Is there a reason for the visarga (अः, or अस्/अर्) changing to an [o] (ओ) sound before voiced consonants or अ in Sanskrit as a part of external sandhi? For example, as an example from Devavanipraveśika, रामस् + गच्छति = रामो गच्छति.

Most sandhi rules make sense intuitively (इ + अ = य, with [i] and semivowel [y]) or historically (अ + इ/ई = ए due to the nature of the diphthong in Vedic), but this rule converting a visarga to an [o] seems to come out of nowhere.


1 Answer 1


This is a bit of a historical mystery. Sydney Allen in his book Sandhi offers some conjectures. The premise that seems not questionable is that the original situation was that the language had z for /s/ before voiced consonants; it is also a fact that Sanskrit had no voiced fricatives, so we may presume repairs modifying any voiced fricatives (which however have to be historically argued for – this replacement and relates change s→r are the evidence for such a rule). A reasonable solution then is to replace /z/ with a glide, and there is evidence for y as a replacement for z. The crux of his argment is this:

From a general phonetic standpoint one could equally well expect the semivowel to be of [y] or [w] quality; and probably there was an original alternation between these, determined by the nature of the following consonant. But before our earliest records v had in fact been generalized in this function (in contrast with the y generalized before an initial vowel: see p. 62).

I don't think [w] is as likely as [y]. Perhaps the fricative could have been influenced by the following consonant, so that the glide was somewhat indeterminate in place, instead being an anticipatory "shadow" of the following consonant, but then there would be a 50-50 distribution of y-like glides and w-like glides (treating velars as 'w-like'), so why would the w-like variant prevail in historical change? Phonologizations choices are phonetically arbitrary.

In partial support of this analysis, Allen notes dialectal differences in Middle Indo-Aryan. The more general development of final /as/ is as [o] regardless of context, thus Skt. putras → Prakrit putto. But Aśokan inscriptions attest a competing outcome dharmas → dhamme, supporting the idea of y~w vacillation. There are also variant outcomes in Vedic ṣoḍaśaṣaṣ daśa, where one would expect a more w-like variable glide before a retroflex consonant. It should be pointed out that there are problems with this hypothesis, for example as observed by Marsh 1941, /z/ before velars and labials becomes [d], and there is a fair amount of evidence that there is no general w trend associated with /z/ regardless of the following consonant.

It is important to note that even in Sanskrit, it is a misanalysis to say that visarga turns into o, since /punar/ changes final /r/ into visarga and thus you have punaḥ, but that visarga does not change to o, instead, /r/ remains. The synchronic change is to be stated in terms of the underlying consonant, and not the pre-pausal pronunciation of it.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.