viz. that under all circumstances initial a is elided, if the preceding word ends in aye, ayah, ave, or avah. As might be expected, however, so large a class must have numerous exceptions, and these can only be collected by quoting every word ending in these syllables, or every passage in which the exceptions occur. Before these exceptions are enumerated, some other more or less general observations are made, providing for the elision of initial a. Initial a, according to Sûtra 142, is to be elided if the preceding word is vah, and if this vah is preceded by â, na, pra, kva, kitrah, savitâ, eva, or kah. There is, of course, no intelligible reason why, if these words precede vah, the next a should be elided. It is a mere statement of facts, and, generally speaking, these statements are minutely accurate. There is probably no verse in the whole of the Rig-veda where an initial a after* vah is elided, unless these very words precede, or unless some other observation has been made to provide for the elision of the a. For instance, in V. 25, 1, we find vah preceded by akkha, which is not among the words just mentioned, and here the Sanhitâ does not elide the a of agnim, which follows after vah. After all these more or less general observations as to the elision of an initial a are thus exhausted, the author of the Prâtisâkhya descends into particulars, and gives lists, first, of words the initial a of which is always elided ; secondly, of words which, if preceding, require under

all circumstances the elision of the initial a of the next word, whatever may have been said to the contrary in the preceding Sûtras. Afterwards, he gives a number of passages which defy all rules, and must be given on their own merits, and as they stand in the Sanhitâ. Lastly, follow special exceptions to the more or less general rules given before. And here, among these special exceptions, we see that the author of the Prâtisâkhya finds it necessary to quote a passage from a Vâlakhilya hymn in which hetáyah occurs, i. e. a word ending in ayah, and where, in defiance of Sûtra 149, which required the elision of a following initial a under all circumstances (sarvathâ), the initial a of asya is not elided; viii. 50, 2, Sanhitâ, satanîkâ hetáyo asya. It might be objected that the Prâtisâkhya only quotes hetáyah as an exceptional word, and does not refer directly to the verse in the Vâlakhilya hymn. But fortunately hetáyah occurs but twice in the whole of the Rig-veda; and in the other passage where it occurs, i. 190, 4, neither the rule nor the exception as to the elision of an initial a, could apply. The author of the Prâtisâkhya therefore makes no distinction between the Vâlakhilya and any other hymns of the Rig-veda, and he would have considered his phonetic statistics equally at fault, if it had been possible to quote one single passage from the hymns viii. 49 to 59, as contravening his observations, as if such passages had been alleged from the hymns of Vasishtha or Visvamitra.

It would lead me too far, were I to enter here into similar cases in support of the fact that the Prâtisâkhya makes no distinction between the Vâlakhilya and any other hymns of the Rig-vedasanhitâ*. But I doubt whether the bearing of this fact has ever been fully realised. Here we see that the absence of the elision of a short a which follows after a word ending in ayah, was considered of sufficient importance to be recorded in a special rule, because in most cases the Sanhitâ elides an initial a, if preceded by a word ending in ayah. What does this prove? It proves, unless all our views on the chronology of Vedic literature are wrong, that in the fifth century B. C. at least, or previously rather to the time when the Prâtisâkhya was composed, both the Pada and the Sanhitâ texts were so firmly settled that it was impossible, for the sake of uniformity or regularity, to omit one single short a; and it proves à fortiori, that the hymn in which that irregular short a occurs, formed at that time part of the Vedic canon. I confess I feel sometimes frightened by the stringency of this argument, and I should like to see a possibility by which we could explain the addition, not of the Vâlakhilya hymns only, but of other much more modern sounding hymns, at a later time than the period of the Prâtisâkhyas. But until that possi

* The Prâtisâkhya takes into account both the Sâkala and Bâshkala sâkhâs, as may be seen from Sûtra 1057.

bility is shown, we must abide by our own conclusions; and then I ask, who is the critic who would dare to tamper with a canon of scripture of which every iota was settled before the time of Cyrus, and which we possess in exactly that form in which it is described to us by the authors of the Prâtisâkhyas? I say again, that I am not free from misgivings on the subject, and my critical conscience would be far better satisfied if we could ascribe the Prâtisâkhya and all it presupposes to a much later date. But until that is done, the fact remains that the two divergent texts, the Pada and Sanhitâ, which we now possess, existed, as we now possess them, previous to the time of the Prâtisâkhya: they have not diverged nor varied since, and the vertex to which they point, starting from the distance of the two texts as measured by the Prâtisâkhya, carries us back far beyond the time of Saunaka, if we wish to determine the date of the first authorised collection of the hymns, both in their Pada and in their Sanhitâ form.

Instances abound, if we compare the Pada and Sanhitâ texts, where, if uniformity between the two texts had been the object of the scholars of the ancient Parishads, the lengthening or shortening of a vowel would at once have removed the apparent discordance between the two traditional texts. Nor should it be supposed that such minute discordances between the two, as the length or shortness of a vowel, were always rendered necessary by the

requirements of the metre, and that for that reason the ancient students or the later copyists of the Veda abstained from altering the peculiar spelling of words, which seemed required by the exigencies of the metre in the Sanhitâ text, but not in the Pada text. Though this may be true in some cases, it is not so in all. There are short vowels in the Sanhitâ where, according to grammar, we expect long vowels, and where, according to metre, there was no necessity for shortening them. Yet in these very places all the MSS. of the Sanhitâ text give the irregular short, and all the MSS. of the Pada text the regular long vowel, and the authors of the Prâtisâkhyas bear witness that the same minute difference existed at their own time, nay, previous to their own time. In vii. 60, 12, the Sanhitâ text gives:

iyám deva puróhitir yuvábhyâm yagñéshu mitrâvarunâv akâri.

This primacy, 0 (two) gods, was made for you two, O Mitra and Varuna, at the sacrifices!

Here it is quite clear that deva is meant for a dual, and ought to have been devâ or devau. The metre does not require a short syllable, and yet all the Sanhitâ MSS. read devă, and all the Pada MSS. read devâ; and what is more important, the authors of the Prâtisâkhya had to register this small divergence of the two texts, which existed in their time as it exists in our own*.

* See Prâtisâkhya, Sûtra 309 seq., where several more instances of the same kind are given.