Prakrit, in order to call it an absurdity that those persons would have spoken prose in Sanskrit and poetry in the Gatha dialect, such as we find in some passages already quoted and in many others. Nor is it absurd, even if we do not believe that Pali is the original language of scripture, to contend that the Sanskrit text of the canonical works is at any rate a translation from some dialect. If the Sanskrit text of the Northern Sütras, in general, were the original one, it would be impossible to account for occasional mistranslations and for the fact that the most palpable dialect forms have been left untouched, whenever the passage by being Sanskritised would have been spoilt. A striking instance is afforded in Lalita-vistara, p. 145. There we read the pronouncing of the letter tha of the Indian alphabet is to be brought in connection with the word thapanïyapra<jna, i.e. a question that should be avoided, set aside, Pali thapanïyapanho. Here the context absolutely opposed itself to the Pali or Prakrit thapanlya being rendered by the Sanskrit sthapanïva, because the initial syllable of this form could'not be made to agree with the letter tha. On the same page of the Lalita-vistara we also meet with a word airapatha', the initial sjllable of which must needs harmonise with the diphthong ai, so that airapatha did not admit of being Sanskritised into aryapatha. From the occurrence of this airapatha I infer that the original text was composed in some kind of Prakrit, and not in regular Pali, because the latter has lost both the primary and secondary diphthong ai, though it may be asked whether forms such as kayira (Sansk. karya), payirupasati (Sansk. pary u pasati), and the like are anything else but instances of inaccurate spelling. 2 This much is certain that thaira occurs in the inscriptions of Agoka, and in these the diphthong cannot but have the value of a short a followed by i.
If we eliminate the Sanskrit, there remain two dialects, Pali and the Gatha idiom. Which of the two can lay claim to being the original language of the Buddhist scriptures or is the nearest approach to it? Pali is intelligible in its phonetics, the Gathas are not. Under ordinary circumstances the comparatively greater regularity of Pali would tend to favour its claims; the case before us is, however, so peculiar that it is not safe to draw inferences from the state in which the Gathas have come to us. It seems to me that
1 Written airapatha, for the Vrddhi vowel denotes the sound of ai in Sanskrit, at least originally; from the same diphthong being used in the Acjoka ediets in thaira (Sansk. sthavira), we must infer that the diphthong was, in the then Prakrit, sounded ai, not ai.
2 That is, kayira was probably pronounced kaira, which cannot be exactly expressed by kaira, because those who were acquainted with the rules of Sanskrit grammar would pronounce this and similar words with the sound of ai.