With regard to a critical restoration of the text of the Upanishads, I have but seldom relied on the authority of new MSS., but have endeavoured throughout to follow that text which is presupposed by the commentaries, whether they are the work of the old Sankarâkârya, or of the more modern Sankarânanda, or Sâyana, or others. Though there still prevails some uncertainty as to the date of Sankarâkârya, commonly assigned to the eighth century A.D., yet I doubt whether any MSS. of the Upanishads could now be found prior to 1000 A.D. The text, therefore, which Sankara had before his eyes, or, it may be, his ears, commands, I think, a higher authority than that of any MSS. likely to be recovered at present.

It may be objected that Sankara's text belonged to one locality only, and that different readings and different recensions may have existed in other parts of India. That is perfectly true. We possess various recensions of several Upanishads, as handed down in different Sâkhâs of different Vedas, and we know of various readings recorded by the commentators. These, where they are of importance for our purposes, have been carefully taken into account.

It has also been supposed that Sankara, who, in writing his commentaries on the Upanishad, was chiefly guided by philosophical considerations, his chief object being to use the Upanishads as a sacred foundation for the Vedânta philosophy, may now and then have taken liberties with the text. That may be so, but no stringent proof of it has as yet been brought forward, and I therefore hold that when we succeed in establishing throughout that text which served as the basis of Sankara's commentaries, we have done enough for the present, and have fulfilled at all events the first and indispensable task in a critical treatment of the text of the Upanishads.

But in the same manner as it is easy to see that the text of the Rig-veda, which is presupposed by Sâyana's commentary and even by earlier works, is in many places palpably corrupt, we cannot resist the same conviction with regard to the text of the Upanishads. In some cases the metre, in others grammar, in others again the collation of analogous passages enable us to detect errors, and probably very ancient errors, that had crept into the text long before Sankara composed his commentaries.

Some questions connected with the metres of the Upanishads have been very learnedly treated by Professor Gildemeister in his essalv, Zur Theorie des Sloka. The lesson to be derived from that essay, and from a study of the Upanishads, is certainly to abstain for the present from conjectural emendations. In the old Upanishads the same metrical freedom prevails as in the hymns; in the later Upanishads, much may be tolerated as the result of conscious or unconscious imitation. The metrical emendations that suggest themselves are generally so easy and so obvious that, for that very reason, we should hesitate before correcting what native scholars would have corrected long ago, if they had thought that there was any real necessity for correction.

It is easy to suggest, for instance, that in the Vâgasaneyisamhîtâ-upanishad, verse 5, instead of tad antar asya sarvasya, tadu sarvasyâsya bâhyatah, the original text may have been tad antar asya sarvasya tadu sarvasya bâhyatah; yet Sankara evidently read sarvasyâsya, and as the same reading is found in the text of the Vâgasaneyi-samhitâ, who would venture to correct so old a mistake?

Again, if in verse 8, we left out yâthâtathyatah, we should get a much more regular metre,

Here vyada forms one syllable by what I have proposed to call synizesis^[39], which is allowed in the Upanishads as well as in the hymns. All would then seem right, except that it is difficult to explain how so rare a word as yâthâtathyatah could have been introduced into the text.

In verse 10 one feels tempted to propose the omission of eva in anyad âhur avidyayâ, while in verse 11, an eva inserted after vidyâm ka would certainly improve the metre.

In verse 15 the expression satyadharmâya drishtaye is archaic, but perfectly legitimate in the sense of “that we may see the nature of the True,” or “that we see him whose nature is true.” When this verse is repeated in the Maitr. Up. VI, 35, we find instead, satyadharmâya vishnave, “for the true Vishnu.” But here, again, no sound critic would venture to correct a mistake, intentional or unintentional, which is sanctioned both by the MSS. of the text and by the commentary.

Such instances, where every reader feels tempted at once to correct the textus receptus, occur again and again, and when they seem of any interest they have been mentioned in the notes. It may happen, however, that the correction, though at first sight plausible, has to be surrendered on more mature consideration. Thus in the Vâgasaneyi-samhitâ-upanishad, verse 2, one feels certainly inclined to write evam tve nânyatheto 'sti, instead of evam tvayi nânyatheto 'sti. But tve, if it were used here, would probably itself have to be pronounced dissyllabically, while tvayi, though it never occurs in the Rig-veda, may well keep its place here, in the last book of the Vâgasaneyisamhitâ, provided we pronounce it by synizesis, i.e. as one syllable.

Attempts have been made sometimes to go beyond Sankara, and to restore the text, as it ought to have been originally, but as it was no longer in Sankara's time. It is one thing to decline to follow Sankara in every one of his interpretations, it is quite another to decline to accept the text which he interprets. The former is inevitable, the latter is always very precarious.

Thus I see, for instance, that M. Regnaud, in the Errata to the second volume of his excellent work on the Upanishads (Matériaux pour servir à l'histoire de la philosophie de l'Inde, 1878) proposes to read in the Brihad-âranyaka upanishad IV, 3, 1-8, sam anena vadishya iti, instead of sa mene na vadishya iti. Sankara adopted the latter reading, and explained accordingly, that Yâgñavalkya went to king Ganaka, but made up his mind not to speak. M. Regnaud, reading sam anena vadishya iti, takes the very opposite view, namely, that Yâgñavalkya went to king Ganaka, having made up his mind to have a conversation with him. As M. Regnaud does not rest this emendation on the authority of any new MSS., we may examine it as an ingenious conjecture; but in that case it seems to me clear that, if we adopted it, we should have at the same time to omit the whole sentence which follows. Sankara saw clearly that what had to be accounted or explained was why the king should address the Brahman first, samrâd eva pûrvam paprakkha; whereas if Yâgñavalkya had come with the intention of having a conversation with the king, he, the Brahman, should have spoken first. This irregularity is explained by the intervening sentence, in which we are reminded that on a former occasion, when Ganaka and Yâgñavalkya had a disputation on the Agnihotra, Yâgñavalkya granted Ganaka a boon to choose, and he chose as his boon the right of asking questions according to his pleasure. Having received that boon, Ganaka was at liberty to question Yâgñavalkya, even though he did not like it, and hence Ganaka is introduced here as the first to ask a question.

All this hangs well together, while if we assume that Yâgñavalkya came for the purpose of having a conversation with Ganaka, the whole sentence from “atha ha yag ganakas ka” to “pûrvam paprakkha” would be useless, nor would there be any excuse for Ganaka beginning the conversation, when Yâgñavalkya came himself on purpose to question him.

It is necessary, even when we feel obliged to reject an interpretation of Sankara's, without at the same time altering the text, to remember that Sankara, where he is not blinded by philosophical predilections, commands the highest respect as an interpreter. I cannot help thinking therefore that M. Regnaud (vol. i, p. 59) was right in translating the passage in the Khând. Up. V, 3, 7, tasmâd u sarveshu lokeshu kshattrasyaiva prasâsanam abhût, by “que le kshatriya seul l'a enseignée dans tous les mondes.” For when he proposes in the Errata to translate instead, “ç'est pourquoi l'empire dans tous les mondes fut attribué au kshatriya seulement,” he forgets that such an idea is foreign to the ordinary atmosphere in which the Upanishads move. It is not on account of the philosophical knowledge possessed by a few Kshatriyas, such as Ganaka or Pravâhana, that the privilege of government belongs everywhere to the second class. That rests on a totally different basis. Such exceptional knowledge, as is displayed by a few kings, might be an excuse for their claiming the privileges belonging to the Brahmans, but it would never, in the eyes of the ancient Indian Aryas, be considered as an argument for their claiming kingly power. Therefore, although I am well aware that prasâs is most frequently used in the sense of ruling, I have no doubt that Sankara likewise was fully aware of that, and that if he nevertheless explained prasâsana here in the sense of prasâstritvam sishyânâm, he did so because this meaning too was admissible, particularly here, where we may actually translate it by proclaiming, while the other meaning, that of ruling, would simply be impossible in the concatenation of ideas, which is placed before us in the Upanishad.

It seems, no doubt, extremely strange that neither the last redactors of the text of the Upanishads, nor the commentators, who probably knew the principal Upanishads by heart, should have perceived how certain passages in one Upanishad represented the same or nearly the same text which is found in another Upanishad, only occasionally with the most palpable corruptions.

Thus when the ceremony of offering a mantha or mash is described, we read in the Khândogya-upanishad V, 2, 6, that it is to be accompanied by certain words which on the whole are intelligible. But when the same passage occurs again in the Brihad-âranyaka, those words have been changed to such a degree, and in two different ways in the two Sâkhâs of the Mâdhyandinas and Kânvas, that, though the commentator explains them, they are almost unintelligible. I shall place the three passages together in three parallel lines:

The text in the Khândogya-upanishad yields a certain sense, viz. “Thou art Ama by name, for all this together exists in thee. He is the oldest and best, the king, the sovereign. May he make me the oldest, the best, the king, the sovereign. May I be all this.” This, according to the commentator, is addressed to Prâna, and Ama, though a purely artificial word, is used in the sense of Prâna, or breath, in another passage also, viz. Brihad-âranyaka-up. I, 3, 22. If therefore we accept this meaning of Ama, the rest is easy and intelligible.

But if we proceed to the Brihad-âranyaka, in the Mâdhyandina-sâkhâ, we find the commentator proposing the following interpretation: “O Mantha, thou art a full knower, complete knowledge of me belongs to thee.” This meaning is obtained by deriving âmah from â+man, in the sense of knower, and then taking âmam, as a neuter, in the sense of knowledge, derivations which are simply impossible.

Lastly, if we come to the text of the Kânva-sâkhâ, the grammatical interpretation becomes bolder still. Sankara does not explain the passage at all, which is strange, but Anandagiri interprets âmamsi tvam by “Thou knowest (all),” and âmamhi te mahi, by “we know thy great (shape),” which are again impossible forms.

But although there can be little doubt here that the reading of the Khândogya-upanishad gives us the original text, or a text nearest to the original, no sound critic would venture to correct the readings of the Brihad-âranyaka. They are corruptions, but even as corruptions they possess authority, at all events up to a certain point, and it is the fixing of those certain points or cbronological limits, which alone can impart a scientific character to our criticism of ancient texts.

In the Kaushîtaki-brâhmana-upanishad Professor Cowell has pointed out a passage to me, where we must go beyond the text as it stood when commented on by the Sankarânanda. In the beginning of the fourth adhyâya all MSS. of the text read savasan, and this is the reading which the commentator seems anxious to explain, though not very successfully. I thought that possibly the commentator might have had before him the reading savasan, or so 'vasan, but both would be very unusual. Professor Cowell in his Various Readings, p. xii, conjectured samvasan, which would be liable to the same objection. He now, however, informs me that, as B. has samtvan, and C. satvan, he believes the original text to have been Satvan-Matsyeshu. This seems to me quite convincing, and is borne out by the reading of the Berlin MS., so far as it can be made out from Professor Weber's essay on the Upanishads, Indische Studien I, p.419. I see that Boehtlingk and Roth in their Sanskrit Dictionary, s.v. satvat, suggest the same emendation.

The more we study the nature of Sanskrit MSS., the more, I believe, we shall feel convinced that their proper arrangement is one by locality rather than by time. I have frequently dwelt on this subject in the introductions to the successive volumes of my edition of the Rig-veda and its commentary by Sâyanâkârya, and my convictions on this point have become stronger ever since. A MS., however modern, from the south of India or from the north, is more important as a check on the textus receptus of any Sanskrit work, as prevalent in Bengal or Bombay, than ever so many MSS., even if of greater antiquity, from the same locality. When therefore I was informed by my friend Dr. Bühler that he had discovered in Kashmir a MS. of the Aitareya-upanishad, I certainly expected some real help from such a treasure. The MS. is described by its discoverer in the last number of the journal of the Bombay Asiatic Society, p.34^[40], and has since been sent to me by the Indian Government. It is written on birch bark (bhûrga), and in the alphabet commonly called Sâradâ. The leaves are very much injured on the margin and it is almost impossible to handle them without some injury. In many places the bark has shrunk, probably on being moistened, and the letters have become illegible. Apart from these drawbacks, there remain the difficulties inherent in the Sâradâ alphabet which, owing to its numerous combinations, is extremely difficult to read, and very trying to eyes which are growing weak. However, I collated the Upanishad from the Aitareya-âranyaka, which turned out to be the last portion only, viz. the Samhitâ-upanishad (Ait. Âr. 111, 1-2), or, as it is called here, Samhitâranya, and I am sorry to say my expectations have been disappointed. The MS. shows certain graphic peculiarities which Dr. Bühler has pointed out. It is particularly careful in the use of the sibilants, replacing the Visarga by sibilants, writing s + s and s + s instead of h + s and h + s; distinguishing also the Gihvâmûlîya and Upadhmanîya. If therefore the MS. writes antastha, we may be sure that it really meant to write so, and not antahstha, or, as it would have written, antasstha. It shows equal care in the use of the nasals, and generally carries on the sandhi between different paragraphs. Here and there I met with better readings than those given in Rajendralal Mitra's edition, but in most cases the commentary would have been sufficient to restore the right reading. A few various readings, which seemed to deserve being mentioned, will be found in the notes. The MS., though carefully written, is not free from the ordinary blunders. At first one feels inclined to attribute some importance to every peculiarity of a new MS., but very soon one finds out that what seems peculiar, is in reality carelessness. Thus Ait. Âr. III, I, 5, 2, the Kashmir MS. has pûrvam aksharam rûpam, instead of what alone can be right, pûrvarûpam. Instead of pragayâ pasubhih it writes repeatedly pragaya pasubhih, which is impossible. In III, 2, 2, it leaves out again and again manomaya between khandomaya and vânmaya; but that this is a mere accident we learn later on, where in the same sentence manomayo, is found in its right place. Such cases reduce this MS. to its proper level, and make us look with suspicion on any accidental variations, such as I have noticed in my translation.

The additional paragraph, noticed by Dr. Bühler, is very indistinct, and contains, so far as I am able to find out, sânti verses only.

I have no doubt that the discovery of new MSS. of the Upanishads and their commentaries will throw new light on the very numerous difficulties with which a translator of the Upanishads, particularly in attempting a complete and faithful translation, has at present to grapple. Some of the difficulties, which existed thirty years ago, have been removed since by the general progress of Vedic scholarship, and by the editions of texts and commentaries and translations of Upanishads, many of which were known at that time in manuscript only. But I fully agree with M. Regnaud as to the difficultés considérables que les meilleures traductions laissent subsister, and which can be solved only by a continued study of the Upanishads, the Âranyakas, the Brâhmanas, and the Vedânta-sûtras.