Translating the text
1. The first stage of the translation process is purely analytical. A passage is translated on the basis of lexicographical resources (e.g., dictionaries) and syntactic rules (e.g., grammars). During this operation, it is important to distinguish what is understood and what still remains problematic. All options should be kept open. Interpretations or eisegetical readings should be rejected. From a practical perspective, I systematically used the Mahāvyutpatti to find the Sanskrit terms behind general Tibetan expressions. For technical terms, I relied on Schmithausen (2014) and the academic research mentioned above.
Our Tibetan text is itself a translation. This somewhat complicates our task since we have to decipher the Sanskrit behind the Tibetan in order to make sense of some difficult sentences or passages. However, this approach is necessary on a lexicographical and syntactic level as can be seen in the following examples: One should read the Tibetan brtsams pa as ārabhya, a Buddhist hybrid Sanskrit expression that has the meaning of “referring to/having to do with” and not “beginning with.”23 Likewise, rab tu phye ba stands for prabhāvita, which in the Saṃdhinirmocana means “consisting in/characterized as/characterized by” and not “distinguished.”24 The problem is even more acute in the case of Sanskrit compounds that have been translated into Tibetan according to the way they were formulated in Sanskrit. As an illustration of this, compounds ending with lakṣaṇa (Tib. mtshan nyid) often make more sense in Saṃdhinirmocana as bahuvrīhis than karmadhārayas or tatpuruṣas, not to mention dvandvas. Lamotte thus reads rtog ge kun las ’das mtshan nyid (sarvatarkasamatikrāntalakṣaṇa) as a bahuvrīhi,25 which I believe is appropriate in the context of the passage in question.
2. Once a “technically correct” translation of the source text has been produced, Billeter advises us to bring together the various elements of a sentence or a passage until we obtain a clear picture of what is said in the source text. This step therefore consists in understanding the meaning of the translated passage by literally seeing (or visualizing) its meaning. In a way, the first operation is about das Erklären (“explanation”), while the second concerns das Verstehen (“understanding”). In this sense, the latter uses the resources of one’s imagination and metalinguistic knowledge to establish connections with a web of meaning that is not restricted to the translated sentence or passage alone.
The second operation is therefore a synthesis, a recognition of relations between meaning units of various orders (one would think here of the idea conveyed by saṃjñā and similar terms in which the upasarga sam- plays a central role), whereas the first phase is analytical (in the sense of vicāra, vicāraṇa, and vijñāna, in which the upasarga vi- expresses the notion of taking apart). In fact, translators translate into the target language their comprehension of the source text. They actually never translate the text itself but their understanding or representation of ideas, situations, and emotions conveyed by a text. To illustrate this point, one could mention the problem, encountered by scholars, of translating terms related to meditative practice. The Maitreya chapter of Saṃdhinirmocana is probably the most difficult to understand if one is not familiar with Buddhist practice. Lamotte translates manasikāra with “réflection,” while Frauwallner uses “Beobachtung” (lit. “observation”). Both these translation choices obfuscate the meaning of what the term “directing one’s attention” actually denotes. This is not in itself a major issue, but if such inaccuracies proliferate in the same passage or chapter, the meaning of the translation becomes unclear, although it may well be technically correct on a syntactic level and lexicographical perspective (at least when it comes to correctly identifying Tibetan technical terms on the basis of the Sanskrit).
In the context of pre-Dignāgean Buddhist logic, chapter 10 represents another case in point. This chapter is indeed replete with abstruse concepts not belonging to the well-researched and documented later systems of Buddhist logic. In this case, translating the many occurrences of the connective particle kyi in long compounds with the English preposition “of” will not help the reader much, though it will certainly give the translators the peace of mind of having produced a “technically correct” translation. However, I believe that translators have only two options here: (a) take a risk and, for example, tell us if they actually understand the connective kyi in the sense of “belonging to,” “resulting from,” “consisting of,” and so on; or (b) admit that they do not understand the source text. In the case of the technical compounds found in chapter 10, I therefore tried to ask myself what these terms actually referred to, what could have been the system of logic presented in these pages. For example, I read gzhan gyi rigs kyi dpe nye bar sbyar ba’i mtshan nyid (anyajātīyadṛṣṭāntopasaṃhāralakṣaṇa)26 as rendering into Tibetan a Sanskrit bahuvrīhi. As a consequence, I translated this long compound with “[The logical proof] characterized by a demonstration through an instance belonging to a different class [of phenomena]”
It goes without saying that these translations are at this stage provisional, as further research on the subject matter is necessary. But in order to translate these technical terms, we cannot just give a technically correct translation of a succession of words. Beyond the first phase of the work, which is purely analytical, we still need to develop a mental representation of the situation presented in the text by establishing relations with a context that might go beyond the text.
3. In the third operation, Billeter insists on the necessity for translators to become writers. They should formulate in the target language their understanding of the source text as accurately and naturally as possible. At this stage, translators should focus on literary elements of the translation, such as idioms, voice, and figures of speech. According to Billeter, difficulties in writing accurately and naturally in the target language are often the direct consequence of not having performed the second operation. The translation might well be technically correct, but it still does not make sense, an experience all translators go through when they fail to understand the meaning (or visualize the situation) referred to by the source text.
4. In the fourth operation, translators should reflect on the role played by linguistic constraints and conventions in the formulation of the source text as well as those imposed by the target language. What options did the author of the text have in terms of expression? How would someone express the same ideas in the target language? As a consequence, the notion of form and pragmatics in the target language becomes central. To detail the various operations leading to an actual translation, Vinay and Darbelnet’s model is useful.27 Translators should first identify the units of translation in relation to the translation process: the lexicon (e.g., semantic values, objective and affective aspects, lexical associations and modulations), the syntactic structure (e.g., transpositions between word classes, supplementation of pronouns or conjunctions, modifications in terms of gender, number, characterization, tenses, voice, modality, and verbal aspects), and the message (e.g., meaning, stylistics, pragmatics, topicalization, figures of speech, metalinguistic aspects, specific segmentation of reality). Then, they should examine the descriptive, affective, and intellectual content of the units of translation in the source text to reconstitute the situation at the origin of the message. These two first steps correspond to Billeter’s two first operations. Finally, translators still have to formulate the message in the target language without omitting any relevant element from the source language.
To achieve this, Vinay and Darbelnet argue that translators have only two methods: direct and oblique translation. Direct translation includes three strategies:
• a. Borrowing: the term in the source language is used in the target language to overcome an insuperable metalinguistic lacuna, or it is used because the term is also commonly used in the target language. For instance, I use the Sanskrit bodhisattva and nirvāṇa in my English translation.
• b. Calque: I translated bodhicitta with “awakening mind,” which is both a lexical and structural calque.
• c. Literal translation: most lists and simple sentences are for instance relatively unproblematic direct translations of the source language.
When a Literal translation fails to render the message, is structurally impossible, or misleads the reader due to the lack of a corresponding expression belonging to the same register, one should turn to an oblique translation method among the following strategies:
• d. Transposition: one replaces a word class by another. For example, the frequent nominalizations of Sanskrit and Tibetan are turned into verb clauses. The highly technical nature of some terms makes it necessary to reflect the Sanskrit as much as possible while “unpacking” what is a condensed compound. As an illustration, I translated tadanyavairūpyopalabdhi with “a perception that does not conform with anything other than the [thing to establish],” in which vairūpya is translated as a verb. Our text is mostly written in the same way as a treatise (śāstra), reflecting what is referred to as the nominal style in Sanskrit, or scholastic Sanskrit, in which the nominalization of verbal clauses by means of compounds or suffixes is common. As is often the case in technical or hyperspecialized environments, processes or conceptual frameworks are encapsulated as technical terms (often nouns) implying a complex or recurring pattern. As an analogy, think of a medical term such as hemiglossectomy standing for a removal of a part of the tongue. The passive impersonal phrase “a hemiglossectomy was performed on the patient at 11 pm” includes the nominalization of an action through a compound (hemiglossectomy). It could be rewritten as “[the surgeon] removed a portion of the patient’s tongue at 11 pm.” As can be seen from the translations of Lamotte and Frauwallner, nominalization seems to be less of a problem in French and German than it is in modern (American) English in which readability is more of a concern. When translating the Saṃdhinirmocana, I therefore tried to turn nominal compounds common in scholastic Sanskrit into English verbal sentences by transposing these compounds into verbal sentences. However, since the text is very technical (particularly from Chapter 7 onward), I decided in some cases to keep nominal compounds that were indicative of a technical term and not just a nontechnical action or state of affairs. For example, the text mentions throughout a “concept” being referred to a “X” (see for instance 7.3–6). Just like the surgical term above, such complex nominal compounds stand for a specific action or concept and are part of a “specialist’s jargon.” Turning these compounds into verbal sentences might have the counterproductive effect of erasing an essential feature of this kind of literature consisting in endless lists of often technical terms. Therefore, in this particular case, it would probably be best to avoid transposition.
• e. Modulation: this strategy implies a change of perspective or standpoint made in order to avoid an awkward rendering of the source language. In its simplest form, translating sla ba ma yin (D, folio 25.b, 7.32) with “it is difficult” is an illustration of an optional modulation. Any change of syntactic subject for the sake of clarifying a sentence would be a modulation. Whether this decision is appropriate or not on the part of the translator is something that one should evaluate on a case-by-case basis. I would like to illustrate this point with issues related to the nidāna of the sūtra, in which topicalization plays an important role. The first paragraph of the prologue is a presentation of the place where the Buddha is dwelling. The topicalization of the temple (khang) is achieved through a succession of compounds, mainly bahuvrīhis. Lamotte’s translation reflects this thematization to perfection. In contrast, Powers fails to topicalize the palace to the same degree. In his translation, the logical subject of the several clauses describing this palace is sometimes ambiguous. In this case, one should consider the fact that the Sanskrit structure of this paragraph is built on a process of topicalization that we can easily render in English. In a word, we have no reason to alter this literary device by inducing a modulation of the translation through a change of perspective induced, for example, by a modification of the grammatical or logical subject in the target language.
• f. Equivalence: the same situation can be expressed both in the source language and the target language in completely different stylistic and structural ways due to the necessity to resort to idioms in order to convey the message of the source text. For example, I translated evam etat (de de bzhin te/no) literally with “so it is” in English, which is a slightly pompous and old-fashioned expression no one would probably use today. Instead, one would probably say in an actual dialogue something like “You are right, Dharmodgata” or “This is true, Dharmodgata.”28
• g. Adaptation: this method aims at replacing altogether a reference to a situation in the source language if it is completely unknown in the target language. I generally try to avoid adaptations while translating, for the simple reason that one has to be certain that, for instance, two different metaphors or examples refer to the same situation or object. To conclude on this point, it seems to me that a number of fixed or technical expressions in the Kangyur could be translated in a systematic way following Vinay and Darbelnet’s approach. This research would establish a set of solid conventions that would improve accuracy and readability.
Returning to Billeter’s schema, in his fourth operation the translator should verify that what has been translated into the target language corresponds to the meaning of the source text. Do the two texts express the same idea? Do they produce the same effect on the reader? To answer these questions, Billeter recommends reading one’s text aloud. During this operation, translators should also check whether the translation fits within a specific cultural register in the target language. Discourses take place within a corpus of existing literature that is culturally determined by centuries of textual production. Some statements from a different cultural background resonate through a web of meaning, discursive practices, or literary figures of speech once expressed in the target language. In fact, the web of meaning of the target language within which the translation is received finds its parallel in the web of meaning within which the source text was produced. Within the source text and culture, concepts, ideas, and references resonate throughout sentences, paragraphs, chapters, works, and genres. For example, some philosophical definitions can represent intratextual and extratextual variations on a theme for which there is no metalinguistic context in the target language. Translators therefore need to understand the text not only as a whole and in relation to its various components, but also in connection with both the source and the target cultures. This is of course particularly true of more “philosophical” texts for which it is essential to evaluate how the translation interacts with the webs of meaning of the source and target cultures. Practically, it is important to cross check the consistency (or lack thereof) of meaning units across the text while keeping in mind that the translation is also obviously culturally situated. This process is fundamental because it facilitates the validation (or invalidation) of translation hypotheses resulting from the two first steps of the translation process.
In the context of Saṃdhinirmocana, the problem is compounded by the fact that the sūtra can be read as a collection of independent texts that would have been put together during the third or fourth century ᴄᴇ. The academic community considers the sūtra as a highly composite compilation lacking coherence from a philological perspective.29 Lamotte explains that the first four chapters represent a Prajñāpāramitā for the reasons mentioned above. He sees chapters 5 through 7 as forming a second group of ideas found in the Prajñāpāramitā literature that influenced the Yogācāra school. Finally, he considers chapters 8 through 10 to be later additions.
It is undeniable that the various recensions in Tibetan and Chinese refer to texts that are quite different in structure. For instance, Paramārtha’s translation includes only the first four chapters, which, according to Lamotte, might have originally formed an independent sūtra. In addition, it is obvious that the ten chapters of Saṃdhinirmocana do not follow a consistent textual pattern. The first six have no title. They are concluded by a few summarizing gāthās and a standard formula indicating the name of the person who questioned the Buddha and the number of the chapter (e.g., “This was the chapter of Guṇākara—the sixth chapter”). Chapter 7 has a whole summary of the chapter in the form of a supplement right after the concluding gāthās, while in chapter 10 the Buddha is questioned on complementary topics once the concluded gāthās have been proclaimed. Chapters 7, 8, 9, and 10 each come to an end with a nītārthanirdeśa (instruction of definitive meaning) on the chapter topic. This nītārthanirdeśa is referred to as a Saṃdhinirmocana and used as the chapter title. On account of this, Lamotte surmises that there might have originally been several independent Saṃdhinirmocanasūtras that came to be grouped together as the text we know today.
One should also note that the dialogue structure of chapters 1 through 7 differs from that of chapters 8 through 10. In the first group, the Buddha elaborates on a topic in the form of a monologue once his interlocutor has questioned him on a specific topic, whereas in the second group a dialogue takes place through short questions and answers. As a consequence of all these philological divergences, one has to conclude that the text is rather composite in nature and probably the result of a succession of additions and adjustments. In a word, I agree with Schmithausen that Saṃdhinirmocana is not an organic whole that would have been composed from the onset in its present form and that its chapters are not mutually dependent.30 However, this hypothesis should ideally be the object of further research by a team following an approach similar to that of Professor Jonathan Silk’s European Research Council project “Open Philology — The Composition of Buddhist Scriptures” at the University of Leiden, to which I have had the good fortune to take part. The multiformity and intertextuality of Mahāyāna sūtras are not the result of a linear development from an Ur-text but the expression of oral-formulaic processes of composition and transmission. For our research program focusing on the Ratnakūṭa collection of sūtras, we have been developing digital and philological tools to identify, analyze, and map the fluidity and modularity of Mahāyāna texts. By using these tools, we could better understand the historical development of the complex textual environment of the Saṃdhinirmocana, which includes several translations and many witnesses of this work.
From the perspective of the narrative and doctrinal content of the Tibetan translation, a somewhat different picture emerges. Even if each chapter does not depend on all others in terms of meaning, there is definitely a progression with regard to the flow of thought in Saṃdhinirmocana insofar as later chapters do depend on the definitions and lines of thinking posited in the former chapters, a central fact for translators of this complex text. We can perceive this continuity in the intratextual cross-references that create a terminological resonance echoing throughout the text. Unraveling these cross-references is as important during the translation process as noting the textual variations indicating a deviation from a specific literary pattern. While translating I thus tried to evaluate the text in terms of regularities and discontinuities in the use of definitions and the flow of meaning unfolding throughout the text. One should therefore temper the impression that the text has been “patched” together on the basis of loosely related texts on the basis of philological arguments whose significance is difficult to assess. For example, the fact that chapter 1 is the only chapter in which a dialogue occurs between two bodhisattvas has never been mentioned by any researcher as a textual inconsistency preventing them from considering the first four chapters as a coherent whole. Minor divergences should therefore not deter us from asking ourselves why these chapters were taught or put together in the first place. I would like to illustrate with a few concrete examples the doctrinal coherence of the text. The term ādānavijñāna in 5.3 is also found in 8.37.1.i; the model of the three kinds of essencelessness (niḥsvabhāvatā) of chapter 7 corresponds to the model of the three defining characteristics (lakṣaṇa) of chapter 6, of which two are foreshadowed in 1.2 through the terms parikalpa (kun tu rtog pa) and *apariniṣpanna (yongs su ma grub pa); the other-dependent defining characteristic (paratantralakṣaṇa) introduced in chapter 6 is mentioned in 7.10; chapter 8 presupposes chapters 5 and 6;31 the concluding paragraphs of 7.33 and 8.41 are almost identical; as noted by Schmithausen, saṃskāranimitta is referred to in similar ways in both 1.5 and 7.25–27;32 *viśuddhyālambana is mentioned with the same function in 4.8, 7.6, 7.25–27, and 8.20; 10.7 refers to the seven aspects of true reality (tathatā) defined in 8.20; 8.21 and 10.7.4.ii contain the same formulation; the famous quote “Whether tathāgatas…” is found in 4.10, 7.9, and 10.7.4.vii.d; 10.9 mentions the enumeration citta, manas, and vijñāna exactly in the way it is expressed in 5.1–6; 10.9 enumerates the domains as in 8.23.
On account of the elements adduced above and with Davidson’s principle of charity in mind, I would like to formulate the hypothesis that there is a good reason why these chapters are found in this order: the structure of the text as we know it today is necessary to provide Mahāyāna practitioners with a systematic teaching on (1) ultimate reality qua basis, which is the nondual inexpressible domain of gnosis (chapters 1 through 4), (2) the path to awakening from the domain of mind to the domain of gnosis (chapters 5 through 9), and (3) ultimate reality qua result of the path, which represents a shift in one’s basis of existence as one attains the domain of gnosis (chapter 10).
Indeed, it seems impossible to deny that, considered as a single text (and not as a succession of independent texts), the Saṃdhinirmocana aims at providing a systematic teaching on the Single Vehicle through the three aspects of basis, path, and result in order to solve seeming contradictions and quandaries in doctrines that were of primary importance for followers of the Great Vehicle (e.g., the two truths in chapter 3 and meditative practice in chapter 8). Now, if we read the Saṃdhinirmocana as a single text, we have to confront the web of meaning found in this text in its entirety with the web of meaning of the target culture in order to avoid potential misunderstandings.
This operation has a major impact on the translation of some key terms, such as vijñaptimātra. Since idealism (in the sense that mind is an unchanging essence) is not an option given the teaching imparted in this sūtra, I tried to avoid any potential confusion resulting from an unfortunate choice of terminology. In a word, I would rather stay on the safe side than insert in my translation a potentially misleading term. As a consequence, I decided to translate vijñaptimātra as “a mere representation” instead of using nominalizations such as “cognition-only.” The first expression is relatively unambiguous in the target culture as it minimizes the risk of misunderstanding the message of the text. Another option would be “just a representation.” These formulations mitigate the risk of superimposing an essence on what is meant by vijñapti.33 The formulation “cognition-only” in the sense of “pure cognition” is in contrast ambivalent. It could also (but not necessarily) signify that only cognition truly exists and by extension, that only mind exists as an essence.
In the last step of the translation process, Billeter recommends that translators perform various operations aiming at polishing the translation, such as replicating the possible effects of semantic resonance throughout the text, improving the connection between sentences and paragraphs, modifying the order of clauses, solving problems of euphony, or editing the translation to make it clearer and simpler by chunking long sentences or eliminating repetitions. To illustrate one of these various tasks in the context of the present project, I decided to review all the terminology pertaining to the semantic field of insight (vipaśyanā) after I had finished translating the entire text. I took as a starting point 8.4, in which vipaśyanā is defined by means of a series of technical terms, such as pratyavekṣaṇa, vibhājanā, pravicaya, paritarka, parimīmāṃsā, nitīraṇa, and vitarka. I first tried to find the best translation for each term in the context of this chapter. Next, I checked the usage of all these terms and other related concepts (e.g., pratisaṃkhyā) throughout the text to standardize the corresponding English terminology. I also tried to minimize the use of square brackets indicating additions to the text when these additions were logically implied by the source text. A typology of such situations would include various operations, such as breaking down a compound, clarifying an abbreviated form corresponding to a well-attested collocation, stating a logical subject, object, or verb that is elided in the source text, mentioning the number of a technical term that usually comes as a list of individual items.
Through all these operations, my aim has been to maximize both accuracy and readability while maintaining the consistency of the very systematic presentation of the Great Vehicle developed in the Saṃdhinirmocanasūtra. This text is important in this spiritual tradition since it condenses all aspects of Mahāyāna. I hope that this translation will contribute to improve our understanding of the interplay between liberation as a path and primordial freedom as the ground of being.