21 (cf. OP nijayam)> OIA nirdyam; II* du$manas> PI, PIA *duz v manas> OIA durmanah. Before r it disappears and lengthens the preceding short vowel. Thus: II *nisrauga-> PI, PIA* niz v rauga-> OIA niroga-. Dialectally in PIA this z v becomes z and is lost before other voiced consonants ; e.g. II *dus v dabha-> PIA *duz v dabha-, duz'dabha->QLA dudabha-( 7 edicy (J) The operation of Bartholomae's Law started at the Indo-Iranian stage and it is perhaps more strictly operative in Proto-Indo-Aryan. The law indicates that a consonant-group consisting of a voiced aspirate and an unvoiced aspirate changes into a group of voiced non-aspirate plus voiced aspirate. The phenomenon is a kind of mutual assimilation and meta thesis combined. Thus : II *dughta-> IA dugdhd-', II *uz'hta-> PIA
- uz'dha->01AiidhJ--,Il *basfhta-> PIA *baz'dha->
OIA badha-; II* trnaz'hti> PIA* trnaz'dhi> OIA trnedhi ; etc. (g) Proto-Indo-Aryan has, like Proto-Iranian, only r and r (for both long and short). In Old Indo-Aryan r has a dialected tendency of changing into I. (This tendency was probably already there in Proto-Indo- Aryan.) I in Sanskrit is a variant of /*, and r (long) is an analogical creation in the written language. As a matter of fact r (long) occurs only in some plural case forms of some nouns, to be exact in nom-acc, pi. nt. and in acc. and gen. pi. of nouns in -r (e.g. datr'ni, filfn,matr [ h and bhratfnam). Thenom.-acc. nt. form is very late ; it does not occur in Vedic and is not generally met with in Classical Sanskrit literature. The acc. pi. and gen. pi. forms are not historical