat, dsadat. This is the normal form of the tf-aorist for roots con- sisting of two consonants and the thematic vowel. Further- more where accent occurs these forms are accented like stems of the first present class. Examples of this are kdras , sdnat , srirat, dars am ( — the Gk. present stem SepKo^ai), gdman , THE VERB 336 sddatam „ sadatam , and the participles sddant sdnant (these have also contaminated the regular type above to some extent, so that forms accented like ruhat occur occasionally), A number of the stems listed here are probably thematisa- tions of root aorists, and not ancient. For instance the a-aorist dgamat appears later in the history of the language than the root aorist agan . On the other hand some are clearly old (e.g. dsadat), and since the type appears also in Greek (eyevero, yevecrdat) it must be referred to Indo-European. The a-aorist has attracted a number of reduplicated forms which did not originally belong to it, namely apaptat {pat- * to fall dvocat (for avavc-, vac - ‘ to speak ') and, with what in the perfect becomes a substitute for reduplication, neiat (nas- ' to perish '). Reduplicated Aorist Active, S. 1 djijanam, 2 dp j anas, 3 djijanat , ... P. 3 djtjanan . Middle, S. 1 djtjane, 2 ajijanathds , 3 apjanata , ... P. 3 djijan- anta . The typical reduplicating vowel of this is i but the redup- lication is subject to the following modifications : (1) If the root begins with two consonants short i is employed in redup- lication ; aciksipat , apisprsat. (2) If the root vowel is u the reduplicating vowel is u or u under the same circumstances : abubudhat, acukrudhat. (3) If the root is a heavy syllable two processes are found : (a) the root is unaltered and the redupli- cation is short : adidiksam , abnbhusam ; where the radical vowel is a followed by two consonants or long d the reduplicat- ing vowel is a : adadaksatn , adadhavam ] (b) in order to pre- serve the rhythm favoured in this aorist heavy roots may be weakened : avtvasam (vas-) f acikradat (krand-). It is in accord- ance with this principle that roots appear either with guna or in the weak form : apjanat but avivrdhat. Verbs which make a causative stem in -dp- substitute -ip- in the reduplicated aorist : atisthipat (sthdpayati). (4) Roots beginning with a vowel are found either to repeat the whole root (1 dmamat ) or, later, to repeat the last consonant with 1 ( drpipai ). In the Vedic language there are a number of non-thematic forms attached to the reduplicated aorist : ajigar , asisnat , didhar , etc. These represent an alternative type of the redup- licated aorist which was early abandoned owing to the preval- ence of the thematic type. THE VERB 337 The reduplicated aorist stands out from the other forms of aorist because it is attached not to the simple verb but to the causative ; ajijanat is the aorist of jandyati , avwrdhat of vardhdyati and so forth. Such an aorist is therefore made from all roots which have causatives, in addition to their normal aorists. This arrangement is essentially a development of Indo- Aryan, though its roots go back to Indo-Iranian. The other IE languages have nothing which can strictly be compared. In form the stem of the reduplicated aorist is related to the re- duplicating present [bibharti), and more closely to the thematic variety of the shme (Usthati ; Gk. ylyvo^iai ; Lat. gignit, sistit) but it has developed features of its own, namely the length of the reduplication and the guna of the root in certain forms ( ajijanat ), The transitive sense which is observable in some of the related stems with this type of reduplication (Lat. gignit , etc.) has been generalised and eventually developed into a full causative. Related Iranian forms provide some interesting information about the history of this formation, because it has also present stems formed in the same way : zizanmti 3 pi., etc. (also non- thematic zizdnti 3 sg"., cf. the Skt. type ajigar). From this it is possible to conclude that originally this was a type of stem forming both present and preterite in the manner of the various formations of the present system. The aorist, it has been observed, came about by the specialisation of certain preterite formations, and this is a case in point. To begin with we may assume two pairs *jijanati : ajijanat and jandyati : djanayat having essentially the same sense. In the further development jandyati comes to be exclusively used in the present and ajijanat originally simple preterite ( — imperfect)" becomes when isolated integrated into the aorist system. There are a few roots in the classical language which take this aorist as part of their primary conjugation, e.g. asisriyat and adudruvat from sri- ‘ to resort ' and dru- ‘ to run There are further examples in the Veda (e.g. dcikradat from brand - ' to roar ') and also some non- thematic forms which are made exactly like imperfects of the reduplicating class : dsisret , ddudrot. Certain reduplicated stems which became attached to the a-aorist have already been mentioned. These continue the IE usage which had reduplicated aorists in primary function and none that were specifically causative : cf. Gk. tne<jvov t tW/cAero, etc. 338 THE VERB The s- Aorist Active, S. i anaisam, 2 dnaisis , 3 anaistt, D. 1 dnaisva , 2 anaistam , 3 dnaisidm, P. 1 dnaisma , 2 dnaista, 3 dnaisur . Middle, S. 1 anesi, 2 dnesthds, 3 dnesta, D. 1 dnesvahi , 2 awe- sathdm, 3 dnesdtam, P, 1 dnesmahi, 2 dnedhvam , 3 anesata. The s-aorist stem differs from other verbal stems in having the vrddhi grade throughout the active, in the dual and plural as well as in the singular. In the middle roots with medial vowel i, w, r appear in the weak form [acchitsi, drutsi , dsrksi) ; also roots in final r (tfArsi) and in the Veda certain roots with final nasal, e.g. dgasmahi from gam- and masiya opt. 1 sg. from man- (cf. Av. mahmaidi) . Elsewhere there is guna. The sub' junctive takes guna in both active and middle (stosdni, etc.) and this is extended to certain injunctive forms ( jesma ). Before terminations beginning with occlusive when the root ends in such the s of this aorist is elided according to the usual phonetic rule : araudhs-ta becomes arauddha, etc. This leads to some confusion between this aorist and the root aorist, but this is largely eliminated in the post-Vedic period by the disuse of the root aorist except in connection with very few roots. Phonetic decay also strongly affected the 2 and 3 sg. active, with the result that both the tense sign $ and the terminations frequently disappear ; abhar for abhdr-s-t and abhdr-s-s, simi- larly araik , asvait, etc. In the post-Vedic period these incon- venient and ambiguous forms are abandoned and their place is taken by new formations taking the connecting vowel -l- : anaistt , dcchaitsit , etc. The s-aorist and the other forms of sigmatic aorist are sharply distinguished from the other classes of aorist in that there are no present-imperfect stems formed in the same w^ay. There are indeed in the Veda certain isolated forms of the present made in this way (stuse, hise, krse ) as well as some anomalous formations containing .9 which cannot be referred to the s-aorist stem (i arcase f rhjase , ii grntse , punisd) but these have the appear- ance of being tentative formations which never developed very far rather than relics of an earlier system. The s-aorist is found in Greek etc.) and Slavonic (vesu, sluchu, etc,). In Latin s-aorist forms have coalesced with perfect forms to make one tense (perf, dixi, diixi, etc.). In Irish injunctive and subjunctive forms of the THE VERB 339 s-aorist are retained (the s-subjunctive). No trace of it appears in Germanic. Radical vrddhi is attested for the vowel e by Latin and Slavonic (Lat. vexl, O. SI. vesu : Skt. dvaksam from vah-) ; for roots in diphthongs there is no clear evidence. In Hittite there is no s-aorist any more than any other kind of aorist, but there are certain preterite forms in the 2 and 3 singular which have final -s : 2 sg. da-a-as ‘ you took tarna-a-s ‘ you put in da-is f you placed pa-is 4 you gave ' ; 3 sg. da-a-as ‘ he took da-a-is ‘ he placed ag-ga-as 4 he died etc. These forms consist of the verbal stem enlarged by the suffix -5 and have no personal ending proper, and beside them there are forms to which the personal terminations have been secondarily added : 2 sg. da-is -ta beside da-is, 3 sg. na-is-ta 1 lead ' beside na-i-is. These forms are compared to the s-aorist of other IE languages, but it seems unlikely that they are simply remains of a fully developed IE s-aorist system. In the first place we have seen reason to believe that the aorist in general has arisen by specialisation out of an undifferentiated preterite, and in this respect Hittite should represent an earlier state of affairs. Furthermore there is some agreement between Hittite and Tocharian on this point, since the latter language has also a certain type of preterite using an s- stem in the 3 sg. : A. prdkds, B. preksa * he asked and this coincidence does not seem to be fortuitous. Bearing these considerations in mind we may per- haps rather explain the IE s-aorist to be a post-Hittite forma- tion based on the extension to the whole paradigm of an s- suffix which was originally restricted to the preterite of certain persons (notably the 3 sg.) of one class of verbs. The is- Aorist Active, S. 1 dpdvisam, 2 dp dins, 3 dpdvit, D. 1 dpdvisva. . . . PL 3 dpdvisur ; Middle S. 1 dpavisi, etc. ... P. 3 dpavisata. This aorist has in the active vrddhi of the root if it terminates in a vowel [dpdvisam, atdrisam, dsdyisam) and guna in the case of medial i, u, r [drocisam, etc.). Both types occur from roots with medial a : dkdntsam ; dvadhit . In the middle the root has normally guna, but occasional forms with weak root occur : nudisthds , (opt.) rucisiya beside rocisiya, gmisiya, idhisTmahi. Just as the s-aorist is founded on certain s- stems which functioned as finite verbal forms, so the is- aorist is formed on stems in the compound suffix -i-s, cf. arocista beside roots THE VERB 340 4 light To judge by the small number of such forms in Iranian (xsnavisd ‘ I will satisfy 6 ovisi 4 I expect ') it was not to begin with very frequent, but it became more common in Indo- Aryan in accordance with the tendency that is observable everywhere with set forms of the verb. Several different but closely related types of formation have come to be associated with the is- aorist. (1) The suffix -is- (as opposed to -is- with short -i~) is used in the case of the root grabh- 4 to seize ' : agrabhisma . (2) Non-sigmatic formations in -i- (compare abravit , etc., of the present system) appear : agrabhtm , avadhtm, agrhttdm, agrhithds . The normal 2 and 3 sg. of the is-aorist is adapted from this formation, since -is-$ and - is-t cannot phonetically produce -is and -it. (3) Some preterite stems in short -i- (compare presents of the type vamiti ) have been incorporated in this aorist ; atdrima , avadiran, bddhithds f avita , etc. The isolated and anomalous vanusania of the RV. is interest- ing because it is formed on the bases of an -us- stem in the same way as this aorist is founded on - is - stems, but unlike the is- aorist it has never developed to form a system. The aorist forms sthesam , sthesur , desma , jnesam, khyesam, etc., from roots in d are best interpreted as regular f.s'- aorist forms (staH-ts-, etc.). There are corresponding middle forms with weak root as in the examples quoted above : asth-is-i , asthisata , etc. With these are associated certain non-sigmatic forms which were mentioned in connection with the root aorist : asth-i-ta , etc. The sis - Aorist This aorist, which is inflected exactly like the preceding (dydsisam, dyasis, dyasit , etc.) arises from a mixture of the two preceding. It is an innovation of Indo-Aryan, unknown to Iranian, and in the RV. it can be quoted only from two roots (gd- 4 sing yd ‘ go '). More examples are found later but it is never very common. In the classical language it is allowed to be made, in the active only, from roots in d and nam- 4 bow yam- 4 hold ram- 4 be content The sa- Aorist The sa - aorist has the normal thematic inflection and the weak root. The accent on unaugmented forms, when it occurs, is on the suffix (dhuksdn) which is in accordance with the apo- phony. It is made only from roots containing a medial vowel THE VERB 341 i, u, r and a final consonant which combines with the s of the suffix to produce -ks- e.g. adiksat , dmrksat, amrksanta, avrksam , aduksat later adhuksai from dis - ' point tnrs- ' touch mrj- 1 wipe vrh- ' tear duh- ‘ milk 1 / It is rare in RV. (examples from 7 roots), which may suggest that it is an innovation, though from Iranian (O. Pers. niyapisam * I wrote down ') it appears to be of at least Indo-Iranian date. Nothing that can be exactly compared is found in the other IE languages. The Passive Aorist in -1 There exists a passive aorist in -i, used only in the 3rd person singular, which is independent of any of the foregoing aorist stems : djhdyi 1 was known adarsi f was seen ', etc. Unaug- mented forms (which appear in both indicative and injunctive use) are always accented* on the root syllable : srdvi, pddi t etc. Roots having i, u, r as medial vowel appear in the guna grade {aceti, dbodhi , asarji) ; elsewhere there is normally vrddhi (1 dgdmi , akdriy dstdvi, a&dyi), more rarely guna ( ajani , avadhi ). The formation is taken by some 40 roots in the RV., to which others are added later. It appears also in Iranian (Av. srdvi, O . Pers. addriy — Skt. srdvi, ddhdri), but not elsewhere in Indo- European. Neglecting the augment, which was a secondary and optional addition to preterite formations in Indo-European, it is clear that these forms are nothing more than old neuter i-stems, without any termination, which have been adapted to the ver- bal conjugation. §11. The Perfect Active, S. 1 cakdra , cakira , 2 cakdrtha, 3 cakdra, D. 1 cakrvd, 2 cakrathur , 3 cakratur , P. 1 cakrma , 2 cakra , 3 cakrur . Middle, S. 1 cakrd, 2 cakrsd , 3 cakrd, D. 1 cakrvdhe, 2 cakrdihe , 3 cakrdte , P. 1 cakrmdhe, 2 cakrdhvd , 3 cakrird . The perfect is formed from the root stem but this is char- acterised by (1) reduplication and (2) a special series of endings. The general principles of reduplication have already been de- tailed (§ 5). Of the types there enumerated the one adopted for the perfect is that which uses the vowel a (IE e) in the reduplica- ting syllable, with the proviso that in Sanskrit (as opposed to Greek, etc.) i and u are substituted before roots which contain such a vowel {tatdna : pipdsa , bubhoja ). Special features of the THE VERB 34^ perfect reduplication, in addition to those mentioned above are as follows : (1) There is a class of roots in the Veda which re- duplicate with a long vowel : ctddhara , jdgara , mdmrjd , plpaya , itlt&va. This is mainly intensive reduplication, but in some cases the a of the normal reduplication has coalesced with an element elsewhere lost before the beginning of the root, e.g. in the per- fects of gr- 4 to awake ’ and mrj- ‘ to wipe 1 ; cf. Gk. iyelpto, ofiopyvvfu. (2) Two roots in u reduplicate with a instead of the usual u and in both this is associated with irregular weak form of root in the singular active ; babhuva , sasiiva from bhu- 4 to be 1 and su- f to give birth (3) Roots beginning with a nor- mally have d- (a + a) in the perfect, e.g., dda , asa from ad- 4 to eat 5 and as - ‘ to be A different type appears in the perfect of anj- 4 to anoint ' and as- 1 to attain * (variant root forms atns-, nas-) where an n which forms part of the root is repeated in the reduplication : dnanja , dnaje ; dndmsa, dnasd (cf. Ir. i-dn-aic ' he came '). This spreads by analogy to other roots including a number beginning with r- : dnarca , dnrcd from rc- or arc- to praise. (4) Roots beginning with i or u reduplicate with these vowels which in the strong forms of the active are pre- fixed to the guna grade of the roots with intervening -y- and -v- and in the w'eak grade coalesce with the radical vowels to form 1 and it : iyesa , isd , uvo'ca, Rce from is- 4 to seek ' and uc- ‘ to be accustomed (5) A similar type of reduplication appears in the case of one root beginning with ya- and a number beginning with va- w'hich reduplicate with i and u . These coalesce with i- and u- in the weak forms of the root to produce i and d : tydja, ijd from yaj- 4 to sacrifice ’ (weak form ij- in pass. ijydte, etc.), uvdca , lice from vac- 4 to speak ' (weak form uc- in ucydte y uktd etc.) ; similar forms from vap- ‘ sow r vad- 4 speak vas- 4 dwell ' and vah- 4 carry (6) Roots having a medial a before a single consonant, and beginning with a consonant that is unchanged in reduplication have the normal reduplica- tion only in the strong forms : tatdna , papdta from tan- f stretch ' and pat- 4 fall ' the weak forms of the perfect are made by substituting -e- for the a of the root : tend, tenur , peed , pecur. This is an innovation of Sanskrit wilich is by no means complete in the Vedic period ; the more original forms which occur are V. paptima , tatne, mamndte , etc. The type originated in certain roots which acquired such form through normal phonetic development, notably sad - 1 to sit ' where sedur stands regularly THE VERB 343 for earlier *$azdur (cf. Av. hazdyat pf. opt.) and roots beginning with y- (yemur for *yaymur). (7) The root vid- * know J has no reduplication, and this is in accordance with the related lan- guages : Skt. veda, Gk. otSa, Goth, wait, etc. There are a few other sporadic cases of unreduplicated perfect forms in the Veda, e.g. taksathur , taksur, skambhaihur , skambhur , and three perfect participles formed without reduplication : ddivds- ' pious tnidhvas- 1 liberal 1 and sahvas- ' overpowering In apophony the perfect follows the normal type of verbal inflection, that is to say the strong form of the root appears in the three singular persons of the active, the weak form else- where. Normally the strong grade is guna (1 vavarta , cikeia, bitbodha) but wherever in the strong stem medial a appears be- fore a single consonant (1 sg. act. tatdpa , bibhdya, cakdra, etc.) vrddhi is substituted in the 3rd person sg., and in the later lan- guage optionally in the 1st ; tatdpa , bibhdya , cakdra , etc. This feature does not appear outside Indo-Iranian ; the majority of the languages show guna only, and by general agreement the o-grade of the root : Gk. yeyovc, etc. Original o-grade is attested also by Sanskrit, since certain roots which have under- gone the second palatalisation retain the original guttural in these perfect forms : cikaya, cikeia f jigdya, jaghana from ci- 1 gather cit- * observe ji - ‘ conquer ' and han- ' slay The weak forms of the root are the normal ones that result from the loss of the guna vowel : cakrd,jaghnd,jagmur , bubudhe , etc. Roots in d lose this vowel altogether in the weak forms as generally (t ia-dh-ur ), but before consonantal endings invariably insert the union vowel i (dadh-i-re) in contradistinction to the present. Samprasarana appears in such roots as vyadh- ' to pierce ’ and svap - ' to sleep ' ( vividhur , susupur ; their redupli- cating vowel goes with this form). In some roots having nasal in the strong form the weak form is marked by the absence of this nasal : cakradd from krand- 1 to roar On weak stems of the type ten- see above. For the personal terminations see § 6. There are not mamy complications. Instead of the normal endings roots in d ter- minate in -au in the 1 and 3 sg. of the active ; dadau t dadhau, tasthdu , jajndu from dd- * to give dhd- ' to place sthd- ' to stand jnd- ‘ to know ' (for IE *deddn-u , etc., with vrddhi be- fore laryngeal). The final u- element, which appears here in place of a personal termination, is found also in Latin, incor- M 344 THE VERB porated into certain perfect forms: novit ‘knew', cf. Skt. jajnau ; plevit f filled cf. paprdu . A feature of the perfect conjugation is the frequency with which the connecting vowel -i- appears before the terminations that begin with a consonant : 2 sg. bubodhitha, 1 du. bubudhivd, 1 pi. bubudhimd, ... 3 pi. mid. bubudhire , etc. In the later lan- guage the -re of the 3 pi. mid. has it always. Before other con- sonantal endings except the 2 sg, act. it is taken by the vast majority of verbs. In the 2 sg. act. the unit forms allowed are more numerous, and a number of roots take optionally either form, e.g. ninciha, nindyitha. Roots in d have this option, but when taking -i- they appear in the weak form and the accent is transferred to the ending : dad&tha , dadithd. In the Veda the use of the union vowel is less extensive than in the classical language. As a general rule it appears after roots ending in a consonant provided the last syllable of the stem is a heavy one : vivdditha, ucimd, paptimd, etc., as opposed to tatdntha, yuyujma, etc. It is also taken by roots in -a (dadintd, dadhimd) but the type dadithd of the 2 sg. is unknown to the early usage. In Iranian the use of the auxiliary vowel is exceedingly rare which makes it clear that in the main its employment in the perfect (as elsewhere in the verbal system) is an innovation of Indo- Aryan. The perfect tense is widely represented in Indo-European, having been dropped only in Armenian and Balto-Slavonic. Perfects common to Sanskrit and other languages may be illustrated by such examples as the following : jajJna (Jan- ‘ to beget ') : Gk. yiyovt ; daddrsa (, drs - ‘ to see ’) : Gk. 8e8 o/>k« ; cicchida , cicchide ( chid - 4 to split ') : Lat. scicidt, Goth, skaiskaip ; diddia, didisd (dis~ ‘ to point out ') : Gk. 8e8et^a, Sc'Say/iai, Goth, ga-taih ; rirdca t ririci (tic- ‘ to leave Gk. AcAom-a, Lat. liqui, Goth, laih ; nineja , ninije (nij- ‘ to wash '), Ir. -nenaig ; tutdda, tutudur (tud- ‘to push'), Lat. tutudi, Goth, staistaut ; vavdrta ( vrt - 1 to turn '), Lat. vorti, vertt , Goth, warp ; dadhdrsa (1 dhrs - * to be bold '), Goth, ga-dars ; jaghina ( han - ‘ to slay '), Ir. 1 sg. -gegon, 3 sg. -gegoin. In some languages, notably Latin and Germanic, the redupli- cation is not an essential part of the perfect formation. It occurs with certain roots and is absent in the case of others. This corresponds more nearly to the original state of affairs in IE. The reduplication was, to begin with, no more an essential THE VERB 345 part of the perfect formation than was the augment of the aorist and imperfect. Its generalisation in Greek and Indo-Iranian is one of the many isoglosses that unite those two branches within the IE family. Even they preserve in vida ' knows ’ the older type of non-reduplicating perfect. Here, as elsewhere, Hittite shows greater divergence from the normal IE type. Hittite has no perfect, but a special type of present conjugation, the hi- conjugation, which has been compared with, and has certain features in common with, the normal IE perfect. At the same time the gap is not easy to bridge, since we have on the one hand a special tense with a sense of its own (state, result) which is made by most roots in addition to the present tense, and on the other hand a variant form of the present taken by certain roots. It is probable that in this matter Hittite is the major innovator, but it is not pos- sible simply to derive the Hitt, /h-conjugation from a system corresponding to the IE perfect, because there are outside Hittite also certain present formations which go with it : notably (i) Skt. formations of the type dduha enumerated above, (2) the Gk. conjugation of thematic verbs (A eyio, A iyeis., Aeyei). The perfect in Sanskrit and Greek conjugates in both active and middle. There is reason to believe that this is a secondary arrangement. In Sanskrit the middle endings of the perfect are in the main obvious imitations of the present, in marked con- trast to the active endings which differ so markedly from those of the present. Furthermore it has already been pointed out that an active perfect not infrequently corresponds to a middle present, and that the endings of the active are more closely related to certain middle endings than to other active endings. All these indications lead us to believe that the existence of two voices in the perfect is of later origin than in other parts of the verbal system. The evidence also points to the conclusion that the perfect did not to begin with have a corresponding preterite. Such forms of this kind as exist in several languages are to be classed as independent innovations. This applies both to Sanskrit and to Greek, and to a greater extent in the former, since while Greek did eventually develop a pluperfect with a meaning of its own, the forms classified as such in Sanskrit are in the main isolated and unstable formations which appear in the Vedic THE VERB 346 language but are not used later. Such are sg. 1 ajagrabham, 3 djagart, du. 2 amunniktam, pi. 3 dbibhayiir, mid. pi. 3 ajag- tniran . The distinction between these forms and the imperfect of the reduplicating class is not very clearly to be drawn. Their sense is normally that of simple preterites. §12. Injunctive and Subjunctive The so-called injunctive is no separate morphological cate- gory, but the term is applied to those unaugmented forms of the imperfect or aorist which are used with the force of subjunctive or imperative (§ 2). In the RV. the augmentless forms are more than half as common as the augmented and they may appear in- differently in preterite or injunctive use. Such forms occur from both imperfect and aorist stems, more frequently from the latter. In the later Atharvaveda the proportion of such forms occurring is noticeably less, and of those that do occur a greater part are formed from the aorist stem. In post-Vedic Sanskrit the injunctive disappears except in one construction. Prohibi- tions continue to be expressed by the use of md in connection with unaugmented aorist forms : md bhaisth ‘ do not be afraid md gdh * do not go etc. The subjunctive stem is formed by the addition of the vowel a to the indicative tense stem, the gunated form of such a stem being employed if it exists : doha juhdva yundja-, etc., from the tense stems doh juhd yundj This a combines with the a of thematic stems to form long d : bhava-, tudd ucyd-. The inflection of the subjunctive from non-thematic verbal stems is illustrated by the following paradigms : Active, S. 1 dyani , dyd , 2 dyasi, dyas , 3 dyati, dyat, D. 1 aydva, 2 dyathas , 3 dyatas , P. 1 dydma f 2 ayatha , 3 dyan . Middle, S. 1 &sai, 2 &sase t dsdsai , 3 asate , asaiai , D. 1 asavahai, asavahe , 2 dsaithe , 3 d$aite f P. x dsamahai , asdmahe, 2 dsadhve, asadhvai, 3 asante, asanta, dsantai. It will be observed that the endings of the subjunctive are partly secondary and partly primary. In 1 du., 1 pi. and 3 pi. of the active secondary endings are employed ; in 2 and 3 du. and 2 pi. primary endings ; in 2 and 3 sg. either primary or secondary endings are used. In the middle forms with secondary endings are rare, appearing normally only in the 3 pi. The 1 sg. active has a special ending of its own, -dni, beside which in the earlier language simple -d appears which is to be THE VERB 347 compared to the -d which appears in Greek in the subjunctive as well as in the active of the i sg. of thematic verbs (dyw). The 1 sg. middle ending -ai arises from the contraction of the a of the subjunctive stem with the -e of the termination. This -ai is then extended to other parts of the middle inflection and such eventually become the normal forms. This type of termination is preceded by the vowel a even in subjunctives of non-thematic verbs. The conjugation of subjunctives from thematic stems is the same as the above but based on a stem in d arising from the combination of the a of the tense stem and the subjunctive a : Active, S. I bhdvdni, 2 bhdvdsi, bhdvas, 3 bhdvati, bhdvdt, D. 1 bhdvdva, 2 bhdvdthas , 3 bhdvdtas, P. 1 bhdvdma, 2 bhdvatha, 3 bhavdn . Middle, S. 1 bhdvai, 2 bhdvdse , bhdvdsai, 3 bhdvdte, bhdvdtai , D. 1 bhdvdvahai, 2 bhdvaithe , 3 bhdvaite , P. 1 bhdvdmahai, 2 bhavddhvai, 3 bhdvdntai. The subjunctive can be formed in the Vedic language from all three tense stems, present, aorist and perfect. This variety of formation is not matched by any variety of meaning, e.g. srndvat, sravai and susravat all mean * he will hear ’ or ' let him hear 1 and no sort of difference related to the tense stem appears between them. Aorist subjunctives are commonest from the root aorist {karat, gamut, yamat, varat ; karati, josati , bhddati , etc.) and from the s-aorist {jdsat , ndsai, matsat ; nesati , parsati, etc.) ; none are found from the sa-aorist. Examples of sub- junctives from the perfect stem ar e jaghdnat, jujosat, paspdrsat ; jujosati, didesati , bubodhati, etc. The subjunctive remained in use during the later Vedic period (Brahmanas and Upanisads), but, apart from the first persons which were incorporated in the imperative, it is extinct in the classical language. A subjunctive corresponding in form and meaning to that of Sanskrit appears in Greek. Here the primary endings are ex- clusively used and the forms with long vowel associated with thematic stems have become predominant. The quality of this vowel (where Sanskrit has d) varies in accordance with the variation in the indicative (ay^/ze^, dyrjrt). Some old short vowel subjunctives are preserved as futures (eSo/Ltat, mo/zeu) and the simplest explanation of the s-future is that it is the sub- junctive of the s-aorist. The Latin future erit corresponds to THE VERB 348 the Skt. subjunctive dsat{i) and both Italic and Celtic have forms deriving from the s-aorist subjunctive (Lat. faxo , Osc. deivasi , Ir, 1 sg, -Has, 3 pi. -tiassat from tiagu ‘ go '). These two groups have also an ^-subjunctive which does not appear in Greek or Sanskrit. The subjunctive is absent over a considerable part of Indo- European, and has the appearance of being a comparatively late formation. It can be plausibly explained as having grown out of the injunctive, certain forms of which developed into an independent system. The hesitation between primary and secondary endings in Sanskrit represents a transition from an earlier system in which the endings were secondary (as in the injunctive) and a new one in which primary endings are applied as being more appropriate to its predominantly future meaning. This transition has been completed in Greek. Of the short and long vowel subjunctives the former is the earlier and more original. The addition of the thematic suffix to a stem already provided with such is without parallel elsewhere in IE stem formation, and its presence here is due to the analogy which created bhavd etc., having the same relation to bhava- as asa- to as-. The short vowel subjunctives with secondary endings (< karat , gamut) have a form w'hich cannot in itself be distinguished from injunctives (augmentless preterites) of thematic stems. That they are subjunctives depends not on the nature of the stem itself, since such stems are commonly used in the forma- tion of present /imperfects, but in their relation to other forms in the system. The absence of any distinctive formative in the most original type of subjunctive is a clear indication of its secondary origin. §13. The Imperative Active, S. 1 bhdvdni , dydni , 2 bhava , iht, 3 bhdvatu , dtu r D. 1 bhdvdva, dyava, 2 bhdvatam, iiam , 3 bhdvatdm , it&m, P. 1 bhdvdma , dyama, 2 bhdvata, itd, 3 bhdvantu , ydntu. Middle, S. 1 bhdvai , dsai, 2 bhavasva , dssva, 3 bhdvatdm , astdm , D. I bhavavahai, asdvahai , 2 bhavethdm , asdthdm, 3 bhdvetdm, asdtdm , P. X bhdvdmahai , isdmahai , 2 bhdvadhvam , addhvam, 3 bhdvantdm, dsatam . This paradigm is composite. The first persons in the three numbers are properly subjunctive forms. Injunctive forms are employed in 2 and 3 du. and 2 pi. Specific imperative forms THE VERB 349 occur in 2 sg. and 3 sg. and pi. In the 2 sg. the stem of thematic verbs functions as imperative without any addition in Sanskrit as in the related languages : bhara, Av. bar a, Gk. fcpz, Arm. her, Goth, batr, Ir. heir ; prcchd 1 ask Lat. posce ; dja ' drive Gk. aye, Lat. age , etc. With non-thematic verbs the ending is -hi, originally -dhi : ihl 'go', Av. idi, Gk. IBi. The original -rf/n appears in Sanskrit after consonantal stems ( viddhi ‘ know Gk. 108 1 , dugdhi ' milk etc.) and occasionally elsewhere, edht for *azdhi (cf. Av. zdl) from as - ' to be juhudht from hu- ' to sacrifice The Veda has further examples : srnudhi ‘ hear gadhi 'go', vrdhi ' cover The forms of the 3 sg. and pi. are made by the addition of a particle -u to the secondary endings : bhdvat-u , etc. Corre- sponding forms occur in Hittite : 3 sg. estu : Skt. dstu, kuendu, Skt. hantu ; 3 pi. asandu, Skt. sdntu ; kunandu, Skt. ghndntii , etc. In the hi-ve rbs which have no ^-ending in the 3 sg. present, this element u appears alone in the 3 sg. impv. : aku, aru from ak - ‘ to die at - ‘ to arrive 3 sg. pres, aki , ari. In the middle the termination of the 2 sg. is -sva. A corre- sponding formation is found only in Iranian : barayuha , Skt. bhdrasva ; kzrzsva, Skt. krsvd, etc. This -sva is considered to be the stem of the reflexive pronoun. The 3 sg. and pi. are made by the addition of -dm to the secondary endings and here too corresponding forms are found only in Iranian, vwpzyatam, xraoszntam . The few verbs which have inflections without t in the 3 sg. active (e.g. duhd , aduha) keep this feature in the 3 sg. impv. : duham, saydm. In the 3 pi. they have -ram with r as in the indicative : duhram , cf. 3 pi. mid. duhre. Compromise forms are duhratdm and seratdm. Beside the normal endings above there appears, particularly in the earlier language an ending -tat. This is indifferent to the distinction between active and middle and it appears most frequently used as 2 sg. : brutat 1 say dhattat 1 put dhavatat

• run vittat 1 know etc. It may also be used for other persons

and numbers : 1 sg. jdgrtdd ahdrn ' let me keep awake 3 sg. raja murdhanam vt pdtayatdt ' let the king cause his head to fall off 2 pi. dpah . . . devdsu nah sukfto brutat ' O waters, announce us to the gods as well-doers In the later language the use as 3 sg. tends to preponderate, but the total of examples as com- pared with the earlier language is small. This form of imperative appears also in Greek (as 3 sg.) and THE VERB 350 in Latin (as 2, 3 sg.) : iana ‘ let him know Skt. viit&t , Sotoj , ltcj, Zaru), etc. ; Lat. vehitd , Skt. vdhatat, poscitd, habeto, esto, etc., O. Lat. estod, etc., cf. Osc. Ukttud , estud . In the Veda there are certain 2 sg. forms in - si with impera- tive value : dhdksi ' burn ydksi ' worship pdrsi ' cross pr&si ' fill srosi ‘ hear J , etc. The termination is identical with that of the 2 sg. indie, pres., but these imperatives are quite clearly distinguished because the presents are differently formed (ddhasi, srnosi, etc.). One corresponding form is quoted from Avestan: doisl * show J from dues-. In the classical language the imperative forms are from the present stem. In the Vedic language imperatives may be made from all three stems, present, aorist and perfect, and, as in the case of the other moods, no difference of meaning appears be- tween them. Examples of aorist imperatives are : krdhi, srudht, gahi , gantu , yuksvd ; sada t sdna, sadatu ; vocatdt , vocatu ; of perfect imperatives, cikiddhi, mumugdhi , sasadhi, didestu , vavrtsva. It should be noted that imperatives are not normally made from the s- aorist. There are a few formations such as 2 sg. nesa , 3 sg. nesatu and 3 sg. mid. rdsatdm which are all thematic formations and therefore cannot properly be attached to the s-aorist. §14. Optative and Precative N on-thematic : Active, S. 1 duhyim. 2 dnhyis, 3 duhy&t, D. 1 duhydva , 2 duhyitam, 3 duhyatdm, P. 1 duhydma , 2 duhyata , 3 duhyur. Middle, S. 1 duhlyd , 2 duhtthas , 3 duhlta , D. 1 duhtvdhi , 2 duhtydtham, 3 duhiyatam, P. 1 duhimahi , 2 duhidhvam , 3 duhlrdn. Thematic : Active, S. I bhdveyam , 2 bhdves , 3 bhdvet, D. 1 bhdvema , 2 bhdvet am, 3 bhdvetdm, P. 1 bhdvema , 2 bhdveta, 3 bhdveyur. Middle, S. I bhdveya , 2 bhdvethds , 3 bhdveta , D. 1 bhdvevahi „ 2 bhdveydthdm , 3 bhdveydtdm , P. 1 bhavemahi , 2 bhdvedhvam , 3 bhdveran. The non-thematic inflection of the optative differs in apo- phony from the usual system. The* strong form of the suffix is not confined to the three singular persons of the active, but extended to all the active with the exception of the 3 pi. That THE VERB 351 this is an innovation is clear from Latin which preserves two grades in the case of the verb ' to be ' : O. Lat. siem, siet for later $im f sit beside slmus. A similar extension of the strong forms was observed in the case of roots in -a : 1 pi. pres, ydmas ‘ we go ’, aor. adhdma 1 we placed In the thematic classes the diphthongal stem of the optative (bhares, bhdret , etc. — Gk. <f)€pcus, <f>epoi, Goth, batrais, bairai) is formed by contraction of the thematic suffix and the weak form of the optative suffix (0+*). The terminations of the optative are mainly the normal secondary terminations. The 1 sg. middle has a special ending which has been noticed (§ 6), and the -ran of the 3 pi. appears in a minority of preterite forms ( dduhran , etc.). The anomalous Vedic 3 sg. duhlydt (after which 3 pi. duhiydn) seems to be based on *duhtyd formed without -t- after the fashion of the indicative {duhd, dduha). In the classical language the optative is formed from the present stem. In the Vedic language it is formed from all three stems, present, aorist and perfect, and, as with the subjunctive, no difference of meaning is attached to this difference of forma- tion. Root aorist optatives are fairly common ; asyam , rdhyam , gamy as, bhuyM , middle asiya , etc. They are rarer in the a-aorist and reduplicated aorist. From the sigmatic aorists optatives are formed only in the middle and the 2 and 3 sg. take invariably the precative s ; masiya {man- ‘ to think '), tnam- sisthcis, mamsista , gmisiya, janisista, ydsisisthds, etc. The perfect optative is common : j agamy dm, riricydm , vavrtyds, nimyat , papatydt , vavrtlya , caksamithds , jagraslta, etc. The oldest type of optative is that attached to root stems, present or aorist. Here the suffix is attached to the root in the same way as in the various present stems, and the normal secondary endings are added to it : gam-yd-m like kri-nd-m . This stem developed on its own lines on account of the special meaning which became associated with it. The main develop* ments which produced the optative in its final form were (1) the incorporation of the optative in the present system {asnuydt replacing asydt , etc,) and (2) the creation of optative forms to thematic stems by combining with them the weak form of the optative suffix. The first process is still incomplete in the Veda. On this theory the optative was to begin with a quite inde- pendent stem and its association with the various tense stems THE VERB 352 secondary. The perfect optatives should be regarded from the same point of view. Reduplication in early IE was a feature liable to turn up in many parts of the verbal system, and on the other hand though it came to be especially associated with the perfect it was not to begin with an essential feature of that system. Bearing this in mind it is clear that a reduplicated optative is not in its origin connected with the perfect system. We have an optative stem (gamyd-m) originally independent of the tense stems, and beside it a reduplicated optative stem (jagamyd-m) originally equally independent. With the incorpor- ation of the optative into the various tense stems these redup- licated optatives became formally attached to the perfect, but in meaning they never acquired any of the characteristics of the perfect. The dying out in the later language of all forms of the optative except those belonging to the present system, resulted naturally from the absence of any distinction of meaning be- tween the different forms. The forms of the Precative in the classical language are as follows ; Active, S, I bhuydsam , 2 bhuyds, 3 bhuydt , D. 1 bhiiydsva 2 bhuydstam, 3 bhuydstam, P. 1 bhuyasma , 2 bhuyasta , 3 bhu- yasur. Middle, S. I bhavislya, 2 bhavisisthds , 3 bhavisistd, D. 1 bhavi - stvdhi , 2 bhavisiydsthdm , 3 bhavislyastdm , P. 1 bhavisimdhi, 2 bhavistdhvdm, 3 bhavistrdn. The active forms are always made directly from the root by the addition of the optative suffix extended by s. The older form of the 3 sg. act. was bhuyds, which is preserved in the Vedic language. The middle forms are formed from the stem of the sigmatic aorist, and the precative s is absent in the first persons and in the 3 pi. The precative is the only modal form from a non-present stem retained by the classical language. The use of the precative is not common in the classical lan- guage and knowledge of its inflection is based on the statements of the grammarians. In the pre-classical language most of the active forms as given by the grammarians are attested with the exception that the older form of the 3 sg. is used. In the middle there is no distinction in the Veda between optative and preca- tive. The optative of the sigmatic aorist invariably inserts the precative s in the 2 and 3 sg. and this s is employed nowhere else in the conjugation. Such optative forms with precative 5 are THE VERB 353 occasionally formed in the Veda from other stems : root-aorist, padlstd, mucista, a-aorist, videsta, reduplicated aorist, ririsista, perfect, sdsahisthds. In the Rgveda , as opposed to the later Vedic literature, very nearly the same state of affairs prevails in the active. Precative forms from the root aorist are numerous in the 2 sg. (which cannot in this case be distinguished from the ordinary optative) and in the 3 sg* In the 3 sg. no non-precative forms are re- corded. Outside these two persons there occur only 1 sg. bhuyasam and 1 pi. kriydsma (once each). Otherwise forms in ~yd only appear outside the 2 and 3 sg. It is clear that the two first person precatives are innovations of the later RV. period and that in the original inflection s was proper only to the 2 and 3 sg., and there always used, as in the middle. The inflection may therefore be compared to that of the Hittite verbs men- tioned above (§ io, S. 1 tarnahun , 2 tarnas , 3 tarnas, P. 1 tar- numen , 2 tarnatin , 3 tarnir) in which s functions as the common termination of the 2, 3 sg. but does not appear in the other persons. In its earliest form the Sanskrit precative, which is not to be distinguished from the optative of the root aorist, or in the middle from that of the sigmatic aorist, preserves this ancient characteristic of one class of secondary endings. On the other hand, in the rest of the conjugation, the optative has adopted the normal secondary endings in these two persons. The precative is thus one of the most archaic formations in Sanskrit grammar. §15. Secondary Conjugation Under secondary conjugation are classified certain forms of present stem which differ from the ordinary series of present stems because (1) they have acquired a special meaning and (2) they are normally taken by roots in addition to their ordinary presents. The four types of secondary conjugation are (1) passive, (2) intensive, (3) causative and (4) desiderative. They are all essentially present formations, and with certain exceptions their inflection in other systems consists of more or less sporadic innovations. I. The Passive The formation of the passive is closely connected with that of the fourth present class. It differs from the middle of that THE VERB 354 class only in the position of the accent. In the passive this is on the suffix yd whereas in the fourth class it is on the root : many ate ' thinks but badhydte 1 is bound This distinction is secondary since roots in the fourth class appear in their weak form (isyati, tsyate) and this indicates original suffixal accent. Furthermore there are a few old intransitives (not passives) which are suffixally accented : mriydte 1 dies ', dhriydte * is stead- fast There is also fluctuation of accent in some passive and intransitive forms, mucyate and mucyate ' is released kstyate and kstyate r is destroyed jiyate and jiyate ' suffers loss pdcyate and pacydte * is cooked The passive in this form is found also in Iranian (Av. kir- yeinte — kriyante), but not elsewhere. It is an Indo-Iranian innovation based on the fourth present class, and its origin was due to the frequency of intransitive verbs in that class, par- ticularly with middle inflection : jay ate ‘ is born pdcyate ‘ be- comes ripe, cooked ', t-dpyate ' becomes hot etc. Since a number of these verbs had differently formed transitive presents beside them (tdpati ‘ heats etc.) they could easily form the nucleus from which the passive system developed. Differentia- tion was made by the retention in the passive of the old accent, for which in the fourth class presents radical accent has been substituted. The examples above in which there is variation of accent are mainly old intransitives which have been adopted as passives (: mucyate r gets loose etc.). The middle inflection is universal in Sanskrit (except for some late and incorrect Epic forms, drsyati ' is seen etc.), but active forms are not uncom- mon in Iranian : Av. hairy eiti 1 is carried O. Pers. Qahydmahy

• we are called etc. Probably to begin with the usage was

uncertain and the exclusive use of the middle later generalised in Indo-Aryan. The passive is inflected only in the present system. In the perfect and future the middle voice frequently functions as passive : cakri 1 was done karisyate * will be done In the aorist there is a passive 3 sg. of independent formation [dkdri) which has already been described. In the immediate pre- classical period there was a tendency, which did not go very far, to extend this by adding other forms. None are found in the Vedic literature, but the grammarians lay down, for roots ending in vowels and graft-, drs han-, special passive aorist forms based on the above, e.g. 1 sg. dddyisi , 3 pi. andyisata, etc. : 3 sg. THE VERB 355 addyi, andyi. This type of stem was further extended to the future : I sg. ddyisye, 3 sg. ghdnisyate, etc. Such forms occur very rarely in classical Sanskrit, and they are all learned forma- tions taken from the grammar. In addition to finite verbal forms the passive meaning could be expressed by the passive participles in -ta and the future passive participle in -1 tavya. In the later history of Indo-Aryan, in the Prakrit period, all forms of active preterite were lost, and their place was taken by passive constructions with the participle in -ta, This process is reflected in the later Sanskrit literature ; the usual construction becomes mayd brdhmano dr$tah 1 the brahman was seen by me 1 instead of ahant brdhmanam apasyam ‘ I saw the brahman AssociatecRvith this is the increasing use of the impersonal passive : iha sthiyatdm ‘ stay here 1 (lit. r let it be stayed here '), tenet bhavitavyam ' it must be him r , etc. This type of later Sanskrit is largely Prakrit in disguise. By such devices the wealth of the Paninean verbal morphology can be mostly ignored, and this simplified Sanskrit was understand- ably popular. II, The Intensive The intensive is a form of present stem which expresses intensification or repetition of the sense expressed by the root. It is of common occurrence in the Vedic language, being- attested from over 90 roots. In the classical language, though allowed by the grammarians to be made from every root, it is of infrequent occurrence. The stem consists of the root preceded by strong reduplication. In the case of roots containing i or u this reduplication has the corresponding guna vowel : 3 sg. act. nenekti , vevetti, mid. nenikti , dedistd ( nij - - wash f , vid- ' know dis- 1 point out ’) ; 3 sg. act. jdhaviti, nonavlti , 1 pi. nonumas ( hu - f call 3 nu- ‘ roar '). Where the vowel is a the correspond- ing long vowel appears in the reduplication : 3 sg. cdkastti, papatlti , 3 pi. ndnadati (has- ‘ appear pat - ' fall nod- ‘ roar ’). When roots contain or terminate in r (/) or a nasal, this con- sonant is repeated in reduplication : 1 sg. carkarmi, 2 sg. dar- darsi , 3 sg. janghanti , calcaliti , (with dissimilation) alarti , 3 mid. nannate {kf- ‘ to commemorate df- * to split han - ' slay cal- ' move ar- ' rise, go nam- ‘ bend As an alternative re- duplication with long d is sometimes used with roots of this form : jdgarti ' is awake etc. An t is frequently inserted beTHE VERB 356 tween the strong reduplication and the root : 3 sg. vartvarti , kanikranti, gamganti , 3 pL davidyutati, bharibhrati (vrt- ‘ turn brand- 1 shout gam- ' go dyut- ‘ shine bhr - 1 carry '). The apophony of the root follows the usual system ; it is strong in the three persons of the active, elsewhere weak : 3 sg. act. nenekti, pi. nenijati, 3 sg. mid. nenikte . When l is inserted after the root in the singular active, the root only has guna where this does not produce a long syllable : johavtti but vdviditi. The accent is on the reduplication in the strong forms and in the 3 pi. (vdvetti, vhidati) ; elsewhere on the terminations according to the classical grammar (vevidmas, etc.), but the Vedic usage fluctuates : 3 sg. mid. neniktd, etc., beside tttikte, etc. The terminations are the normal ones (with -ati in the 3 pi, as in the reduplicating class). A common feature is the use of the connecting vowel f. This was observed also in the root class (brdvUi, etc.), but it is much more common in the intensive : johavtti , tartariti , dardanti, etc. It is never used when the same kind of t appears after the reduplication. It is employed in the three singular persons and once in the dual (tartarithas) . In the 3 sg. mid. the ending -e occurs about as frequently as the ending -te : Ukite , joguve, yoyuve, etc. The 2 and 3 sg. of the imperfect suffer the usual phonetic mutilation ; adardar for 2 sg.

• adardar-s and 3* ddardar-t f etc. The connecting vowel -h

appears in the imperfect in the 3 sg. ( djohavit ) and once in the 3 du. (dvavasUam) . The termination of the 3 pi. act. is -ur as in the reduplicating class : djohavur. - The intensive forms subjunctives commonly, but almost ex- clusively with secondary endings. The root has guna only when this does not make a long vowel : 3 sg. janghanat, bobhavat ; carkrsat , davidyutat. Imperatives are not uncommon : 2 sg. dardrhi, carkrdhi , 3 sg. vevestu , dadhartu , 2 pi. jdgrtd ; with -tat, 2 sg. carkrtat, jagrtdt. There are a few forms with the auxiliary vowel t : janghamhi, johavitu. The optative is exceedingly rare (vevisydt AV,). There exist a few intensives with perfect inflections. These are not perfects to the above, but an alternative type of present, in accordance with the old sense of the perfect. Such are davidhava , nondva , dodrdva, lelaya with the ordinary meaning of the intensive present. There exists a second type of intensive formation which reTHE VERB 357 duplicates in the same way as the above, but forms its stem by the addition of the accented -yd- suffix and inflects exclusively in the middle : marmrjydte, dedipydte , dodhuydte, etc. (tnrj- ‘ wipe J , dip- 1 shine dhu- ‘ shake '). This is rare in the Vedic language, but in the classical language it is commoner than the basic type. Intensive formations corresponding to the Sanskrit basic type were common in Old Iranian : cf. Av. zaozaomi, carvkdrvmahi , (opt.) dardairydt , daeddiU, (thematic) naenizaiti, corresponding to the Sanskrit intensive bases johav -, carkar dardar-, dedi& and nenij Though not recorded outside Indo-Iranian the formation is evidently ancient in Indo-European. The fact that it does not appear elsewhere is due to the general abandonment of non-thematic verbal inflection in the majority of IE lan- guages. In contradistinction the second type of intensive formation (i dedipydte ), though rare in early Sanskrit, has parallels elsewhere, particularly in Greek : noppvpoj 1 be in un- easy motion ' (Skt. bhur-) t ira^paivto 4 shine brightly SapSanrcu ' tear asunder p,app,atpuj 4 glitter etc. III. The Causative The causative is the most productive of the secondary con- jugations from the early period onwards. The stem is formed by the addition of the suffix -dya- to the root, which normally appears in its strengthened form, and it is identical with the stem of the tenth class of verbs. There are a considerable number of verbal formations in -dya- t particularly in the early language which have no causative function. Some have a fre- quentative sense (patdyati 4 flies about etc.) which from the comparative evidence is ancient (Gk. 7 tot€o/xch). The causative is only one of the uses attached to the dya - stem, but in course of time it becomes the predominant one. There is in the earlier language a distinction between causatives with strengthened root and non-causatives with weak root : dyutdya ruedya - 4 shine ' : dyotdya-, rocaya- 4 illuminate etc. ; similarly be- tween guna and vrddhi in patdya- 1 fly about pdtaya - ' cause to fall V The distinction is not absolute since there are forma- tions with weak root having a causative sense ( drmhdya - ' make firm ') and conversely formations with strengthened root having a non-causative sense {muddy a- 1 get intoxicated '). In the later language the bulk of the non-causative forms die out, and what 358 THE VERB remain are combined with formations of a more denominative character to form the tenth present class. In the causative the root always has guna where this produces a long syllable : tarpayati, vardhdyati , kalpdyati , bodhdyati , ceidyati from trp- ' to be satisfied vrdh- ' to increase kip - 1 to arrange budh- ' to be aware ' and oil- 4 to observe Roots which in their strong form insert a nasal have this in associa- tion with guna in the causative : manddyati 4 gladdens sramsdyati 4 causes to fall etc. Where the guna form produces a short, syllable ( kar etc.) vrddhi is most commonly employed in the causatives : kdrdyati ‘ causes to do trasayati 4 terrifies nasdyati 4 destroys cydvayati 4 causes to fall etc. Rut a number of such roots retain guna : gamdyati 4 causes to go tvardyati 1 makes to hasten namdyati 4 causes to bend etc. The non-strengthened form of the root appears normally only with roots that have no other form (guhdyati from gith- 4 to conceal J ), only rarely elsewhere (girayati, sphurdyati). In dusdyati 4 spoils ' (intr. dusyati , sb. dosa-) the long vowel serves as a substitute for the normal strengthening. The anomalous purdyati 4 fills ' is influenced by the form of the past participle passive (purnd-). Roots in a commonly insert -p~ before the causative suffix : dapdyatiy sthapdyati, mdpdyati, etc,, from da- 4 to give stha - 4 to stand md- 4 to measure J , etc. This -p- is an old suffix or enlargement which is known from comparative evidence to have been associated with certain of such roots (Lith. stap$tis 4 to stand still J ), and it has been extended to the whole class in the causative. It is further applied to the root r- [arpdyati) and to a number of roots in -i : adhydpayati from adhi -M 4 to study etc. In ropdyati 4 plants 4 (ruh- 4 to grow J ) it replaces the final consonant of the root. In Middle Indo-Aryan the popularity of this form of causative grew until it replaced the normal kind. A number of such Prakritic formations appear in later Sanskrit (knddpayati 4 causes to play * t jivdpayati 4 causes to live etc.). There are a few other miscellaneous insertions before the causative suffix, namely -l- : pdlayati 'protects' ( pa -), -n- in prinayati 4 pleases ' (pn-, prinati), in Wnsayate 4 frightens 4- in ghdtayati 4 has slain ' (hart -) . Formations outside the present system are made more com- monly from the causative than from the other forms of second- ary conjugation. The future ( vardhayisyati , etc.) appears only THE VERB 359 very rarely in the Rgveda, but later is regularly made. For the aorist, as already observed, the reduplicated aorist has been adapted to serve for the causative. Besides this a few sporadic is- aorist forms occur in the early language ( avadayisthas , etc.). For the perfect the periphrastic form is used : gamayam cakdra > gamayam dsa. The passive is made by suffixing the passive yd directly to the form of the root as it appears in the causative : karydte, sthdpydte, etc. {simple passive kriydte , sthiydte). Nominal forms from the causative are : participle in -td, kdrita gerundive, kdrayitavya kdrya-, karamya infinitive, karayitum , gerund, karayitvd . The suffix of the gerund in -ya is added directly to the root when this is strengthened in the causative {-kdrya), otherwise to the ay of the causative suffix {-gamayya). The present formations in -aya are closely related to the nominal t-stems [roci-jrocay- : rocdyati). The causative stem consists of a thematic enlargement of this suffix, of a type which occurs, though very rarely, in the nominal formations. The formation, since it contains a series of guna vowels, is not likely to be very ancient in Indo-European, but it occurs fairly widely : Gk. Tpoirla >, orpoptuj (tp4ttco, orpit jfrcu), Lat. spondeo (: Gk. crnevScu), moneo , Goth, nasjan, drausjan (ga-nisan, driusan), etc. The meaning is frequentative, as usually in Greek, or causative. The latter meaning is normal in Germanic [nasjan 1 to save ga-nisan ' to be saved ’) and in Slavonic. Where Verneris law operates Germanic confirms the position of the accent on the suffix, as in Sanskrit ; wait pan : ft award] an , It has also participial forms corresponding to the Sanskrit participles in -ltd [frawardips) which are thus shown to be ancient. In Slavonic there is a series of causatives with vrddhi as in Sanskrit, and this is one of the special features which con- nects the two families with Indo-European : O, SI. saditi ' to plant cf. Skt. sadayati 1 makes to sit, settles slaviti ' to praise cf. Skt. iravdyati ' makes to hear, be heard IV. The Desiderative The desiderative stem is formed by means of the suffix -sa associated with reduplication. The vowel of the reduplicating syllable is normally i but u is employed when that vowel occurs in the root : bibhitsati , titrpsati, but yuyutsaii ( bhid - ' to split trp - ‘ be satisfied yudh- ‘ to fight ’). Long l occurs in a very THE VERB 360 few cases : mimdmsate 1 investigates ' (man- 1 to think The accent rests always on the reduplication. The root appears normally in its weak form, but a final i and w are lengthened ; jigisati ‘ desires to conquer ' (ji-), juhusati ' desires to sacrifice ' (hu-), Final r of a root becomes ir or ur before the desiderative -sa : ciklrsati ' desires to do Utirsati ‘ desires to cross mumursati ' is about to die This is phon- etically justifiable only in the case of roots in -f, he. those originally having final h (tf~, tar{n)~ : Utirsati) and from these it is extended to the rest. A number of roots form an abbreviated stem in the desidera- tive in which the reduplication and the root are contracted into one syllable. An example is dipsati from dabh- 4 to injure Corresponding to this Av. has diwzaidyai (inf.), and from a comparison of the two an Indo-Iranian stem dibzha - emerges. This represents a simplification of the original consonant group which occurred when the vowel of the root was elided in its weak form, i.e, di-dbh-sa a regularly formed desiderative. In the same way siksa- and siksa- appear from sak- and sah-, later dkiksa ripsa lipsa etc. (dah- f rabh-, labh-) ; to these are added Ipsa - and irtsa- from roots beginning with- a vowel (dp- ' to obtain rdh- r to prosper r ). The roots dd and dhd make respectively ditsa- and dhitsa - in which the a of the root has regularly disappeared in the weak form (di-d-sa- t di-dh-sa -). The roots van - 4 to win ' and san- ‘ to gain * make the desider- ative stems vivdsa- and sisdsa- with d out of -kh- as in other derivatives. Roots in d t apart from those mentioned above, generally keep the strong form in the desiderative : yiydsa-, pipdsa- f from yd- 4 to go pd « r to drink This, like the other forms with strong root below, is a Sanskrit innovation, as is clear from the pre- servation of ancient stems like ditsa - and the existence of Vedic pi-p-i-sa- beside pipasa . Anomalous strong forms appear from certain roots terminating in a nasal : jighdmsa -, jigdmsa- (be- side jigamisa-) from han gam-. When the desiderative suffix appears as isa with the union vowel a final i, u, 1 of a root necessarily and a medial i, u, r optionally appear in the guna grade : sisayisa-, ninartisa etc., but also rurudisa-. These and similar forms are laid down by the grammarians, but they , do not occur in the earlier language. Like other verbal formations the desiderative sa may be augTHE VERB 361 merited by the union vowel i and appear as -is a-. In the early language there occurs only didh-i-sa - (dha~, beside dhitsa-) ; also i in certain cases where this enlargement has produced what is in practice an alternative form of the root (pxptsa-> jihisa -). The numerous classical formations in isa (which have normally guna of root as noted above) are complete innovations. From the desiderative there are made, though not in the earliest language : a future in isya : tiiiksisye , an is- aorist, dcikirsisam and a periphrastic perfect, ipsdm cakdrajdsa . Nominal derivatives from the desiderative stem occur earlier ; the most common are an adjective in -u (titiksu-) and an abstract noun in d [ntTmamsd). In most desiderative stems the meaning (' wish to do some- thing 'J is straightforward and clear, though sometimes it is rather 4 to be about to do something ’ (tnuntilrsaU) . In the case of a few roots the desiderative stem has developed a special meaning : cikitsa - 4 cure jugupsa - 4 despise titiksa - 4 endure bibhatsa- (bddh-) 4 abhor rnimdmsa- 4 investigate susrusa- 4 obey The antiquity of the desiderative in Indo-European is attested by the reduced forms (ditsa-, dipsa-) which have been affected by the old apophony. Nevertheless it is not widely represented, a fact which must be due to loss in the individual languages. The only branch of Indo-European outside Indo-Iranian where a comparable formation occurs is Celtic. Since there are no close relations between these two members of the family this is itself an indication that the formation is ancient. The Celtic formation to be compared is the Old Irish reduplicated s-future : 1 sg. -ninus <*niniksd (nigid 4 washes ', cf. Skt. nij-, des. ninths ati), 2 sg. -riris {con-rig 'binds'), 3 pi- lilsit from ligid 4 licks 1 (Skt. lih- t des. liliksati). §16. Denominative Verbs Denominative verbs are those that are formed on the basis of a noun stem. Ultimately, as already observed, all verbal stems are not to be distinguished from the corresponding noun stems, but they have acquired independence. The denominative proper is a formation by which verbal stems continue to be made from the nouns existing in the language. The suffix employed in making denominatives is accented -yd- y the same suffix which forms one of the "primary verbal classes ( divyati ). THE VERB 362 The only difference is that the denominatives preserve the original accent of the suffix which in the primary verbs has been replaced by radical accent. The denominative is of IE origin and among the other languages it is particularly well repre- sented in Greek : rtK^aipw 4 determine " (for - aryo from the neut. noun ri^piap) cx&up ai ‘ hate ", ovopuiivta ‘ name ", uoXttL^cj f trumpet " (crdkTny£), K7]pva<jw ‘ proclaim " (/ojpuf), etc., etc. Similar formations in other languages are Lat. custodio, finio (cast os, finis ) , Goth, glitmunjan 4 glitter lauhatjan r shine ", etc. They are also common in Hittite, which gives a greater antiquity to the denominative formation than might otherwise have been expected : irtnaliya - * to be ill ' [irmalaS ill kusaniya- ' hire ' ( kusan * pay J ), lamniya - 1 to name * (laman ' name ", cf. Gk. ovopaLv w), etc. Denominatives in -yd- are formed from all the various nominal stems and they may conveniently be classified accord- ingly. Stems in r : vadharydti ' hurls a weapon ", cf. vddhar f weapon f . This type of nominal stem is practically obsolete, and the result is that there are a number of such denominatives where the corresponding nouns-stem has been lost : sratharydti ‘ becomes loose ", sapary ati 4 attends to, worships ", ratKarydti ' rides in a chariot ", adhvarydti * performs a sacrifice ", vithurydti ' staggers Stems in n : Denominatives formed from w-stems are krpanydti 4 solicits ", turanydti * is speedy ", damanydti 4 sub- dues ", bhuranydti 4 is active ", saranydti 4 hastens ", dhisanydti 'pays attention", ruvanydti "roars', huvanydti ' calls ", etc. This type is based on the old neuter n-i^tems, likewise mainly extinct, and corresponding nominal stems are either non- existent or take the form of thematic derivatives : krpand- y turdna etc. Stems in s : These are well preserved and denominatives are frequent ; apasydti 1 is active namasydti ' reverences ", canasydti ' is pleased ", manasydti ‘ is mindful of ", etc. In some cases the corresponding s-stem is not preserved, e.g. irasydti 4 is jealous ", daiasydti 1 renders service to ", In other cases -asya- is extended to become an independent suffix, with a desiderative meaning : vrsasydti * desires the male ", stanasyati * desires the breast From the compound stems is and us are formed avisydti 4 is eager to help " (the identity of this form with the THE VERB 363 future in - isya - should be noticed), tarusydti 4 strives to over- come etc. Rare examples of denominatives formed from stems in occlusive appear in bkisajydti 4 acts the physician ' and {from a stem not otherwise preserved) isudhydti ‘ implores ' (Av. iSuidya -) . It will be observed from the examples quoted above that the denominatives in yd are normally from the neuter consonantal stems. Denominatives from the masculine (agent-noun) stems are rare: e.g. vrsanydti 'acts like a male*. A few such for- mations are based on the nominative singulars : rajaydte 4 is kingly vrsdydte ‘ acts like a bull svdmiydti 4 treats as master Stems in i and t : janiydti 4 seeks a wife * ( jdni-) t tavistydte 4 is strong ' (tdvist). The form with long l is usually extended to stems in short i (kavtydti f acts like a wise man sakhiydti 4 de- sires friendship aratiydti 4 is inimical '), but such forms are shortened in the pada text. The suffix -iya- develops to some extent independently, with a desiderative sense, and is applied to other than f-stems : putriyati 4 desires, a son mdmsiydti 4 craves flesh etc. Stems in u and u ; Here also the long form of the suffix is applied to both types of stem, though short u is restored in the pada text : rjuydti 4 is straight vasuydti 4 desires wealth ', etc. In some cases there is no noun-stem and -uya- functions as an independent verbal suffix : asuyati 4 grumbles ankuydti 4 moves crookedly stability dti 4 stands firm Stems in d : prtandydti 4 fights ducchundydte 1 desires mis- chief ', mandydte 4 is well disposed Denominatives from a- stems are an ancient IE type, though more frequently elsewhere formed without the addition of ya : Hitt, newahhun 4 I re- newed Lat. novdre , Gk. vedv . This type is found in Sanskrit only when a is incorporated in the root {triti), otherwise the -ya- denominative is used. The suffix - dya - early became an independent suffix, and there are a number of roots which in- flect in this way without there being any corresponding d- nouns : mathdyati 4 stirs irathdydti 4 loosens musaydti 4 steals etc. These stems commonly alternate with stems of the ninth class : mathnati, irathniti , musnati , etc. As a result of the close association of the two types the denominative yd is sometimes appended to ninth class stems ; hrnayd hrniyd -

• be angry THE VERB

364 In the Vedic language the denominative in - dya - is commonly extended beyond its proper field and it is used to form denomin- atives from thematic stems beside the regular forms in -aydti : aghdyati 4 plans mischief ’, asvdydti 4 seeks for horses firiyaydte 4 holds dear Thematic Stems : amitr aydti 4 acts like an enemy devaydti 1 cultivates the gods, is pious vasnaydti ' bargains etc. This is the latest type of denominative formation. The addition of a further suffix to a final thematic suffix is against the principles of IE stem formation. It appears here in the denominative purely by analogy, deva-yd-ti, etc., being created after the pattern of brahman-yd-ti, etc. The resulting stem is similar to the causative, differing only in accent, but the origin and analysis are quite different. On the one hand vve have an f-stem with thematic extension (analyse vardhdy-a-) f on the other hand a thematic stem with the mechanical and analogical addition of the denominative -yd- (analyse deva-yd-). The similarity of the two forms gave rise to some confusion, and there are stems, apparently denominative in origin which have the causative accent : arthdyate 1 desires mantrdyate 4 takes council mrgdyate ‘hunts', etc. These are normally classified in the tenth present class. In the later classical language most of the old denominatives made from consonant sterns disappear. The thematic type re- mains living and takes two forms : (1) in the active the normal -ayati is used, kalusayaii 4 makes turbid tarunayati 4 rejuven- ates (2) in the middle, with intransitive sense, -dyate is used, kalasdyate 4 becomes turbid y tar undy ate 1 is rejuvenated It was noted above that the -dya- stem, properly a derivative from the nominal a-stem, was commonly used in the Veda to make denominatives from thematic stems, with the result that there are two alternative formations. In the later development of the language those two are specialised in different uses as just stated. Forms outside the present system from denominative stems occur with the utmost rarity. There are a few isolated is - aorist (avrsdyisata) and future forms ( kanduyisyati ). Parti- ciples in -ta {kanduyitd- 1 etc.) are somewhat more frequent. In the Vedic language abstract nouns in a iyasuya) and adjectives in -u (vasiiyu-) l made like the similar formations from the de- siderative stem, are common, but the type in general dies out later. THE VERB 365 § 17, Infinitives The difference between Vedic and classical Sanskrit is no- where more marked than in the infinitive. The classical lan- guage has only one form of infinitive, in turn, which is added to the gunated root ( kdrtum ), and which, like other verbal forma- tive s may be provided with the connecting vowel i (bhdvitum). In the Vedic language this formation is exceedingly rare, but there exists a whole series of other forms classed as infinitives which do not survive in the later language. These Vedic infini- tives consist of a variety of verbal action nouns inflected in various cases, namely : (i) Accusative , from root stems and stems in -tu ; pratiram ' to prolong datum 4 to give The former may be compared with the Qscan-Umbrian infinitives in - omjum : Umbr. erom, Osc. ezum ' to be Osc. edum f to eat *, etc. The latter, which eventually becomes the sole form of infinitive, has parallels in the Latin supine (datum) and in Balto-Slavonic (Lith. deh{, O. SL detu ' to place '). (ii) Dative, much the most frequent type. These infinitives are made from root stems (dr&i ' to see bhujd 4 to enjoy ’), from stems in -as ( dyase 4 to go arhdse ‘ to be worthy of '), from stems in -i (drsdye 4 to see ', yudhdye 4 to fight ’), from stems in -ii (vttdye 4 to enjoy sdtaye 4 to win '), from stems in -tu (Have 'to go’, ydstave ' to sacrifice ') from stems in -lava (dtavai ' to go ’), from stems in dhya ( duhadhyai 1 to milk sdhadhyai 4 to overcome ’), from stems in man (dimane 4 to give ’) and van (ddvdne 1 to give ’). Of these the infinitive in -tavai is remarkable in having a double accent (a phenomenon which has not been explained) , and in always being followed by the particle u (Hava u). This infinitive, and the one in -dhyai are also distinguished in being formed from stems which are not otherwise in active use, and also in preserving the older form of the dative singular which has been replaced by -dya in the declension of nouns. (iii) Ablative-Genitive , from root stems and stems in -tu : avapddas 4 falling down ’, samprcas 4 coming in contact 1 ; itos 4 going nidhdtos 4 putting down (iv) Locative , from root stems ( samcaksi 1 on beholding ’), stems in - san - (nesdni 'to lead 1 ), in -tar- (vi)dhartdri 4 to support (bestow) sotari f in the pressing THE VERB 366 The Vedic language (with Old Iranian) represents most accur- ately the state of affairs in Indo-European. The infinitive as an independent category is not yet fully developed. The forms classed as infinitives are various cases of verbal action nouns, in which as a general rule the case has its normal force : Acc. vdsti drdbham ' he desires to begin, wants a beginning Dat. avis tanvam krnuse dr id kam 1 you reveal your body for seeing AbL sd vm mahim dhunim dtor aramndt ‘ he stopped the great river from flowing A curious feature of the Vedic language is that the noun which is logically the object of the infinitive is placed in the same case as the infinitive, so that for instance ' to see the sun ' is expressed drsdye surydya, lit. 1 for seeing, for the sun ' ; simi- larly, with ablative, trAdhvam kart&d uvapddak * save us from falling into a pit lit. ' save us from a pit, from falling down In the normal usage of the Vedic infinitive there is not a great deal to distinguish it from an ordinary verbal noun in- flected in an oblique case. One of the few things that places these formations in a special category is the fact that the majority of verbal noun stems which appear in this usage are not otherwise used, nor in other cases. Taking the neuter s- stems as an example, there are many regular nouns so formed ( ydsas ' fame etc.), but there are in addition a large number which appear only in the dative case, in this infinitival use. Many such dative infinitives are a] so distinguished formally, since they are given an accent (jivdse) which is different from that of the neuter nouns. The infinitives in the Veda which are most removed from ordinary nominal formation are those formed from stems which are no longer used in the formation of ordinary nouns. Such are the dative infinitives in -dhyai and the comparatively rare locative formations in -sani and -tari. Another feature differentiating infinitive from verbal noun, one only partially developed in the Vedic language, is that it governs the accusative like a verb instead of the genitive like a verbal noun, e.g. mahi ddvdne ' to give something big ' as opposed to gotrdsya ddvdne 1 for the giving of a herd In the classical language where the infinitive in -turn has re- placed all others, the infinitive has become quite independent of the nominal formation. It also takes over the sense of the dative infinitive ( avasthdium sthdndntaram cintaya 1 think THE VERB 367 of another place to stay in ') so that its original force as the accusative of a verbal noun is obscured. In one respect it re- tains a trace of its nominal origin, because it can be compounded, like a noun-stem, with kama- and manas : yastukdma- 1 desirous of sacrificing vaktumanas - ' minded to speak The Sanskrit infinitive, in its final form, is much less developed and integrated into the verbal system than the infinitives of Latin and Greek. The latter languages have developed special forms for various tenses (esse, fuisse) and for the voices ( agere , agt), by a process of adaptation which took place independently in the two languages. Nothing of this kind appears in Sanskrit. There are in the Veda a few forms where the infinitive appears attached to special tense-stems (pusydse

• to flourish grmsdni ' to praise -pfccham ' to ask ' ; from

the perfect vavrdhddhyai ' to strengthen '), but these tentative formations came to nothing. The system by which the infini- tive is formed only from the root prevailed, and the syntactical use of the infinitive is correspondingly, wide. In particular it has to function not only as active and middle indiscriminately, but also, when the context demands it, as passive: kartum drabdhah * began to be made *, etc. This usage is particularly frequent with the passive forms of sak- : kartum na sakyate 4 cannot be done etc. §18. Active and Middle Participles Like the infinitives these participles are in origin purely nominal forms and as such have been treated in the chapters concerning the formation and declension of nouns* They be- long to the verb inasmuch as they have become integrated into the verbal system. This integration goes further than in the case of the infinitive in Sanskrit (though not as far as in Greek), and the process started earlier. The various participles are attached to particular tense stems, and they are divided, like the finite verb, into active and middle. The active participle in -ant- is in Classical Sanskrit entirely, and in the Vedic language mainly, formed from the present stems of the verb. In the Vedic language there is a small number of such participles which are attached to the root aorist stem (krdnt-, gmdnt - from kr- * do gam- 1 go ') and to the a-aorist stem (trpant- f vrdhdnt - from trp- ' be satisfied ' and vrdh- ‘grow ’). This association is mainly superficial, since such THE VERB 368 formations are not different from typical adjective formations with accented suffix added straight to the root. Before their integration into the verbal system the -ant- formations were ordinary adjectives (of which some examples remain, brhdni - 1 tall etc.), and the original type, derived straight from the root and having the adjectival accent, is preserved in these aorist participles. The adaptation of ant- adjectives to make participles began early, since there are ant- participles also in Hittite. But at the time of the separation of Hittite the ant- participle had not settled down into its final role since in that language the ant- participles are used in a passive sense, as opposed to the active sense in the rest of Indo-European. The specialisation of the formation in ant as an active participle was followed by its transference to the present system. The radical formations were replaced by formations made from the various types of present stem ( krdnt - by krnvdnt etc.). In the Veda this process is almost complete and the number of aorist participles is already small. By the classical period the process is complete. In Greek the same process began but ended differently, since there the appearance of present participles (^evyuiv) beside the older aoristic (i.e. suffixally accented type) <f>vywv led to the evolution of a twofold system in which these two types of participle, like the moods associated with the two tenses, express different kinds of action (punctual and durative). Greek has further extended the formation of this participle to the s-aorist stem, where it is to all intents and purposes non- existent in Sanskrit, as it was in Indo-European. The association of the active participle with the present system had the result that its accent (originally on the final, as an adjective) came to correspond to that of the verbal stem to which it was attached. It appears on the suffix in the case of the suffixally accented thematic class (tuddnt-) and in non- thematic verbs ( duhdnt -, srnvdnt-, etc.). On the other hand the radically accented thematic verbs keep this accent in the parti- ciple : bhdvant-, etc. The reduplicating verbs have accent on the reduplicating syllable associated with weak form of the participial suffix even in the strong cases : nom, sg. hlbhrat t acc. sg, bibhratam. The formation and morphology of the active perfect parti- ciple in -vasjus have already been detailed. The existence of a THE VERB 369 separate participle for the perfect is in accordance with the view already recorded that the difference between present/ aorist and perfect is the most original division in the verbal tense system. The perfect participle has the perfect sense (as opposed to the aorist participle which has no aorist sense, and as opposed to the moods of the perfect), cakrvds - ‘ one who has done etc. The accent is on the participial suffix and the per- feet stem appears in its weak form. The union vowel i ( tenivds , etc.) appears under much the same conditions as in the rest of the perfect. In the middle the participle used is in -anidna for thematic verbs ( bhdvamdna vi&dmdna- 1 cintdyamana-) and in -ana for non-thematic verbs ( duhand -, sunvdnd -, yunjand-, etc. ; accent final except in the 3rd class and intensives : juhvdna -, cekitdna-, etc.). The adaptation of these formations as participles is pro- bably later than that of the active participles in -ant, since com- parable forms are not widely spread in Indo-European. Corre- sponding to - anidna - Iranian has -amna- and Greek -ofievos, the actual forms varying in each case. No other IE languages have such participles, and where similar formations appear (Lat. alumnus > etc.) they are purely nominal. The participle in - ana is found only in Indo-Iranian, and only rare formations in the nominal derivation can be compared to it elsewhere (Lat. col onus, etc.). The middle usage of the participle is through adaptation, and it is certainly much later than the existence of middle forms in the finite verb. How the adaptation came about is no longer clear, since there is nothing about the related mal- formations of the noun that is connected with the middle, and in particular the Greek infinitives in -finviai), which have also become part of the verbal morphology, have an active, not a middle sense. What was said above about the integration of the active participle into the present system applies also to the middle participle. Like other derivatives based on the simple ^-suffix and the compound w^n-suffix these were originally made from the root, and after their adaptation as participles the present stem came to be used instead. In the classical language the aorist formations ( drsdnd -, vrdhand sucdmdna -) which incor- porate what remains of the old radical formations are replaced in favour of the present tense. In contradistinction to the active there is no special participial suffix for the perfect in the 370 THE VERB middle and the form - ana of the non-thematic verbs is used : cakrdnd-, jajndnd-, etc. This is in accordance with the fact, noted before, that the middle is later in the perfect than in the presen t-aorist system. §19. The Past Participle Passive This participle is most commonly made by the addition of the suffix -id to the weak form of the root (srutd- 4 heard etc.), and like other verbal derivatives it frequently employs the auxiliary vowel -2- (paiitd- ' fallen ', etc.). The meaning is passive except in the case of intransitive verbs (gatd- 1 gone etc.). The forma- tion is ancient in Indo-European as is clear both from the fact that it appears in large proportion of the languages, and be- cause it is subject to the old IE apophony. At the same time it does not appear to go back to the period wffien Hittite separated, since in that language the passive participle is expressed differ- ently, by the suffix -ant. In contradistinction to the active and middle participles it is not associated with particular tense stems but formed directly from the root both in Sanskrit and other IE languages, A minority of roots form their past passive participle in nd instead of -id. This is particularly the case with roots in -f (kirnri- 'scattered', glrn a 'swallowed'), roots in -2 [ksvnd- 4 wasted away '), roots in -d (bhinna- * broken *, chinnd - 4 cut ') and it is found in a number of roots in -j ( bhugnd - bent bhagnd- 'broken'). Very occasionally other suffixes are so used, notably pakvd- 4 cooked, ripe The details of these formations have been systematically treated in the section dealing with the formation of nouns, and need not be repeated here. The importance of the past participle passive increases in the later language, and still more so in Prakrit, on account of the change that took place from active to passive construction. It becomes customary in later time to express past actions not by active preterites but by the past passive participle associated with the instrumental : sa mayd drstah 4 he (was) seen by me ' for ‘ I saw him This resulted in middle Indo- Aryan in the elimination of the old preterites, and in modern Indo- Aryan all the tenses expressing the preterite are based on the old past participle passive. THE VERB 371 The past participle passive could be extended by the addition of the possessive suffix -vant : krtdvant- ' one who has some- thing (or things) done and this naturally assumes the func- tions of an active past participle. This is a creation of Indo- Aryan and the first purely participial formation of this character appears in the Atharva-veda : asit&vaty dtiihau ‘ one's guest having eaten Later the participle in tavant {-navant- when roots take -na in this participle) comes to be used independently, the copula being understood, in place of an active preterite : na mam kascid drstavdn 1 no one has seen (saw) me In the classical language this is the common usage and it forms another alternative to the use of the preterite tenses in addition to the passive construction mentioned above. §20. Gerundives or Future Passive Participles Classical Sanskrit has three verbal adjectives of identical function and having the sense of the Latin gerundive : kdrya-, kartavya karaniya- * to be done, faciendus Of these the first is the only one to be found in the Rgveda , where it is common. The suffix is normally to be pronounced - iya . Formations of this kind are found with all three grades of root : guhya- 1 to be hidden dvSsya - 4 to be hated vdcya - * to be said Final d of a root coalesces with the suffix to produce -eya : diya-

• to be given Roots in i, u, r commonly take the augment

before this suffix : srutya - ' to be heard The accent is normally on the root, but there are some exceptions : bhavyd adyd-. The formation in - tavya first appears in the Atharvaveda ( janitavyd - 4 to be born himsitavyd - 4 to be injured '). It be- comes commoner in the period of the Brahmanas, and in the classical language it is freely formed from all roots. The accent of the above two examples is the only type that occurs in accented texts. The grammarians allow also acute accent of the penultimate. In origin the formation is a secondary adjectival derivative from the action nouns in -tu. The gerundive in -antya (a secondary adjectival derivative from the verbal nouns in -ana) is likewise first recorded in the Atharvaveda (upajivamya) . It remains rare in the Brahmanas, but is common in the classical language, though not as frequent as the type in -tavya. THE VERB 372 In addition to the three types o f gerundive that appear in the classical language there are several formations in the same function that appear only in the Veda : (i) in -tva (normally pronounced as two syllables, -tuva) with accent and guna of root : kdrtva- 1 to be done jdtva ndntva vdktva etc. (ii) in -enya (- eniya -) ; drsenya , yudhinya , vdrenya , etc. (iii) in -dyya (trisyllabic) : panayya- ‘ to be praised, praise- worthy daksdyya- , sravayya etc. §21. Gerund or Indeclinable Participle This form of participle was analysed above as being a kind of adverbially used action noun. The type of participle is not familiar elsewhere in Indo-European and although the form is explicable through the normal IE processes of stem formation, its adoption in this particular syntactic use is in the main a development of lndo-Aryan. In the classical language the gerund is formed by means of the suffix -tva when the verbal root is uncompounded by pre- position, otherwise by the suffix -ya. In the Rgveda the latter suffix is in the majority of instances long (-yd) and this no doubt is the more original form, the suffix -yd making verbal abstracts being used adverbially in the same way as -tva. Other instances have been noted of final -d (-an) appearing as short a due to special circumstances of sentence sandhi (atra, etc.). Beside -tva the Rgveda also has a form -tvdya which appears to be a contamination of the two alternative forms just mentioned. It also has a form in -tv% (hitvi ' having left ', etc.) terminating in the suffix -i, wdiich is used in the same conditions as the -tva form and is commoner than it. This tvi- form though absent in Sanskrit, is continued in certain Middle lndo-Aryan dialects of the North-West and West, and was clearly a local dialectal feature of Old lndo-Aryan. Extended gerund forms in -tvdnam and -tvinam are mentioned as Vedic by the grammarians but examples of them have not been found in the extant literature. Middle lndo-Aryan has a common gerund in -tuna (gantiina) which appears to contain the same elements as - tvdnam , but with different apophony. The accusative of verbal action nouns in -a is used adverbially in constructions that resemble the gerund : imdny dngdni THE VERB 373 vyatyasam sete ' he lies down changing the position of these limbs etc. This usage does not occur in the earliest literature (RV., AV.), but it is common in the pre-classical prose. In the later classical prose it is comparatively rare, being used chiefly where the form is repeated : dariam-dariam 4 continually see- ing srdvam-srdvam 4 continually hearing CHAPTER VIII LOANWORDS IN SANSKRIT § i. Non-Aryan Influence on Sanskirt In the preceding chapters the history and development of the Sanskrit language has been described, from its remote Indo- European beginnings until it received final and definite form in India. The process was one of continual linguistic change, and when Sanskrit was artificially stabilised by the grammarians, this process was continued in the popular speech to produce first the Middle Indo-Aryan languages and finally the Modern Indo- Aryan languages. So far we have dealt only with developments that affected the inherited linguistic material which constitutes the basic texture of the language. But this is not all that has to be taken into consideration, since there are to be found in addition many elements in the language whose origin is to be sought elsewhere, namely in the influence of the various non-Aryan languages in contact with which Indo-Aryan developed. Such influence affected mainly of course the vocabulary of the language. In more general terms such influence is'seen in the phonetic development of a new series of occlusives, the so- called cerebrals. To begin with cerebrals appear in pure Aryan words as a result of phonetic changes affecting these ( nizda -> nizda-> mda-) and although such a development is a part of the processes taking place within Indo-Aryan itself, it can hardly be an accident that it should occur in the only branch of Indo- European which was in contact with languages possessing such sounds. In grammar the rapid loss of the Indo-European grammar in the stages subsequent to Sanskrit was very likely accelerated by the acquisition of Aryan speech by peoples who spoke originally different languages. On the other hand foreign influence in matters of detail is always difficult to establish. One feature in Sanskrit which may perhaps be assigned to such influence is the use of the gerund or conjunctive participle. In form these adverbial participles are of course purely Indo- 374 LOANWORDS IN SANSKRIT 375 European in origin, and their structure has been analysed above. On the other hand thic type of formation is not used to make such participles elsewhere in Indo-European, and the employ- ment of such adverbial forms to make a type of participle not familiar elsewhere is one of the special characteristics of Indo- Aryan. The same type of participle with the same kind of syntactic usage happens also to be a noteworthy feature of Dravidian. It may well be that the extensive use made of this formation in Sanskrit is partly due to the influence of Dravidian usage. Although a few points of this kind may profitably be exam- ined from the point of view of foreign influence, it is mainly in the vocabulary that detailed confirmation of such influence must be sought, and it is to an examination of this side of the problem that the present chapter is devoted. The basic vocab- ulary of Sanskrit is Indo-European, and it is this which has appeared in the preceding chapters, but in addition there exist large numbers of words which are without Indo-European etymology. In the very earliest language such words are few, but they progressively become more numerous. In the Middle Indo-Aryan period there is a further growth of new vocabulary, and again in the Modern Indo-Aryan languages there appears an abundance of words which are unknown to the earlier stages of the language. The tendency to substitute new words for inherited IE words has been permanently active in Indo-Aryan, Among common examples in Sanskrit we may note g hot aka- 4 horse ' which appears beside asva- in later Sanskrit and supplants it in the later history of Indo-Aryan, Similarly svdn- * dbg s gives way to kukkura- and its derivatives. It is not unusual to find pairs of names in Sanskrit, used equally commonly, of which one is non-Aryan, e.g. mar jar a- * cat ' ( mrj --) beside biddla vyaghrd-

• tiger ' beside sardula -, fksa - ' bear * beside bhalluka-. Some-

times the number of synonyms is much greater. The common word for elephant is hastin- (‘ possessed of a hand ’), but beside it, all in common use, we find gaja-, kunjara - p ibha ndga- and mdtanga Similarly beside mahisd- * buffalo ' we find other terms such as kdsara-, luldya -, sairibha - and heramba-. In such cases what are clearly local words, belonging originally to different languages have been adopted into Sanskrit, and the multiplicity of the Sanskrit vocabulary reflects an original LOANWORDS IN SANSKRIT 376 linguistic complication in India which has receded before the advance of Indo-Aryan. These few examples serve to illustrate the composite nature of the Sanskrit vocabulary, and the total number of such extraneous words is very large. Their source is mainly to be found in pre-Aryan languages of India. It is likely that there existed in India various linguistic groups which have been totally extinguished by the advance of Indo-Aryan, and in so far as Sanskrit has drawn words from such sources, their origin must remain for ever unknown. On the other hand those non- Aryan languages which have maintained their independent existence form a valuable source for the investigation of the extraneous elements in Sanskrit. It will therefore be conven- ient to enumerate the various groups involved, and to examine what contribution each has to make to the investigation of the problem. On its northern and eastern frontiers Indo-Aryan is con- tiguous with Tibeto-Burman languages and a number of such dialects' are spoken within the political frontiers of India. In spite of this contact no evidence of influence from this side on Indo-Aryan has been produced. This linguistic family has always remained essentially external to India proper. Further- more it appears that on the Eastern frontiers of India these peoples have displaced earlier Austro-Asiatic populations and that their contact is not very ancient. It is possible that a few Sanskrit words may eventually be traced to this origin, but at present no satisfactory evidence of such influence is available. In the extreme North-West of India there is found Buru- shaski, a language which so far stands on its own. A connection between Burushaski and certain of the Caucasian languages has been suggested, but not proved, though it is possible that further work in this direction might be fruitful. An earlier form of this language must have existed in this region before the Aryan in- vasion, and it is likely to have occupied a more extensive terri- tory. Evidence that Sanskrit has been influenced from this source has not been produced. It is a border language and such influence would only have been possible in the very earliest period of Indo-Aryan, since from the Vedic period it must have existed much as it does now, an isolated unit in a remote mountain tract. In Eastern India there is found a family of languages which LOANWORDS IN SANSKRIT 377 is of considerable importance from this and other points of view. The Munda or Kolarian languages as they are variously called, have in no instances achieved the status of literary languages but they are important scientifically firstly because there is definite evidence that Indo- Aryan had been influenced from this source, and secondly because of their connection with Mon, Khmer and other languages east of India. The most important centre of this family is the Chota Nagpur Plateau, where San tali, Mundari, and a number of fairly closely related dialects are spoken. In Orissa, not far from the above area occurs Juang, and further south, on the Orissa Madras border, Savara, Gadaba and two other dialects which form a special group within the Munda family. Of these Savara is particularly well preserved and less overlaid by Indo-Aryan than most members of the family. The most western Munda tribe is that of the Kurkus, who occupy the Satpura and Mahadeo hills in Madhya Pradesh. The most important linguistic family in India outside Indo- Aryan is the Dravidian family. Four members of this family have achieved the status of literary languages — Tamil, Malay- alam, Telugu, Kanarese, and in the case of Tamil the literary tradition goes back for at least two thousand years. Besides the major languages there are numerous minor non-literary Dravid- ian languages spoken in various pacts of India, namely : (i) Southern : Tulu, Coorg, Toda, Kota. (ii) Central : (a) KoIamFNaiki, Parji, Dravidian Gadba ; (5) Gondi, Konda, Pengo, Kui, Kuvi. (iii) Northern : (a) Kurukh, Malto ; (6) Brahui. The existence of the last member of the family in Baluchistan, far away from the main concentration of Dravidian is consistent with the theory that before the Aryan conquest Dravidian occupied a much greater area including considerably portions of Northern India. We shall see that the extensive influence of Dravidian on Sanskrit, beginning at an early period, also seems to point to this conclusion. In addition to the above-mentioned linguistic groups it has recently been established that the Nahali language still spoken by a small number of people in the Nimar district of Madhya Pradesh constitutes a further independent unit, unconnected with any of the groups previously mentioned. This language was briefly treated in the Linguistic survey of India where it LOANWORDS IN SANSKRIT 378 was wrongly assigned to the Munda or Kolanian family. Examination of the material provided by the Linguistic Survey itself rendered this assertion extremely doubtful, and recent field researches, by greatly increasing the material available, have made it clear that Nahali is the last surviving remnant of what must originally have been a quite independent family. Yet others may have existed of which nothing is now known, and the possibility that Sanskrit (and also later Indo- Aryan) has drawn on such sources is always to be borne in mind. Remains of an ancient language of India have been unearthed in the Indus cities of the third millennium b.c. So far no serious progress has been made in its decipherment, since no key to the solution is available. There is at present no means of knowing what kind of language is represented in these documents, which might be connected with one of the linguistic groups known in India, or be something quite different. Nor is there any means of knowing whether or not Sanskrit may have been influenced from this. There is only the possibility that some day, with the discovery of further information, a new chapter may be con- tributed to the linguistic history of India. From this brief survey it is clear that there are two practical sources where the origin of the non-Aryan element in Sanskrit may be sought, namely the Munda and Dravidian languages, and in both these directions progress has been made. As far as the Munda languages are concerned the main difficulty is that many of them have been inadequately explored. A necessary basis for the study of their influence on Sanskrit is a proper comparative study of the languages themselves, but this cannot be undertaken until adequate grammars and dictionaries exist for all the independent members. At present the most detailed information exists for the Northern group (San tali, Mundari, etc.) but this happens to be the one which has been most pro- foundly influenced by Indo- Aryan. Consequently in the absence of full comparative evidence it is often difficult to decide which way the borrowing has taken place. There^isTitso the question of the relation of Munda and Mon-Khmer. The evidence of this is clear enough to be decisive, but it has not been worked out in proper detail. This will eventually be necessary both for the comparative study of the Munda languages themselves, and for the special question under discussion, their influence on Indo- Aryan. LOANWORDS IN SANSKRIT 379 The connection between Munda and Mon-Khmer, etc., as members of a larger Austro-Asiatic family, has normally been assumed by those who have investigated this section of the Sanskrit vocabulary. Such etymologies are in some cases only available from Austro-Asiatic languages outside India* For instance one of the words for elephant mentioned above, mdtanga- has been explained as Austro-Asiatic for ' animal with a hand ' (cf. hastin-), but the forms with which it may be com- pared (tang 4 hand maintong ' elephant ') are quoted not from India but from the Malay peninsula* The same is the case with Skt. angand 1 women ' which is explained as containing a com- mon Austro-Asiatic word for woman with prefix an - : cf. Khmer kan, Mon k’ha , etc., with prefix a-, Bahnar akan , with prefix en~ f Nicobar enkdna , The bird known in Sanskrit as kulinga- fork-tailed strike ') has apparently an Austro- Asiatic name (Khasi khltn ‘ kite, eagle Khmer khlen , Stieng klin ' kite '), but forms are not quoted from Munda. Common Austro-Asiatic words may have ceased to be current in Munda, or not known through defective documentation, and conse- quently etymologies based on languages outside India may be consistent with Sanskrit having acquired the words in India, In some cases the source of a word is definitely to be sought out- side India, e.g. in the case of imported plants. Such is the case with lavanga- 4 cloves where the origin of the plant as well as the name (Javanese lawan , etc.) is to be sought in Indonesia. The following is a short list of words for which, with reason- able plausibility, a Munda, or more widely, Austro-Asiatic source has been claimed : aldbu ' bottle-gourd ' : cf. Malay labu, labo, Khmer Ibow, Batak labu , etc. unduru - 1 rat f : with prefix un- ; cf. Khmer kandor with dif- ferent prefix, Savara guntur- ' rat further Savara ondreri- ‘ rat kadall * banana ' : cf. Sakai telui, kelui, Nicobar talui, Khmer tut taloi , Palaong kloai ‘ plantain ' ; Savara kin-ten- f banana karpasa- 4 cotton ’ (> Gk. Kapnacros) : cf, Malayan kapas t etc. An unprefixed form appears in Grail pag t bag , Stieng patic, which may be reflected in Ta. panci, Ka. panji 1 cotton and possibly in Skt. pica id jamb ala- * mud ’ : cf. Santal jobg, etc., 4 damp Kharia LOANWORDS IN SANSKRIT 380 jobhi ‘ swampy ground jubild 4 a wet field Ho, jobe * mud Savarayoft^- ' id jim-, jemati ' to eat ’ (late ; common in Mod. TA, Hi. jevnd

• to eat jimdnd 4 to feed Mar. jevne 1 to eat etc.) : cf.

Santal jam, Kurku jome, Juang jim t Savara jvm, etc. tambiila- 4 betel ' : prefixed form ; cf. Alak baht, Khmer mluo , Bahnar bblbu, etc. ; various prefixes, Mon jablu , Halang lamlu , etc. No form is quoted corresponding exactly to San- skrit, but the same radical element is shared by all. marica- ' pepper ' : cf. Mon mrdk, Khmer meted * id In the Munda languages there are some forms corresponding to Skt. marica-, but the opinion now is that these are loans from Sanskrit. Idngala- f plough Pa. nangala : cf., with varying prefixes, Khmer ankal, Cam lanal, lunar , Khasi ka-lynkor, Malay tengala , tangdla, Batak tingala, Makassar nankala . In Munda there is Santal naheL This word is interesting because Dravidian has borrowed independently from the same source : Ta. nancil t Ka. negal , etc. A non-prefixed form with the change k>h characteristic of the northern group of Mun<Ja languages, appears in Sanskrit as hala- 4 plough sarsapa- 'mustard': Pkt. sdsava-; cf. Malay sesawi, etc. Old Tamil aiyavi, if form *sasavi also belongs here. This short selection of words is sufficient to show the import- ance of Austro-Asiatic as a source of Sanskrit words. When the languages concerned have been properly studied and properly compared it is expected that more will be available, and that there will be greater certainty about the detailed history of the forms concerned. At present such studies are in their infancy, so that it is not possible to estimate how much of the Indo- Aryan vocabulary will eventually prove to be derived from this source. The most important source of the foreign element in the Sanskrit vocabulary is to be found in the Dravidian languages. Although the comparative study of the Dravidian languages is still in its infancy, the position is much better than with the Munda languages. Full lexicographical material is available for the major literary languages, .and although much work remains to be done in the first-hand study of the minor lan- gauges, more is known about them than about the majority of the Munda languages. More work has been done on the influLOANWORDS IN SANSKRIT 381 ence of Dravidian on Sanskrit and more abundant results have been achieved. It is now possible to draw* up a considerable list of words in Sanskrit which can be traced either with certainty or with a high degree of probability to a Dravidian origin. This is illustrated by the following list: agurti 4 fragrant aloe wood ' : Ta. Ma. akil, Tu. agilu 4 id ankola- 4 Alangium hexapetalum ’ : Ta, arincil , Ma. arifinil

• id

anala- 4 fire-’: Ta. anal 4 fire; vb. to burn Ma. anal 1 fire ’ Ka. analu 4 heat arka - 4 Calotropis gigantea 1 : Ta. erukku, Ma. erikku, Ka. erke, ekke 4 id. . ulapa - ' bush, shrub, a kind of soft grass, a creeper ’ : Ta. ulavai 4 green twig with leaves on it, branch of a tree, grove, n. of various shrubs ulupin- 4 porpoise: Ka, unaci , Te. uluca, ulusa. ulukhala - ‘ mortar Ta. ulakkai 4 pestle Ma. ulakka , Ka. olake 4 id Te. rdkali 4 a large wooden pestle eda- 4 sheep, ram, wild goat ’ : Ta. ydtu, dtu 4 goat, sheep Ka. ddu 4 goat Tu. edu 4 id ', Te. eta 4 ram ', Go. etl 4 she* goat Brah. hot 4 id kanka- "heron': cf. Ta. Ma. Ka. kokku "crane, stork, heron Tu. kotngu 4 crane ’, Te. konga , Kuvi kongi 4 id kajjala- 4 soot, lampblack ': Ta. karical 4 blackness katu- 4 pungent, acrid, sharp ': Ta. katu 4 severe, pungent, sharp Ma. katu 4 extreme, impetuous, fierce katukka 4 to grow hard, sharp J , Ka. Te. T11. kadu 4 severe, intense etc. kathina - ' hard, firm, stiff ’ : Ta, katti 4 anything hardened, coagulated Ka. kadugu 4 to become hard ', gatti 4 firmness, hardness Tu. gatti 4 firm, hard Te. kaitidi 4 hard-hearted gatti 4 hard, firm kanra- 4 shoot of bamboo cf. Ka. karile 4 bamboo shoot J , Tu. kanile, Pa. karri, Kur. kharra 4 id Brah. kharring 4 to sprout kdka- 4 crow ': Ta. kdkkai , Ma. kdkka , Ka. kdke , Pa. kdkal , Kur. khdkhd. Malt, qdqe, Brah. khdkhd 4 id kdca- t kdja- 4 carrying yoke ' ; Ta. kd 4 id ’, kdvu 4 carry with yoke Pa. kdcal 4 carrying yoke r , kdn- 4 to carry with yoke Kui kdsa , Kuvi kdnju 4 carrying yoke kdncika-, kdnjika- 4 rice-gruel ’ : Ta. kanci, Ma. kanni, Ka. Tu. Te. ganji 4 id LOANWORDS IN SANSKRIT 382 kdnana- 4 forest 1 : Ta. kd 4 forest kan 4 id. ', kdnam 4 wood- land, grove kdnal 4 grove or forest on the seashore Ma. kdvu 'garden, grove kdnal 4 dry jungle Ka. kd 4 forest J . kdla - 4 black ' : Ta. kar 4 blackness ’, Ka. kar 4 id kargu 4 to turn black kuta - 4 pot ’ : Ta. Ma. kutam , Ka. koda, Ko. korm 4 id kuti - 4 hut, house ' : Ta. Ma. kuti, Ka. Tu. Te. gudi 4 hut, house, temple ', Kui kiiri 4 hut kutila- 4 crooked J : Ta. kotu 4 crooked kuta 4 curved, bent ' kutavu 4 bend, curve Ma, kotu, Ka. kudu 4 crooked kutt- 4 to pound J : Ta. kuti- 4 cuff, strike with the knuckles Ma. kuttuka 4 to pound, cuff ', Ka. kuttu 4 to beat, pound kunda- 4 hole in the ground, pit ’ : Ta. kuntu 4 hollow, pool, pit ', Ma. kuntu 4 hole, pit Ka. kunte, kunda . gundi 4 hole, pit *, etc. kundala- 4 ring, earring, coil of rope ; : cf. Ka. gunda, gundu 4 round Tu. gundu 4 anything round gundala 4 an ear orna- ment ', Te. gundrana 4 roundness gundrani 4 round kudddla- 4 kind of spade or hoe ' : Ka. guddali 4 kind of pick- axe, hoe ', Tu. guddoli , Te. guddali 4 id Ko. kuddy 4 hoe Malt, qodali 4 id ' : cf. Ka. guddu 4 strike, pound ', etc. kuntala - 4 hair of head 4 : Ta. Ma. kuntal , Ka. kudal 4 id kurula - 'curl': Ta. kurul 4 /to curl; a curl', Ma. kurul 4 curls Ka. kurul , Te. kurulu 4 id kulattha- 4 Dolichos uniflorus J : cf. Ta. kol , Ma. kollu , Tu. kudu , Pa. kol 4 id.’. kuvalaya- 4 lotus ' : Ta. kuvalai, Ka. kbmale , koval , kble 4 id kupa- 4 mast Ta. Ma. kumpu , Tu. kuve 4 id ketaka - 4 Pandanus odoratissimus ’ : Ta. kaitai, kaital, Ma. Ka. kedage, Te. gedage 4 id kemuka- (also kevuka kecuka-, kacu-, kacvi) 4 Colocasia anti- quorum ’ : Ta, Ma. cempu , Tu. tevu, Ka. kesavu , kesave, Te. 4 id J . kotara - 4 hollow, cavity': Ka. gotaru, gotru 4 hollow, hole (in wall, tree, etc.), cf. godagu 4 id Som* 4 corner J : Ta. 4 crookedness, corner, angle Aonw 4 to be bent, crooked ', Ma. kon * corner, angle J , konuka 4 to bend Ka. kon, kone, Tu. Aowe, Te. Aona 4 corner koraka- 4 bud J , Ta. kurai 4 sprout, shoot Kui koru 4 new shoot or bud Kur. khbrna 4 to shoot out new leaves ’, khbr 4 leaf-bud, new leaves Malt, qoroce 4 to sprout LOANWORDS IN SANSKRIT 383 khala- ‘threshing-floor': Ta. Ma. kalam ‘threshing-floor, open space ', Ka. kala y kana ‘ threshing floor', Te, kalanu , Pa. kali, Kui Jdai ‘ id khala - * a rogue Ta. kal ' to steal kalvan ‘ thief kalavu ' theft, deception Ka. kalla ‘ thief Te. kalla ‘ deceit kallari ‘ a rogue etc. ganda- ‘ lump, excrescence, boil ’ Ka. gadde ‘ mass, lump, concretion Te. gadda ' lump, mass, clot, boil guda- ‘ globe, ball ' : Ka. gudasu 1 anything round guddu ‘ eyeball, egg Te. guddu ' id ' ghuka- ‘ owl* : Ta. kukai, Ka. gwgi, gwge, gwii, Te. gwfoz, giibi ' id '. candana- ‘ sandal wood ' : Ta. cdntu ‘ paste, sandal paste cdttu ‘ daub, smear Ma. cdntu sandal paste Ka. sadu ‘ a fragrant substance Te. cddit 1 to rub into a paste capetd * slap with the open hand ' : Ka. capparisu ‘ to slap cappali 1 clapping the hands Te. cappata 1 a clap of the hands cikkana- ‘unctuous, viscid': Te. cikkd-badu ‘to become thick or inspissated cikkani ‘ thick or inspissated Ka. cigil jigil ' to be viscous, glutinous curnb- ‘to kiss': Ta cuppu ‘to suck', cumpu ‘to suck, fondle with the lips Tu. jumbuni ‘ to suck etc. cudd ‘ tuft of hair, crest: Ta. cutu to wear on the head hair tuft, crest Ma. cutuka ‘ to wear on the head cuttii ‘ cock's comb I£a. sudu, etc. talina - ‘ thin, fine, slender, meagre ': Ka, tel ‘ thinness fine- ness tellane , tellanna ‘ thin, delicate Te. tellena ' thinnish ', etc, tddaka tdla- f tdlaka - ‘ lock, bolt ' : Ta. tar ‘ bolt; bar tdr- kkol ' id Ma. Ka. tar, Tu. tdrkolu ‘ id tdmarasa- ‘lotus',: Ta. tdmarai , Ma. tdmara Ka. tdmare, Te. tdmara , Pa. tdmar ‘ id tala- palmyra palm Ka, tar, Te. tddu ‘ id tubari ‘ Cajanus indicus ' : Ta. tuvarai , Ma. tuvara, Ka. togari , tovari , Tu. togari , togare ‘ id tuvara- ‘ astringent ': Ta. ‘ to be astringent; astrin- gency tuvarppu ‘ astringent taste ', Ka. tuvara , tovara, togari , iogaru 1 astringent Kui tor pa ' to be astringent ' cotton, down ': Ta. Ma. ' feather, down etc. fiakra- ' alligator Ka. negar, Tu. negani , Te. negadu ‘ id nirgundV Yitex negundo ' : cf. Ta. noca, Tu. nekki Ka. nekki f lekki , /aMt ' id LOANWORDS IN SANSKRIT 384 nira- 4 water Ta. Ma. Ka. nir, Tu, niru, Pa, nir 4 water ', Kui niru ' juice, sap, essence Brah. dir 4 water patola - * Trichosanthes dioeca': Ta. putal, puialai , Ma. puttal, pittal , Ka. Te. potla ‘ id panda- 4 eunuch, effeminate man J : Ta. pen , 4 woman ', pettaiyan , e/i 1 hermaphrodite Ka. pen, penda 4 woman Te. 4 woman pedi 4 eunuch etc. palli ‘ house lizard 1 : Ta. Ma, Ka. Tu. palli 4 lizard / Te. balli 4 id palli 4 small village Ta. Ma. Ka. palli 4 hamlet, settlement, small village Te. palli , palliya 4 id punkha- 4 feathered part of arrow ' ; Ta, puruku 4 arrowhead J , Ka. pilukuy pilku ‘ feathered part of arrow puttikd 4 the w'hite ant or termite cf. Ta. purru, Ka. puttu, Te. puUa, Kur. puttd , Malt, pule ‘ white anthill This Drav. word also appears in Skt. as puta- in pipilikdputa- 4 anthill punndga Calophyllum inophyllum f : Ta. punnai , Ma. punna , Ka. ponne, punnike , Tu. ponne , Te. ponna 4 id baka- 4 crane ' : Ta. vakkd, vankd 4 white stork ', Te. vakku 1 crane bala - 1 strength Ta . val ‘ strong valam ' strength Ka. 6a/ 4 strong balume , baluhu 4 strength J , Tu. balu 1 big, power- ful ', Te. vali, valudu 4 id etc. bila - * hole, cave ' : Ta. vilava 4 cleft, crack vil, villu 4 to crack, split Ma. villu 4 to crack, burst open villa ! 4 a hollow, rent villu ' a crack, aperture bilva- 4 Aegle marmelos': Ta. vild , vilavu, cellil ‘ Feronia elephantum Ma. vild, Ka. belaval. Te. veldga 4 id manku- 4 confused, stupefied ' (Buddh. Skt. madgu -) : cf. Ta. makku 4 to become dull; dullness manku 4 to grow dim, lose lustre Ma. mannuka id Ka. manku dimness, obscurity V maggu ‘ grow dim or faint 1 . mayura - 4 peacock r : Ta. mannai , mayil , Ma. mayil, Tu, mairu , Pa. manil 4 id mallikd 'jasmine': Ma. mullai, Ma. mulla , Ka. molle, Te. tnolla 4 id tnasi- 4 ink, lampblack ’ : Ta. mai 4 blackness, ink, lamp- black Ka. tnasi 4 dirt, impurity, soot, ink J , Tu. map 4 coal, black powder, ink 1 Te. masi blackness, soot, charcoal, ink maid 4 wreath, garland ' : Ta. mdlai , Ka. male, Ma. Te. mala garland ’, Ta. malai 4 to wear as a garland LOANWORDS IN SANSKRIT 385 mm a- 4 fish ' : Ta. min, Ka. min, Te. minu, Go. min , Malt. minu 4 id mukuta - ‘crest, diadem': Ta. Ma. mukatu 4 top, highest part, head Ta. mucci 4 crown of head Ma. mukal r top, summit, ridge of roof Ka. Te. tnogadu 4 rudge of roof Tu. mugili 4 turret ',00, mukur 4 comb of cock mukula - 4 bud ' : Ta. Ma. mukir 4 a bud Ta. mukai 1 to bud; a bud Ta. mokkul 4 a bud Ka. mugul ' a bud; to bud m °gg u > m °gg e ‘ a bud Kui mogo ' bud mukta 4 pearl ’: Ta. muttu , muttam , Ma. Ka. Tu. muttu 4 id muraja- 4 drum Ta. muracu 4 drum J , mracam 4 id ', mural 4 ton sound ', Ka. more 4 to hum Te. morayu 4 to sound '. tnurungi Moringa pterygosperma 1 : Ta. murunkai, Ma. murinna , Ka. nuggi, nugge , Tu. nurige , nurge , Te. munaga, Pa. m-ulrjga 4 id/, etc. ZaZ<z 4 saliva, spittle': Ma. woZa, nola, Tu, noli, none, Ka. Idle 4 id'. valaya- ‘ bracelet ' : Ta. i/aZas 4 circuit, bracelet vb. to bend, be round, surround Ka. bale 4 bracelet balasu 4 to go round, encircle, encompass valli 4 creeper ' : Ta. Ma. valli , Ka. balli, Te, valli 1 id sakala - 4 scales of ‘fish; bark': Ta. cekil 4 skin or rind of fruit .fish-scales Tu. caguli 1 rind of fruit Malt, ceglo 4 shell of fruit simikd ‘ant ': Te. cima , Kol. sima, Kuvi sima 4 ant hintdla- 4 the marshy date tree J : Ta. intu 4 date pam ', incu , iccam-panai 4 id Ma. Ufa, itlal , Ka. Teal, Toil, Tu. tncilu , icilu, Te. idu, idadu 4 id hudukka- 4 small drum J : Ta. utukku , utukkai 4 a small drum Ma. utukka , Tu. uduku, Te. uduka 4 id heramba - 4 buffalo ' : Ta. erumai 4 buffalo Ma. erima , Ka. enirne, Tu, erme, Go. ermi, armi 4 id Concerning the date when these words were taken into Sanskrit it may be observed that the majority are post-Vedic. On the other hand it is important to note that there is a small nucleus already found in the Rgveda, Such are: ulukhala-, katuka-, kunda , khala-, bala-, bila-, mayura The number added in the later Samhitas (e.g. AV tula-, bilva-, VS kanka-) and in the Brahmanas (e.g. SB arka-, manku-) remains com- paratively restricted. The large majority first appear in the classical language, but in its early stage, being first recorded in LOANWORDS IN SANSKRIT 386 Panini, Patanjali, Mahabharata, Srautasutra, etc. The ma- jority appear also in Pali, which is important for dating since these canonical texts take us back to a period from 500- 300 b.c, The number that occur first only in later Sanskrit literature is again comparatively small. It is clear that as far as Sanskrit is concerned the active period of borrowing from Dravidian was well over before the Christian era. In Prakrit there are some new borrowings from Dravidian, but they are a good deal less numerous than those recorded above for the early Sanskrit period. They form only a small percentage of the new vocabulary of Prakrit. The common vocabulary of Modern Indo-Aryan has further new elements as opposed to Prakrit, but it is only rarely that any of these can be shown to be Dravidian. It is evident from this survey that the main influence of Dravidian on Indo-Aryan was concentrated at a particular historical period, namely between the late Vedic period and the formation of the classical language. This is significant from the point of view of the locality where the influence took place. It is not possible that at this period such influence could have been exercised by the Dravidian languages of the South. There were no intensive contacts with South India before the Maury a period by which time the majority of these words had already been adopted by Indo-Aryan. If the influence took place in the North in the central Gangetic plain and the classical Madhya- desa the assumption that the pre-Aryan population of this area contained a considerable element of Dravidian speakers would best account for the Dravidian words in Sanskrit. The Dravidian languages Kurukh and Malto are preserved even now in Northern India, and may be regarded as islands surviving from a once extensive Dravidian territory. The Dravidian words in the Rgveda attest the presence of Dravidian in North- Western India at that period. Brahui in Baluchis tan. remains as the modern representative of north-western Dravidian. It follows that the problem of Dravidian loairwords in San- skrit is somewhat different from what is usually met with in loanword studies, since the particular dialects or languages from which the borrowings took place have vanished leaving no record behind, and the major Dravidian languages of the South, with which mainly the comparisons must be made, are separated by great distances geographically and by anything up to a LOANWORDS IN SANSKRIT 387 millenium or over in time. Fortunately the differences be- tween the various Dravidian languages are not so great as to render dubious the reconstruction of the primitive form of the language and the form of words met with in the loanwords in Sanskrit does not differ materially from that which is arrived at by the comparative study of the existing Dravidian languages. It is a characteristic of the Dravidian languages that they have not evolved with the same rapidity as lndo- Aryan, and con- sequently the classical Dravidian languages and even the minor spoken languages recorded only in modern times can be used profitably to trace the Dravidian origin of Sanskrit words which were borrowed before any of these languages are themselves recorded, and from other ancient Dravidian dialects which have themselves disappeared. §2. Loanwords from Greek and Iranian The Sanskrit vocabulary acquired a limited number of Greek words, partly as a result of the rule of the Bactrian Greeks in North-West India in the second and first centuries B.C., and partly through contacts in respect of trade, etc. with the Graeco-Roman world. Words that can be ascribed to the Bactrian Greeks are khalina- ' bridle ’ (yaAlvo?), surungd 1 underground passage ' (avpiyg) and pari stoma- ‘ coverlet, blanket' (neplcnpwpa), and possibly kunla- 4 lance ' ( kovtos ). The Sanskrit lexica have preserved a word kenika 1 tent 1 which can be explained as a Prakritic adaptation of Gk. uKi^vip In common with other Hellenistic rulers the Greeks of Bactria and India adopted the title * saviour ' (rendered tratara- in their coins). This title, not in its precise sense, but as an honorific epithet gained currency in the local Prakrits as sottra-jsodira-, and thence was adapted into Sanskrit as sautira- * hero, noble and generous man Other words of Greek origin are better accounted for as having been acquired through trade, etc., e.g. kastira- ‘ tin ' ( KaaoiTepos ) meld ‘ ink * (fieAav), marakata- c emerald ’ (/xapaySos, opapayhos) kimpala- 'kind of musical instrument' (KvpfioAov) . Some words of Greek origin which only appear very late in Sanskrit have probably been taken in the first place from some intermediate language, e.g. kalama- ' pen ’ (xraAa/xo?) dramma- ‘ a coin ' (Spay/x^). The word kramelaka- * camel ' (kci/x^Ao?) is also late, and disguised by popular etymology (as if from kram-). LOANWORDS IN SANSKRIT 388 It is possible that harimantha- * chickpea ’ has been similarly adapted from Gk. epefiivBos, though if so it has been even more effectively disguised. There is a special class of Greek words occurring as technical terms in astronomical literature. The development of Indian astronomy in the early centuries of the Christian era was pro- foundly affected by Greek astronomy and a considerable number of its technical terms were adopted. Such words are heli- ‘ sun ’, hord ' hour kendra - ' centre of a circle ’ jdmitra-

• diameter ’ (Gk. rjXivs, utpa, Kevrpov, StdpLerpov). Some of these

also were distorted by popular etymology, e.g. hrdroga - ‘ the zodiacal sign Aquarius J (Gk. vSpoxoos). Contact with the Graeco-Roman world introduced one Latin word only into Sanskrit, namely dmdra- ‘ name of a coin * (cdenarius). Following on the Greek invasions the northwestern part of India was the scene of a series of invasions by Iranian tribes, Pahlavas, Sakas, Kusanas and finally Hunas. This state of affairs prevailed during most of the first six centuries of the Christian era, and resulted in the establishment of a number of powerful dynasties of such origin, principally in Northwestern and Western India, but sometimes extending their sway further afield. This resulted in the adoption of a number of Iranian terms, first into Prakrit (principally the North-Western Prakrit), and eventually into Sanskrit. Examples from Indian inscriptions in Kharosthi and Brahrni scripts are horaka- ' donator J , and bakanapati- ' shrine attendant * (Khot. haur hot- ‘ to give Sogdian fiynpt). In the North-Western Prakrit used in Central Asia these are quite frequent. In Sanskrit itself there are a not inconsiderable number which have entered the language at various periods. One of the earliest is lipi ‘ writing J (occurring in Asokan also in the form dipi-) from OPers. dipi-. This word was borrowed at the time writing was introduced into India. Later a word divira- ‘ scribe 1 of the same origin was introduced, though it had restricted .currency. The word mudrd ' seal ’ is likewise attributable to Old Persian, although.it happens not to be recorded in Iranian until later (Pahl. NPers. muhr ), and karsa- 1 n. of a particular weight ' has been con- sidered to derive from OPers. karsa-. Among the other Iranian loanwords the following may be listed : ksatrapa - ‘ satrap ’ used as a title by certain of the above LOANWORDS IN SANSKRIT 389 mentioned Iranian rulers, varahana - 4 armour, mail * {01 r.

• varopdna- 4 protecting the breast f ), khola- 4 helmet, a kind of

hat J (Av. xaoSa-, PaSt. xol), jagara- 4 armour ’ (cf. Pa§t. zgara 4 armour ’), mddhl 4 armour mail * (MPers. mdSi[k)) t tin ( a kind of arrow 1 (NPers. tty) parydna- 4 saddle 4 (Ir. *paridana - in NPers. pdldn 'pack-saddle*, etc.), pllu - 4 elephant 4 (NPers. pit) bandx 4 prisoner, captive ’ (NPers. bandah), gola- 4 ball 4 (NPers. gdy <*gauda-) } ganja- 4 treasury’ (NPers. ganj), As can be seen from this list the words borrowed from Iranian have mainly to do with military affairs and equipment, which is in accordance with their military superiority over many centuries. Of words not connected with w r ar, mention may be made of pustaka- 4 book ’ ultimately derived from Iranian (Pers.) post 4 skin ' and mihira - 4 sun ’ (Pers. mihr OIr. midra-). Some Iranian loanwords are attested only in writers from Kashmir (e.g. ganja-, divira-) where they derive from the North-Western Prakrit. There are others which are found only in Buddhist Sanskrit, e.g. kdkhorda- 4 wizard ' (cf. Av. kax'arha-) mocika- 4 shoemaker ’ MPers. mocak) whence Hindi mod 4 id These words are also derived through the North-Western Prakrit, since it was the Buddhist writers of the North-Western schools who were responsible for their adoption. APPENDIX TO THE THIRD EDITION Supplementary Notes and Bibliographical References CHAPTER I § i . For a detailed discussion of the term arya- and its cognates see H. W, Bailey, Iranian arya and daha , Transactions of the Philological Society, 1959, pp- 71-115. For arya- as applied to the language of the Indo-Aryans, cf. Saiikhayana Aranyaka 8, 9: yatraryd vdg vadati, and Aitareya Aranyaka 3, 2, 5: yatra kva ca arya vaco bhasante. For similar usage in Buddhist Sanskrit, see Bailey, op. cit. p. 102. The term Samskrta is late in appearing, and it is not used by Panini or Patanjali. The earliest recorded occurrence is in Ramayana, 3, 10, 54, after which it becomes quite common (Nafyasastra, Su£ruta, Kavyadarsa, etc). § 3. The mutual relations of the Indo-European languages, as well as the question of their original home, have continued to be the subject of active discussions. Among recent works dealing with these problems the following list is a selection: W. Porzig, Die Giiederung des indogermanischen Sprachgebietes , Heidelberg, 1954; P- Thieme, Die Heimat der indogermanischen Gemeinsprache, Mainz, 1953; H. Krahe, Sprache nnd Vorzeit , Heidelberg, 1954; H. Hencken, Indo-European languages and archaeology (American Anthropological Association, Memoir 84) 1955; P. Bosch-Gimpera. Les Indoeiiropeens: problemes archeologiques , Paris, 1961; G. Devoto, Origine indoeuropee , Firenze, 1961; M. Gimbutas, The Indo-europeans ; archaeo- logical problems (in American Anthropologist, 65, pp. 815 ff.), 1963; V. Georgiev, Introduzione alia storia delle lingue indo- europee, Rome, 1966; G. Cardona, H. M. Hoenigswald and A, Senn (ed.), Indo-European and Indo-Europeans, Philadelphia, 1970; R. A. Crossland, Invaders from the North (in Cambridge 390 APPENDIX TO THE THIRD EDITION 39I Ancient History, pp. 824-876), 1971- Opinions as to the original Indo-European homeland continue to differ; sugges- tions include the region south of the Baltic sea (Thieme), the steppes north of the Caucasus (Gimbutas) and the Danube basin and surrounding regions (Georgiev). The last alternative seems to fit in best with the historical distribution of the languages and dialects., § 4. The inport&nce of the connections between Indo- Iranian and Baltic in particular has been stressed by H. W. Bailey (BSOAS 21 (1958), pp. 42 ff.) in connection with such words as navanita- 4 butter', netra- ' churning string 1 (cf. Lett. niju, nit ' to make a circular movement, churn '), btja- ' seed ' (with 6-<w-; cf. Lith. miezys 'barley-corn', Lett, miezis), rip- * ascent, elevation ' (Lith. lipti 4 to rise, ascend '). Further Baltic comparisons, to be added to the list given in this section are as follows: land - 'to be weary, slothful', tandri ‘ sloth ' : Lith. tandus 4 lazy, slothful ' ; murkhd- 4 fool ' : Lith. mulkis ; tvis- 'to sparkle, glitter': Lith. tviskdti 'to lighten'; sakala- ‘ chip, fragment, splinter': Lith. Sakalys ' chip of wood ' ; pula - ' bunch, bundle Lett, pulis ' heap bull- * anus Lith. bulls vala - ‘ the hair of an animals tail (particularly a horse's tail) ' : Lith vdlas 4 hair of horse's tail § 5. The Indo-Aryan (and Indo-European) loanwords in the Finno-Ugrian languages are listed by B. Collinder in Fenno - ugric vocabulary; an etymological dictionary of the Uralic languages , pp. 129-141, Stockholm, 1955. § 6. Since 1955 a small amount of new Aryan ^material from the Near East has turned up. The documents from Nuzi have revealed certain colour adjectives applied to horses, in Hurrian form paprunnujbabrunnu , pinkarannu , paritannu , correspond- ing to Sanskrit babhru - pihgala - and palitd The change of original -Z- to -r- in parita- and pinkara-, a change which had previously been noted in SuriaS, is of considerable importance, since it shows that this change, characteristic of Iranian and the Rgvedic dialect of Old Indian had already taken place before 1500 B.C. Other words that have been noted in this connection are makanni 4 gift maninnu 4 neck ornament cf. Skt. mani -, maghd and possibly urukmannu corresponding to Skt. rukmd392 APPENDIX TO THE THIRD EDITION

• bright ornament V If the connection between mistannii
• reward ’ and Skt. mtdhd - proposed by. Mayrhofer is accepted,

then we have a pre-Vedic Indo-Iranian form ( *mizdha -) repre- sented. A complete bibliography to the end of 1965, and a summing up of the subject, is provided by M. Mayrhofer in Die Indo- Arier im alien Vorderasien; mil einer analytischen Bibliographic , Wiesbaden, 1966. There is also a detailed discussion of the subject by A. Kammenhuber in Die Arier im vorderen Orient, Heidelberg, 1968. The work contains valuable discussions, but carries scepticism too far (on which see M. Mayrhofer, Die vorderasiatischen Arier , Asiatische Studien (fitudes Asiatiques) XXIII, pp. 139-154), Bern, j 9 6 9 * The Aryan gods of the Mitanni treaties are discussed in an important article by P. Thieme in JAOS, 80, pp. 301-317, ig6o. CHAPTER II § 2. On the subject matter of this section see further: M. B. Emeneau, The dialects of old Indo-Aryan, in H. Birnbaum and J. Puhvel (ed.), Ancient Indo-European dialects , Berkeley and Los Angeles, 1966. § 3. The study of Panini and the Sanskrit grammarians has been actively pursued during recent years. The following are the most important publications since 1955: P. Thieme, Panini and the Pdnintyas, JAOS, 76, pp. 1-23, 1956; L. Renou, Terminologie grammaticale du Sanskrit (2nd ed.), Paris, 1957; Y. Ojihara and L, Renou, La Kdsikd-vrtti {Adhyaya I, Pada i), Paris, 1960-67; K. V. Abhyankar, A dictionary of Sanskrit grammar , Baroda, 1961; B. Shefts, Grammatical method in Panini; his treatment of Sanskrit present stems , New Haven, 1961 ; K. Birwe, Der Ganapdtha zu den Adhydyas IV and V der Grammatik Pdninis , Vet such einer Rekonstruktion , Wiesbaden, 1961, and Studien zu Adhyaya III der Astddhyayt Pdninis , Wiesbaden, 1966; V. N. Misra, The descriptive technique of Panini, The Hague, 1966; R. Rocher, La tkeorie des voix du verbe dans Vecole panininienne, Brussels, 1968; G. Cardona, Studies in Indian grammarians , I : the method of description reflected in the Sivasutras, Philadelphia, 1969. APPENDIX TO THE THIRD EDITION 393 § 7 (p. 61). The use of mixed Sanskrit in inscriptions, par- ticularly of the Kushanas, is discussed, and illustrated with examples by E. Lamotte in Histoire de la Bouddhisme indienne, pp. 640-41, Louvain, 1958. On Buddhist Sanskrit see further H. W. Bailey, Buddhist Sanskrit , JRAS, 1955, pp. 13-24; J. Brough, The language of the Buddhist Sanskrit texts, BSOAS, 16, 357-375, 1954; V. Ragh- avan, Buddhist Hybrid Sanskrit , Indian Linguistics, 16, 313-322. (pp. 61-2). The vocabulary of Jaina Sanskrit has now been dealt with by B. J. Sandesara and J. P. Thaker in Lexico- graphical studies in Jaina Sanskrit , Baroda, 1962. § 8. For Sanskrit in Indonesia see now J. Ensink and J. A. B. Buitenen, Glossary of Sanskrit from Indonesia , Vdk , no. 6, Poona, 1964. CHAPTER III § 3. There is now a detailed study of the surd aspirates by R. Hiersche : Untersuchungen zur Frage der tenues aspiratae im Indogermanischen , Wiesbaden, 1964. Hiersche rejects the laryngeal explanations, and considers the aspiration to have developed mainly in combinations with sibilant (sthdgati, as opposed to Gk. oriyca, e tc., a phenomenon to which reference was made above, p. 72). Initial surd aspirates are explained by assuming loss of mobile 5- ( phena OPruss. spoayno , etc.). Unfortunately Hiersche does not deal at all with those cases of sonant aspirates where such an explanation is impossible : e.g. rdtha-, sapha-, sahkhd-, iAkha. For the opposing view see F. B. J. Kuiper, Indo-Iranian Journal, 9, pp. 218-227. If seems however that only a portion of the instances can be explained by the laryngeal theory, and that for others (e.g. phena-) an explanation on the lines proposed by Hiersche is preferable. §5. The statement (p. 75, 1. 21) that the satem languages have uniformly abandoned all trace of the labial element needs qualification in one respect as far as Sanskrit is concerned. As first pointed out by O. Szemerenyi in a paper f The problem of Indo-European labio- velars * read to the Philological Society in March, 1952, roots in f have a weak form in urjur when the original was a labio- velar, just as happens in the case of the 394 APPENDIX TO THE THIRD EDITION labials, whereas when the initial was a pure velar the weak form is in ir/ir. This is seen most clearly in the derivatives from the root gf-jgur 4 to welcome ' (original labio-velar) on the one hand, and gffgir - ' to sing * on the other. See further my article Sanskrit gfjgur - ‘to welcome' (BSOAS, 1957, pp. 133-144), and Szemerenyi, Einfuhrung in die vergleichende Sprachwissen- schaft, pp. 60-61, § 10. The most recent comprehensive discussion of IE mobile s is by F. Edgerton, IE s movable , Language 34, pp. 445^453- He takes it to be a sandhi-phenomenon due to the great fre- quency of final -5 in Indo-European. § 11. The subject matter of this section is dealt with in detail in my two articles, On the phonological history of Sanskrit ksdm- 4 earth fksa- 4 bear * and liksd 4 nit ', and Sanskrit ksi-: Gk. pQivio , in Journal of the American Oriental Society, 79, pp. 85-90 and 255-262, 1959. The subject is treated differently, but beginning from the same standpoint (IE dheghom- 4 earth etc.) by W. Merlingen in Mv-q^ x^P LV (Gedenkschrift Kret- schmer)II, 49 ff. 1957, and Die Sprache, 8, pp. 74-76, 1962. See also O. Szemerenyi, Einfuhrung in die vergleichende Sprach - wissenschaft, pp. 46-47. § 13. The literature on IE h and its varieties (‘ laryngeals ') since 1955 is extensive. A useful survey and summing up is to be found in W. Winter (ed.) Evidence for laryngeals, The Hague, 1965; especially to be recommended is E. Polome's intro- ductory survey in this volume: The laryngeal theory so far. A critical bibliographical survey (pp. 9-78) . § 17. On Fortunatov's law see my article A reconsideration of Fortunatov's law , BSOAS, 35, pp. 531-544, 1972, which con- tains a brief account of the disputes concerning this subject, as well as a defence of Fortunatov’s theory. As regards spontaneous cerebralisation it was made clear by H. W. Bailey in a series of articles from 1952 onwards that this has taken -place to a much greater extent than previously recognised. I have discussed this subject, adding to Bailey's material, in my article Spontaneous cerebrals in Sanskrit , BSOAS, 35. pp- 538-559. 1971- There are also some cases where the occurrence of cerebral -5APPENDIX TO THE THIRD EDITION 395 after a and a can be accounted for by special reasons; see my articles, Sanskrit jdldsa in W. B. Henning Memorial Volume, pp. 89-97, T 9 > I 9 70, and Sanskrit sdspa - and baspa -, JRAS, 1969, pp. 112-117. § 18. Under miscellaneous changes reference should be made to H. W. Bailey's remarks on the alternation of b and m (Skt. bija- : Ir. miz-, Lith. miezys ) in his articles Iranian missa-, Indian bija-, BSOAS, 18, pp. 32-42, and Missa suppletum, ibid . 2i, pp. 40-47. See also my article, Sanskrit amoda - ‘ fragrance , perfume Indological Studies in honour of W. Norman Brown, pp. 23-27, New Haven, 1962. § 24. There have recently been some investigations into the rendering of the Vedic accents by traditional reciters of the Veda in S. India. See J. E. B. Gray, An analysis of Nambudiri Rgvedic recitation and the nature of the Vedic accent, BSOAS, 22, 499-430, and J. F. Staal, Nambudiri Vedic Recitation, The Hague, 1961. CHAPTER IV § 5 (p. 131). On the IE suffix - men - (and the Sanskrit suffix -man-) see now J. Haudry, Le sufiixe I.E. -men, BSL, 66, pp. 109-137,1972. § 13. The Sanskrit s-suffixes have been studied in- detail by J, Manessy-Guitton in a number of books and articles; Les substantifs en -as dans la Rksqmhitd ; contribution a V etude de la morphologie vedique, Dakar-Paris, 1961; Rechtrches sur les derives nominaux a bases sigmatiques en Sanskrit et en Latin , Dakar, 1963; Les adjectifs simples en -as du Rk-samhitd II J, 7, 258-283; Les noms Sanskrits en - nas , Ilj, 8, 171-196, 1965. § 14 (p. 168). On the Sanskrit stems in -ti see the detailed study by G. Liebert, Das nominal suffix -ti im altindischen, Goteborg-Lund, 1947. (p. 173). On the Vedic abstract nouns in -tat and^ -tdti see now L. Renou, Les derives abstraits en -tat et -tdti du Rgveda, BSL, 55, 10-19, 1960396 APPENDIX TO THE THIRD EDITION CHAPTER V § 9 (p. 256)* The instrumental singular termination - ena of the a-declension is the subject of a detailed study by C. Hauri: Zur V orgeschichte des Ausgangs - ena des Instr. Sing, der a- stamme des Altindischen , Gottingen, 1963. CHAPTER VI § 1. The most recent comprehensive study of the Indo-Euro- pean numerals is by O. Szemerenyi: Studies in the Indo- European system of numerals , Heidelberg, 1960* § 2 (p. 268, 11 . 33-34). The question as to whether Prakrit se is to be directly compared with the Iranian words quoted in this section has long been a matter of dispute. The subject has recently been exhaustively discussed by M. Seheller; Das mittelindische Enklitikum se , KZ, 81, pp. 1-53. He comes to the conclusion that Pkt, se is a secondary development from un- accented asya. CHAPTER VII § 2 (p. 298 ff.). On the Indo-European (and Sanskrit) moods, see J. Gonda, The character of the Indo-European moods , Wies- baden, 1956. (p. 299). The use of the Vedic Injunctive has been exhaus- tively examined and discussed by K. Hoffman: Der Injunktiv im Veda, Heidelberg, 1967. § 3 (P* 302). The question of multiple presents formed from Sanskrit roots has been discussed by J. Vekerdi: On poly- morphic presents in the Rgveda, Acta Orientalia Acadentiae Scientiarum Hungaricae 12, pp. 249-287, Budapest, 1955. §9. On the periphrastic future see now J. Gonda, A critical survey of the publications on the periphrastic future in Sanskrit, Lingua, 6; 15&-179, §10. The Sanskrit aorist has been studied by.T. J. ElariAPPENDIX TO THE THIRD EDITION 397 zenkova, Aorist v Rigvede , Moscow, i960, and the sigmatic aorist in particular, very fully and exhaustively, by J. Nartem Die sigmatischen Aoristen im Veda, Wiesbaden, 1964. Cl further on the reduplicated aorist, M. Leumann, Der altin - dische kausative Aorist ajijanat , Indological Studies in honour of W. Norman Brown , pp. 152-159, New Haven, 1962, and on the sa -aorist, S. Insler, The Sanskrit sa-Aorist, Munchener Studien zur Sprachwissenschaft , 26, pp. 43-50, 1969. § T 3 (P- 35°) ■ Concerning the imperatives in -si, it should be stated not only that they are not normally formed from roots having root presents, but also that in the majority of cases they are formed from roots which make an 5-aorist. From this it is clear that these forms are to be attached to the 5-aorist stem, and they may be most simply classified as s-aorist imperatives. This is the conclusion reached by G. Cardona in a recent study of this problem; The Vedic imperatives in -si f Language , 41, pp. 1-18, 1965. On the other hand O. Szemerdnyi (Language, 42, i-6 f 1966) prefers to regard them as syncopated forms of original subjunctives (darsasi>darsi). This appears to be less satisfactory, not only on account of the phonetic difficulties involved, but also because the primary and predominant use of these forms is as imperatives. The few cases, to which Cardona and Szemerenyi draw attention, in which they are used in sub- ordinate clauses, are probably to be accounted for as misuse by later poets of a form which had become obsolete. § 14 (p. 352). The termination -s of the third person singular active of the root aorist optative/precative is found also in Iranian, for which see my article The Sanskrit Precative in Asiatica (Festschrift Weller), pp. 35-42, 1954. § 15 I. On the passive see J. Gonda, Remarks on the Sanskrit passive, Leiden, 1951. § 17. The latest study of the Vedic infinitives is by P. Sgall, Die Infinitive im Rgveda, Orientalia Pragensia I, pp. 137-268, Prague, 1958. CHAPTER VIII § 1 (p- 374)* O n Sanskrit ghotaka- 'horse ' see now my observa- tions in International Journal of Dravidian Linguistics I, pp. 398 APPENDIX TO THE THIRD EDITION 20 ff. (1972) where it is argued that it is in origin a middle Indo- Aryan descriptive epithet, (p. 380). On marica - in the Munda languages see now N. H, Zide, Current Trends in Linguistics , Vol, V, pp. 420-421. In the same paragraph he also queries the origin of the verb jim- from Munda, on the ground that the vowel is different in the Mun<Ja languages (Santali jom- t etc.), and further remarks that ■ the identification of words in Indo-Aryan and Dravidian as Mun<Ja loans, even when this has been done by careful scholars, is not often convincing, particularly in the light of newer data As regards sarsapa - ' mustard the Austro-Asiatic origin of this word has now been rendered doubtful by an article by W. B. Henning (Istituto Orientale di Napoli, Annali , Sezione Linguistica, VI, 29-47, x 9 & 5 )> where he treats of the Iranian words for ‘ mustard ' going back to an original *$in&apa-> and obviously related to the Sanskrit word. In view of this Malay se$awi> etc., should probably be regarded as loanwords from Indo-Aryan rather than vice-versa. SELECT BIBLIOGRAPHY Allen, W* S. Phonetics in ancient India, Oxford 1953. 1 Sandhi; the theoretical , phonetic and historical bases of word- junction in Sanskrit , 's-Gravenhage 1962. Arntz, H. Sprachliche Beziehungen zwischen arisch und Balio-slavisch, Heidelberg 1933. Benveniste, E* Origines de la formation de$ noms en indo- europien , I, Paris 1935. Bloch, J. L’indo-aryen du Veda aux temps modernes , Paris 1934 * Brugmann, H. Grundriss der vergleichende Grammatik der indo-germaniscken Sprachen, 2nd ed., Strassburg 1897-1916 (Reprint, Berlin 1967). Kurze vergleichende Grammatik der ind o-ger m anis chen Sprachen, Strassburg 1902-4 (reissue, Berlin 1922),. BChler, J. G. Ihdische Paldographie , Strassburg 1896. Chatte r j 1, S. K. Indo-Aryan and Hindi , Ahmedabad 1942. DelbrUck, B. Altindische Syntax , Halle 1888. Edgerton, F. Buddhist Hybrid Sanskrit, I Grammar, II Dictionary, Newhaven 1953. Friedrich, J* Hethitisches Elementarbuch , I, Heidelberg 1940. Geiger, W. Pali Literatur und Sprache , Strassburg 1916. Ghosh, B. Linguistic Introduction to Sanskrit , Calcutta 1937 - Gonda, J. Old Indian (Handbuch der Orientalistik, II Abt. Bd, I, Abschnitt 1), Leiden/Koln 1971. Hendriksen, H. Untersuchungen fiber die Bedeutung des Hethitischen fur die LaryngaUheorie , Copenhagen 1941. Jacobsohn, H. Arier und Ugrofinnen , Gottingen 1922. Kent, R. G. Old Persian Grammar , Texts , Lexicon , Newhaven 1950. K u r Y L o w I c z, J. U accentuation des langues indo-europdennes , Cracow 1952. Uapophonie en indo-europe'en, Wroclaw 1956. 399 SELECT BIBLIOGRAPHY 400 Macdonell, A. A. Vedic Grammar , Strassburg 1910. Mansion, J. Esquisse Tune histoire de la langue sanskrite, Paris 1931. Mayrhofer, M. Kurzgefasstes etymologisches Wdrterbuch des Altindischen (in progress), Heidelberg 1953 ff. — — Sanskrit-Grammatik (mil sprachvergleichenden Erlailter- ungen), Berlin 1965. Meillet, A. Les diahctes indo-europdennes, Paris 1908 (1922). Introduction a Vdtude comparative des langues indo-euro- pdennes , 8th ed., Paris 1937 (1949). Pedersen, H. Hettitisch und die under en indoeuropdischen Sprachen, Copenhagen 1938. Tocharisch von Gesichtspunkt der indoeuropdischen Sprach- vergleichung, Copenhagen 1941. Pisani, V. Grammatica del Antico Indiana (incomplete), Rome 1930-33. Pischel, R. Grammatik der Prakrit-Sprachen , Strassburg 1900, Pokorny, J. Indogermanisches etymologisches W or ter buck, Bern 1959. Reichelt, H. Avestisches Elementarbuch, Heidelberg 1909. Renou, L. Grammaire sanscrite , Paris 1930. Grammaire de la langue vedique, Paris 1952. — — Histoire de la langue sanskrite , Lyon and Paris 1956. Specht, F. Der Ur sprung der indogermanischen Deklination, Gottingen 1944 (1948). Speyer, J. Sanskrit Syntax, Leiden 1886, — — Vedische und Sanskrit-Syntax , Strassburg 1896. Sturtevant, E. H. The Indo-Hittite Laryngeals , Baltimore 1942. A Comparative Grammar of the Hittite Language (Revised Edition), I, New Haven 1951. Sz em ere ny i, O. Einfuhrung in die vergleichende Sprach- wissenschaft, Darmstadt 1970. Thumb, A. Handbuch des Sanskrit . Dritte, stark umgear- beitete Auflage von Richard Hauschild; I Teil: Gram- matik, Heidelberg 1958-59, Uhlenbeck, C. C. Kurzgefasstes etymologisches Wdrterbuch der altindischen Sprache , Amsterdam 1898-99. Wackernagel, J. and Debrunner, A. Altindische Grammatik , vols. I— III, Gdttingen 1896-1954. SELECT BIBLIOGRAPHY 401 Walde, A. V ergleichendes Worterbuch dcr indogermanischen Sprachen , heratisgegeben und bearbeitet von J . Pokorny , 3 vols., Berlin and Leipzig 1927-32. Whitney, W. D. A Sanskrit Grammar , 5th ed., Leipzig 1924. INDEX a°, 104, 283 Angara-, 22 a-, 272, 285 anguri-, 157 amsu 180 dngya -, 185 amsa-, 68, 163 dnghri-, 157 amhati-, 169 dcchA, 287 dmhas 73, 96, 223

• i-, 25, 73, 148, 304, 329

anihti-, 73, 182 20 akrstapacyd 213 ajagara-, 126 akravihasta 160, 177 ajalomd-, 211 a£s- f 290 ajdvayah , 218 aksa-, 80 ajina-, 21 a&si, 19, 82, 177, 221, 224, 247 ajmd- f 175 aksnayd, 153 djman 131, 175 agdsti , 164 102, 148, 185 aguru 381 185 agndyi, 18 1, 251

• $}-, 327> 342

agndvisnu, 181, 251 an j ana 136, 138 157, 184, 223, 233, anjasd-, 163 245 ££., 252 anjivd 187 agnimdni- , 147 a/-, 98 agnivdnl 147 atani-, 98 agnimindhd 214 amyaskd 197 agra-, 136, 286 dnu-, 184 agratds, 167 flta's, 277 agrimd-, 176 atasd-, 163 agriyd-, 185 atasi-, 164 agrw-, 225 a#, 285, 286 185 dtithi-, 137, 188, 196 197 atimdtra-, 217 aghdydti, 364 39* 197 ankd-, 76 attar-, 147 ankasd 163 dtyamhas-, 216 ankura-, 149 atrd -, 147, 151 ankuydti , 363 fl/m, 132, 277, 280 ankola 381 tf7n-, 157 angfl, 284 athari-, 192 angand, 379 dtharvan 4 402 INDEX 403 ad-, 112 , 295, 319, 342 adali, 194 addnt m, 145, 226 adds, 276 ff . adand-, 155 dddbhya-, 109 adga-, 94 addhd, 94 addhati 169 admard-, 149 adyatana-, 153 adhamd 176, 275 ddhara-, 129, 275 adharahanu- , 210 adhardt, 283 adhds, 279, 286 adhdstdt, 280 285, 286 adhivasd-, 288 adhydpayati, 358 ddhvan-, 134, 148 adhvara-, 148 adhvarydti , 362 adhvaryu-, 187 an-, 102, 190, 321, 329 ana-, 20, 272, 274 anadvdh -, 221 dnatidbhuta- , 190 anapta-, 167 anarvis-, 128 anala-, 381 anas-, 103, 158 anilayd , 213 248, 285, 286 anudatta-, 115 anudrd 216 annsthii, 279 dntaka -, 197 dntama-, 176, 275 an£ar, 132, 278, 285, 286 dntara-, 150, 275 antara , 285, 286 antarhita 288 85, 102, 279 antra-, 137 and has-, 39 annddd 126, 213 anya-, 27 5 anyatas, 279 any dir a, 132 anydthd , 280 anvavasarga-, 54 a/>- [dp as), 221 a/>a, 68, 102, 286 dpaciti -, 1 14, 169, 288 dpatighni, 194 apamd -, 275 dpara-, 275 aparahnd 114 dpas-, 39, 103, 140, 158, 203 apds-, 160, 203 apasydti , 368 apam ndpdt , 4 a/n, 285, 286 aptur-, 139 dpnavdna 156 dpnas -, 159 apratd, 250 apvi , 19 1 ah jit-, 90 abht, 285, 286 abhidyu-, 216 abhidnth-, 288 abhidrohd - f 125 abhipitvd 171 abhimdtl 170 abhimdUn-, 170 abhiratha-, 29 abkydse, 286 abhrd-, 104, 136 dbhri-, 157 am-, 322 dm a-, 39, 272, 275 dmati-, 169 amativdn-, 142 dmatra 151 amanyuta-, 167 dmavati , 194 404 INDEX amd, 275 amitraddmbhana- , 213 amitrayati , 363 arnind 152 dtnivd, 1 91 araii-, 276 if. amurhi y 281 amrtatvd -, 17 1 dynbhas 136 aya~, 125 ay am, 246 if. ayas-, 102 ayas*. 159, 224, 365 aytt/a-, 57, 124 ayuta-, 262 ardni 157 aratnU, 83, 99, 157 a'ram, 174 ardtiyati, 363 ardtivdn-, 142 an'-, 40, 185, 222, 245, 247 aruna- t 152 aruntuda 160 arusa 163 a'r«s, 1 6a 124, 381 , 385 argAfl-, 25, 75, 78, 124 arc- ; 342 area-, 124 aredtri-, 157 are/-, 177, 179 arcin', 143 arc/s-, 160, 177, 225 drjuna-, 85, 152 arnavd 189 drnas, 159 arnasd 163 ar/V, 289, 364 drtha-, 196, 286 arpayati, 358 drbha-, 22 arbhaka 197 arya- t 40 arya-, 15, 185 aryamdn 4, 143 arvan-, 229 drvant-, 39, 120, 229 arvak, 286 arsas, 158 arsasdna-, 155 arA-, 78 arha-, 124 arhdnd , 157, 191 arhdse, 365 a7am, 85, 174 a/a£«, 379 ava, 279, 284, 285 aua-, 272, 274 avatd- t 98 ava/a-, 21, 39, 98, 153 avatara-y 150 avadhra 147 avapddas, 365 avapanna-y 288 avapdna-, 288 avar, 128, 148, 278 dvara-y 148, 285 avalokayati, 331 as/as, 279, 286 avasd-, 163 avahittha-, 57 av/-, 68, 88, 102, 177, 222-3, 2 45 aviki, 197 avisydtiy 189, 362 avisyd, 189 avisyU', 189 avirat-, 165 dvya-, 87, 185 avyathisyaiy 164 as-, 290, 324, 342 as///-, 91, 261 as'n/-, 157 diman- , 77, in, 135 asmanta-, 130, 153 asmard-y 149 astnari, 135 asr/-, 157 INDEX asm, 157, 178 diva-, 102, 120, 187 , 375 asvatara-, 150 dsvavant -, 120 asvavdla 211 asvavid-, 213 a'sVi, 103 asvaydti, 363 asddha 95 astama-, 176, 262 33, 91, 104, 259-60 as/ra, 25, 97, 156 as-, 94, 96, in, 295, 319, 349 asamkhyeya 262 asand , 157 as/-, 15, 104, 184 asikni , 99, 161 dsita-, 41, 99, 161 as«-, 180 asura-, 5, 24, 40 asuyati , 363 asiiryampasya-, 213 asr/-, 85, 128, 221 asa«, 226 ff. astrd-, 137 as/W, 68, 85, 103, 177, 221, asmd- t 264 ff. asmarfiya-, 269 dsvapnaj 198 a/*a, 284 ahdm, 20, 88, 263 ff. aTtar, 127, 222, 226 a'Ai-, 184 ahordtra 218 ahndya, 283 285-7 akhara 290 akhii-, 290 dgas-,7 6, 158 agniveii 199 agnidhra 200 angirasd 199 angusd-, 163 dcaryajdya, 211 ajarasa-, 216 a;/-, 179 180 adhyd-, 98 dndd-, 22 dtithya 137 dtmdn-, 134, 269 adityd-, 186 adyd-, 200, 371 ddhorana -, 129 ddhra-, 39 anusak „ 278 antra-, 137 dpathi 216 ^as, 158 apt-, 39. 184 dmatara 151 dmoda-, 395 ayasa-, 163 iy«, 160, 178-9, 183-4 183-4 dyus-, 160 ira, 25 irtm, 98 7 drtvijya-, 199 irya-, 15, 40, 390 drseyd 199 avam, 263 au/s, 279, 287 5 s'a, 191 asind-, 152 dsistha 162 asw-, 103, 146, 182, 184 aiumant-, 146 asvina-, 200 as-, 94, 295, 219-20 5 s-, 103, 123 dsandi, 192 dsdnnisu-, 215 dsanvdnt 146 asutivald 149 asya-, 221 406 ahuti-, 169 INDEX 103, in, 290, 291, 319, 324, 349 ijyi, 109 idadadha-, 214 itara-, 275 279 itthdm, 277, 280 z#Aa, 280 z/ya, 1 91 itvara 148-9 270, 276 ff. zrfa, 277, 280 idanTtn , 277, 280 indra-, 5, 28 218 indrasatru 215 indrasend, 21 1 indr old 28, 214 iddhagni-, 215 idhmd~ } 175 Vra/-, 290, 324 ifi/za-, 375 iyant- } 144 irasydti, 362 is-, 326, 330, 331, 342 isdni, 133 isiVa-, 39, 149 isu-, 180 isudhydti, 363 i$td-, 78 istani-, 157 istdpurta 218 172 iWf 33, 45, 7°> 2 77> 281 tksenya 152 185 360 irmd-, 39, 87 3 20 isvard-, 28, 148-9 fsrf/, 279 w, 284 uktd-, 91 ukthd-, 196 ukthasamsin 213 w£s-, 220 uksdn-, 143, 220, 222, 225 uksavehdt 21 1 ukha , 19 1 ugra- } 76-7; 147 ugrdtama 176 ugrdbdhu 215 342 ucdtha-, 196 WCCtf, 232 uccais, 282 uccaistardm, 282 56 w/tf, 284 uttamd-, 176 utiara 150, 285 utsa-, 163 330 wd, 285-6 udaka-, 208, 221 udaktat, 280 udan-, 103, 208, 221 udaniya - t 88, 186 udanyu 187 udapdtra 21 1 udayana 288 udara-, 136 uddtta 1 14 uira-, 118, 120, 121, 147, 204, 206 udvana-, 152 unduru-, 379 279, 286 upakaksd 216 upajivaniya-, 371 upatapat, 164 upatrnya 216 upanadam , 217 upabdd 73, 90, 109 upara-, 148 INDEX 407 upardtat-, 173 upardjam, 217 updri, 128, 286 upasruti 288 upasarga 48 upasthd-, 164 updnasa 163 uptd-, iog ubhri-, 258 wms-, 158 wrw-, 182 urusydti , 133, 189 urunasd 216 urvdrd, 87, 156 ulapa 381 ulupin-, 381 ulukay atii- , 21 1 ulukhala-, 381, 385 'wi&a, 191 ulkabhihata - , 214 usij- t 39, 198 MS-, 328 ww-, 8x, 128, 135, 222 MSflV, 94, 207, 222 usdsa-nakta, 217 ustd-, 168 ustra- f 15 1 usnd-, 15 1 usrd-, 150 78, 91, 95 168, 232 udkar, 25, 102, 127, 222, 232 tlma-, 175 4 r/- f 101 urjasvala-, 149 Mrwfi, 87, 98, 191 289 urdhvd 87, 186, 286 urdhvdnabhas-, 216 urdhvasfind-, 156 urmi-, 87, 185-6 wrmy«-, 185 186 Usman- , 134 r-. 323, 33° rkvan-, 142, 145, 206, 229 fkvant 145 82, 92, hi, 163, 375, 394 , °rksard- t 150 rc-, 123, 221 reuse, 159, 224 rjisd-, 163 rju- t 182 rjuydti, 363 rjrd- t 152 rnavdn-, 142 rtadhaman-, 28 rtayd, 282 rtdvan 142, 227 rtavari, 142, 194 rtuthd , 280 rtupati -, 57 rte', 286 rdh- t 302, 360 fddhi-, 168 rbhuks-, 229 rbhvan-, 227 fbhvas-, 160, 227 rsabhd 198 rsz~, 184 rs#-, 4 rsva-, 186 rhdnt-, 144 eka-, 30, 258, 275 ekataya 263 eka-vartana 28 ekavtrd-, 210 ekddasa , 261 ekonavimsati 261 98, 381 eia-, 167 etarhi, 281 171, 365 etavai, 1 71, 365 etos, 365 edh-, 292 o 408 cdhatu-, 170 edhas 103, 158 cna-, 270, 274 cm, 167 ema-, 175 email -, 175 evd, 283 evdm , 283 dsa-, 124 124 262, 270 okas-, 141 okivds-, 160 57 djas-, 103 ojistha 162 ojmdn-, 134, 207 owl-, 192 170, omdn-, 134-5 ostha-, 21 ausijd-, 199 13, 20, 75, 270, kakud 198 kakubh-, 70, 198 kakubha-, 70 kakuhd-, 70 kdksa-, 68, 76, 92 kanka-, 381 kacu-, 382 kacvi, 382 kacchapa- r 93 kajjala-, 381 katu-, 381, 385 kathina-, 98, 381 kadevara-, 98 kanduyd-, 304 katamd-, 274-5 katard-, 274-5 AaW, 168, 274 katidhd r 280 kathdm, 276, 280 INDEX to/m, 276, 280 kadalt, 379 kadd, 20, 276-7 kadrii 192 kdnistha-, 162 kanlyas 16 1 kanydna , 156 kanydla, 156 kanyd, 77, 193, 195, 227 kapand, 18 1 kapi 184 kapild-, 184 kapisa-, 197 kdprth-, 128 kapdia 167 kapha-, 26 331 kdrana-, 136, 138, 151 karand-, 135, 136, 138, 151 karaniya-, 371 katas- , 158 kardsna-, 153 karira-, 381 kdrtsa-, 163 272-3 karttna-, 138 kdrfidati, 101 karkata-, 75 kartdr-, 139, 188 kartave , 171 kartdvya-y 172, 371 kdrtum , 163 kartr {nt,)> 170 kdrtos , 171 kartr a- , 137 kdrtva 171, 372 karpata-y 57 karpasa 379 karmakard-y 213 karmatha-y 150 kdrman-, 131 karmdnta- , 154 karmdra-y 141, 154 kdrvara-, 136 karsa-, 136 karhi, 274, 281 kalatra 138 kalama-, 387 kalusaydti , 364 kalusdydte , 364 kalydni, 254 kavatnu-, 158 kavdsakhd-, 181 kavi- , 76, 181, 184-5 kaviyati , 363 kavyd 89, 185-6 kascana, 283 kasyapa- t 93 kastam, 284 kastlra 387 kdka-, 381 kakhorda-, 389 kaca-, 381 kdja-, 381 kdncika-, 381 kanjika-, 381 kdnvdyana 199 kdnana-y 382 katana- , 33, 97 kdraniya-, 359 karayitavya-, 359 kdrayituMy 359 kdrayitva t 359 kari-, 184 kdrita-, 359 kdru-y 40 kdrotara-, 150 kdrya-y 359, 371 kdrsivana-y 149 A;aia-, 77, 382 kavyd- , 200 355 kdsara-, 375 Wm, 75, 273 kimpala-, 387 kiyant- y 144, 274 Ai/a, 284 itft-, 184 Hrntf-, 87, 370 INDEX 409 kUala-y 41 kivant 274 Lukkura-y 375 kunjara-y 375 kuta-, 382 98, 382 kutila-y 382 katt-y 382 kuthdra-y 98 kuni-y 98 kuntha-y 98 kunda-y 382, 385 kundala-y 382 Awtas, 274 kutra, 274, 280 kudddla- , 382 kunta-y 387 kuntala-y 382 kup-y 330 kupaya- t 186 kupitd-, 189 kumarakd-y 197 kuntdrika, 197 kurula-y 382 kulattha-y 382 kulampuna-y 214 kulihga-y 379 kuliia-y 98 °kulva-y 76, 83, 187 kuvalaya-y 382 274 kustha-y 57 20, 274, 281 kucity 274 98 kdpa- t 27 , 382 fy-, 79 , no, 178 , 294 , 325 , 332-4, 34i. 354, 358 - 6 o, 365 krcchfd-, 136 , 150 krkadasU-y 194 krnvant-, 368 fcfi-, 48 , 76 , 291 , 294 , 328 , 330 109 krtavant-y 371 410 INDEX krtdkrta -, 218 krdtii-, 41, 170, 223, 230, 245 krte i 286 krdtvdmagha 215 krlnu-, 158 krdnt-, 367 -8 krtrima 176 krand-, 292, 343, 356 kftvan -, 142 hr am-, 329 kftvas , 163 kramelaka-, 387 ArtM, 172 kravis 39, 68, 76-7, 106, 160, ArM, 172 177 krtsnd -, 133 kravya-, 180 kfdhu -, 182 kriyd, 190 krntdtra-, 138 An-, 75-6, 291, 326 kfp-, 105 knd-, 94, 358 krpana 138, 362 knda, 1 91 krpanydti, 362 ■ A rudh-, 330 kfmi-, 75 krorfa-, 57 krsd-, 124 A rosand-, 15 1 Ars-, 302 kland-, 292 Am-, 179 klam-, 76 A rstvald-, 146, 148 A Ionian-, 83, 99, 134 krsnd-, 21, 153 Am, 274 krsnatd, 172, 191 A satrd-, 4, 92, 138, 292 krsnasakuni-, 210 A satrapa-, 388 krsnasarpa -, 55, 210 Asarf-, 292-3 A rsara-, 150 Asan-, 82, 290, 292, 325, 334 A?-, 81, 330, 355 Aso/>- f 92, 192, 221 kip-, 105 ksapd, 191, 193 kecuka 382 ksam-, 2gi, 295 kenikd, 387 ksam-, 82, 220, 234 keta 124 ksam d, 1 91 ketaka 382 ksdmya-, 185 kendra-, 387 Aw-, 82, 95 kemuka-, 382 A sdmd-, 95 kevuka 382 ksdy-, 95 Aesa-, 99 A51-, 82, 92, 323, 334, 354, 394 kesava-, 99, 187 As/fa'-, 168 kesavant 145 Asi^-, 100 A esavardhana-, 213 ksipanu-, 148 A esasmasru-, 218 ksiprd 147-8 kesara-, 99 kstnd-, 95, 370 kotara-, 382 ksird-, 136 Aofo'-, 98, 262 ksudrd-, 80, 147 kotha- , 57 ksudh-, 92, 292, 295 kona 382 ksumdnt -, 100 AoZa-, 57 ksumd, 176 INDEX ksurd 149, 324 ksepnu 148 ksetra-, 92, 137 kse'ma ~, 175 ksont-, 192 ksnu-, 289, 324 khanj -, 81 khan-, 290 khdnitra-, 188 kharvd-, 187 khala 383, 385 khaltna 387 khaln, 284 khd-, 72, 290 khdrt, 192 khuddaka 57 khola-, 389 ^y«-. 335 375 ganja-, 389 ganda-, 383 gafri-, no, 168, 370 104, 169, 245, 252 gatvd, 172 gatal, 172 gdntum , 170 gdbhasti 164 gam-, 75, 79, 93, no, 290, 329 334. 339. 356, 358-60 gdmadhyai, 198 gdmana no gamitd-, 189 gdmbhan 128 gatnbhdra-, 128, 136 gay a-, 124, 290 gam^Aa-, 162 garutmani-, 166 gardabhd 198 gdrmut 166 garA- r 91 gdvya-, 185 gavydya-, 186 4x1 gdhvara 136 ga'-, 290, 291, 330, 331, 334 ga^H-, 39, 170 gathd-, 196 gm-, 21, 87, 184 girnd-, 370 gaia-, 98, 383 361 75 guru-, 46, 75, 86, 116, 182 35 ^ giihya-, 185, 371 342 grmsdni, 133, 361 grdhnu-, 148, 158 grdhrd 148 gfbhi-, 184 gr-, 22, 75, 87, 290, 358 go-, 75, 104, 117, 181, 251 goghnd 126, 213 gomant-, 223 gomin 144 gold-, 389 govinda 214 gospada 212 gawH, 253 39.75-6, 205 gnaspati 212 gmdnt-, 367 grath-, 292 grapsa 163 grabh-, 70, 326, 340 gras-, 290 graft-, 70 grabhd 126 gramavdsa 217 grdmyd 185 grdyan-, 15, 75, 134 grdhi-, 200 graa, 20, 191 gnsmd 175 graivya 186 331 giant-, 157 fKDEX 412 gha, 284 ghand-, 124 gharmd-, 75, 78, 175 ghas-, 109 ghasmard 149 ghdtuka-, 197 ghclsd-, 109 ghuka-, 383 ghrnd‘, 151 gkrni-, 157 gArta-, 15 ghfsu-, 187 gA/sfli-, 187 ghotaka-, 375, 398 ghndnt 144 ghramsa 162 329 ca, 77, 104, 113, 284 cafervas-, 161, 224, 369 cakrd~ f 76-7, 83 caferf, 192 cakrdnd-, 370 cafes-, 92, 320 cdksase , 159 cdk$us- t 161 cdtasras , 14 1, 259 caturthd 196, 262 caturdasa, 260 caturdhd, 280 catustaya-, 263 cafe, 263 catvara-, 136, 141, 259 catvdras , 77, 136, 141, 259 catv&ri , 238, 259 catvdrimsd 263 catvdrimsdt-, 261 catias, 39 canasydti , 362 candana-, 383 candrdmas 210 capeta, 383 cam-, 329 camasa-t 163 camasddhvaryu- , 217 camti-, 192, 254 cay ana 138 car-, 325, 328 cardni-, 157 cdrase, 159, 224 carisnu 158 earn-, 77 carman 13 1 2 ca/-, 355 earn-, 204 carutara 151 a-, 191, 322, 343 cikitvit, 166, 279 cikitsa 361 cikkana 383 ciif-, 291, 293, 297, 320, 331, 343. 358 c#a, 172 179 cirti-, 232 cintdyamdna- , 369 cintaydna 156 ezra-, 282-3 cumb-, 383 car-, 331 ctW«, 3S3 cdkitdna 369 corfa-, 124 coditar 190 caurabhaya-, 21 1 358 cyautnd - 39, 153 cAaj 5 a, 57 cfearf-, 331 chardts, 160 cfea-, 292, 328, 331 chdga-, 26-7 chayd } 93, 191 cAzi-, 93, 292, 321 chidura-, 149 chinnd 370 chinndpaksa 215 INDEX 4^3 jaks-, 109, 190, 321 jagatkartar 55 j agar a 389 jagmust , 194 jagdha-, 95 jdjjhat-, 95, 109 jajndnd 370 jathara-, 97 jada-, 98 ydfo/, 78, 178 192 jatilkd , 197 jatru-, 178, 183 j™-, 79, 321, 344, 354 janamejayd- , 213 janas , 39 39. 247, 363 janiya 88, 186 janitavya-, 371 janiydti, 363 jambdla 379 jamb ha-, 73 jdmbhya-, 185 125 jay us 1 61 jdyya-, 185 jaranti, 157, 191 jar ant-, 73, 84 jar as-, 159 191 jarinidn-, 134 jar hi-, 98 jaldsa-, 395 jdlpi-, 179 java-, 124 jdvana 152 javistha 162 jdsu-, 180 jdsuri 157 jdsvan 142 305, 327, 355 jagrvi 187 /ata, 170 janasruteyd-, 199 yam*, 73, 178-9, 242 jimarya-, 129 jdmdtar-, 140 jdmt-, 184 jdmitra 387 y«>^ 7 , 191 jay dp ati, 218 jdyii-, 182 y<‘, 79 . no, I 9 °. 2 9 °> 301 1 324. 326, 343, 354 jig was-, 161 jigtsi, 207 jighatsu-, 96 jtghdmsanl 145 71^-, no jitvara-, 148 jinv-, 290, 324 jim-, 380, 398 jisnu-, 158 jihmd 175 jihvd, 191-2, 195 ymii-, 45 fiv-, 290, 327, 358 jtvd-, 77, 103, 186, 324 jivapitrka- , 216 jivdputra 215 jivdse, 159 jivatu 170 juvas-, 158 y««-i 73, 80,-103, 328 jiisti-, 168, 232 juhvdna 369 /«-, 291 y«^d", 291 y«tf-, 169 jurni 157 jf-, 301-2 jetva-, 1 71, 372 jenya-, 39 jeya-, 185 jaitra-, 151, 199 7’asa-, 74, 80, 125 jostdr 74 johittra 15 1 414 jna-, 340, 341, 343 jndti-, 169 jyd-, 291 jya, 77 jyani-, 157 jyesthdlati 173 jyrfAi 279 jydtis-, 160, 225 jyoiiskft -, 166, 213 jrayasand 155 /f&a, 99 damara-, 99 rff-, 98 dhakkd , 99 /a-, 19, 269 ff. tdku-, 186 takmdn-, 134-5 /a&s-, 72, 80, 92, 94, 329 tdksan-, 82-3, 241 331 /a/a-, 167 iatanusti-, 164 /a/as, 167, 279 /a'/i, 168-9 /a/m, 280 tdthd, 280 tadd, 20, 270, 280 tadantm , 280 iaddhita 48 325 , 34 I “2 /a'«-, 127, 193 tana-, 127 tanaya-, 186 tdnus-, 192* tanu-, 192-3, 204, 225, 253 ff., 269 tanilsubhra-, 2x4 tantu-, 170, 207 tantri -, 192 land-, 391 tandri-, 157, 192 tdnyati , 81 INDEX tanyatu-, 189 tap-, 330 tapani-, 192 tdpas-, 158 tapasvin -, 144 ldpus~, 160-1 tapyatu 170 tamasd 163 tdmisrd, 138, 156, 191 /aV-, 140 /am-, 125 tdrana-, 138 tardni-, 157 lardddvesa -, 216 tar as-, 160 /an, 192 tarty as-, 162 tansdni, 162 tdruna 152 tarunayati, 364 tarundydte, 364 tarutar-, 190 tdrutra 151 tarns-, 160 tarns a-, 190 tamsydti, 363 tarj~, 292 tdrdman-, 132 /aVAt, 281 talina-, 383 talpeiaya-, 213 lavas-, 159-60 tavisd-, 163 tavisi, 164, 192, 363 tavistyate, 363 tdvyas-, 16 1 230, tdsara 136 /as/Aifas-, 161 tddaka -, 383 tamarasa-, 383 tdmbula 380 tdyu-, 81 /am-, 125 /am, 81 INDEX 4 J 5 tar ana- # 138 tdla{ka-), 383 talu, 178 tdvaka 269 tigitd-, 167 tigmd-, 76, 81, 175 Si, 356, 361 iitau-, 39, 178 tiiiksu 361 87, 159, 279, 285, 287 tilamisra-, 214 tisrds, 81, 141, *203, 259 tiksnd 153 tlrnd-, 87, in tirtvd, 172 tirtha-, 45, 196 iwrd-> 149 fr#, 284 tuccha 93 tucchyd-, 21 W-, 81 , 301, 329 tuddnt 145, 368 tubari, 383 tumra 148 330 /wrrf-, 125 turdna-, 362 turanydti, 362 tarty a-, 262 turya-, 262 tuvara 383 tuvisvdni 157 /us-, 330 321 turni-, 157 turv-, 190, 290, 329 turvdni-, 157 turvdne, 131, 224 ttirvaydna 166 turvi-, 184 /w/a-, 383, 385 tusnim, 21 tfna-, 25, hi, 328 trnodakd-, 218 trtiya-, 259 Mtyasavand-, 210 hi, 328 trdild-, 149 typ-, 99 , 324. 358 trpdnt-, 367 182 trstd-, 167 trsndj 198 tfsnd, 156, 1 91 95 81, 87, 106, in, 178, 190, 290, 291, 302, 329, 360 tejas 158 tejasvin-, 144 tenivds 369 tokd-, 39 tokman-, 39, 175 tosas-, 160 £ya-, 210, 272-3 tydjas 160 tnipu, 178 trayd-, 263 trdyastrimsat 261 trayodasa , 260 tow-, hi, 290, 293, 329, 358 trasddasyu 216 tra- t 291, 331 /«-, 68, 80, 259 trimsd 263 trimsdt -, 261 tridhdy 263, 280 tridhatUy 280 trivatsd-, 216 trivft-y 221 Ms, 263 tretd , 201 traitand- f 201 210, 272, 274 ivaks-, 39, 92 tvadiya 269 264 /Dflr-, 358 tvarand- t 151 INDEX 416 tvarita 46 tvdstar -, 100 to*?-, 391 tvisi-, 179 ddmstra 147, 15 1 damsdna -, 138 daksa~, 163 daksas-, 159 daks ay y a- , 372 ddkstna 20, 80, 92, 152, 275, 285 daksinatds, 167, 279 daksinatra , 280 daksini, 282 daksinagni 210 daksu-, 164 daksus-, 161 dagdhd -, 78, 90, 91 dattd-, 106 datra-, 138 2i, 177, 221 ddnt-, 90, 1 12 71, 95, 109, 324, 360 dam-, hi ddma-, 39, 68, 103, 125 damanydti, 362 ddmiya-, 82, 186 191 darad- } 198 daridrati , 321 dartnu 158 darma 175 darmdn-, 143, 175 darvi-, 187 darsata 112, 161 darsana 138 da/a-, 83 <£asa, 68, 74, 173, 258, 260 dasdt-, 165, 168, 261 dasatt-, 165, 168 ddsataya 186 dasama-y 176, 262 dasasydti , 362 dasma 39, 175 dasyu-, 37, 41, 187 dasra-, 39 71, 78, 329, 360 20, 89, 90, 106-7, 302, 322, 330 - 1 , 334, 343, 354, 358, 360 ddksayana-, 152, 199 ddtdr-, hi, 121 ddtave y 17 1 dativdra 216 da/w, 170 203, 254 dddhrvi-, 187 dana-, 68, 128 danavd 200 danu, 39, 178 danupinva-, 213 dantd-, 87, 21 1 daman-, 134-5, I 43^ 20 7 damane , 131, 224, 305 dava-, 125 davdne, 131, 159, 226, 365 °ddvari, 142 178-9, 200, 222, 227, 247-8 daruna -, 152 darsanika- t in disuri- t 157 ddsvds-y 39, 160, 343 dasd-y 25, 41 ddsyahputra-y 212 dx/a-, 189 dina-y 292 33 div-y 330 divasd-y 103 dfva, 282 divdtara- f 150 divU{mant-)y 166 divvy oni-y 215 divira-y 388 divodasa-y 212 divyd-y 185 INDEX 417 dis-, 73, 95, 220-1, 329, 341, 355 dis-, 193 dtsd, 193 dih 70 dt- t 98 dtdivi 187 321 dlndra 387 dip-, 292 dirgha-, 68, 76, 87, 198 dirghatd , 172 dirghdsmasru- , 215 ducchund , 93 ducchunaydte , 363 duradabhnd 214 durond-, 152 duvasana 153 duvas-, 158, 160 duvasyu 187 rfws-, 321, 358 70, 78, 95, 320, 332, 341 duhddhyai , 365 duhdnt-, 368 duhand-, 155, 369 duhitdr 88, 140, 243 duddbha-, 94, 101 diltd-, 161 duty a- 1 180 duraka 197 duramgamd- , 126 durvd, 1 91 355 drmh-, 294, 357 drmhitdr 140 4 *-. 73, hi, 178, 341, 343, 354 drsati-, 169 dr s dye, 179 drsand 369 <rsV, 123, 365 drienya 372 drsad-, 198 drs/a-, no, 168 rffsjtoi, 172 371 devd-, 5 ,' 103, 230, 255 devakilbisd- , 21 1 devatd, 172, 194 devatdt{i)- } 173 devdtta-, 214 devatra, 132, 280 devatvd 17 1 devaydjana-, 213 devaydti , 187, 363 devayu-, 37, 187 devdr-, 103, 141 devavanda 213 devasds , 281 devast ut-, 166 devdkiti 212 194-5, 253 ff. destrd-, 137 desna-, 153 70/163, 192-3, 225, 253 daivya 199 191, 282 tio's, 133, 158, 221 dohada 56 dohala 56 dohas-, 158 dohdse , 159 daurkrda-, 56 dyuksd 162 292, 331, 356-7 dyut-, 123 dyumdnt 146 dyumnd-, 146, 152 rfyo-, 104, 117, 181, 230, 234, 250 dyotana 152 dyotand , 157 dyotani 157 drapsd-, 39, 163 dram-, 290-1 dramma-, 387 dravd-, 125 dravdt f 279 dravard- t 148 INDEX 418 drdvina-, 138 drdvinas 159 dravitnu 158, 188 drdvya 180 drastdr 188 drahydt, 279 dra-, 290-1 drdghmdn 175 dfu-, 290, 291, 337, 356 drugdhd-, 78 drughand 21 1 dr up add - 1 21 1 druh-, 78 , 292 druh-, 123 druhvan-, 142, 145 dvaya-, 263 dvddasa , 260-1 dvddasa-, 263 dvar-, 221 rfw-, 258 dvitd, 194 dvitiya 258, 262 dvipad 258 dvisas , 281 96, 319 rfpxs-, 93, 119, 123 263 dvesas-, 119 dvesya 371 19, 258 dhdksu 164 dhdna - f 138 dhanajit 213 dhananjayd-, 213 dhdnasdti-, 212 dhanin-, 143 dhdnu , 180, 192 dhanutdr-, 190 dhdnus 160, 192, 227-8 dhanu 192 dhdnvan 130, 134, 160, 192, 223, 227-8 dham- y 22 dharuna-, 152 dharnasi-, 164 dharni 157 dhartari , 132, 365 dhdrtra-, 136 dhdrma - r 175 dharman - f 142 69, 70, 90, 106, no, 290 322, 333-4, 343, 360 dhatu-, 170 dhdna , 22 dhdman 135 dhdyase, 159, 224 dhdra, 156 dhdru-, 157 dhdrmika 199 dAiv-,'294, 329, 356 dhdsi-, 164 dhik , 283 dhiyamjinvd-, 213 dhisanydtiy 362 dAf-, 221, 230, 231, 254 dhiti-, 169 dhuni- t 180 290, 302, 324, 351 dhuti-, 169 dhuma-y 69, 103, 175 dhumrd-, 149 dhurv 290 dhftrvaney 13 1 dhusara-y 150 290, 341, 354 dhfti-y 168 iArs-, 80 , 324 dhrsdj-, 198 dhfsti-y 169 dhrsnu - r 158 dhrsnuyd, 283 dhend, 157 dhenu-y 158, 245, 252 dhenukdy 197 dhenumatl , 194 dhenustarfy 21 1 dhaurita-y 129 INDEX 419 dhya-, 339 narmatha 150 dkrdji 179 nary a-, 185 dhru- , 292 nava- f n., 81 dhruvd-, 186 ndva-, adj,, 68, 126, 261, 283 dhrtlvi 187 wai>a, nuin., 260 dhvani 179, 180 navati-, 261 dhvards -, 160 navaddvd-, 210 dhvas-, 290 ndvanUa -, 391 dhvasdni -, 148 navamd 260, 262 dhvasmdn 135 ndvd, 86, 194 dhvasrd-, 148 ndvisti-, 164 ndvtyas 161, 194 wa, 20, 283 navy as 161 nagnd-, 151 nas-, 290, 334, 358 waTtf-, 39, 75, 165, 168 nasvara 148 naktdn-, 132, 227 was-, 329 ndktam, 282 was-, 102, 221 naktaya , 282 nasvdnt-, 146 ndkii-, 168 waA-, 70, 330 nakra-, 383 ndhus -, 39, 139, 161 naks-, 290 375 ndksatra -, 138, 196 nadyd 200 nakhd-, 196 nand, 284 wa/a-, 56, 98 ndndnam, 284 nada-, 98 ndndi , 192 nadd-, 98, 124 napitd-, 167-8 nadi-, 184 ndbhi-, ii, 177, 179, 201 wail, 192 ndman 19, 24, 48, 68, 223, naddhd-, 70 226, 232 ndntva-, 372 nayd-, 125 napat, 103, 140, 165 nay aka-, no naptdr-, 140 ndsatya -, 5, 28 nabhanu 152, 192 wz, 285-6 ndbhas 68-9, 158 nims-, 321 nabhasa 163 nikasd , 285 ndbhya-, 177, 180 niktd-, 78, 168 340, 358 wzAs-, 290 nan as and-, 155 w/- f 78, 322, 355 namasyati, 114, 362 nityam , 282 ndmasvant-, 228 wii-, 329 naV-, 141, 243-4 nidaghd-, 78 nansid, 164 nidht-, 288 narmd-, 175 nind-, 329 ndrtnan 175 nindd, 191 420 INDEX nip at a-, 48 paksati-, 169 nimnd -, 152 pdksas 159 niyut 165 pdksman 13 1 nirdyana 288 paksmald 149 niravasita-, 54 pakstn 143 ntrasta 92 pankit 260, 261 nirukta 41 pac- } 77, 294, 329, 342, 354 nirgundt, 383 pacchas, 93 nirjhara-, 95 pajra-, 136 nirhasta 137 panca, 68, 77, 259, 260 nisadvard-, 148 pancathd-, 262 ms, 286 pancamd-, 262 nihaka, 81 pahedsa-, 263 m-, too , no, 338 pancasat-, 261 meats, 282 panjar a-, 136 »hW-. 33 , 45 , 94 pata-, 97 mta'-, no patala -, 97 mra-, 383 patola-, 383 nilalohitd-, 219 pana-, 98 nllotpald ~, 210 panda-, 384 nihiira-, 81 panda, 98 283 pandita-, 98 mi, 283 pat-, 68, 188, 329, 336, 342, nutana -, 153, 283 357 nutna-, 153, 283 pataga 198 nundm , 283 patangd 198 179 pdtatra 138, 196 nrtu-, 182 pdtana-, 138 nr pair-, 2 13 patara-, 148 nrmnd 152 paturu 158 nrvdnt-, 146 I02 > i6 9 > 22 3 > 2 47 > 2 49 netrd-, 137, 391 patti-, 165, 169 nedistha-, 94, 162 paint , 99, 247 39, 272, 275 pdtman-, 131 nesdni, 133, 365 pdtvan -, 130, 142 nestar-, 137 pdtsala 138 nestrd -, 137 path-, 72, 229 nairhastd 137 pad-, 90, 103, 109, j 12 104, 246-7 padd-, 127, 174 na« p 268 paddti-, 165, 169 padvant-, 146 paktha -, 262 pan-, 39, 190 pakvd 186, 370 panayiyya -, 190 paksa~, 163 panayya-, 372 INDEX 421 panitdr 190 pdnisti 164, 190 p any as-, 161 pay as-, 21 paramd-, 176 parasu-, 83 paras, 279, 286 pardstdt, 286 285-6 pardri , 42 84, 279, 285-6 paritas, 167 paripanthin - , 216 parisadvald- , 149 paristoma-, 387 parihastd 216 par in as-, 159 parinasd-, 163 42, 119, 234 pdrus 40, 160, 227 parusd-, 40, 163 pdreta 288 parnadhvas- , 96 84, 329 paryana 389 pdrvata-, 153 pdrvan-, 40, 130, 160, 227 pdrsu-, 180 parsdni, 133, 157 palasti-, 164 palali, 192 paldva 83 pdlikni, 99, 167 palitd-, 83, 99, 164, 167, 391 384 palvald-, 83, 148 pdvirava-, 189 pavtru-, 187 81, 330 ^>asw, 77, 178, 222, 225, 245-6 pasu-, 81, 102, 179, 222, 245-6 paiumdnl 147, 220, 224, 228 pasca, 232 pascal , 233, 283, 286 pdsas 158 pasty a, 39 72, 85, 103, hi, 329 pamsu 22 pamsurd 149 pdmsula 149 pdjas 40, 158-9 pancajanya 186 patrahasta 215 pdda 122 pdparddhi 57 papmdn-, 134-5 pdman-, 134-5 pdmand-, 135, 149 pamara 135, 149 pdyu- 40, 183 pirsni -, 84, 157 palitya-, 199 pdsdnd-, 98 pi-, 289, 324 pingala 391 picu, 379 pinda 98 . pitar-, 105, 106, hi, 140, 203, 220, 222, 230, 234, 243 pitdputrau , 218 39, 179, 222, 226, 246 pitrya-, 185 pinv-, 289, 324 pipilikdputa- , 384 pisuna-, 152, 291 />*>-, 80, 292, 327, 341 ptd-, 94, 292 pita-, in 172 ply-, 291 piyatnu-, 158 ptyii-, 182 ptytisa-, 133, 163 pilu-, 389 ptvan 103, 149, 187 ptvara-, 149 ptvan , 142 ptvas-, 1 18 INDEX 422 pwasd-, 163 pums-, 228 punkha-, 384 pnta- y 97 p u ny aLaksnnka- , 216 puttika t 384 putYa-> 1 51 putrakd-> 197 putrdvant- , 145 putrlydti , 363 punar , 278 punarmanyd-, 213 punndga 384 83, 87 purandard-, 213 purds t 87, 159, 279, 280, 286 purdsiat , 280, 286 194, 286 pur and - , 155 purtsa-, 163 puru-, 87, 182, 207, 230, 245, 250-2 purusa 163 purusdta, 172 purusairdy 132, 280 purusamrgd-, 2x0 purusarajd-y 217 pulaka-, 164 pulasti- t 164 133 pusti-y 169 pustaka-y 389 105, 291, 339 pUtd-y 291 pUU-y 103, l69 puru-y 163 purusa- } 163 purnd-y 87, in, 151 291 purndmdsa-y 210 purva-y 18 7, 275 pida-> 391 pusana-y 153 pusdn-y 133, 143 pusary d-y 133 prksd-y 162 prtandydtiy 363 pfthavdna-y 156 prthu-y 72, 104, 182 prthuka-y 71 prthvi, 194, 254 pfdaku-y 197 pfsm-y 157 pfsat- s 144, 164, 167 pr sat d-y 167 pf-y 83, hi, 120, 291, 301-2, 322, 326, 358 petva-y 1 71 polar- y 137 potrd-y 137 paurukutsi-y 199 pauruseya-, 199 pyd-, 291 pyuksna-y 153 1 1 4, 279, 286 prach- f 94 , 104, 289, 330 prajipatisfsta-y 214 pratard-y 150 ^ra^*, 177, 279, 285, 287 pratidivan- } 141 pratidosdm, 282 pratinisaniy 217 pratiram , 365 pratisara-, 198 pratnd-y 153 pratndthdy 280 pratyagniy 217 pratydnc-, 221 prath- t 72, 292 prathamd- , 262 pratkimdn-y 131, 135, 188, 227 prdthtyas-y 161 pradaksinity 16 7, 279 prapitvd-, 171 prapyasd-, 163 prabhangin-y 288 prayatadaksina-, 215 pray as-, 158 prayuta-y 262 INDEX pravand-, 152 prasnd-, 91, 289, 330 p™~> 344 prdkrta -, 1 praktdt , 280 prajdpatya-, 186 prdtdr, 128, 132, 234, 278 pradus , 279, 287 prdyas, 279 prdsii-, 194 priyd-, 125 priyadhama -, 175 priydyate , 304 p™-, 35 s pru-, 83, 285, 328 prus-, 22 preni-, 157 premdn -, 134 plthan-, 81, 83, 134 plu-, 83, 99, 328 pluti-, 169 plusi-, 84, 164, 184 291 pharvara - t 136 phalgu 40, 192 phala 71, 81 phena-, 20, 71, 81, 393 baka-, 384 badhira-, 149 389 bandh-, 70 bdndhu -, 70, 180 bandhutd , 172 babhuvas -, 161 babhru 192, 391 babhrusa-, 197 bar hand, 157 barhis-, 21, 160 384, 385 baladavan-, 213 195, 225, 253 bahirdhd , 280 423 20, 102, 131, 279, 280, 286 bahu-, 145, 182 bahudha, 280 bahuvrihi-, 215 204 190, 329, 367 191, 235 6as/)a-, 395 bdhu~, 70, 103, 116, 180 bahya -, 102 biddla -, 375 bibhrat -, 368 384, 385 Wfoa-, 384, 385 391, 395 buddha-, 90 buddhi-, 90 20, 21, 104, 328, 335 budhd-, 125 budhnd 70 bubudhana -, 155 bull- ,391 brmh-, 22 brhdnt-, 144, 368 bfhaspati-, 209, 212 bddha-, 125 bradhnd -, 22 brahman -, 119, 121, 131, 136, 139, 206 brahman-, 15, 120, 121, 136, 142, 206, 223, 230 brahmabhuya-, 180 320-1 bhakta 78, 92 bhakti-, 169 bkaks-, 92, 290 bhagna-, 370 bhahga -, 22 bhahgurd-, 149 33, 78, 92 bhanj-, 100 bhata-, 98 INDEX 424 bhatta-, 56 bhaUdra(ka-) , 56, 98 bhadrapdpa -, 218 bhayd-, 174 bhdra 21 bharatd 167 bharddvaja- , 37, 216 bharant 145 bhdrantf, 203 bharisd-, 163 bhdrman 131 bharv-, 290 bhala, 284 bhalluka 375 bhdvana-, 118 bhdvant-, 223, 269, 368-9 bhdvamdna-, 369 Mas-, 98 Mas-, 291 bhasad 198 Ma-, 70 bhajana -, 138 bhdnu -, 158 bhama-, 175 bhdra-, 125, 207 bharata - (-f), 1 bhdrvara -, 149 bhdrman , 200 bhavyd 200, 371 Mas-, 98 bhdsura-, 149 bhasvard -, 148 M*i-, 327, 359 bhidurd 149 bhidelima-, 176 bhinnd 15 1, 370 bhiyas 159, 202, 222, 225 bhiydsana -, 155 bhisaj-, 198 bhisajydti, 363 Mi*-, 22, 290, 297, 302, 358 bhimd 174 bhirti-, 158 M*sa, 164, 191 bhugnd-, 151, 370 M«/-, 100, 295, 327 bhuji- r 179 bhujisya-, 164 bhujmdn 143 bhujyii-, 187 tor-, 291 bhuranydti, 129, 362 bhurij -, 198 bhurvdni -, 149, 157 bhuvana 138 bhuvanti-, 129 bhiivar, 128 M*7-, 30, 333, 342, 365 Mfi-, 221, 231, 254 bhutvi , 172 bhumdn-, 134-5, 22 7 bhUmi 176, 185 bhumyd- f 185 bhtiyas 161 bhiiyasi, 204 bhuri-, 157 bhfirja 73, 87 M«$-, 331 Mr-, 15, 102, 1 16, 322, 328, 338. 356 bhfgavana 156 bhrta-, 167 bhrti-, 169 bhdga-, 124 Mo^a-, 125 bhrdtar-, 103 maks- f 25 mdksa, 25 maksika , 25 maksd, 80, 92, 196, 279 makhd -, 196 maghd-, 391 maghdvan-, 142, 222, 229 manku 384 ma;)’-, 94, 329 majjdn 21, 134 raaw{-, 98, 184, 391 INDEX 425 manignvd 214 mdtasna-, 138 mail-, 169, 207, 230, 233, 246 matka-, 269 matsard-, 133, 150 math-, 229, 326, 363 mad-, 331, 357 madird-, 149 madiya-, 269 maderaghu-, 214 madgu-, 94 madvan-, 142 mddhu, 24, 69, 88, 102, 177-8, 184, 194, 222-3, 22 7> 2 3 2 > 245 ff. mddhujikva 215 madhurd-, 149 tnadhula -, 149 madhuhastya -, 216 madhya -, 102, 174, 177, 180, 286 madhyamd-, 174 man-, 234, 360, 361 mdnas 158 manasd 163 manasydti, 362 man&vt, 181, 251 mantsd, 164, 191 mdnu-, 180, 201 manus-, 139, 161 manojighra- , 214 manoyuj 213 mdntu 1 70-1 mantrdyate, 364 mandaydtsakha-, 216 mandasdnd-, 133, 155 mdnman-, 131 manyu 37, 187 moywra-, 384 mayuraroman- , 215 marakata-, 387 marica-, 380, 398 ntdrici 166, 197 marut 29, 166 marditar 190 mdrya 40 maryakd 197 marsa-, 80 mallikd, 384 malvd-, 186 mast-, 384 mastiska-, 164, 197 mdstu, 170 mdhar , 128-9 mahdnt 144 mdhas-, 160 mdha- t 88, 195, 229 mahagramd-, 210 mahdvird 210 mahddkand-, 210 mahi, 229 mahitvand-, 17 1 mahimdn 131, 134 mahisd 163, 375 wa-, 68, 322-3, 358 m 4 , 102, 299 mamsiydti , 363 ntddht t 389 mdtanga-, 375, 379 matdr-, 19, 68, 103, 116, 140, 203, 243 matarisvan 41 matftamd , 176 mdira, 156 mana-, 125 mdnava 199, 200-1 manasd-, 163 mdmaka-, 199, 269 mdmakina-, 269 mdy&, 19 1 mdruia-, 199 mar jar a-, 375 mdrdlka 199 mala, 384 mas-, 94, 103 mahina-, 46

• »*-, 292, 324

miks-, 290 INDEX 426 mil, 124, 165 mitrd 4 , 28, 15 1 mitradha , 280 mitravarund , 217 292 mithati 169 mithds f 159, 279 mithu 279 mithund 152 miyedka 40, 94 misrd-, 21, 290 miA-, 95 mihira 389 Mi-, 291 midhd 40, 94, 391 midhustama-, 176 mtdhvds-, 160, 343 mina-, 385 mimdmsd, 361 mukuta- K 385 mukula 385 mukta , 385 mukhatds , 387 mwo, 22, 330, 354 mwrfm ,-388 muni-, 184 muraja 385 murungt, 385 363 mustvdn 142 muhu, 130, 279 wwAwr, 130, 153, 279 muhurtd-, 130, 253 murkhd-, 391 murdhdtt-, 134, 222, 226, 234 ?nds~ t 68, 80 «r-i 354, 360 mrks- t 290 mr gay ate, 364 wf, 78, 290, 320, 342, 357 wri-, 190, 331 mrdayaku-, 190 mrdika 40, 91, 94, 197 mrtd-, 104 mrtyu-, 187 mrtyubdndhu 2 1 1 mrtsna-, 153 wrd-, 291 mrdu-, 104 mrdh-, 292 mfdhas - f 158 wrs* ( 329 rars-, 80 mrsti, 194 mrstd-, 78 wf-, 291, 326 meghd-, 69, 76 medhi-, 33, 94 medhird-, 149 me/a, 387 nwcika 389 mnd- f 291 mn'/-, 291 mrddlyas-, 161 inland 151 mine-, 292 mlecchd 1 _ya-, 272, 274 jyaAr/, 68, 127, 164, 220, 221, 226 yaksd 163 ydksma - f 175 ydksman 175 va;-, 34, 78, 102, 109, 294 yajatd-, 112, 167 ydjatra 151 yajdtha 196 ydjamdna - t 155 yajas-, 160 ydjus-, 160 yajna-, 4, 102, 151 yajnddhtra - , 214 yajyu-, 37, 187 ydjvari, 142 yatamd-, 275 yatard 274-5 168, 274 INDEX 427 ydtra , 274 ydtha , 274 274, 280 yantur-, 139 329, 330, 343 yamd- t 4 ydrhi, 281 ydvamant-, 147 yaWsa-, 163 ydsas-, 120, 160, 206 yaids-, 120, 160, 206 yastukdma 367 jyas-, 329 yahu~, 40, 120, 186 yahvd 40, 186 yahvdnt -, 120 ya- t 291, 330, 360 yaga-, 79 292 ydcitdr 188 jyatana, 147 yi<ar-, 140, 147, 203 yavant 274 °ydvari, 142 jyw-, 292, 320, 322 yuktd-, 78, 89, 91 yukti- f 169 jywga-, 68, 89, 127, 173-4. 232 yugapad , 278 yugmd-, 173, 232 yuj-,78, 89, 91, 104, 292, 327 yuiijdnt 222 yuhjdna 369 yudh-, 330, 359 yudhaye , 179, 365 yudhenya 372 yudhmd- t 175 yudhvan 142 yuvatl ”, 229 yuvan-, 68, 141, 222, 229 yuvasa 197 yuvaku, 197, 227 yuvadatta , 267 yuvdtn, 264 yusmdka 267 yusmddatta 267 yuthd -, 196, 198 yUpa~ t 198 yupaddru, 21 1 ywyam, 264 68, 221 .yoga-, 124 yodhd 125 yodhya 185 152, 185 141, 166 15, 40, 158 rdmhi -, 179 raks-, 82, 290 rdksas-, 92, 160 raksds 160, 224, 230 raghit 182 raghuyd , 283 rajatd-, 73, 167 rajatapdtra 210 rajas-, 40 rajasd 163 rd;V-, 179 rana- } 138 rant;#-, 186 rdnvan 186 ra/i-, 168 rdtna-, 99, 138 ratnadhatama 176 ra*Aa-, 71, 72, 84, 129, 196, 393 rathary dti f 362 rathaspdti -, 256 rdthiya-, 87, 186 rathird-, 149 194-5, 205, 253 radavasu 216 rafiA-, 360 rdbhyas-, 161 raw-, 336 rayi-, 179, 222, 226, 246 raiand , 157 rasmdn 134, 175 INDEX 428 rasntt 176 rasi, 1 91 rdhas , 281, 283 r&-, 323 r£j-, 15, 68, 92, 103, 123 rdjan-, 15, 68, 84, 103, 123, 128, I4I-2, 200, 21 1, 223, 234 rajaputrd- , 21 1 , 215 r ajar si-, 21 1 rd jay die, 363 rajjva 187 rajni, 205, 254 rajya-, 15, 180, 185 rati-, 109, 170 rdtin 170 157 rdyaspdsa-, 212 rastrd-, 137 rd$tradip$u~, 213 rasabhd 198 nAfro, 75 rikthd -, 196 no, 75, 80, 103, 327, 335 ffl-, 165 rip-, 83, 292 rip’, 39 1 ripu-, 182 riprd-, 13ft n‘A-, 83, 95 n-, 328 no, 292 rukmd-, 175, 391 ruksd-, 162 n*o, 291 rwc<Z-, 125 ruci-, 177 > J 79 > i8 9 rucira- t 149 m;-, 292, 329 ru/a-, 124 292-3, 321, 340 rudrd-, 57 ruddhd 90 70, 327-8, 338 rudhird 68, 70, 84, 103, 149 rup-, 85, 292 ruvanydti , 362 rusant-, 76, 144, 292 70, 328, 333, 358 ruksd-, 163 rupakd -, 197 284 reknas-, 118, 159 renu-, 180 retas-, 159 resmdn-, 134 rat-, 15, 86 r oka-, 124 raga-, 124 roca-, 124 roct-, 177, 189 racis-, 106, 160, 177, 189 yocmtw-, 188 rodasi, 40 ra/>o, 179 roman-, 131 romasd 197 rohita-, 70, 85 rohisyai, 164 laksd-, 262 laksmt-, 176, 192 %Aw-, 68, 75, 83, 182 labdhd-, 90 /fl&A-, 295, 360 ZanZ-, 61 lavanga 379 lavana-, 81, 152 langala-, 380 laid, 385 ZiAsa, 394 83, 292, 330 Z#*-, 388 Wo, 73. 7 8 > 8 3 i 9 r , 95 Zi-, 292, 326, 356 Udhd-, 78 Z^-, 85, 292, 330 lubdhaka-, 57 INDEX 429 lubh-, 68, 83 luldya 375 loka-, 104 lohila 70, 85 vamsaga-, 198 vdktva 171, 372 vakrd-, 147 vaks-, 143, 290 vaksdtha-, 196 vagvand 152 vagvanu 158 vac-, 80, 92, 336, 342 vac ana-, 15 1 vacas-, 116, 118, 120, 158 vacasa-, 163 vdjra-, 24, 26, 148 vdjrabahu 215 vanij 198 vatsd-, 162 vatsatard 150 vai-, 106, 190, 342, 359 vai/r-, 339 vadhd 125 vddhatra -, 138, 196 vadhdnd, 128 vadhar , 128-9, 147, 362 vadharyati, 362 vadhasnd-, 153 vadhasnu -, 158 vadhU-, 192 vddhri- t 40, 157 van-, 340, 360 van-, 123, 127 vana-, 123, 127 vandd-, 198 vanana , 157 vdnanvant 128 vanar°- t 148 vanargu-, 128 vdnas~ t 158 vdnaspdti 212 vanisthu-, 164 VflMWS-, l6l vanusydti, 189 vdnya 87, 185-6 vap- } 109, 342 vdpra-, 148 vdpus-, 160 -1 vdpusa-, 201 vara-, 301, 321 vamrakd-, 197 vaydm , 264 vayana-, 138 vara-, 124 varatra, 198 varahd-, 25, 198 varimdn 131, 135 varimdt 132 vdrivas 159, 160, 188 vdrlman 16 1 vdruna 28 varutra 324 varutha 196 varcnya-, 152, 372 vartani-, 157 vartis-, 160 vdrtman-, 131 vdrdha 125 vardhitdr 188 vdrdhra 198 vdrman-, 131 varsfyas-, 80 varsman So, 134 valaya- t 385 valgu-, 182 va/Zt, 385 vas'-, 91, 97, 264, 319 vdsti- t 169 vas- f 93, 96, 267, 295, 320-1, 329. 330 , 342 vasaii-, 169 vasana-, 25 vasantd - f 129 vasar° , 148, 222, 226 va'sis^Aa-, 100 va'sxyas-, 100 vasw, ioo, 178, 184 430 INDEX vasuid, 172 vasutati-, 173 vastly dti , 363 vasuya , 364 t’astiyw-, 187, 364 vdstu, 170, 248 vdstra-, 137 vasnd 40 vasnaydti , 363 v asm an-, 131-2 vdsyas-, 1 61 20, 68, 78, 92, 95, 294, 329 vahdt-, 165 vakatu-, 170 vahni-, 40, 157 fa, 284 vagura, 156 vdgmin 144 vaghdt 165 vac-, 68, 92, 103, 123, 207, 221 vacya-, 185, 371 40 vajayu-, 37 vajina 138 vatyd, 192 vanara-, 128, 148 vapusa 163, 201 267-8 vdmd- t 175 vayavyd-, 199 l>aya:sa-, 163 123 aara-, 125 uaraftawa-, 43, 389 van, 177, Z23, 227, 245, 248 varya-, 200 391 284 vavrdhddhyai , 367 vahd-, 125 vdhas-, 158 msa-, 125 vdsantika 199 vasard 148, 200 158, 200 170, 242 vasteya-, 199 v/j 285, 287 zu-, 181, 251 vimsd 263 vimsati-, 261 vimsatitama- , 176 viklava 99 viksipti 57 vicchitti-, 57 vt)’-, 292 vithurydti, 362 wirf-, 68, 71, go, 103, 116, 325, . 327, 330. 335. 349. 362 viddtha 196 vidddvasu-, 216 vidurd 149 vidmanapas 214 vidmdne , 131, 159, 224 vidyd, 191 vidvald 148 vulvas 228-9 vidhavd , 69, 187 vidhu-, 182, 187 I/JM, 20, 285-6 vip-, 292 vipanyd, 129, 148 vipanyu-, 129 vipds-, 92 vtpula 42, 84 vipra-, in, 148 vipnis-, 91, 93-4, 96 viplava 99 vibhavasu-, 161 vibhrtra 151 vibhvan -, 141 virdsit , 100 virukmant 146 vivd$van- t 130, 145 vivdsvant - 7 4, 145 ms-, 322, 329 wtf-, 33, 73, 91- 2 , 242 INDEX visamdna-, 369 vrsdyate, 363-4 visya-, 186, 201 wjf-, 97 vispati-, 22, 209, 21 1 vrsnf-, 157 visva-, 275 vetana 138 visvdkarma 175 vetasd-, 163 visva jit-, 166 125 visvdthd, 280 40 visvaminvd- , 213 Dmi, 191 viivdha , 281 vepdthu 196 znsa-, 80, 126 138 visuna 152 vesantd-, 129, 153 vista-, 81 veiaV, 160 vz&i-, 40 vesdna-, 138 vitaye t 365 vehdt-, 165 vtra-, 85, 103 va/, 284 vlrdjdta 212 vaiydpriya-> 57 324, 334 vaiyavrtya 57 vr&a-, 19, 25, 104, hi, 116, uauyfl-, 201 126, 196, 232 uyao, 323 vrkatdt 173 vyath-, 292 vrkdti- y 169 vydthis 160 195, 225, 252 vyadvara-, 148 vrkkd-> 25 vyaiA-, 330, 343 162 vydghrd-, 375 vrjana-, 138 vratd- f 22 vrjind 152 vratacdrin ~ ( 213 wtf-, 68, 84, hi, 295, 322, 328, 344. 356 sdmstar 137 vft-; 165 s«A-, 324, 335, 360 vritd-, 104, 168 iakala-, 385 vrtra-, 151 sakuna-, 22, 152 vrtrahdn 213, 220-1 iakuni-, 153 vrddhd- t 90 sakunta 153 vfddhi 168 sa'Ar/, 127, 164, 221 wtfA-, 294, 358 sdkti-, 168 vrdhd 125 sagmd-, 175 vrdhdnt 367 73 vrdhasdnd- , 155 sankd, 191 vrdhdnd 369 Sanku 180 vrsanyati , 363 sankhd 72, 196, 393 143 saa, 192 vrsabha 198 iacwasw-, 161 vrsasydti , 362 satagvin 216 vrsakapi-, 21 1 £atatama- t 263 431 INDEX 432 sdtdm, 13, 24, 73, 104, 262 satasds, 263, 281 sdtru-, 223 satrutvd-, 17 1 sad-, 102 sdnakais, 282 sdntdti-, 173 santivd-, 187 sapatha-, 196 saphd-, 71, 197, 393 saphara-, 22 saw-, 321 saw, 123 sama-, 40 sawi, 177, 192 192 s'ayw-, 182 iayutrd , 132, 280 sardd-, 74, 198 sarabhd 25 sdru- i 180 sdrdha-, 77 sardhanjaha- , 214 iarvd-, 186 s dr van, 142 salabha 198 ialaka , 27 s'a/fta(/a-), 386 salyakrnta 214 savasdnd-, 155 savasin-, 143 ^az>asl, 164 sasa-, 99 sasajya-, 145, 186 ^asfjyas-, 145, 186 iasvant 145, 186 sas^a-, 395 s'as-, 99, 290, 292 sasfra-, 137 sa-, 322-3, 331 sdka-, 22 idkind 132 sakvard-, 148 sa£Aa, 72, 191, 393 santd-, 87 sdnti-, 168 sdpa-, 22 sdrdula-, 375 s'as-, 94, 99, 320 sdstrd 137 sikvd-, 186 sikvan-, 186, 227 sikvas-, 160, 223 situ-, 106, 189 sithird-, 46, 149 sithild 46, 149 siprci, 40 siphd , 19 1 szras-, 87, 133, 158, 221, 227 sirognvd-, 218 sisaya 106 sxsa-, 187 sisna-, 292 sighra-, 198 sttosna 219 stbhatn, 198 sirsdn-, 87, 227,. 234 sukrd-, 148 sukld-, 148 sue-, 291, 321 sued-, 124-5 sueddratha-, 215 sucamdna 369 suci-, 186, 204, 230, 245 sucitvd-, 1 71 sund-, 40, 93 sundhyu 187 swfcA-, 291, 331 subhri-, 157 subhvan 142 susukvand 152 susukvdni-, 157 susruvds-, 161 suska-, 80, 100, 197 swsma^fl-, 186 SW-, 291, 331 suka- t 25-6 sudrd-, 40 INDEX 433 sunyd-, 74 stira-, 148 susd-, 163 ifnga-, 198 srnvdnt -, 368 sr-, 326 soka- t 124-5 soct'-, 177 socts-, 160, 177 sobhdse , 159 sosa-, 103 sautira 387 • snath-, 292, 321 smdsrn -, 77, 99, 178 smasrund-, 149 smasruld-, 149 sydma 22, 175, 186 sy&vd-, 22, 175, 186 syetd-, 40, 161 syend 82 sye'm, 167 srath-, 292, 363 sratharydti t 129, 362 sraddhd 15 sraddkivd -, 187 sr am-, 331 srdvana 109, 152 srdvas-, 73, 83, 158 sravdyya 372 sn-, 83, 291, 337 sra-, 22, no, in, 290-1, 293, 30X, 324, 341, 359, 361 srutd-, 103, no, in, 116, 167, 370 srutya-, 371 sr«s-, 290 142 stent-, 157 srenisds, 281 srejyas-, 228 stand-, 152 ifdm'-, 73, 83, 157 srotta-, 137 Stomata-, 138 srosamdna- , 77 slaksnd-, 83, 99, 153 s/ts-, 99 slesmdn 9g slesmana-, 148 slesmald 149 sldka-y 84, 197 slona-, 152 73, 141, 222, 375 svanin 143 svdbhra-, 136 svdsura-, 77, 99, 149 svasni 192, 253 suas-, 190, 321 sort's, 279 m-, 291 22, 291 svitlci-, 197 ivitna-, 151 svitrd-,. 148 svetana, 129, 148 svetra-, 129 svattan , 129 sa'/, 100, 260 saddha , 95 sasti-, 261 sas^ 7 ama-, 263 sasthd 262 sodasa, 260 sodhi, 95 Sfl, 268 ff. samvatsard 150 samskrta 1 saA/ 7 , 258 sdktu-, 1 jo sakthi , 177, 234 saAsfl-, 163 saksdni-, 157 sdkhi-, in, 137, 180, 181, 185, 196, 223, 230 sakhyd 137, 180 sakhitvand 171 INDEX 434 sakhTydii , 363 sanitvan-, 130 sagh-, 291 sanisydti, 189 samgamd-, 125 samsw-, 189 sac-, 77, 93, 109, 295, 329 sdnutar 190 sacdtha-, 196 sanutdr, 132, 150, 278 sdca, 194, 286 sdnutara-, 150 sdci, 177, 185 sannatara-, 115 saciva 187 sdnyas-, 161 sfl;-, 322 40, 129 sattrd-, 90, 137 safiadi, 283 satyd-, 185 sapary ati, 40, 129, 362 satydtati-, 173 104, 173, 259, 260 satyavddin- , 213 saptati 261 satydnrtd-, 218 saptathd 262 saty avail-, 142 saptamd - f 176, 260, 262 sdtvan-, 28, 143 sdpti-, 169 satvand -, 152 sapsara-, 150 satvdnt 145 sabardugh 128 88, 94, 109, 178, 329, siw, 285, 287 335 - 6 , 342, 359 sama-, 272, 275 sddana-, 138 samd-y 258 sadandi-, 198 samajyd, 191 sddam, 174 sdmana-, 138 sddas-, 109, 158 samanas -, 258 194 samdni, 285 saddprnd 214 samayd, 285 sadivas, 179 samaha, 275, 281 sadman-, 131, 142, 174 samand- t 155. 258 sadmdn-, 142 samit-y 165, 168 sadyds, 279 sdmiti 168 sadhdstha 88 samipe, 286 sddhis 197 samusyald 149 san-, in, 325, 360 samrdjni, 194 sana- t 68, 102, 126 sar-, 81, 291 sanakd-, 197 sardgh-, 198 sandj-, 198 saranydti , 362 sanat, 166 sardtham, 285 sanaya-, 186 sarasi, 164 sdnara-, 136 sarit-, 166, 291 sandt, 233, 283 sarird-, 149 sanatina 153 sarga-y 79 saw/-, 179 sarpis 83, 160 sanitur , 278 sdrma 176 sanitva-, 17 1 sdrva-, 187, 275 INDEX 435 sarvdtas, 167 sarvdtat(i-) , 173 sarvada, 280 sarvandman -, 48 sarvavedasd-, 216 sarsapa 380, 398 salild-, 149 sdvilave, 171 savyd-, 21 sasco/-, 165 sas-, 40, 319, 322 sasthavan -, 149 sasw, 130, 279 sasvdrtd, 194 sa/z-, 78, 360 saAa, 281, 285 sahadhyaiy 198, 36* sdhas-, 158 sahasand-, 155 sahdsra ~, 262 sahasra&ds , 281 sdhuri 157 sahobhdri -, 213 sdhyas-, 161 sdhyu -, 187 5a-, 85, 106, 291, 331 sdkdtn, 285 sdksat, 283 sddha-, 78 sa/a-, 87, hi saJtaye, 365 s«ia-, 125 sa^A-, 324 sddhu-, 182, 324 sdnasi-, 164 sanw, 178-9, 200, 247-8 sdndra-, 22 samand-, 135 sdmi- f 102, 279 samavipra 214 sdmidheni , 152 sdrdhdm, 285 sat>a'-, 125 saAa-, 126 sdhvds-, 140, 343 si- (see also sa-), 189, 291 sic-, 292-3, 330, 333 si/a-, 41 si/a-, 106, 189 stndhu 72, 88, 180, 197 simd-, 275 sxw, 268 simdn-, 134 simdnta -, 153 sitnikd, 385 sw-, 289, 323 sutarman-, 27 sutvan, 142 sunvdnt-, 222 sunvdna -, 369 sw^/a-, 75 sumdnas-, 137, 203, 230 sumnd-, 152 sumndya , 282 sayasas-, 216 Siam-, 41 swra, 25, 156, 191 surungd , 387 suvasas-, 122 susisvi-, 187 susthu t 284 3 20 » 342 sa/a-, 167 sfltoc, 17 1 sw/w-, 170 sutra- f 137 sun-, 157 simzi-, 68, 104, 158, 245 251 s«o-a- ( 29, 83, 185 suryatejas -, 215 surmi, 176, 192 sfkvan-, 130 srkvin-, 131, 177 sV;-, 78, 329 sr«i-, 157 . i8 5 srwi, 192 sfnya-, 185 233, INDEX 436 sftvan-, 142 srtvara-, 148 srp-, 68, 84, 295, 329 srmard 149 srstd-, 78 25, 170 sww. 4 » x 57 , 252 sewz, 1 91 sainbha 375 sod fid-, 78 safari, 365 sota-, 170 soma-, 4, 176 soman-, 143 somapavan-, 213 so'masw^-, 212 stwya-, 185 saumanasd 137, 201 sautnyd-, 199 skabh-, 326 s Jihad-, 81 xtan-, 106, 321, 328, 330 stdna-, 100 stanayitnu-, 81, 106, 158, 189 stanasydti , 362 stanitd -, 189 stabhuyati , 363 star-, 140 star*, 192 sidy ant-, 81 stayw-, 81 sfig/*-, 40, 69, 76 sttrnd-, 291 s/«-, 94, 292, 320 s/wAa, 198 s£&£-, 124, 165 state'-, 169 stubh 292 stubhvan-, 142 stupa-, 198 s/f-, 293:, 323-4 stana-, 81, 151 stated, 137 stoma-, 176 sty ana-, 46 sthag- , 68,71-2, 76, 329, 393 s i hap a t i 169 sthdla-, 136 star?-, 71-2, 103, 106, 107, 190, 298, 322, 329, 330, 334, 343, 358-9

• slhalar, 132

sthdtdr-, 139, 206 sthdtra-, 137 sthavand-, 148 sihdvard-, 148 sthitd-, 106, 189 slhiti-, 106 sthird 106, 149 sthUnd , 98, 157 sthurd 148 sthuld-, 148 sthulabhd - r 198 stheman 106 s they as-, 106 stta-, 103 sndtva, 172 sndyu, 178 130 snihiti 106 snwsff, 103 snuh-, 70 serfs'-, 81, 92, 123, 221 spurdhdse , 159, 226 s£r-, 292 sprdh-, 292 s£rs-, 329 s£r/^ 292, 331 s^rAa, 1 91 sphatika-, 97 sphigt , 192 sphij-, 198 sphira 149 sphuia 97 sphulihga 97 sphiirj 71 sphyd-, 71, 81 swa, 284 INDEX 437 smut, 279 syd, 270, 272-3 syand-, 72, 88, 197, 292 syUman-, 13 1, 134 syona 40, 152 sram'-, 137, 360 sragvin 144 srdj-, 198 sravat-, 165 sravatha-, 196 srastara-, 137 srdma-, 22, 176 sm-, 84, 291, 328 srutd-, 168 sruti-, 169 sruvd-, 186 srctfas-, 1 18, 159 sva-, 141, 268 svadhdvan 142 svflw-, 290, 328 190, 292, 321, 343 svapand 15 1 svapna-, 46, 75 svapnayi, 282 svaydm, 268 svar-, 290 smr, 129, 226 svari-, 184 svarita 115 svdru-, 180 swzvos-, 228 ig, 25, 203, 234, svadistha-, 162 svddu-, 103, 182 svdduksadman 213 svddman-, 131, 134 svadmdn 134 384 hamsd -, 102, 163 Aafc*-, 104, no, 168 Aaw-, 32, 70, 75, 78-9, 1 3 J 9> 343> 355, 360 hdnu-, 180 hdntar-, 284 hantavai, 17 1 hdntos , 17 1 hdntva 172 hdnman 13 1 haras-, 40, 78, 158 hriri-, 139 harit-, 166 hdrita 166-7 harind 152 hdrint, 167 harimdn 134 harimantha- , 387 harmula 150 harmyd 150 hary-, 40 hdva-, 125 Havana no havirdd- t 213 havirddya 180 Gain's, 160, 193, 253 hdvlman-, 135 hdvya- t 185 havyd 87 havyavah-j 221 Aa/a-, 380 hdliksna 153 Aas-, 109 hasani , 157 hdsta-, 74 hastakrta 214 243 Aastfn-, 375 hasra-> 147 Aa-, 290, 322-3, 361 Aas**, 290 hdtaka 97 hardi, 177 hdsa-, 109 W-, 324 At, 284 himsd , 191 io-i, hiffisitavyd -, 371 At/a-, 70 Aital, 372 INDEX 438 hintdla 385 Ae/w-, 170 hinv 324 he'man , 103 /iw-, 3 2 » 73 > *75 hemantd 73, 103, 153 htmd, 176 heramba 375, 385 Mr any a-, 4, 25, 87 AeZi-, 387 hiranyakesya 216 hesas-, 158 hiranydya-, 186 Adfor-, 4, 137, 140 hiranyarathd- , 21 1 hotrvurya 180 hid-, 94 hotrd-, 137 Aw-, 322, 360 hdtrd, 156 hudukka 385 homa- t 176 huvanydti , 362 homan-y no M-, 22, 79, IIO, 329, 331, 355 Aora, 387 hutd- } no Ajya's, 82, 279 hr nay a-, 363 hyastana-, 153 hrmya- , 363 hr add-, 100 hrttds , 279 hrasvd-, 186 hr day a-, 180. hraduni 157 hrdaydvidh 213 An-, 322 hrdroga 387 Arw/-, 165 330 Avar-, 323