THE 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 Elsewhere this has been replaced by innovating forms with guna or vrddhi of suffix in these cases (Gk.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 In Sanskrit and partly in Avestan the - ars has been replaced by -rs (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
