[Image] Lines from a Buddhist Sutra
British Library Most Buddhists will be familiar with the problem of finding two different translations of a text they are inspired by and discovering that the two are inexplicably different. This experience was partly what motivated me to learn Pāli and then Sanskrit (and to dabble in Chinese) in the first place. I remember reading the Bodhicāryāvatara in two translations and being puzzled at the differences. I did not realise at the time that one was a direct translation of the Sanskrit and the other was a secondary translation from the Tibetan translation, which helped to explain some of the major differences. 
If we aren't motivated to learn a scriptural language in order to see for ourselves what the text is saying, presuming it is possible to understand it, then we have limited choices. What most people seem to do is make an aesthetic judgement on which English rendering appeals more. I often hear people say that they prefer this or that translation with no reference to the source language. A monoglot Buddhist will say that some translation captures the meaning and some other translation more literal, with no apparent irony. How does one assess the success, let alone the literalness of a translation when one cannot read the language it was translated from?
Another approach I commonly see is to seek out as many translations as possible and hope to triangulate what the underlying text says. One sees quite elaborate attempts at new renderings of texts with no reference to the Sanskrit or Pāḷi, for example. I've even seen these referred to as a new 'translation'. An old friend used to study the Karaṇīya Metta Sutta by giving each participant in the study group a different translation to read from. Sometimes this is successful and other times not.
Thus we Buddhists make choices between translations on superficial and subjective bases, and we probably think of the translation we are familiar with as "the text". Do we ever stop to wonder what "the text" means if "the text" can be rendered 20 different ways in English? Aren't the different translations in fact different texts?


Critical Editions

But the situation is almost unimaginably worse than this scenario. Because most translations are from critical editions. In the process of making a critical edition one collects up all the surviving 'witnesses' (manuscripts, inscriptions, and earlier editions) and examines each one, possibly correcting scribal errors. Typically each witness is different from all the others, even when they are copies of the same 'original'. Scribes inadvertently introduce errors, large and small, and editors deliberately make amendments, subtractions and additions. Then choosing the best manuscript (best can be judged on any number of bases) one notes all the variations from the best one in the other manuscripts. Traditionally this is first done on a large grid. To produce a critical edition one selects from the variations to produce a text that is consistent and coherent. And if this does not produce a comprehensible or likely reading an editor can suggest an unattested reading that fits better (hopefully with notes to explain the logic of their choice). The editor tries to reconstruct the text as it was first transmitted, or as the author intended it to be. The result is a single text with all the variations footnoted and usually extra notes on amendments (though one of the great problems of Indian textual studies is the practice of silently amending non-standard Sanskrit forms thus obscuring dialectical variants).

And it is these critical editions which end up being translated. In the case of the Heart Sutra for example, Conze consulted more than two dozen sources all different from each other. And he made a number of decisions about the author's intention that in retrospect look doubtful at best or were simply wrong (as discussed in my series of essays on the text earlier in 2013). So each translation hides complexity, sometimes vast complexity, and an industrious process of simplification that is fully subject to human foibles. 
But still worse, some Indian texts can now only be understood by reference to commentaries, often centuries removed from the composition of the text and written by sectarians. Again in the case of the Heart Sutra the commentaries disagree on how to interpret the text along sectarian lines.For example tantrikas treat the text as tantric because it contains a dhāraṇī. And more often than not the commentary itself must undergo textual criticism in order to reconstruct the author's text because it too is subject to all the processes of change that affect a text. 

There is no Diamond in the Diamond Sutra.
Take the Sanskrit Vajracchedikā Prajñāpāramitā Sūtra as an example. For a start the title ought not to be translated as Diamond Sutra or even Diamond Cutter. This was a choice made by Max Müller in 1884 and has been slavishly repeated ever since. But as Conze remarks in the notes to his 1957 edition, the word vajra was very unlikely to be understood as meaning "diamond" by its audience. In that milieu vajra almost certain meant 'thunderbolt' (that wonderfully unscientific word that combines thunder and lightning). Really, we ought to translate vajra as 'lightning bolt'.
Chedikā is from √chid 'to cut off, amputate; cut, hew, split'. A noun form is cheda 'cutoff; cut' and the adjective is chedaka 'cutter, cutting' and in the feminine chedikā. Sandhi rules dictate that initial ch is doubled to cch when preceded by a vowel. Then we ought to ask what kind of compound vajracchedikā is.  Other compounds with -ccheda suggest that it is the first member of the compound which is cut off - i.e. guṇaccheda 'cutting the chord' or dhyānaccheda 'interruption of meditation'. These are tatpuruṣa compounds. Monier-Williams lists no other compounds ending in the feminine -cchedikā. Since "cutting off the lightening" is an unlikely rendering and it is in the feminine gender following prajñāpāramitā which is also feminine, we must suspect a bahuvrīhi compound (i.e. it is an adjective describing prajñāpāramitā): "the perfect wisdom that cuts like lightening". I think this is probably what it means. So really we should refer to it as the [Cuts like] Lightning Sutra, though it's extremely unlikely that the facts will result in a change. 

The Manuscript Tradition and Editions.
Paul Harrison and Shōgo Watanabe have provided us with a detailed account of the history of editions of the Vajracchedikā (Vaj). There are now ten published editions, including Harrison & Watanabe. The first of these was produced in 1881 in Devanāgarī by the redoubtable F. Max Müller. Müller had four witnesses of which two were copies of the same original and two were Chinese block prints. All of these witnesses post-date the composition of Vaj by at least 1500 years. They are copies of copies of copies and each copying introduced errors. It was Müller who introduced the system of breaking the text into sections. His numbering has been retained in subsequent editions, but they do not occur in any manuscript.
Not long after Müller produced his edition a number of manuscripts of Vaj were found and began to be published. Aurel Stein discovered a Central Asian ms. in 1900 that was published by F. E. Pargiter in 1916 (P). This manuscript is thought to date from the late 5th or early 6th century (though dating on palaeographic grounds can be doubtful). Five of the nineteen folios had been lost and many others were poorly preserved. The Pargiter text appears to be similar to the Chinese translation by Kumārajīva (401 CE).
A partial manuscript was found in 1931 as part of a cache of texts discovered near Gilgit (G). The seven surviving folios are dated to the 6th or 7th century. This ms. was not published until 1956 in a Roman script edition. A facsimile edition was published in 1974. Another Roman script version was published by N. Dutt in 1959 which used portions of Müller to fill in the gaps. However none of the Roman script editions were entirely reliable and in 1989 Gregory Schopen published a new edition which corrected the many mistakes. Schopen's edition is available online from the Gretil Archive.
Amongst several editions of the complete Vaj brought out after these finds, only Conze's 1957 publication has attracted any attention. Conze based his edition on Müller's, but presented it in Roman script and included amendments based on the published versions of P and particularly G. Conze introduced a number of innovations such as western punctuation and hyphenated compounds. "However, Conze did not use M consistently as his base text, occasionally making changes to the wording in which he conflated his various witnesses arbitrarily. He also failed to list the differences in his witnesses exhaustively." (Harrison & Watanabe 92). Never-the-less Conze's edition has become, as it were, canonical and most subsequent studies and translations have been based on his edition and this means, for example that "philosophical questions have also been addressed on less than solid foundations..." (92). 
In 1961 P. L. Vaidya produced yet another edition based on Müller but, as per Conze, with "improvements" based on G as it was then (unreliably) published. This text is widely available on the internet via the Digital Sanskrit Buddhist Canon and the Gretil Archive for example. And yet Harrison & Watanabe conclude it "can safely be set aside" (92). Similarly the edition by Joshi simply rearranges the text of previously published editions. 
Finally we have an incomplete ms. (MS 2385) discovered in the Schøyen Collection dated to the 6th or 7th century, and recently published by Harrison & Watanabe  (2006). This text is missing it's ending. Fortunately the Schøyen ms. (S) is very similar in character to the Gilgit ms. (G). Indeed S and G are closer to each other linguistically than either is to the edition of Pargiter (P). Both contain a number of similar Prakritic features (see Harrison & Watanabe (97-99) S contains sections 1-16c; whereas G contains sections 13b-14e and 15b-32b. And thus, while they are not identical where they overlap, together G and S make up a reasonably consistent single text (see below).
In addition a total of twelve identifiable fragments of Vaj have been discovered in Central Asia. Other texts have been catalogued but are presently lost somewhere in the Nepalese National Archives it seems!
So to sum up the most widely used edition of the Sanskrit Vaj is unreliable; the most widely available to those outside academia is also unreliable. An important problem in the history of this text is that the sources available to Müller are considerably longer than P, G or S. Do we treat this as one text that was added to, or do we treat this as one text in at least two recensions, one shorter and one longer? 
One of the weird things about Vaj is that it suggests that anyone who recites "even one verse of four lines" (catuṣpadikām api gāthāṃ) stands to benefit. But this text is not in verse. There's no evidence that it ever was in verse except this phrase. Is it a stock phrase that was used unthinkingly? Or did the text once exist as verse? As far as we know only one Prajñāpāramitā text is in verse: the Ratnaguṇasamcayagāthā.
So far we have a Sanskrit text, available in multiple recensions and versions which may well not point back to a single point of origin, and known far and wide by the mistranslated title. The situation in Chinese is almost as complex with seven different translations of texts which vary in length and quality. 

The Text in Translation
When we read a translation it is almost always the case that this background complexity is completely suppressed or at best highly compressed. 
When it comes to translations we are similarly blessed with many options. Max Müller published his translation in 1894. Conze has published three versions of his English translation with only the most recent being widely available. As with the Heart Sutra, Conze's edition has become standard amongst Buddhists, but when examined it is problematic. My preliminary assessment is that Conze's translation of Vaj suffers from his beliefs getting in the way, just as in his Heart Sutra. Conze in particular embraces paradox and nonsense because it fits his preconceptions about Prajñāpāramitā, but this causes him to mistranslate and to obscure the ways in which the text does make sense.

Schopen has published both a translation of the Gilgit ms. and a complete translation. And translations have also appeared by Mu Soeng, Red Pine  and Richard H. Jones. Now we can add the translation by Harrison of the combined S and G manuscripts. Apart from Schopen and Harrison all the available translations are based either on Müller's or Conze's Sanskrit editions with all their faults. As one might expect there are a number of translations from Chinese also, mainly from Kumārajīva's translation.
Unfortunately the translation by Harrison is relatively inaccessible, though it is based on by far the most carefully constructed edition. There is in fact one interesting and useful presentation of the translation on the web based at Oslo University's Bibliotheca Polyglotta. Though the website in theory makes the text available to everyone, I don't think many Buddhists will find the site, and many won't feel comfortable with the presentation in multiple languages and versions, it is not formatted for easy printing for off-line study, and it lacks all the extensive discussion and notes from the publications mentioned. It would be advantageous to have a popular publication with the Sanskrit text and Harrison's translation (with notes) side by side.

One development mentioned briefly above is worth drawing attention to. Promoted as "a new translation" (it is not) the Diamond Sutra website, by one Alex Johnson, is an extreme example of using English translations found on the internet to try to triangulate the underlying text and produce something more comprehensible, though in this case he has singularly failed to find the text. What the author has done, essentially, is to produce a collage of all the versions. No attention is paid to which text has been translated into English - though translations from Chinese are invariably from Kumārajīva's version and from English from Müller or Conze. At times it strays very far from the Sanskrit and/or Chinese text as the elaborations of previous translators are incorporated to produce a rather bloated and turgid rendition of little doctrinal or literary merit (though clearly Johnson has laboured long to produce this, he'd have been better to spend his time learning Sanskrit or Chinese). Nor is any attention given to the context of the sutra. A single example should suffice: in Section 5 he has the Buddha say, "When you see that all forms are illusive and unreal, then you will begin to perceive your true Buddha nature." But "Buddha nature" is entirely anachronistic and out of place here. It is never mentioned in the text. This late Buddhist idea has been crowbarred into the text in a most inelegant way. The Sanskrit text here is "hi lakṣanālakṣanataḥ tathāgato draṣṭavyaḥ" (Harrison 115). This says: "For a Tathāgata should be seen from the non-characteristic of characteristics.” [As ever arguing against naive realism and reification of sense data] Reconciling Johnson's purple prose with this statement is impossible, and I would say, pointless. And yet if you search "Diamond Sutra" what do you find? 

Complexity
The purpose of this account based on the examination carried out by Harrison & Watanabe is to highlight how complex the manuscript traditions are and how the processes of textual production in the present suppress complexity at every stage, thus to some extent falsifying the witness statements. Vaj is actually not a complicated case, but it highlights a problem that Buddhists simply don't think about. As I said with respect to the Heart Sutra, it is not so much a "text" as a tradition with multiple, competing, variously unreliable, texts. I don't want to go down the road of post-modern textual criticism and deny the existence of the text altogether. For one thing I don't know enough about post-modernism to be credible. But we are obliged to think more about what we mean by "the Diamond Sutra". The production of the text we read is a process in which various scribes and editors have been involved. Many decisions have been made to prune the tangled mass of the tradition in order to present us with reading matter and ideas as homogeneous and simple as possible. Reality is somewhat different: "... we ought to expect multiple branching of the manuscript tradition, with enlargement and other textual changes not fully present in some of the branches, despite the late date of their witnesses. This presents the editor of texts like this with considerable problems which cannot be gone into here, but to put it in a nutshell, the idea that the wording of any Mahāyāna sūtra can be restored to some original and perfect state by text-critical processes must be abandoned: all lines do not converge back on a single point." (Harrison 240. Emphasis added) So according Harrison there might not be a (discoverable) single point of origin, a single authoritative text. And this is an argument against criticisms of Conze. That fact that Conze's version is popular with Buddhists is what makes it authoritative, however uncritical those Buddhists have been. Perhaps we have to consider that his version, with all it's faults, is no less valid than other versions? But wouldn't this be rather too defeatist? Ought not errors of reading and translation be repaired? Awkward and infelicitous, not to say inaccurate, translations can be improved on. Though experience does suggest that given the choice Buddhists will cling to a familiar corrupt text rather than embrace a repaired new one.

Conclusion
In the last twenty years I have gone from naive follower to engaged reader, to published scholar. I've discovered along the way that editors and editions can be unreliable. In my education as a Buddhist I was inculcated with the greatest respect for Dr Conze. My Buddhist teacher dubbed him one of the great Buddhists of the 20th century. But as a scholar his methods left much to be desired and his particular Buddhist beliefs seem to have hampered his scholarship. Most of his work is problematic and all of it needs redoing. I hope to do this for the Heart Sutra in the English speaking world (by formally publishing the material I've been blogging) and clearly Harrison, Watanabe and Schopen have done so for the Diamond Sutra. The Sanskrit edition of the Aṣṭasāhasrika-prajñāpāramitā-sūtra is apparently good enough, so we only require a re-translation of that text (several partial translations have been produced, but as yet no one has undertaken the whole task).
But all of this is simply to play the same old game and something about it nags me. A standardised text is almost a lie. It rests on the idea, drawn from Classical scholarship, of a single author sitting down and composing a text that was then corrupted by scribes over time. But Buddhist texts don't seem like this. They almost always seem to be the product of local traditions (plural) preserved in local dialects and languages.
Clearly Buddhist texts are not like Vedic texts. They are not revelations of eternally unchanging texts. They have not been preserved with the kind of fidelity that Vedic oral texts have. Given that we live 800 years after Buddhism died out in India, the home of Sanskrit text production, we must wonder how much or how little of the variation has survived the burning of Buddhist libraries. If we have this many variations now, how many more were lost? 
Buddhists are often fundamentalist when it comes to texts. We have a 'cult of the book' as Gregory Schopen terms it. The book itself becomes an object of worship (I know of at least a couple of Buddhist shrines that have never-read books on them). The book itself symbolises knowledge, but is in conflict with the anti-intellectual injunction against the written word as definitive. In this view, wisdom cannot be put into words except as nonsense and paradox. So even though the Diamond Sutra is a sacred text, it need not be read, though it is chanted from memory in many monasteries and widely studied
The Buddhist tradition is strangely hostile to complexity at times. We are always trying iron out wrinkles, usually with unintended consequences. It begins to seem a little quixotic to insist that our texts are unitary phenomena. Was the Vajracchedikā composed as single text? Did it once stop at what Conze calls "The First Ending" (§13a) only to be restarted by a latter author? How did the later authors justify adding words, lines and sections? Were they like Alex Johnson, i.e. well meaning but incompetent editors trying to resolve textual variations without really understanding the text? If Harrison is right and the lines do not converge then which Vajracchedikā do we take to be authoritative. In China it's usually the translation by Kumārajīva that is authoritative if there is a choice (though as discussed, this is not true in the case of the Heart Sutra) 
Practising Buddhists often resolve these conflicts and contradictions by changing the frame of the discussion and invoking the authority of personal experience. Which is to say they sidestep the textual issues by trumping the authority of the text with a higher authority. Only in doing so they retain the text as object of worship as the (ultimately faulty) encapsulation of "perfect wisdom". On the other hand historically merely hearing the Vajracchedikā is said to have brought about miraculous conversion: in ancient times for example for Huineng the patriarch of Zen and in modern times by Sangharakshita who, aged 17, both realised he was a Buddhist after reading an early translation from the Chinese and also had a series of mystical experiences that shaped his approach to Buddhism (and indeed to life) subsequently. 
The other frame change we like to invoke is to cite "the Absolute", a term drawn from German idealism but applied to Buddhism especially by Conze. Sometimes the term non-dual is used instead though the meaning is more or less the same. Modern Buddhists frequently believe that there is a viewpoint that stands outside the framework altogether and sees things as they are - though heaven forbid that we call this the god perspective! The Absolute is beyond words and concepts and yet encompasses all words and all concepts. And crucially the Absolute can be invoked to resolve all doubts and all disputes. If one cannot think through a problem to a satisfactory conclusion that is because not all problems are amenable to thought or reason. Some problems and doubts are only resolved by adopting the godlike perspective of the Absolute.  This is the viewpoint which insists that wisdom cannot be put into words except as nonsense and paradox.  Unfortunately credibility is strained at times when people who clearly do not have access to this perspective, use nonsense to silence questions and stifle discussion. 
So, is there any such thing as 'a text'? I spend my time reading and studying and creating texts. However, the sacred Buddhist text as a unitary object with well defined boundaries is a fiction. With a tradition like the Prajñāpāramitā we have a number of texts which represent the tradition in different ways at different times, but are themselves far from stable or fixed. The modern day obsession with fidelity of transmission does not seem to have been shared by our Indian antecedents. Texts were changed as expedient. Mistakes were as likely to be conserved as correct readings were. Better to think of a text as a sketch of a tradition from a particular place and time, seen after several generations of copying. It may be clear and focussed and relatively helpful in understanding the tradition which produced it, or it may be obscured and blurred and unhelpful. Sometimes it's hard to know which. Most Buddhist texts in fact seem to continue to be composed over a considerable period of time that may only have stopped with the destruction of Buddhism in India.


~~oOo~~