________________
342
Amrita
in the Indo-Iranian stage and continue to be so in the old Indo-Aryan. Moreover, the original IE series of velars (including the labio-velars) has developed two types, one palatal and the other velar, the first of which is later represented by affricates in Sanskrit. The explanation of the rise of the palatal series (č, j) from the original velars (k, g) is sought in the effect produced by the following front vowels like e, as against the back vowels like a or 0. It is also known that the change of the velars into the palatals preceded the merger of e with a along with a similar merger of o with it. This sequence is necessary simply because, if the vowels are regarded as merging earlier, the very reason for the development of the palatal sounds would be lacking. The question of the relative chronology of the mergers of o and e with a need not concern us here.
The phonemic statements about these changes and their comparison will bring out an important limitation of the procedure. The two divergent sounds of the velar stops, one more advanced before e (say k”) and one more back before a and o (say k) would not be phonemically distinct as long as the difference between the vowels following them remains phonemic. The phonemic split between these two as // and /k/ can be regarded as having occurred only when they contrast in the same environment, which in this case would be the same phoneme /a/ coming from all the three sounds of the earlier stage. We may graphically refer to these changes in their phonetic concreteness as follows:
III [ke]
[k'e-1] [ko]
[ko-]
1
[ka]
When we try to put these changes in a phonemic form, we have an alternative analysis available for the second stage. One way would be: /ke/ ke/
/ca/
/ko/
/ko/
/ka/
in which the first two stages are identical. Another possible way of analysis would be :
/ke/
/ca/
/ca/
/ko/
/ka/
/kd/
in which case, the last two stages become identical. The difference