Philological Criticism and Critical Editions of the Upanisads
Since the nineteenth century there has been among western scholars a pervasive mistrust of ancient Indian interpreters and commentators, especially the much-maligned Sam kara, as reliable guides to understanding ancient Indian texts. Early scholars were confident – to modern eyes, overconfident – of their ability to uncover “original” meanings through philological acumen unmediated by native gloss or comment. The observation of Dwight Whitney (1890: 407) in his review of Otto Bohtlingk’s (1889a,b) editions and translations of the two major Upanis. ads, the Br. had aran. yaka and the Ch andogya, typifies this attitude: “And the translation is of that character which I pointed out in a paper in this Journal1 some years ago as most to be desired – namely, simply a Sanskrit scholar’s version, made from the text itself, and not from the native comment, and aiming to represent just what the treatises themselves say, as interpreted by the known usages of the language.” Although due to theoretical advances in many fields, we are today, on the one hand, less confident of our ability to recover “original” meanings of ancient documents and, on the other, more aware of the importance of the history of the reception, understanding, and interpretation of texts within the native traditions, the bias against commentators persists. To some degree this bias is justified, not because of some perversity or ignorance on the part of the commentators, but because their goals were different from those of modern scholars; they were primarily theologians and apologists struggling to discover theological truths in their authoritative scriptures. To criticize them for what they did not set out to do is misplaced,2 but they may not always be suitable guides for the work of the textual scholar. The distrust of the commentators’ interpretations, however, spilled over into doubts about the reliability of the textual transmission mediated by these commentators and more broadly into a mistrust of the scribal tradition as such. This mistrust is most evident in the case of the Upanis. ads, a group of texts that came under close scholarly scrutiny both because of their centrality in later Indian theological discourse and because of the perceived philosophical importance of their message. To restore these texts to their presumed pristine state prior to the corrupt ing intervention of scribes and commentators, European Sanskritists of the nineteenth and early twentieth centuries undertook to create “critical editions” of several Upanis. ads, the most famous of which were Bohtlingk’s editions of the Br . had aran. yaka and the Ch andogya. Strange though it may seem, none of these so-called “critical editions”3 used manuscript material; even where manuscript differences are noted, the editions were not based on a careful sifting and collation of all the available manuscript evidence and the application of recognized editorial principles. There is no evidence that even a thorough search for manuscripts was ever undertaken. Indeed, even modern “critical editions” of the Upanis. ads, such as those of Limaye and Vadekar (1958), Frenz (1968–69), Maue (1976), P erez-Coffie (1994), and Oberlies (1995), are not based on a methodical search for and collation of manuscripts. Most are based on comparisons of previously printed editions, often perpetuating printing errors and conjectural readings of these earlier versions. Frenz’s edition of the Kaus. ıtaki Upanis. ad, for example, uses only a single manuscript conveniently located in Tubingen; all other “variants” are derived from previously printed editions. Maue’s (1976) edition of the first chapter of the Br. had aran. yaka Upanis. ad (K an. va recension) has greater claim to be a critical edition. He uses six manuscripts (three from India and three from Europe) but seven printed editions; the limited use of manuscript material may be explained in part as due to his use of only manuscripts containing accentuation. P erez-Coffie (1994) follows Maue closely in his “critical edition” of the second chapter of the Br. had aran. yaka (K an. va recension) and adds only a single manuscript to Maue’s six. In both these editions, however, there is no evidence of a thorough search for manuscripts especially in India. The “critical” in many of these so-called “critical editions”, especially the older ones of Bohtlingk (1889a,b), Hertel (1924), and Hauschild (1927), consists principally in the application of philological acumen to editorial reconstruction. Philological conjectures are an important part of scholarly investigation of texts. The question I want to address in this paper, however, is whether it is legitimate to incorporate such conjectures into the very body of the edited text. I will show that such emendations, especially when they are not clearly demarcated as conjectures, create a new and unfaithful textual transmission that often misleads later scholars and stifles scholarly debate about the received texts. As Salomon (1991: 48) says, “Whether a true critical edition would clarify the textual and linguistic questions about this [i.e., Pra sna] Upanis. ad (and, again, other Upanis. ads as well), or whether such an edition is even feasible, remains to be seen.” Yet, the feasibility of a critical edition cannot be determined unless a manuscript search and collation is undertaken.4 If after such a search one finds only a few variants of significance, that itself will open important and interesting questions regarding the textual history and the transmission of the Upanis. ads.
Bohtlingk’s 1889 editions and translations of the Br . had aran. yaka and the Ch andogya Upanis. ads were a landmark in Upanis. adic scholarship and exerted considerable influence on later scholars. He followed an unconventional – the less charitable but more accurate term may be outrageous – editorial practice, however, especially in the Ch andogya. He gave within the body of the edited text all his conjectural readings and philological improvements to the received text. The corresponding readings in the received text, on the other hand, he placed either in footnotes at the bottom of the page or, even more problematically, in endnotes wedged between the edition and the translation.5 The text itself contains no mechanism to warn the reader that the editor has emended the received text. This editorial practice is especially problematic in the case of wholesale changes that Bohtlingk made to certain words. Thus, for example, in the Ch andogya he changes aitadatmyam to etadatmakam ; saumya to somya; and adhidaivatam to adhidevatam throughout the document without any marker or note to indicate that these are his own conjectures. Even the philological reasons offered in support of these wholesale changes are dubious, as Whitney (1890: 412) observes: “to say that such forms ‘make their first appearance in the epics’ sounds curious; it is equivalent to saying that they are not to be retained in the Upanishads because they do not occur in the Upanishads”. Bohtlingk’s treatment of iti (marking quotations and direct speech) in the Ch andogya Upanis. ad is also idiosyncratic. Sometimes he drops it even though it is found in the received text, and at other times he inserts it where it is absent in the received text. The reason for this practice appears to be the editor’s own judgment as to whether a particular passage calls for a concluding iti. These and similar editorial practices caught the eye of Whitney (1890: 409): “Least of all to be approved, perhaps, is the tampering with the traditional text ::: without any note to inform the reader of the change.” Although such tampering reflects an arrogantly imperious attitude toward the text, at least these changes for the most part do not seriously affect its meaning. In other instances, however, Bohtlingk’s tampering not only mutilated the text, but frequently also misled later scholars who used his edition. It is not possible, nor is it necessary, to give an exhaustive list of such tampering. I will cite only a few egregious examples. Thus, for instance, Ch andogya 1.4.1 reads: om ity etad aks. aram udg ıtham upas ıta (“OM. – one should venerate the High Chant as this syllable”). Bohtlingk’s edition drops udg ıtham and explains the omission not in a footnote, where it would be readily noticed by the reader, but in an endnote, saying that manuscripts and previous editions insert udg ıtham possibly under the influence of Ch andogya 1.1.1. Senart (1930), who followed Bohtlingk’s edition closely, reproduces B ohtlingk’s emended version. He probably did not read the endnote and accepted it as the received reading, apparently unaware that Bohtlingk had emended the text here. So in his translation, Senart is forced to add a parenthetical remark: “Il faut connaitre que om est la syllable (par excellence ou l’«imp erissable»)”, reflecting his discomfort that there was no qualification of aks. ara. Senart, who had already written an article on the topic (Senart, 1909), knew that the verb upas - in the vedic literature does not simply mean “to venerate” but indicates equivalence, i.e., that something is equivalent to something else; indeed, in the Upanis. ads this is the term of choice to indicate various correspondences existing between the bodily, ritual, and cosmic spheres. Accordingly, at Ch andogya 1.1.1 Senart translates the identical phrase as “Il faut savoir que la syllable om est l’udg ıtha” and adds the footnote: “C’est- a-dire: est equivalent a l’udg ıtha lui-m eme.” The absence of the term at Ch andogya 1.4.1 in Bohtlingk’s edition, which Senart accepted as the received text, thus misled him and caused a somewhat contorted translation. The insertion of conjectural readings into the body of the edited text becomes even more troubling when the editor later happens to change his mind. Bohtlingk recorded these changes of heart in a series of articles written in the somewhat obscure journal Berichte uber die Verhandlungen die koniglich s achsischen Gesellschaft der Wissenschaften zu Leipzig , Philologisch-historische Classe, articles that were unavailable to or unnoticed by later scholars, who continued to rely on his editions. A couple of examples will suffice. The traditional reading of Ch andogya 8.15.1 is: ac aryakul ad vedam adh ıtya yathavidh anam . guroh. karmati sesena abhisamavr tya – “Having returned from the teacher’s house, where he learned the Veda in the prescribed manner during his free time after his daily tasks for the teacher”.
Bohtlingk (1889b) found the expression karmati ses ena unacceptable and changed it to karma kr. tvavi ses ena. 6 Senart (1930) is troubled by the reading of the received text but uncertain of Bohtlingk’s emendation. He ends up keeping kr. tva and putting atises . en. a within brackets, noting: “Le texte est ici troubl e. Je ne puis consid erer comme admissible l’interpr etation de yathavidh anam . guroh. karmati ses . en. a que sugg ere C¸ ankara [et d’apr _ es laquelle le disciple n’ etudierait qu’ a ses instants de loisir]. Il faudrait au moins que toute la locution fut ramass ^ ee en seul compos e, et cela m^eme serait difficilement admissible. J’ai donc traduit avec l’addition de kr. tva , introduit par Bohtlingk; mais je ne saurais dire que l’alt eration suppos ee d’une lecture si facile me semble plausible. En tout cas cette hypoth ese implique la correction de atic¸es. en. a qui ne donne aucun sens en avic¸es. en. a; mais cet avic¸es. en. a est lui-m^eme bien faible et bien superflu, et devrait en tout cas beaucoup plutot porter ^ sur la suite: bref, je ne puis rien faire ni de atic¸es. en. a ni de avic¸es. en. a abhisamavr . tya” (Senart, 1930: 121). Senart’s observation about how such a simple reading as karma kr. tva could be changed into such a difficult one points to a principle problem with many of Bohtlingk’s emendations: he rejects difficult readings in favor of easy ones, violating the cardinal principle of lectio difficilior, the bedrock of textual criticism that more difficult readings (in terms of grammar, meaning, orthography, etc.) are to be preferred over easier ones. What Senart did not know, however, is that, in an article published eight years (Bohtlingk, 1897a: 92) after his edition of the Ch andogya, Bohtlingk had changed his mind and returned to the traditional reading. The reason for this turn around is probably because Bohtlingk found the same reading karmati ses . en. a also in the Gautama Dharmasutra (3.6). 7 Although Stenzler’s edition of Gautama appeared in 1876, thirteen years before Bohtlingk’s edition of the Ch andogya Upanis. ad, he probably saw it only after its preparation, indicating once again the perils of philological hubris. Likewise, in the Ch andogya passage (3.11.6) etad eva tato bhuya iti, Bohtlingk drops iti in his edition (1889b) but wants to retain it in a later article (1897a: 82); iti is clearly required here because the phrase bracketed by it gives the reasons for the previous statement: a man should not impart the teaching to anyone even if he is given the whole world, “because that (formulation) is far greater than (all) that”. Senart (1930), once again, was unaware of Bohtlingk’s change of mind and follows his edition in omitting iti. The verbal form pradhmay ıta occurring in Ch andogya 6.14.1, which I have translated “he would drift about” (Olivelle, 1996: 155), did not please Bohtlingk, who gives the conjectural reading pradhaveta in his edition (Bohtlingk, 1889b; rejected by Whitney, 1890: 413, who prefers the traditional reading) but proposes something totally different, prahvay ıta, in a later study (Bohtlingk, 1897b: 128). 8 A conjecture first suggested by Deussen (1897) took a somewhat different route to enter the modern textual transmission as a variant reading. The traditional reading of Ch andogya 4.9.2 is bhagavam . s tv eva me kame br uy at – “But, if it pleases you, sir, you should teach it to me yourself”. Bohtlingk (1889b) found kame (locative singular) to be unsuitable and Sam . kara’s explanation unacceptable9 and presented a clever but suspect emendation by dividing the words differently:10 bhagavam . s tv evam eko me bruy at (“Jedoch konnte der Erhabene, aber auch nur er allein, mir es auf diese Weise verkunden”). Deussen (1897: 124, n. 1), likewise, was dissatisfied with Sam . kara’s explanation but could not accept Bohtlingk’s emendation. So, quite appropriately in a footnote, Deussen suggested emending kame to kamam . , 11 translating the term as “bitte” (please). Senart (1930: 53) also rejected Bohtlingk’s emendation, but could not decide between Deussen’s suggestion and the traditional reading. In 1958 Limaye and Vadekar published an edition of eighteen Upanis. ads that has become the standard edition used by scholars today.12 Within the body of the text, they give the traditional reading kame , but they add a footnote giving kamam not as Deussen’s conjecture but as a variant reading. Now, Limaye and Vadekar’s edition is not based on manuscript evidence;13 their variants are gleaned from published material and emendations suggested by scholars. Indeed, immediately after they note kamam as a variant, they give Bohtlingk’s conjecture. It appears to me very probable that they got kamam from Deussen’s suggestion. Now, in 1980 Deussen’s German translation of the Upanis. ads was retranslated into English by Bedekar and Palsule. They reproduce (1980: 126) Deussen’s footnote about kamam , to which they add a further note of their own: “In the later edition [sic], e.g. in the Eighteen Upanis. ads edited by R. D. Vadekar and V. P. Limaye the reading is ‘kamam ’.” This completes an interesting and instructive circle of modern textual transmission, where a scholarly suggestion in a footnote becomes finally the “accepted” reading. The verses contained in the older prose Upanis. ads and in a special way the later metrical Upanis. ads pose a different type of problem for the editor and the scholar when the traditional text violates the meter. It is a prima facie rule of editing that the original text could not have violated metrical rules, and therefore editors regularly adopt one manuscript reading and reject another on the basis of meter (metri causa). This principle is also frequently invoked in conjectural emendations to the traditional reading of a verse, emendations not based on manuscript evidence. We see this principle applied repeatedly in the “critical editions” of the Mun. d. aka and Svet a svatara Upanis. ads by Hertel (1924) and Hauschild (1927), respectively. Yet, it is well to remember the salutary warning of Max Muller (1879: lxxii): “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 there was any real necessity for correction.” The problems inherent in restoring the meter through conjecture are exemplified by the variety of scholarly opinion on how to restore the meter of the third pada of Kat. ha Upanis. ad 1.19: etam agnim. tavaiva pravaks. yanti janasah . – “People will proclaim this your very own fire”, a pada that contains fourteen syllables instead of the required eleven. Muller (1884), B ohtlingk (1890: 134), and Charpentier (1928–29) take tavaiva to be an interpolation distorting the meter; but this emendation is rejected by Garbe and Alsdorf (1950: 627). Alsdorf, followed by Rau (1971: 173), drops agnim. and eva. Likewise, the second half-verse of Kat. ha Upanis. ad 2.11 is metrically incorrect: stomamahad urugayam . pratis. t. ham . dr. s. t. va dhr . tya dh ıro naciketo ’tyasraks . ıh. – “Great and widespread praise is the foundation. These you have seen, wise Naciketas, and having seen, firmly rejected.” Bohtlingk (1890: 142) proposes deleting dr. s. t. va to restore the meter, whereas Alsdorf (1950: 628) thinks that Bohtlingk has made the “wrong choice” and suggests deleting the next word, dhr. tya . The problem is not such scholarly disagreement, which is fruitful; but when a particular conjecture is reproduced in the edited text, especially if it is an influential edition, then the conjecture becomes accepted passively by later scholarship and preempts that very scholarly give and take. An issue peculiar to verses has to be weighed in any philological reconstruction of broken meters, and that is pronunciation in terms of both sandhi and pr akr. tic ways of pronouncing words.14 Alsdorf (1950: 623) has shown that numerous irregular meters can be regularized by simply dropping external sandhi. The first two padas of Kat. ha Upanis. ad 1.7, for example, read vaisv anarah . pravisatyatithir br ahman . o gr. han (“A Brahmin guest enters a house as the fire in all men”), with seven syllables in the first pada , instead of the regular eight. This can be remedied by dissolving the sandhi between the two padas and reading pravisati atithir . This operation is not a textual emendation because strict sandhi rules are applied consistently mostly in printed texts; most manuscripts are quite free with their sandhi, and when speaking or chanting sandhi is broken whenever there is a pause. Then there are the pr akr. tic pronunciations of certain words that would restore the meter. Some common words listed by Alsdorf (1950) include iva, iti, bhavati, which were pronounced as va, ti, and bhoti. Thus, Kat. ha 1.9 sasyam iva martyah. pacyate sasyam ivaj ayate punah . (“A mortal man ripens like grain, and like grain is born again”) has an extra syllable in both padas . This can be eliminated by using va for iva (sasyam. va). Likewise, Kat. ha 3.5 yas tv avijn~anav an bhavati (“When a man lacks understanding”) has an extra syllable, which can be eliminated by reading bhavati as bhoti.
I have already made reference to the editorial principle of lectio difficilior. Simply stated, this principle holds that more difficult readings (in terms of grammar, meaning, orthography, etc.) are to be preferred over easier readings. This is based on the reasoning that scribes and readers down the centuries were more likely to replace readings they found difficult or did not understand with easier readings; the inverse, namely the replacement of easy readings with more difficult ones, is much less likely. This principle holds especially in the case of “ungrammatical” forms. After all, Indian pandits probably knew their Pan. inian grammar a bit better than modern western scholars. It would have been a simple matter for them to replace the suspect readings with the correct forms. But we find repeatedly that they did not, and this is a clear indication that ancient Indian commentators did not take liberties with their received Upanis. adic texts; they were, by and large, faithful transmitters. The importance of the principle of lectio difficilior and the reliability of the texts transmitted by Sam . kara have been highlighted by the recent work of Richard Salomon (1981, 1991). He has studied the linguistic peculiarities of two late Upanis. ads, the Mun. d. aka and the Pra sna, and seen in them the vestiges of a Ks. atriya dialect of Sanskrit. Arguing for the superiority of Sam . kara’s15 version of the Pra sna Upanis. ad, Salomon (1991: 49) remarks: “The example of pra/pratis . t. hante, 16 just cited, for example, clearly points toward a lectio facilior in the Rangar _ am anuja text,17 suggesting the superiority of the Sa nkara version. And indeed, _ it is precisely the large number of nonstandard forms in the latter text which is the strongest argument for its integrity. For as a matter of general principle, such nonstandard forms – especially when they can be corroborated by parallels in other texts, as is most often the case – should be considered a priori as stronger readings than the ‘grammatically correct’ variants. ::: In other words, the fact that such unexpected forms as pratis . t. hante or avedis. am, which could have been so easily emended to the normal forms, are preserved in the vulgate says much both for its reliability and for their originality.” When European scholars change an unusual form to its “correct” grammatical form, we lose much of the dialectical variations evident in old Sanskrit. A case in point is Mun. d. aka 3.2.2, which contains the non-standard instrumental plural kamabhih . (“with desires”). In his edition of this Upanis. ad, Hertel (1924) emends the term to the regular karmabhih. (“with actions”) thereby not only changing the meaning radically but also obliterating this significant dialectical variant (see Salomon, 1981: 94). The replacement of difficult readings with more standard conjectures also preempts further scholarly meditation on those difficult readings. At Mun. d. aka 3.1.4, for example, we have the expression vijanan vidv an bhavate nativ ad ı, translated by Hume (1931) as “Understanding this, one becomes a knower. There is no superior speaker.” Bohtlingk (1901: 8) emends the second half to bhavati tenativ ad ı, following the reading of Ch andogya 7.15.4. Following the traditional text, however, and without a need for emendation, Rau (1965) has pointed out a far superior reading: bhava tenativ ad ı, 18 according to which the phrase can be translated: “Be a man who perceives, who knows this, and thereby a man who out-talks” (Olivelle, 1996: 274). Svet a svatara Upanis. ad 6.11 contains the word ceta , which Sam . kara19 and, following him, nearly all modern scholars have taken to be an agent noun derived from the root cit-, “to perceive, to observe”. If that were the case, the standard nominal derivative should have been cetta , and Hauschild (1927),20 in his edition of this Upanis. ad, emends ceta to the “grammatically correct” cetta . It is important to note that even though Sam . kara explains ceta to be an agent noun of cit-, nevertheless he preserved the non-standard form, whereas Hauschild changes the form in his edition. Rau (1964), however, relying on the traditional text, sees ceta as an agent noun derived not from cit- but from the root ci- “to avenge” and offers the very plausible translation “avenger”, an interpretation that fits the context better and that would have been foreclosed if Sam . kara, like Hauschild, had emended his text.
The Upanis. adic texts as traditionally handed down also preserve expressions that make no linguistic sense; they are “nonsense” phrases. One example is tajjalan , which occurs in Ch andogya 3.14.1: sarvam. khalv idam. brahma tajjalan iti s anta up as ıta. 21 Sam. kara found the expression jalan problematic. Instead of emending it, however, he gives an explanation that takes it to be an acronym. According to his interpretation, the term means that everything proceeds from (ja = jan, “to be born”), dissolves into (la = l ı, “to dissolve”), and lives by (an, “to breathe”) Brahman. A modern scholar may not agree with Sam . kara’s overly theological explanation, but at least the theologian did not deliberately change the text. In contrast, Bohtlingk (1889b) emends the expression in his critically edited text to read taj jan an ıti s anta upas ıta, “Der zur Ruhe Gelangte verehre es als das, was er kennen mochte.” He does not explain, however, how such a common verb form as jan ani (first person singular present subjunctive) could have become garbled into jalan ; again Bohtlingk has rejected the lectio difficilior in favor of the easier reading.22 Even when commentators encounter an ungrammatical form, they usually preserve it. Generally they explain the irregularity as a vedic form (chandasa ); it was believed that certain Sanskrit forms found in the vedic texts were unusual or irregular and did not conform to Pan. inian grammar.23 This belief, I think, and the innate conservatism (in the best sense of the word) of Brahmin scholars especially with regard to vedic texts prevented them from changing the received text deliberately. They did not, however, naively take every aberrant form to be simply vedic; they also suspected that some forms may be the result of textual corruption. To give but one example, Ch andogya 1.12.3 reads: tan hov acehaiva m a pr atar upasam ıyateti , “He told them: ‘Come and meet me at this very spot in the morning’.” The long “ ı ” of upasam ıyata is irregular; the standard form is upasamiyata . Sam kara notes this irregularity and explains it: dairghyam. chandasam ::: pramadap at . ho va , “The long ( ı) is either a vedic form or an erroneous reading [i.e., a reading caused by carelessness].” In contrast to S.amkara’s careful attention to manuscript evidence,24 Bohtlingk (1989b) summarily changes this word in his edition to the standard upasamiyata
Western, especially European, philologists, we have seen, were often less faithful transmitters of Upanis. adic texts than the Indian scribes and commentators they so often criticized. Native commentators and theologians did not, as often assumed, carelessly or deliberately change the received texts to suit their doctrinal or grammatical tastes. Indeed, it is the modern philologists who are often guilty of changing the texts to suit preconceived notions of correctness, whether grammatical or otherwise
Lest I be misunderstood, my criticism is not against philology as such but against the substitution of arm-chair philology for the tedious but important examination of manuscript evidence and the resultant improper application of philology to mutilate texts. Philology is the indispensable bedrock of any serious study of texts for any purpose, especially for the purpose of historical reconstruction. But philology is not a substitute for critical edition of texts based on manuscript evidence; it is not possible to reconstruct texts by brain power or philological training alone. We have seen confusion reign especially when scholarly conjectures are introduced into the body of edited texts, especially those calling themselves “critical editions”, often with little warning to readers that they are dealing with conjectures and not received readings. In ancient and medieval India texts were transmitted and preserved by copying onto manuscripts and by memorization, which is a lost art today. The copying of texts introduced errors through negligence or misreading the exemplars and sometimes through deliberate emendations. The existence of numerous manuscripts of a single work, however, permits the careful editor to detect such errors and emendations. Error or emendation was limited to a single manuscript and to others for which it served as exemplar. In the modern age the printed book (and today its electronic counterpart) is the unique medium of textual transmission and preservation. The printed book comes in thousands of identical specimens. An error or emendation introduced into a printed edition, unlike its manuscript counterpart, is reproduced in every single specimen. Given the expense of publishing, moreover, once an ancient text has been published, it is unlikely that a new edition would be forthcoming soon or ever. The responsibility, therefore, of a modern editor to ensure the faithfulness of transmission is a thousand times greater than that of a scribe. Philological hubris, I have tried to demonstrate in this paper, has made many modern editions of the Upanisads unreliable. Modern editors may take note of this set of verses often appended by Indian scribes at the end of their manuscripts.
bhagnaprstikat. igr ıvah. stabdhadrstir adhomukhah. /
kastena likhitam. grantham. yatnena pratipalayet
With great trouble I have written this book, My head bent low, with unwavering eyes, I have broken my back, my hips and neck; So be diligent and take care of it.
yadr . sam . pustakam. dr. s. t. va t adr . sam . likhitam. maya / yadi suddham a suddham . va mama dos . o no vidyate //
I copied exactly What I saw in the book; Whether it’s right or wrong, I am not to be blamed
Bohtlingk’s 1889 editions and translations of the Br . had aran. yaka and the Ch andogya Upanis. ads were a landmark in Upanis. adic scholarship and exerted considerable influence on later scholars. He followed an unconventional – the less charitable but more accurate term may be outrageous – editorial practice, however, especially in the Ch andogya. He gave within the body of the edited text all his conjectural readings and philological improvements to the received text. The corresponding readings in the received text, on the other hand, he placed either in footnotes at the bottom of the page or, even more problematically, in endnotes wedged between the edition and the translation.5 The text itself contains no mechanism to warn the reader that the editor has emended the received text. This editorial practice is especially problematic in the case of wholesale changes that Bohtlingk made to certain words. Thus, for example, in the Ch andogya he changes aitadatmyam to etadatmakam ; saumya to somya; and adhidaivatam to adhidevatam throughout the document without any marker or note to indicate that these are his own conjectures. Even the philological reasons offered in support of these wholesale changes are dubious, as Whitney (1890: 412) observes: “to say that such forms ‘make their first appearance in the epics’ sounds curious; it is equivalent to saying that they are not to be retained in the Upanishads because they do not occur in the Upanishads”. Bohtlingk’s treatment of iti (marking quotations and direct speech) in the Ch andogya Upanis. ad is also idiosyncratic. Sometimes he drops it even though it is found in the received text, and at other times he inserts it where it is absent in the received text. The reason for this practice appears to be the editor’s own judgment as to whether a particular passage calls for a concluding iti. These and similar editorial practices caught the eye of Whitney (1890: 409): “Least of all to be approved, perhaps, is the tampering with the traditional text ::: without any note to inform the reader of the change.” Although such tampering reflects an arrogantly imperious attitude toward the text, at least these changes for the most part do not seriously affect its meaning. In other instances, however, Bohtlingk’s tampering not only mutilated the text, but frequently also misled later scholars who used his edition. It is not possible, nor is it necessary, to give an exhaustive list of such tampering. I will cite only a few egregious examples. Thus, for instance, Ch andogya 1.4.1 reads: om ity etad aks. aram udg ıtham upas ıta (“OM. – one should venerate the High Chant as this syllable”). Bohtlingk’s edition drops udg ıtham and explains the omission not in a footnote, where it would be readily noticed by the reader, but in an endnote, saying that manuscripts and previous editions insert udg ıtham possibly under the influence of Ch andogya 1.1.1. Senart (1930), who followed Bohtlingk’s edition closely, reproduces B ohtlingk’s emended version. He probably did not read the endnote and accepted it as the received reading, apparently unaware that Bohtlingk had emended the text here. So in his translation, Senart is forced to add a parenthetical remark: “Il faut connaitre que om est la syllable (par excellence ou l’«imp erissable»)”, reflecting his discomfort that there was no qualification of aks. ara. Senart, who had already written an article on the topic (Senart, 1909), knew that the verb upas - in the vedic literature does not simply mean “to venerate” but indicates equivalence, i.e., that something is equivalent to something else; indeed, in the Upanis. ads this is the term of choice to indicate various correspondences existing between the bodily, ritual, and cosmic spheres. Accordingly, at Ch andogya 1.1.1 Senart translates the identical phrase as “Il faut savoir que la syllable om est l’udg ıtha” and adds the footnote: “C’est- a-dire: est equivalent a l’udg ıtha lui-m eme.” The absence of the term at Ch andogya 1.4.1 in Bohtlingk’s edition, which Senart accepted as the received text, thus misled him and caused a somewhat contorted translation. The insertion of conjectural readings into the body of the edited text becomes even more troubling when the editor later happens to change his mind. Bohtlingk recorded these changes of heart in a series of articles written in the somewhat obscure journal Berichte uber die Verhandlungen die koniglich s achsischen Gesellschaft der Wissenschaften zu Leipzig , Philologisch-historische Classe, articles that were unavailable to or unnoticed by later scholars, who continued to rely on his editions. A couple of examples will suffice. The traditional reading of Ch andogya 8.15.1 is: ac aryakul ad vedam adh ıtya yathavidh anam . guroh. karmati sesena abhisamavr tya – “Having returned from the teacher’s house, where he learned the Veda in the prescribed manner during his free time after his daily tasks for the teacher”.
Bohtlingk (1889b) found the expression karmati ses ena unacceptable and changed it to karma kr. tvavi ses ena. 6 Senart (1930) is troubled by the reading of the received text but uncertain of Bohtlingk’s emendation. He ends up keeping kr. tva and putting atises . en. a within brackets, noting: “Le texte est ici troubl e. Je ne puis consid erer comme admissible l’interpr etation de yathavidh anam . guroh. karmati ses . en. a que sugg ere C¸ ankara [et d’apr _ es laquelle le disciple n’ etudierait qu’ a ses instants de loisir]. Il faudrait au moins que toute la locution fut ramass ^ ee en seul compos e, et cela m^eme serait difficilement admissible. J’ai donc traduit avec l’addition de kr. tva , introduit par Bohtlingk; mais je ne saurais dire que l’alt eration suppos ee d’une lecture si facile me semble plausible. En tout cas cette hypoth ese implique la correction de atic¸es. en. a qui ne donne aucun sens en avic¸es. en. a; mais cet avic¸es. en. a est lui-m^eme bien faible et bien superflu, et devrait en tout cas beaucoup plutot porter ^ sur la suite: bref, je ne puis rien faire ni de atic¸es. en. a ni de avic¸es. en. a abhisamavr . tya” (Senart, 1930: 121). Senart’s observation about how such a simple reading as karma kr. tva could be changed into such a difficult one points to a principle problem with many of Bohtlingk’s emendations: he rejects difficult readings in favor of easy ones, violating the cardinal principle of lectio difficilior, the bedrock of textual criticism that more difficult readings (in terms of grammar, meaning, orthography, etc.) are to be preferred over easier ones. What Senart did not know, however, is that, in an article published eight years (Bohtlingk, 1897a: 92) after his edition of the Ch andogya, Bohtlingk had changed his mind and returned to the traditional reading. The reason for this turn around is probably because Bohtlingk found the same reading karmati ses . en. a also in the Gautama Dharmasutra (3.6). 7 Although Stenzler’s edition of Gautama appeared in 1876, thirteen years before Bohtlingk’s edition of the Ch andogya Upanis. ad, he probably saw it only after its preparation, indicating once again the perils of philological hubris. Likewise, in the Ch andogya passage (3.11.6) etad eva tato bhuya iti, Bohtlingk drops iti in his edition (1889b) but wants to retain it in a later article (1897a: 82); iti is clearly required here because the phrase bracketed by it gives the reasons for the previous statement: a man should not impart the teaching to anyone even if he is given the whole world, “because that (formulation) is far greater than (all) that”. Senart (1930), once again, was unaware of Bohtlingk’s change of mind and follows his edition in omitting iti. The verbal form pradhmay ıta occurring in Ch andogya 6.14.1, which I have translated “he would drift about” (Olivelle, 1996: 155), did not please Bohtlingk, who gives the conjectural reading pradhaveta in his edition (Bohtlingk, 1889b; rejected by Whitney, 1890: 413, who prefers the traditional reading) but proposes something totally different, prahvay ıta, in a later study (Bohtlingk, 1897b: 128). 8 A conjecture first suggested by Deussen (1897) took a somewhat different route to enter the modern textual transmission as a variant reading. The traditional reading of Ch andogya 4.9.2 is bhagavam . s tv eva me kame br uy at – “But, if it pleases you, sir, you should teach it to me yourself”. Bohtlingk (1889b) found kame (locative singular) to be unsuitable and Sam . kara’s explanation unacceptable9 and presented a clever but suspect emendation by dividing the words differently:10 bhagavam . s tv evam eko me bruy at (“Jedoch konnte der Erhabene, aber auch nur er allein, mir es auf diese Weise verkunden”). Deussen (1897: 124, n. 1), likewise, was dissatisfied with Sam . kara’s explanation but could not accept Bohtlingk’s emendation. So, quite appropriately in a footnote, Deussen suggested emending kame to kamam . , 11 translating the term as “bitte” (please). Senart (1930: 53) also rejected Bohtlingk’s emendation, but could not decide between Deussen’s suggestion and the traditional reading. In 1958 Limaye and Vadekar published an edition of eighteen Upanis. ads that has become the standard edition used by scholars today.12 Within the body of the text, they give the traditional reading kame , but they add a footnote giving kamam not as Deussen’s conjecture but as a variant reading. Now, Limaye and Vadekar’s edition is not based on manuscript evidence;13 their variants are gleaned from published material and emendations suggested by scholars. Indeed, immediately after they note kamam as a variant, they give Bohtlingk’s conjecture. It appears to me very probable that they got kamam from Deussen’s suggestion. Now, in 1980 Deussen’s German translation of the Upanis. ads was retranslated into English by Bedekar and Palsule. They reproduce (1980: 126) Deussen’s footnote about kamam , to which they add a further note of their own: “In the later edition [sic], e.g. in the Eighteen Upanis. ads edited by R. D. Vadekar and V. P. Limaye the reading is ‘kamam ’.” This completes an interesting and instructive circle of modern textual transmission, where a scholarly suggestion in a footnote becomes finally the “accepted” reading. The verses contained in the older prose Upanis. ads and in a special way the later metrical Upanis. ads pose a different type of problem for the editor and the scholar when the traditional text violates the meter. It is a prima facie rule of editing that the original text could not have violated metrical rules, and therefore editors regularly adopt one manuscript reading and reject another on the basis of meter (metri causa). This principle is also frequently invoked in conjectural emendations to the traditional reading of a verse, emendations not based on manuscript evidence. We see this principle applied repeatedly in the “critical editions” of the Mun. d. aka and Svet a svatara Upanis. ads by Hertel (1924) and Hauschild (1927), respectively. Yet, it is well to remember the salutary warning of Max Muller (1879: lxxii): “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 there was any real necessity for correction.” The problems inherent in restoring the meter through conjecture are exemplified by the variety of scholarly opinion on how to restore the meter of the third pada of Kat. ha Upanis. ad 1.19: etam agnim. tavaiva pravaks. yanti janasah . – “People will proclaim this your very own fire”, a pada that contains fourteen syllables instead of the required eleven. Muller (1884), B ohtlingk (1890: 134), and Charpentier (1928–29) take tavaiva to be an interpolation distorting the meter; but this emendation is rejected by Garbe and Alsdorf (1950: 627). Alsdorf, followed by Rau (1971: 173), drops agnim. and eva. Likewise, the second half-verse of Kat. ha Upanis. ad 2.11 is metrically incorrect: stomamahad urugayam . pratis. t. ham . dr. s. t. va dhr . tya dh ıro naciketo ’tyasraks . ıh. – “Great and widespread praise is the foundation. These you have seen, wise Naciketas, and having seen, firmly rejected.” Bohtlingk (1890: 142) proposes deleting dr. s. t. va to restore the meter, whereas Alsdorf (1950: 628) thinks that Bohtlingk has made the “wrong choice” and suggests deleting the next word, dhr. tya . The problem is not such scholarly disagreement, which is fruitful; but when a particular conjecture is reproduced in the edited text, especially if it is an influential edition, then the conjecture becomes accepted passively by later scholarship and preempts that very scholarly give and take. An issue peculiar to verses has to be weighed in any philological reconstruction of broken meters, and that is pronunciation in terms of both sandhi and pr akr. tic ways of pronouncing words.14 Alsdorf (1950: 623) has shown that numerous irregular meters can be regularized by simply dropping external sandhi. The first two padas of Kat. ha Upanis. ad 1.7, for example, read vaisv anarah . pravisatyatithir br ahman . o gr. han (“A Brahmin guest enters a house as the fire in all men”), with seven syllables in the first pada , instead of the regular eight. This can be remedied by dissolving the sandhi between the two padas and reading pravisati atithir . This operation is not a textual emendation because strict sandhi rules are applied consistently mostly in printed texts; most manuscripts are quite free with their sandhi, and when speaking or chanting sandhi is broken whenever there is a pause. Then there are the pr akr. tic pronunciations of certain words that would restore the meter. Some common words listed by Alsdorf (1950) include iva, iti, bhavati, which were pronounced as va, ti, and bhoti. Thus, Kat. ha 1.9 sasyam iva martyah. pacyate sasyam ivaj ayate punah . (“A mortal man ripens like grain, and like grain is born again”) has an extra syllable in both padas . This can be eliminated by using va for iva (sasyam. va). Likewise, Kat. ha 3.5 yas tv avijn~anav an bhavati (“When a man lacks understanding”) has an extra syllable, which can be eliminated by reading bhavati as bhoti.
I have already made reference to the editorial principle of lectio difficilior. Simply stated, this principle holds that more difficult readings (in terms of grammar, meaning, orthography, etc.) are to be preferred over easier readings. This is based on the reasoning that scribes and readers down the centuries were more likely to replace readings they found difficult or did not understand with easier readings; the inverse, namely the replacement of easy readings with more difficult ones, is much less likely. This principle holds especially in the case of “ungrammatical” forms. After all, Indian pandits probably knew their Pan. inian grammar a bit better than modern western scholars. It would have been a simple matter for them to replace the suspect readings with the correct forms. But we find repeatedly that they did not, and this is a clear indication that ancient Indian commentators did not take liberties with their received Upanis. adic texts; they were, by and large, faithful transmitters. The importance of the principle of lectio difficilior and the reliability of the texts transmitted by Sam . kara have been highlighted by the recent work of Richard Salomon (1981, 1991). He has studied the linguistic peculiarities of two late Upanis. ads, the Mun. d. aka and the Pra sna, and seen in them the vestiges of a Ks. atriya dialect of Sanskrit. Arguing for the superiority of Sam . kara’s15 version of the Pra sna Upanis. ad, Salomon (1991: 49) remarks: “The example of pra/pratis . t. hante, 16 just cited, for example, clearly points toward a lectio facilior in the Rangar _ am anuja text,17 suggesting the superiority of the Sa nkara version. And indeed, _ it is precisely the large number of nonstandard forms in the latter text which is the strongest argument for its integrity. For as a matter of general principle, such nonstandard forms – especially when they can be corroborated by parallels in other texts, as is most often the case – should be considered a priori as stronger readings than the ‘grammatically correct’ variants. ::: In other words, the fact that such unexpected forms as pratis . t. hante or avedis. am, which could have been so easily emended to the normal forms, are preserved in the vulgate says much both for its reliability and for their originality.” When European scholars change an unusual form to its “correct” grammatical form, we lose much of the dialectical variations evident in old Sanskrit. A case in point is Mun. d. aka 3.2.2, which contains the non-standard instrumental plural kamabhih . (“with desires”). In his edition of this Upanis. ad, Hertel (1924) emends the term to the regular karmabhih. (“with actions”) thereby not only changing the meaning radically but also obliterating this significant dialectical variant (see Salomon, 1981: 94). The replacement of difficult readings with more standard conjectures also preempts further scholarly meditation on those difficult readings. At Mun. d. aka 3.1.4, for example, we have the expression vijanan vidv an bhavate nativ ad ı, translated by Hume (1931) as “Understanding this, one becomes a knower. There is no superior speaker.” Bohtlingk (1901: 8) emends the second half to bhavati tenativ ad ı, following the reading of Ch andogya 7.15.4. Following the traditional text, however, and without a need for emendation, Rau (1965) has pointed out a far superior reading: bhava tenativ ad ı, 18 according to which the phrase can be translated: “Be a man who perceives, who knows this, and thereby a man who out-talks” (Olivelle, 1996: 274). Svet a svatara Upanis. ad 6.11 contains the word ceta , which Sam . kara19 and, following him, nearly all modern scholars have taken to be an agent noun derived from the root cit-, “to perceive, to observe”. If that were the case, the standard nominal derivative should have been cetta , and Hauschild (1927),20 in his edition of this Upanis. ad, emends ceta to the “grammatically correct” cetta . It is important to note that even though Sam . kara explains ceta to be an agent noun of cit-, nevertheless he preserved the non-standard form, whereas Hauschild changes the form in his edition. Rau (1964), however, relying on the traditional text, sees ceta as an agent noun derived not from cit- but from the root ci- “to avenge” and offers the very plausible translation “avenger”, an interpretation that fits the context better and that would have been foreclosed if Sam . kara, like Hauschild, had emended his text.
The Upanis. adic texts as traditionally handed down also preserve expressions that make no linguistic sense; they are “nonsense” phrases. One example is tajjalan , which occurs in Ch andogya 3.14.1: sarvam. khalv idam. brahma tajjalan iti s anta up as ıta. 21 Sam. kara found the expression jalan problematic. Instead of emending it, however, he gives an explanation that takes it to be an acronym. According to his interpretation, the term means that everything proceeds from (ja = jan, “to be born”), dissolves into (la = l ı, “to dissolve”), and lives by (an, “to breathe”) Brahman. A modern scholar may not agree with Sam . kara’s overly theological explanation, but at least the theologian did not deliberately change the text. In contrast, Bohtlingk (1889b) emends the expression in his critically edited text to read taj jan an ıti s anta upas ıta, “Der zur Ruhe Gelangte verehre es als das, was er kennen mochte.” He does not explain, however, how such a common verb form as jan ani (first person singular present subjunctive) could have become garbled into jalan ; again Bohtlingk has rejected the lectio difficilior in favor of the easier reading.22 Even when commentators encounter an ungrammatical form, they usually preserve it. Generally they explain the irregularity as a vedic form (chandasa ); it was believed that certain Sanskrit forms found in the vedic texts were unusual or irregular and did not conform to Pan. inian grammar.23 This belief, I think, and the innate conservatism (in the best sense of the word) of Brahmin scholars especially with regard to vedic texts prevented them from changing the received text deliberately. They did not, however, naively take every aberrant form to be simply vedic; they also suspected that some forms may be the result of textual corruption. To give but one example, Ch andogya 1.12.3 reads: tan hov acehaiva m a pr atar upasam ıyateti , “He told them: ‘Come and meet me at this very spot in the morning’.” The long “ ı ” of upasam ıyata is irregular; the standard form is upasamiyata . Sam kara notes this irregularity and explains it: dairghyam. chandasam ::: pramadap at . ho va , “The long ( ı) is either a vedic form or an erroneous reading [i.e., a reading caused by carelessness].” In contrast to S.amkara’s careful attention to manuscript evidence,24 Bohtlingk (1989b) summarily changes this word in his edition to the standard upasamiyata
Western, especially European, philologists, we have seen, were often less faithful transmitters of Upanis. adic texts than the Indian scribes and commentators they so often criticized. Native commentators and theologians did not, as often assumed, carelessly or deliberately change the received texts to suit their doctrinal or grammatical tastes. Indeed, it is the modern philologists who are often guilty of changing the texts to suit preconceived notions of correctness, whether grammatical or otherwise
Lest I be misunderstood, my criticism is not against philology as such but against the substitution of arm-chair philology for the tedious but important examination of manuscript evidence and the resultant improper application of philology to mutilate texts. Philology is the indispensable bedrock of any serious study of texts for any purpose, especially for the purpose of historical reconstruction. But philology is not a substitute for critical edition of texts based on manuscript evidence; it is not possible to reconstruct texts by brain power or philological training alone. We have seen confusion reign especially when scholarly conjectures are introduced into the body of edited texts, especially those calling themselves “critical editions”, often with little warning to readers that they are dealing with conjectures and not received readings. In ancient and medieval India texts were transmitted and preserved by copying onto manuscripts and by memorization, which is a lost art today. The copying of texts introduced errors through negligence or misreading the exemplars and sometimes through deliberate emendations. The existence of numerous manuscripts of a single work, however, permits the careful editor to detect such errors and emendations. Error or emendation was limited to a single manuscript and to others for which it served as exemplar. In the modern age the printed book (and today its electronic counterpart) is the unique medium of textual transmission and preservation. The printed book comes in thousands of identical specimens. An error or emendation introduced into a printed edition, unlike its manuscript counterpart, is reproduced in every single specimen. Given the expense of publishing, moreover, once an ancient text has been published, it is unlikely that a new edition would be forthcoming soon or ever. The responsibility, therefore, of a modern editor to ensure the faithfulness of transmission is a thousand times greater than that of a scribe. Philological hubris, I have tried to demonstrate in this paper, has made many modern editions of the Upanisads unreliable. Modern editors may take note of this set of verses often appended by Indian scribes at the end of their manuscripts.
bhagnaprstikat. igr ıvah. stabdhadrstir adhomukhah. /
kastena likhitam. grantham. yatnena pratipalayet
With great trouble I have written this book, My head bent low, with unwavering eyes, I have broken my back, my hips and neck; So be diligent and take care of it.
yadr . sam . pustakam. dr. s. t. va t adr . sam . likhitam. maya / yadi suddham a suddham . va mama dos . o no vidyate //
I copied exactly What I saw in the book; Whether it’s right or wrong, I am not to be blamed
Comments
Post a Comment