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
