The existence of aindra grammar which apparently declined in due course due to pANini’s efforts has been discussed extensively in the past by Burnell in his monograph on the topic. While things might have changed in certain directions in our understanding of the evolution of formal grammar of saMskR^ita, the existence of the aindra system as gleaned by Burnell is not in much doubt. In our opinion there is also not much doubt that the earliest Tamil grammar, the tolkAppiyaM was modeled after the aindra vyAkaraNa of saMskR^ita. However, there have been a number of related controversies on these issues that keep coming up: 1) Doubts on the age of the tolkAppiyaM – most of its major commentaries are medieval. Hence, some have declared that it is not at all an early text but actually a late text closer to the medieval period. 2) Age of the aindra grammar itself – here again there are claims that it is a post pANinian innovation rather than predating him as Hindu tradition generally holds.
As we mentioned earlier durgasiMha, learned commentator of the nirukta was clearly aware of the aindra grammar and uses it in his analysis which is generally consistent with sAyaNa’s own knowledge of it. Hindu tradition also links the transmission of the aindri vAk to the god kumAra (related to the famous tale in the bR^ihat-kathA/kathA-sarit-sAgara). Apocryphal versions of this (still not without value) are also offered by the learned Tibeto-Mongol Lama tAranAtha in his history: He states that there was a brAhmaNa named saptavarman, who a contemporary of nAgArjuna and kAlidAsa [a typical example of tAranAtha’s pseudo-history]. He asked ShaNmukha to reveal the aindra grammar to him, and upon skanda merely uttering the first sutra of the grammar “siddhovarNasamAmnAyaH |” the rest of the grammar of 25,000 verses was miraculously transmitted to him. He also mentions that a brAhmaNa named indradhruva obtained the same from indra. The enormous size mentioned by the Lama tAranAtha is also repeated by other late Hindu writers like devabodha who compares the aindra to the sea whereas pANini to the depression of left by the cow’s hoof in size. tAranAtha is also conversant with the apocryphal belief common in India that the pata~njali of the mahAbhAShya was the avatAra of the snake sheSha (likewise the eponymous author of the yoga sUtra is also considered an incarnation of the same snake). Thus, on one hand we see a lay belief, prevalent in greater India at the time, being transmitted by the Tibetan Lama to Mongolia. On the other hand, little earlier than tAranAtha, the great vedic commentator sAyaNa preserved the one crucial piece of evidence that firmly places the origins of the aindra vyAkaraNa in the brAhmaNa layer of the shruti. This lost brAhmaNa cited by sAyaNa appears corrupt (at least in my copy) but is beyond doubt a brAhmaNa passage of vedic provenance (especially given that the majority of sAyaNa’s brAhmaNa citations are entirely verifiable): vAg vai parAchy-avyAkR^itA .avadat te devA indraM abruvann imam no vAcham vyAkurv iti | so .abravId varaM vR^iNe mahyaM chai .evai e.Sha vAyave cha saha gR^ihyAtA iti tasmAd aindravAyavaH saha pragR^ihyate | tAm indro madhyato .avakramya vyAkarot | tAsmAd iyaM vyAkR^itA vAg udyata iti |
A good candidate is the lost brAhmaNa section of the charaka saMhitA of the kR^iShNa yajurveda from which sAyaNa has made other citations. It appears that even in the brAhmaNa period the general assumption was the speech or language was originally not systematized via grammar. The deva indra generated a grammar that allowed the analysis of speech. So the idea of the aindra-vyAkaraNa existing is indeed pre-pANinian and present in the shruti itself. Such a system might actually go back to an earlier Indo-European model, which at least predates the split of Greek+Balto-Slavic+Indo-Iranian and might be related to the earlier Indo-European concept of the divine speech which survives in the Indic world in the form of saMskR^ita being deva-bhAShA.