Gandhari and the early Chinese Buddhist translations reconsidered:

the case of the Saddharmapundarikasutra

by Daniel Boucher

The Journal of the American Oriental Society

Vol.118 No.4

Oct 1998


Copyright by American Oriental Society

            It has for some time now been assumed that many if not most of the 
            early Chinese Buddhist translations derive from originals written in 
            Northwest Middle Indic. A number of scholars have attempted to show 
            that the reconstructed pronunciation of many of the Chinese 
            transcriptions of Indian proper names and Buddhist technical terms 
            in these translations reflect a Prakrit source text that has much in 
            common with, and perhaps is even identical to, a language now widely 
            known as Gandhari. 
            While there can be little doubt that the Chinese translators often 
            heard recitations of Indic texts that were heavily Prakritized, 
            containing a number of features that coincide with what we know of 
            the Gandhari language, it is not as certain that they saw such 
            texts. This is to say, what has not been sufficiently taken into 
            consideration is the fundamentally oral/aural nature of the 
            translation process in China. This paper is an attempt to take such 
            a process into account and to raise some caveats with regard to our 
            understanding of the underlying Indian language of these 
            Until quite recently, there were few thorough examinations of the 
            early Chinese Buddhist translations. With the exception of a few 
            brave Japanese souls, scholars of both Indian and Chinese Buddhism 
            have generally been put off by the difficult if not at times 
            impenetrable language of these texts. Moreover, there has been 
            little to attract scholars to these abstruse texts. While the 
            translations of the first few centuries of the Common Era had 
            considerable impact on the gentry Buddhism that emerged after the 
            collapse of the Han dynasty, they were subsequently eclipsed by the 
            translations of Kumarajiva and his successors. It was these later 
            translations that had a greater impact on the development of the 
            indigenous schools of Chinese Buddhism. 
            From the other side of the Himalayas, Indologists have generally 
            questioned - with good reason - the reliability of these first 
            attempted translations as documents for the study of Indian 
            Buddhism. The majority of our historical data - prefaces, colophons, 
            early bibliographies, etc.-paint a rather dismal picture of the 
            earliest translation teams in China. The Indian or Central Asian 
            missionary is frequently described as having little or no skill in 
            Chinese; it is virtually certain that practically no Chinese of this 
            early period commanded any Indian literary language; and it is not 
            at all clear how these texts were copied, transmitted, or preserved. 
            As a result, it has been universally accepted that the translations 
            of later Indian-trained specialists such as Xuanzang, as well as the 
            very literal renderings in Tibetan, are far more trustworthy in 
            absence of an Indic original. 
            Be that as it may, the early translations are currently enjoying an 
            upsurge of scholarly attention. This newfound interest has come from 
            two camps. Sinologists, led in the West by Erik Zurcher, have sought 
            to mine these texts as repositories of early Chinese vernacular 
            language. The fundamentally oral/aural nature of the translation 
            process in China - a process that will be discussed in detail below 
            - has left remnants of what appears to be the spoken idiom of 
            Luoyang during the first few centuries C.E.(1) Indologists, on the 
            other hand, have been drawn to these texts as early representatives 
            of Mahayana Buddhist sutras drafted at a time thought to be rather 
            close, by Indian standards, to that of their composition. In fact, 
            these early translations predate our oldest Sanskrit manuscripts by 
            as many as four or five centuries and may well reveal an earlier 
            redaction of the Indian textual tradition. In addition, it is also 
            believed that these early translations may contain clues concerning 
            the Indic language of transmission. Given the fact that almost all 
            of our extant Indic language materials date from a period when 
            Sanskritization had already profoundly reshaped their idiom, these 
            early Chinese sources may be one of our few windows into their 
            earlier Middle Indic stage. 
            Already in 1914 Paul Pelliot had surveyed the transcriptions of 
            proper names in the Chinese translations of the Milindapahha in 
            order to reconstruct their underlying Indic forms? While Pelliot had 
            noted similarities between some of the names in the Chinese texts 
            and forms originating in Northwest India, as well as the possibility 
            of Iranian influence, this was, in his own words, "une etude 
            In the early 1930s Friedrich Weller and Ernst Waldschmidt turned 
            their attention to the early fifth-century Chinese translation of 
            the Dirghagama.(3) Weller examined thirty-six transcriptions from 
            the fifteenth sutra of the Dirghagama, noting that their 
            reconstructed pronunciation showed many features closer to Prakrit 
            than to Sanskrit, though he hesitated to label the specific idiom. 
            Waldschmidt investigated an even larger body of transcriptions from 
            the nineteenth sutra (the Mahasamajasutra).(4) He was perhaps the 
            first to notice similarities between the reconstructed language of 
            these Chinese transcriptions and the language of the Dutreuil de 
            Rhins manuscript of the Dharmapada that had been discovered in the 
            late nineteenth century.(5) Nevertheless, there were unresolved 
            problems that kept Waldschmidt from drawing firm conclusions 
            concerning the nature of the underlying Prakrit. 
            The first attempt to identify and describe the features of the 
            Middle Indic idiom that appears in some of these early Chinese 
            transcriptions as well as in a number of Central Asian languages is 
            the groundbreaking article by H. W. Bailey entitled "Gandhari," by 
            which name scholars have continued to identify this Northwest 
            Prakrit.(6) For Bailey, this Middle Indic language encompassed the 
            Asokan kharosthi edicts from Shahbazgarhi and Mansehra,(7) the 
            various donative inscriptions from northwest India,(8) the 
            Dharmapada found near Khotan (the Dutreuil de Rhins manuscript),(9) 
            the documents from the ancient Shanshan kingdom found at Niya and 
            Loulan,(10) and the miscellaneous traces preserved in Central Asian 
            and Chinese sources. 
            Since the publication of Bailey's article, attention paid to this 
            language has steadily increased. In 1962 John Brough published a 
            masterful study of the Gandhari Dharmapada which thoroughly 
            discussed all aspects of the discovery, publication, and language of 
            the manuscript as well as its relationship to other versions of the 
            text. In discussing the broader role of Gandhari Prakrit in the 
            transmission of Buddhist texts, Brough also advanced the growing 
            consensus that some early Chinese translations may have been 
            translated from originals written in Gandhari.(11) Brough was 
            prudently cautious in his remarks, recognizing that very few texts 
            had been systematically studied with this problem in mind. However, 
            within three years - and with no further studies undertaken to my 
            knowledge - he was able to state: "Sufficient evidence, however, has 
            now accumulated to establish that the originals of these early 
            Chinese translations were mostly, even if not exclusively, texts 
            written in the Northwestern (Gandhari) Prakrit."(12) While Brough's 
            newfound certainty is indeed curious, it is noteworthy that his 
            conclusions concerning the role of Gandhari Prakrit have been 
            regularly repeated by subsequent scholars, generating what I call 
            the "Gandhari hypothesis." 
            Franz Bernhard, in an oft-cited article published in 1970, 
            reiterated the now firmly established Gandhari hypothesis: 
            Phonetic transcriptions in early Chinese translations of Buddhist 
            texts make it clear that Gandhari was the medium in which Buddhism 
            was first propagated in Central Asia, the medium through which 
            Indian culture was transmitted from the northwest across Central 
            Asia to China.(13) 
            Bernhard describes Gandhari as "the Buddhist missionary dialect par 
            excellence," a kind of lingua franca comparable to ecclesiastical 
            Latin of the European Middle Ages. 
            It is difficult to know what would constitute evidence for a lingua 
            franca in Central Asia on the basis of the rather scant extant 
            records.(14) There can be no doubt that Gandhari had a noticeable 
            impact on other languages it encountered in Central Asia,(15) and 
            most scholars have assumed that it had been most widely influential 
            during the height of the Kushan empire in the first few centuries of 
            the Common Era.(16) Whether this impact can be described as the 
            impact of a lingua franca, a common language shared by speakers of 
            diverse language groups for the purposes of commerce, 
            administration, or religious intercourse, is far more uncertain.(17) 
            Bernhard would like to see the Dharmaguptaka school as primarily 
            responsible for this spread of Gandhari in Central Asia.(18) Some of 
            Bernhard's evidence indicating such a role for the Dharmaguptakas, 
            however, has recently been shown to be problematic.(19) Furthermore, 
            it is well known that the Sarvastivadins had the most substantial 
            presence in Central Asia, at least as discernible from the preserved 
            remains of Buddhist literature in this region and from the reports 
            of Chinese pilgrims passing through. And, not insignificantly, the 
            Sarvastivadins are specifically connected with the Sanskritization 
            of canonical literature.(20) Nevertheless, some connection with the 
            Dharmaguptakas is not entirely without basis. The Chinese 
            translation of the Dharmaguptaka-vinaya refers to the recitation of 
            the arapacana formulary(21) and this formulary has now been 
            convincingly shown to be the syllabic order of Gandhari Prakrit in 
            kharosthi script.(22) Moreover, as mentioned above, the Chinese 
            translation of the Dirghagama, widely believed to belong to the 
            Dharmaguptaka school, has been repeatedly cited as derived from a 
            Gandhari original.(23) 
            Since Bernhard's article, the Gandhari hypothesis has been repeated, 
            more or less intact, by Indologists(24) and Sinologists(25) alike, 
            usually without any substantial increase of data. Sinologists have 
            generally sought to use the transcriptional data to aid in the 
            reconstruction of Ancient Chinese. Indologists have, conversely, 
            used the reconstructed pronunciation of Chinese to determine the 
            underlying Indian language of the translation. The circularity of 
            this process becomes immediately evident and has not gone unnoticed 
            by some of the principal investigators: 
            Since a good deal is known about the sound systems of various Middle 
            Indic dialects and the ways they differed from that of Sanskrit, the 
            Chinese forms sometimes allow us to guess whether the original 
            language of a particular text had a certain feature in common with 
            Sanskrit or was more similar to one or more of the Prakrits. When 
            care is taken to avoid circularity, information obtained in this way 
            can, I believe, be safely used in the reconstruction of BTD 
            [Buddhist Transcriptional Dialect(s)].(26) 
            This brief overview of the development of the "Gandhari hypothesis" 
            should make clear that the evidence marshalled to date concerning 
            the role of this Northwest Middle Indic language in the transmission 
            of Buddhism to China is rather meager. It has in general been 
            founded upon a small body of transcriptions, principally from a few 
            sutras in the Dirghagama only. And the conjectures concerning the 
            underlying Indic language of these transcriptions have been repeated 
            sufficiently to qualify now as "facts." 
            But there are other problems. From the Indian side, this hypothesis 
            has gained so much credibility as to inhibit the consideration of 
            other Prakrits or mixtures of Prakrits as possible source languages. 
            It is, of course, possible, perhaps even probable, that texts 
            composed in Central Indian Prakrits were funneled through the 
            Northwest language on route to China. Such a transmission could have 
            imprinted upon these texts a number of orthographic and dialectical 
            features of the Gandhari language. But at the very least this would 
            have resulted in texts that were linguistically mixed in some very 
            complicated and difficult-to-discern ways. I will return to this 
            issue again at the end of this paper. 
            On the Chinese side, scholars have typically assumed that the 
            transcriptional evidence accurately reflects the Indian source 
            language. This takes for granted that the Chinese scribes - and it 
            was almost always Chinese scribes who took down the final text - 
            were able accurately and consistently to distinguish the Indian 
            phonemes and find suitable equivalents for them with sinographs - 
            all with no real knowledge of Sanskrit or Prakrit. Some of the 
            evidence gathered below will call this into question, at least with 
            regard to one of the early translation teams. More importantly, 
            however, even if the Chinese did for the most part accurately record 
            the sound of an Indic word, that does not demonstrate that the word 
            was written in the Indian manuscript as they heard it. This problem 
            has been summarized by Heinz Bechert: 
            [W]e can only view with the greatest scepticism any attempts to come 
            to conclusions about pronunciation on the basis of orthography, 
            since we must never lose sight of the broad spectrum of possible 
            divergences between orthography and pronunciation that we are 
            familiar with from our knowledge of the development of other 
            languages and from examination of later stages in the evolution of 
            the Indic languages themselves.(27) 
            Thus on the Chinese side we have to consider the problem in reverse: 
            evidence for a particular pronunciation of an Indic locution does 
            not ipso facto indicate the language in which that text was written. 
            It is this problem that I will attempt to explore in more detail in 
            this paper. 
            In light of the problems discussed above, I shall attempt a somewhat 
            different approach to examining the influence of Middle Indic - 
            particularly Gandhari - on the early Chinese Buddhist translations. 
            I will, first of all, restrict this investigation to one text. We 
            know all too well that Indian Buddhist texts were not transmitted to 
            China in a single installment. They were brought over a period of 
            several centuries by an ethnically diverse group of missionaries(28) 
            who themselves hailed from a variety of Indian and Central Asian 
            locales. In this way I hope to avoid generic statements about "the" 
            linguistic medium of transmission. 
            Furthermore, rather than focusing upon the Chinese transcriptions of 
            Indian names and terms, which, as I have suggested, raise a number 
            of problems not all of which can be controlled, we shall look 
            instead at mistakes in translation that were due in all probability 
            to phonological confusions caused by a Prakritic or Central Asian 
            pronunciation of the text. It is my contention-to be fleshed out 
            below - that the fundamentally oral/aural nature of the translation 
            process in China led to a number of problems of interpretation for 
            Chinese assistants on these teams who had limited skills in Indian 
            For this purpose we are very fortunate to have the recent and 
            brilliant study by Seishi Karashima,(29) whose work has broken 
            entirely new ground in the study of these early translations. He has 
            meticulously combed through the earliest Chinese translation of the 
            Saddharmapundarikasutra - that of Dharmaraksa,(30) whose translation 
            is dated to 286 C.E. - and has provided a point-by-point analysis of 
            the agreements and disagreements of Dharmaraksa's translation with 
            all of the extant Sanskrit manuscript remains as well as with the 
            fifth-century version of Kumarajiva. In so doing he has also offered 
            ingenious explanations of some of the discrepancies between 
            Dharmaraksa's text and those of the various Sanskrit manuscripts 
            which may stem from confusions caused by a more Prakritic - and, as 
            I will argue, oral/aural - transmission of the text. 
            The advantages of concentrating our attention on the 
            Saddharmapundarikasutra (hereafter SP) then are manifold. We possess 
            extensive manuscript finds with considerable divergences among them 
            that often allow us to differentiate translation mistakes from 
            redactional variations, something that can seldom be done with most 
            Indian Buddhist texts.(31) Nevertheless, despite this quantity of 
            manuscript material, it cannot be assumed that we can always proceed 
            with full knowledge of the Indic "original" underlying Dharmaraksa's 
            translation. We will return to this problem throughout the paper. 
            The fact that the earliest translation of the SP is by Dharmaraksa 
            is also helpful for this examination. Besides the fact that he was 
            one of the most prolific of the early translators during the 
            formative period of Buddhism in China, we have a fair amount of 
            information concerning his life and translation procedures that will 
            bear upon our consideration of the range of forces operating in this 
            translation. He is, for example, one of the first of the foreign 
            translators who is reported by Chinese biographers to have been 
            fluent in both Sanskrit and Chinese as well as the full range of 
            Central Asian languages. Our evidence for mistakes in the 
            translation, then, will provoke us to reexamine these reports from 
            the native hagiographies as well as provide clues concerning the 
            actual dynamic among the participants on the translation teams. 
            In the evidence amassed below, I have in general followed 
            Karashima's lead in the analysis of the philological problems 
            presented by Dharmaraksa's translation. Nevertheless, there are a 
            number of places where I cannot accept Karashima's readings - places 
            where I believe he may have pushed the Prakritic explanation further 
            than is warranted. I have, therefore, despite Karashima's huge body 
            of evidence, cited only what I view to be valid examples of 
            confusions based upon a more heavily Prakritic transmission of the 
            text. Then, having looked at such phonological problems, I will turn 
            to an examination of two colophons to the translation that reveal 
            much about the process by which it was rendered into Chinese as well 
            as some of its early life in China. I will follow this with a look 
            at other kinds of evidence from the translation that expose in 
            different ways the complexity of the data for evaluating the 
            underlying Indic language. It is hoped that such a problematizing of 
            an early Chinese translation will provide some important caveats for 
            the use of these texts by both Sinologists and Indologists. 
            Evidence for Gandhari Prakrit Underlying Dharmaraksa's Translation 
            of the SP 
            In this section I will draw upon Karashima's study in order to 
            highlight specific mistakes in Dharmaraksa's translation that may 
            have been due to the misinterpretation of words or phrases whose 
            forms, though distinct in Sanskrit, would have coalesced in Prakrit, 
            making them more difficult to distinguish for the translation team. 
            I will begin by giving examples that could be construed as providing 
            evidence for a transmission of this text specifically in Gandhari 
            Confusions Related to Vowels 
            Dharmaraksa's translation exhibits frequent confusions between long 
            and short vowels. This would be especially understandable if his 
            Indic text were written in kharosthi script, which does not 
            ordinarily mark vowel length.(33) 
            KN 13.8: balan sahayan parivarjayitva aryesu samsargaratan samahitan 
            having avoided foolish company, they take pleasure in association 
            among the Aryans 
            Dh 65a.10: in the company of strong and close friends (Krsh, 32) 
            Dharmaraksa has confused bala (childish, foolish) for bala (strong); 
            other examples include KN 48.7: balah; Dh 70c.18: (Krsh, 53); KN 
            99.4: balana etadrsa bhonti (they are fit for fools); Dh 79c.26: 
            (intent upon the faculties and powers) (Krsh, 81).(34) This 
            confusion occurs in the opposite direction as well: 
            KN 54.12: karunya mahyam balavantu tesu I have great compassion for 
            Dh 72a.29-b.1: I manifest great compassion and take pity on these 
            fools (Krsh, 59)(35) 
            KN 120.5: pratipatti darsenti bahuprakaram sattvana [the Buddha] 
            teaches good conduct to beings in multifarious ways 
            Dh 83a.24: thus the brilliance of a great lamp illuminates the 
            innumerable masses (Krsh, 91)(36) 
            Karashima has proposed a confusion between pratipa(tti) (good 
            conduct) and pradipa (lamp). Note also the confusion between voiced 
            and unvoiced intervocalic stops, a widespread Prakritic phenomenon. 
            KN 54.6: samharsayami vividhair upayaih I gladden through various 
            Dh 72a.20-21: what [sentient beings] love in their hearts has many 
            Karashima has proposed that Dharmaraksa incorrectly divided these 
            two words, causing a confusion between -dhair upa- and rupa (Krsh, 
            58). This proposal is not certain since Dharmaraksa's rendering 
            could be an attempt to translate vividhair upayaih with sexiang, 
            though this would be an extremely unusual rendering for upaya in his 
            corpus of translations.(37) 
            KN 3.5: Aniksiptadhurena (name of a bodhisattva: "whose burden is 
            not abandoned") 
            Dh 63a.28: "not put down far away" (Krsh, 27). 
            There appears to be a confusion here between -dhura (burden) and 
            -dura (long distance); note also the confusion between aspirate and 
            non-aspirate consonants which will be discussed below. We should 
            also mention that this rendering of this bodhisattva's name occurs 
            in the works of previous translators, for example, Zhi Qian's early 
            third-century translation of the Vimalakirtinirdesasutra (Taisho 
            474, vol. 14, 519b.15). Thus we must always allow for the 
            possibility that such a name could have been drawn from an 
            established lexicon of translation equivalents and would not 
            therefore represent evidence for the underlying language of this 
            Indic original.(38) 
            Besides these there are also a number of other vocalic confusions, 
            but many are confusions of quality rather than length and are either 
            common in many Prakrits or represent problems of a different nature. 
            Thus they cannot be used to indicate a Gandhari source. 
            Confusions Related to Consonants 
            There are quite a number of mistakes in Dharmaraksa's translation 
            that appear to be due to confusions between aspirated and 
            unaspirated voiced consonants in both initial and intervocalic 
            position. Weakness of aspiration - discerned from occasional 
            interchange of aspirated and unaspirated stops in Gandhari texts and 
            inscriptions - is frequently cited as a defining feature of Gandhari 
            among the Prakrits and is especially common among the consonants 
            g/gh and d/dh (Burrow 1937, [sections] 24-27; Brough 1962, [section] 
            49; Fussman 1989, [section] 35). 
            KN 15.1: ghantasamuhai with multitudes of bells(39) 
            Dh 65b. 12: there being a large quantity of incense (Krsh, 34) 
            Here a confusion appears to occur between ghanta (bell) and gandha 
            (incense); there is also the interchange of voiced and unvoiced 
            stops as well as dental and retroflex consonants. 
            KN 56.8: aham pi samksobhi imasmi darune utpanna sattvana 
            I too have arisen in this dreadful commotion [i.e. the world], in 
            the midst of the impurities of beings(40) 
            Dh 72c.3: At that time I was a bhiksu,(41) and I too came forth 
            among the masses of men in order to uphold this dharma. (Krsh, 60) 
            The relationship between the Sanskrit and the Chinese is not 
            entirely clear. It may be that Dharmaraksa confused darune 
            (dreadful) with dharana (preserve, uphold) or perhaps even with 
            dharma. The word daruna seems to have given Dharmaraksa particular 
            problems as he often made mistakes in its interpretation: 
            KN 253.11: sudarune extremely creel 
            Dh 104c.15: able to receive ( dhrama or drama (with concomitant 
            confusion of u and a); or that druma was pronounced with an 
            epenthetic -a- [daruma] in which the unaccented -u- was heard only 
            weakly, thereby making its pronunciation nearly indistinguishable 
            from that of dharma. We should also note that the following kimnara 
            king in the list presented here is Mahadharma, accurately rendered 
            by Dharmaraksa as [Chinese Text Omitted]. It is telling that two 
            completely different Indic words in such close proximity could be 
            translated with the same sinographs. Unable to hear a difference 
            between druma and dharma, Dharmaraksa's translation assistants may 
            have logically assumed that Dharma- would precede Mahadharma-. 
            b/bh and -t-/-(d)dh- 
            The following groups of examples illustrate several different 
            problems that occurred simultaneously. Therefore I will discuss them 
            together while also attempting to distinguish the various 
            phonological developments at work. To begin with, there are many 
            examples of a confusion between a form of the verb [-square root of 
            bhu] (to be, become) and bodhi (enlightenment). 
            KN 283.6: sukhasthito bhoti sada vicaksanah the wise one is always 
            at ease 
            Dh 108b.27: the wise always dwell at ease in enlightenment (Krsh, 
            KN 57.15: loke utpadu bhoti purusarsabhanam there is the appearance 
            in the world of the bulls of men, i.e. the buddhas 
            Dh 72c.26-27: there is a buddha in the world, a great saint and 
            sage, who manifests noble enlightenment (Krsh, 60) 
            Dharmaraksa has confused bh- and b- as well as -t- and -dh- in these 
            examples. We might expect that the latter confusion was heard as no 
            more of a difference than that between -t- and -d-, which are 
            interchanged in other contexts as well.(43) While weakness of 
            aspiration in Gandhari could be cited in both cases, it is 
            nevertheless astounding that the translator(s) would have produced a 
            text that so completely departs from the Indic version.(44) 
            KN 45.9: ye bhonti hinabhirata those who are engaged in lowly 
            Dh 70a.23: those who do not delight in full enlightenment (Krsh, 50) 
            KN 99.4: balana etadrsa bhonti gocaras such [worldly books] are the 
            domain of fools 
            Dh79c.26: and they practice toward enlightenment focusing upon the 
            faculties and powers (Krsh, 80) 
            In these examples Dharmaraksa's translation also ignores the nasal 
            present in the third person plural form; mistakes regarding nasals 
            will be discussed in more detail below. In the latter example, he 
            has also interpreted gocaras from its etymological root [-square 
            root of] car rather than in its more standard Buddhist sense of 
            "range, sphere, domain, association." 
            KN 336.5: bodhisattvas ca ye bhonti caritah kalpa-kotiyah and which 
            bodhisattvas who have practiced for kotis of aeons 
            Dh 116c.12: if bodhisattvas seek enlightenment, they [should] 
            practice for kotis of aeons (Krsh, 191) 
            Karshima records an important Sanskrit variant here: bodhi caritva 
            (instead of bhonti caritah). This variant comes from a Central Asian 
            fragment in the Otani Collection that was transcribed by N. D. 
            Mironov and whose readings are preserved in the notes to N. Dutt's 
            1953 edition of the SP. Since Dharmaraksa, as we have seen, has a 
            propensity to confuse bho(n)ti with bodhi,(45) it is difficult to 
            draw conclusions about his conformity to one or another manuscript 
            tradition in this example. In fact, this is a very good illustration 
            of a problem one is regularly faced with in these early Chinese 
            testimonies to Indian redactional histories. 
            There are instances in which bhuta appears to have been confused for 
            KN 45.14: vadami yeneha ca bhutaniscayam by which I will speak here 
            about true resolve 
            Dh 70b.1: for which reason [I] can speak on what the Buddha has 
            decided (Krsh, 50) 
            KN 200.3: bahubhis ca bhutair gunair abhistuto lauded for his many 
            genuine qualities 
            Dh 95c.1: brilliantly glorified and praised the virtues of the 
            buddhas (Krsh, 124) 
            There also seem to be instances in which Dharmaraksa's translation 
            team misinterpreted -(d)dh- as having been derived from -t-: 
            Kash 47a.4-5: . . . evaham saradvatiputra 
            buddhajnana-(darsana)samdarsaka it is I, Saradvatiputra, who display 
            the exhibition of buddha-knowledge(46) 
            Dh 69c.8: [I] manifest the knowledge of truth (Krsh, 47) 
            KN 330.13: maharsina prakasayanten'ima buddha-bhumim by the great 
            seer who reveals this buddhahood 
            Dh 116a.4: the great saint . . . makes a detailed revelation and 
            establishes this true stage (Krsh, 190) 
            In these two examples, Dharmaraksa took buddha- in the beginning of 
            compounds as bhuta-. Since the normal Prakritic development is 
            clearly from unvoiced to voiced stops, we might speculate that 
            Dharmaraksa's translation assistants, hearing an intervocalic voiced 
            dental stop, perhaps pronounced with considerable friction, deduced 
            it to be derived from an unvoiced stop, despite the fact that such a 
            reading could not have been represented in writing in the underlying 
            Indic text.(47) However, we are ahead of ourselves here and should 
            continue with an examination of the linguistic data before setting 
            forth hypotheses about how this translation acquired its current 
            Sanskrit -th- and -dh- are both generally represented by -dh- in 
            Gandhari, as well as occasionally by -d- (e.g., yada  v is a widespread Prakritic phenomenon (cf. 
            Pischel 1955, [section]199; von Hinuber 1986, [section]181). 
            Kash 121a.3: bahuprakaram pravadanti dharmam they declare the dharma 
            in many ways(50) 
            Dh 83a.28: they bring about decline to the manifold dharma (Krsh, 
            Karashima has suggested that pravadanti was misconstrued as 
            prapatanti (lit., they fall down), though we might expect the verb 
            here to have been understood as a causative (prapatenti). Thus both 
            -v- and -d- were taken as derived from unvoiced originals (-p- and 
            KN 398.4: adavati (a word within a dharanimantrapada) 
            Dh 130b.3: ("a sentence for wealth") (Krsh, 237) 
            Ada is confused with adhya (wealthy, rich, opulent); we would assume 
            a derivation from addha through assimilation of the consonant 
            conjunct along with weakened aspiration. -vati appears to have been 
            confused with -pada; an original -p- would have been assumed for the 
            -v- and the -t- may have been voiced. In both of the cases cited 
            here, as well as others cited elsewhere, an existing -v- was 
            interpreted incorrectly as deriving from -p-, despite the fact that 
            it is unlikely to have been so represented in writing. 
            The alternation of -m- and -v- is quite common in the Prakrits 
            (Pischel 1955, [sections] 248, 250) "but is rare in Gandhari sources 
            other than the Dharmapada" (Brough 1962, [section]36). The most 
            probable explanation is that -m- serves as a notation for an 
            allophone of /v/ in nasalized contexts (see Pischel 1955, 
            [sections]251, 261; Brough 1962, [section]36; von Hinuber 1986 
            [sections]209-11). Thus we find words in the Gandhari Dharmapada 
            such as bhamana'i ( 
            abha-loka-svara: -v- > [[Beta]] > -bh-) or being devoiced as in 
            Tokharian (e.g., durgandhi understood as durgati: -(n)dh- > -t-). It 
            should be clear by now that the oral/aural nature of the translation 
            process must be treated with as much consideration as the linguistic 
            data itself. Furthermore, there is a considerable body of other 
            kinds of evidence that may provide even more details about the 
            underlying language of the Indic text and the roles of the 
            translation participants. 
            IV. ADDITIONAL DATA 
            Double Translations 
            One of the most unusual features of Dharmaraksa's translation idiom 
            and one to which I have alluded already is the occurrence of what I 
            call double translations. These are cases in which an Indic term is 
            rendered twice in close proximity, presumably because two different 
            words had collapsed together in pronunciation, at least as recited 
            by Dharmaraksa.(84) His translation assistants, unable to decide 
            between two or more possible options, offered both possibilities 
            despite the fact that such a rendering almost always resulted in 
            nonsense. We will look at several examples of this phenomenon below. 
            KN 162.5: lokavidu one who understands the world (epithet of a 
            Dh 89b.13: sagely father of the world (Krsh, 108-9) 
            KN 193.1: yatha vayam lokavidu bhavema just as we will become 
            knowers of the world 
            Dh 93b.23-24: we will become wise fathers of the world (Krsh, 119) 
            Dharmaraksa appears to have rendered both -vidu (wise) and -pitu 
            (father). While there are a number of instances of an interchange 
            between p and v in kharosthi documents and inscriptions - if that 
            were the script of Dharmaraksa's manuscript - it is obvious that 
            both words could not have been represented in the same place. Such a 
            mistake suggests that the pronunciation of these two words (-vidu 
            and -pitu) had coalesced, and therefore, Dharmaraksa's translation 
            assistants, unable to determine the proper reading, deduced that two 
            voiced consonants here (-v-, -d-) could have been derived from two 
            unvoiced consonants (-p-, -t-). It is also possible, as I have 
            mentioned several times now, that Dharmaraksa's pronunciation habits 
            were influenced by a Tokharian idiom in which -v- and -d- were 
            devoiced, which would also account for the uncertainty of 
            KN 301.6: svakaras caiva te sattvah and these beings of good 
            Dh 111a.6: beings who have good causes/rooms (Krsh, 176) 
            It appears here that Dharmaraksa and/or his assistants understood 
            both akara (ground, reason, cause, disposition; cf. BHSD, 86) and 
            agara (dwelling, house, room). Of course we have already seen 
            several examples of confusions between voiced and unvoiced 
            intervocalic consonants.(85) What is astounding here though is that 
            a decision was not made between the two possibilities, resulting in 
            an incoherent translation. 
            KN 231.3-4: tathagatapaniparimarjitamurdhanas ca te bhavisyanti they 
            will have their heads stroked by the hand of the tathagata 
            Dh 101b.16: to seek the water of the tathagata and aspire to be in 
            the Buddha's palm - this is the result of the practice of former 
            vows (Krsh, 140) 
            Clearly Dharmaraksa did not adequately convey this line to his 
            assistant, if he even understood it himself. It appears that there 
            was a confusion between -pani-(hand) and pani(ya) (lit., 
            "drinkable," hence "water"), exhibiting the interchange between n/n 
            that we discussed above. As one can plainly see, by not adequately 
            differentiating the possibilities presented by the recitation of the 
            text, Dharmaraksa's translation assistants produced an utterly 
            nonsensical rendering of the passage. Even if Dharmaraksa himself 
            were responsible for some of the confusions, for example, by having 
            indicated alternative "possibilities" of interpretation, it is clear 
            that he could not have fully understood and accepted such a 
            rendering in Chinese. In fact, data such as these make it especially 
            difficult to believe that a single person with adequate knowledge of 
            both Indic languages and Chinese could have left such translations 
            intact. Much of our evidence suggests to the contrary that 
            Dharmaraksa's recitation of the Indic text was mediated by someone 
            with a modest command of Sanskrit/Prakrit vocabulary and a rather 
            poor grasp of Sanskrit grammar. This points to the semibilingual 
            intermediaries that our colophons speak of and in the case of the 
            SP, Nie Chengyuan, in particular. 
            It has been known for some time that there are two widely occurring 
            versions of the title of the buddha of the "second vehicle": 
            pratyaya-buddha (awakened from [external?] causes) and 
            pratyeka-buddha (awakened on one's own). The alternation of these 
            two terms has led to a number of folk etymologies in Buddhist 
            literature, as well as in modern scholarship.(86) The best 
            discussion of the philological problems related to this figure is by 
            K. R. Norman.(87) Norman convincingly demonstrates that the 
            available Pali, Prakrit, and Sanskrit evidence in Buddhist and Jain 
            texts and inscriptions points to pratyaya-buddha as the original 
            form of the word, and that pratyeka-buddha represents an incorrect 
            back-formation (as would the Jain patteya-buddha). 
            Translations reflecting one or the other form of the term occur 
            throughout the Chinese Buddhist canon, including those of 
            Dharmaraksa as well. Yet there are also instances in which 
            Dharmaraksa (or perhaps his assistants) was unable to decide between 
            the two: 
            KN 10.4: pratyekayanam ca vadanti tesam they speak to them about the 
            solitary vehicle 
            Dh 64b.4: furthermore they are able to obtain the vehicle of the 
            conditioned-solitary buddhas 
            Dharmaraksa's rendering reflects an underlying 
            pratyaya-eka-[buddha-]yana, clearly nonsensical in any context, but 
            cognizant, interestingly, of the two possible words that could have 
            collapsed in Prakritic pronunciation. We should also note that this 
            particular double translation predates Dharmaraksa. In his 
            translation of the Vimalakirtinirdesasutra, Zhi Qian (ca. 220-52) 
            has the following: [Chinese Text Omitted] (furthermore [I] will 
            establish others in the practice of the sravakas and of the 
            pratyaya-eka-buddhas) (Taisho 474, vol. 14, 522a.26). Thus, as in 
            previous examples, we must consider the possibility that Dharmaraksa 
            and his team borrowed well-known locutions from previous 
            Besides a number of alternations between these two words - cases 
            where Dharmaraksa reads yana when one or more of the Sanskrit 
            manuscripts reads jnana and vice versa - there are several instances 
            in which a Chinese rendering of both terms was provided for one or 
            the other Indic word.(88) 
            KN 49.2: ekam idam yana(89) dvitiya nasti this is the only vehicle; 
            there is no second 
            Dh 71a.2: as for wisdom/vehicle, there is one, never two (Krsh 1993, 
            KN 49.7-8: sarve ca te darsayi ekayanam ekam ca yanam avatarayanti 
            ekasmi yane paripacayanti acintiya prani-sahasrakotiyah 
            All [buddhas] have manifested but a single vehicle, and they 
            introduce one vehicle only. With this one vehicle they bring to 
            maturation inconceivably numerous thousands of kotis of living 
            Dh71a.8-10: For the sake of beings everywhere, [the buddhas] 
            manifest one vehicle; therefore they teach this path to liberate the 
            unliberated. They always teach for the sake of men the equanimous 
            path/knowledge, converting hundreds of thousands of millions of 
            kotis of beings (Krsh 1993, 143) 
            Note that besides the double translation here (daohui), this verse 
            also clearly establishes the semantic equivalence of sheng (vehicle) 
            and dao (path). 
            KN 189.1-2: ma khalv ima ekam eva buddhajnanam(90) srutva dravenaiva 
            pratinivartayeyur naivo-pasamkrameyuh 
            bahupariklesam idam buddhajnanam(91) samudanayitavyam iti 
            These [beings], having heard this one and only buddha-knowledge, 
            should not casually turn back and not go all the way [thinking]: "To 
            acquire this buddha-knowledge is fraught with too many 
            Dh 92c.14-15: Furthermore, the Buddha taught from the beginning that 
            there is one vehicle; having heard the Buddha teach the dharma, 
            [these beings] do not accept the path/knowledge (Krsh 1993, 140) 
            What is especially striking about this example is that in both lines 
            of this verse we have buddhajnanam (or in the case of the Kashgar 
            MS, buddhayanam) represented in the Sanskrit, but two different 
            renderings in the Chinese, the latter a double translation. While 
            Dharmaraksa's strict adherence to four-character prosody certainly 
            motivated the use of a two-character equivalent here, this example 
            would suggest either a certain amount of indecision on the part of 
            the scribe, or perhaps an intentional attempt to indicate the 
            ambiguity of a Prakritic locution. 
            There are several instances in which jnana is rendered as dhyana and 
            vice versa in Dharmaraksa's translation. Such an interchange 
            presumably would have taken place, as Karashima rightly suggests, 
            through a Prakritic development jhana > jana (or jana) jj > j and dhy > 
            jh with loss of perceived aspiration). This is further confirmed by 
            the fact that jnana is also confused with jana (people): 
            Kash 31a.2-3: samadapeti bahu bodhisatva(m) acintika kotisahasra 
            jnane(93) [the buddha] inspires inconceivably many thousands of 
            kotis of bodhisattvas toward knowledge 
            Dh 66c.2-3: [the buddha] will encourage and develop innumerable 
            bodhisattvas and inconceivable hundreds of thousands of kotis of men 
            (Krsh 1993, 148)(94) 
            To complicate matters further, there are also instances in which 
            Dharmaraksa confused dhyana with dana (giving): 
            KN 13.10: dhyanena te prasthita agrabodhim they set out for highest 
            enlightenment through meditation 
            Dh 65a.13: setting their thoughts on giving, they seek the noble 
            enlightenment (Krsh, 32)(95) 
            KN 24.13: sarvasvadanani parityajantah forsaking donations of their 
            whole property 
            Dh 67a.3-4: no meditation gives rise to causes [sic!] (Krsh, 40)(96) 
            These examples would appear to represent a confusion between dry and 
            d. We should expect the assimilation of dental stops with y 
            conjuncts to palatal stops, though there are exceptions among the 
            Prakrits (cf. Pischel 1955, [section]281). Moreover, we can never 
            rule out the possibility of redactional differences playing a role 
            in the disparity between the Sanskrit and Chinese versions. To give 
            one relevant example, KN 14.4 reads: desenti te pranasahasrakotinam 
            jnanena te prasthita agrabodhim (they [the buddhas] teach thousands 
            of kotis of beings, and these (beings) set out for highest 
            enlightenment by means of knowledge). As Karashima points out (Krsh, 
            33), the Central Asian MS fragment from the Stein Collection (H3 
            Kha. i 24, fol. 4a.8; Toda 1981, 265) also reads jnanena here. But 
            the Kashgar MS (Kash 20a.6) and the Gilgit MS(97) both read danena. 
            Dharmaraksa appears to follow the latter. 
            The examples we have examined in this section provide us with some 
            information concerning Dharmaraksa's pronunciation of his Indic 
            text. Confusions between jnana and jana suggest that the jn- 
            conjunct had been assimilated to j- ( jana), pronounced in Northwest fashion as [z], would have 
            been confusable with [[Delta]].(98) 
            But even if this hypothesis be accepted - and it is certainly not 
            clear that it should be - the underlying language of the text is 
            still not determined. As Edgerton has convincingly demonstrated, 
            Buddhist Hybrid Sanskrit orthography can be quite misleading as an 
            index of actual pronunciation.(99) It is clear from an examination 
            of the verse portions of BHS texts that these sutras were originally 
            pronounced with far more Prakritic features than are now preserved 
            in the manuscripts. For example, the most widely occurring meter in 
            the SP is the tristubhjagati, which requires that the third, sixth, 
            seventh, and ninth syllables be light. Thus a consonant conjunct 
            occurring initially in a following syllable would have to be 
            pronounced as assimilated even if it were not resolved 
            orthographically. In one of the examples just cited, acintika 
            kotisahasra jnane, the ninth syllable, -sra must be metrically 
            light, though orthographically it is heavy by position, being 
            followed by the conjunct jn - which must therefore have been 
            assimilated in actual pronunciation. 
            What is not as clear from the Indic texts, however, is exactly how 
            such conjuncts would have been assimilated. In the case of jn- there 
            are a number of possibilities: j ( v. We would also have to presume the insertion of 
            an epenthetic -u-, here under the influence of the labial semivowel, 
            again a fairly common Prakritic development (cf. von Hinuber 1986, 
            [section]155). While this explanation may seem to stretch 
            credibility, it is difficult to discern an alternative. In addition, 
            the syntax of Dharmaraksa's translation, generously strained in my 
            own rendering, suggests that he did not perceive both kotisahasran 
            bahavah and sadabhijnan as referring to mahabhagan (literally, 
            "those possessed of a great share," thus the highly fortunate, 
            illustrious, and in religious contexts, the virtuous and holy).(103) 
            This verse then provides yet another piece of evidence for the 
            erratic - to put it charitably - knowledge of Sanskrit grammar of 
            Dharmaraksa's translation team. 
            Mistaken Division of Words 
            We have already noted several examples above in which Dharmaraksa or 
            his assistants misconstrued a passage by dividing the words in the 
            sentence improperly. In one case, for example, Dharmaraksa took 
            vividhair upa-yaih as vividhai rupa(yaih). I will note two other 
            apparent cases of such a mistake. 
            KN 120.3-4: anuvartamanas tatha nityakalam nimitta-carina braviti 
            dharmam dharmesvaro isvaru sarvaloke mahesvaro lokavinayakendrah 
            The lord of the dharma, lord over the whole world, great lord, chief 
            of the leaders of the world, always preaches the dharma in 
            conformity with those who follow [mere] appearances. 
            Dh 83a.21-23: In consoling and urging on [others] always at just the 
            right time, he has never engaged in acts out of hope for merit; in 
            the whole world he is the venerable of the dharma, and is considered 
            by all as the great lord, the supreme tathagata. (Krsh, 91) 
            While there are several interesting problems in this verse, the one 
            that principally concerns us here is the fact that Dharmaraksa or a 
            member of his translation team has mistakenly interpreted the first 
            line in the negative, presumably by taking the -na of the gen. pl. 
            nimittacarina as the negative marker na.(104) We have already noted 
            above that the Gandhari sources differ in their treatment of these 
            two nasals. Obviously such a mistake plays havoc with the 
            understanding of the verse and cannot be attributed merely to 
            phonological confusions. 
            KN 27.12: pujam ca tesam vipulam akarsit he performed extensive 
            homage to them 
            Two Nepalese MSS (one [K[prime]] brought from Tibet by E. Kawaguchi 
            and preserved in the Toyo Bunko in Tokyo(105) and MS no. 3/672 
            [?678] preserved in the National Archives of Kathmandu, Nepal) read 
            vipulam aharsit (

y (Pischel 1955, [section] 236), then we could 
            speculate that he misread the kharosthi y as s - two of the most 
            graphically similar aksaras in this script - and understood asiti 
            ("eighty"). What is curious in this case is that this name was read 
            correctly, both in transcription (cf. note 101) and in translation, 
            several times in nearby passages. But here Dharmaraksa not only 
            misread the text, but produced a translation that is transparently 
            incoherent. Thus, when the colophon states that this translation was 
            proofread by a Kuchean layman and an Indian sramana, such mistakes 
            remind us to take such information cum grano salis. 
            Moreover, there is no reason to assume that Buddhist Hybrid Sanskrit 
            manuscripts were not transmitted in kharosthi script. Among the 
            kharosthi documents discovered at Niya are two that are written in 
            Sanskrit: document no. 511 is composed in a mixed Buddhist Sanskrit 
            with a number of Prakritisms and document no. 523 is in pure 
            classical Sanskrit, replete with long vowels, visarga, virama, and 
            proper sandhi.(118) These documents were certainly composed by 
            someone conversant with the brahmi script as indicated by the fact 
            that the verses are numbered in both documents with brahmi numerals. 
            In all probability, the modifications to the kharosthi script that 
            made correct Sanskrit possible would have occurred under a brahmi 
            In addition to mistakes based upon phonological confusions, we have 
            also found evidence for grammatical misunderstandings, mistaken 
            division of words, and connotative misrenderings - all of which 
            again point to a translator with limited skill in Indic languages. 
            We discovered that context seems to have played a significant role 
            in Dharmaraksa's or his translation assistant's arriving at an 
            accurate rendering of certain lexical items.(120) Of course, we 
            cannot presume that all of these mistakes are the result of Nie 
            Chengyuan's misunderstandings. It is likely that Dharmaraksa himself 
            would have sometimes misread his Indian manuscript, which could have 
            itself been fraught with scribal errors of indeterminable types. 
            Furthermore, Dharmaraksa may have provided glosses to his assistants 
            that would have been misleading. Nevertheless, the predominance of 
            correctly translated items in much of the text (e.g., lokavid when 
            it occurs within the standard list of epithets) side-by-side with 
            occasional mistakes, even when context demanded a narrower reading, 
            suggests a rendering by someone whose understanding of the Indic 
            text was imperfectly mediated. Unless the Indic text contained 
            unusually irregular variants of the same words, or Dharmaraksa's 
            understanding and recitation of the text fluctuated in erratic ways, 
            the most probable explanation, it seems, rests with the middlemen: 
            the Chinese assistants who were responsible for receiving the Indian 
            text with a severely limited arsenal of linguistic tools and who 
            transformed their understanding of it into a semi-literary Chinese 
            Lastly, we have taken notice of an interpolation in Dharmaraksa's 
            translation that perhaps more than any other piece of data points to 
            the strong likelihood of a native Chinese source. In this passage we 
            observed two locutions that were doubtless derived from the 
            contemporary Chinese literary vocabulary, reflecting an attempt by 
            the Chinese members of the translation committee to narrow the gulf 
            between the Indian and Chinese religious worlds. 
            In short, what this rather sizable mass of data would seem to 
            indicate is that the evidence for the underlying Indic text of this 
            translation is in fact evidence for the Chinese reception of the 
            Indic text. And this reception, as we have seen, suffered at times 
            from rather severe limitations in expertise.(122) Thus the attempt 
            to see Gandhari Prakrit specifically beneath our extant Chinese 
            translation must take into account the complex interaction between 
            an orthographically indeterminable Indic text, its recitation by a 
            Yuezhi monk trained by an Indian master at Dunhuang, and its 
            transmission to a linguistically underprepared Chinese upasaka. 
            In addition, the linguistic complexity of the underlying Indic text 
            cannot be underestimated. Even if we want to suppose the existence 
            of a considerable number of Buddhist texts written in the Gandhari 
            language, most canonical texts used in the northwest would have 
            originated from central Indian Prakrits. And the process of turning 
            such Prakrits into Gandhari would have decidedly shaped and perhaps 
            significantly altered the final text. K. R. Norman, for example, has 
            argued: "It cannot be emphasized too much that all the versions of 
            canonical Hinayana Buddhist texts which we possess are translations, 
            and even the earliest we possess are translations of some still 
            earlier version, now lost."(123) Heinz Bechert, on the other hand, 
            has suggested that translation - a linguistic transfer between 
            mutually unintelligible languages or dialects - is too strong a 
            characterization of this process: 
            Some scholars believed that this transformation was a real 
            "translation" of texts which at that time already existed as written 
            literary texts. Others think - and I agree with them - that the 
            transposition was no formalized translation. It was another kind of 
            transformation from one dialect into another dialect, that took 
            place in the course of a tradition, which was still an oral 
            tradition, but had already entered the process of being formalized 
            linguistically . . . .(124) 
            However, these positions are not necessarily as sharply opposed as 
            they might first appear. Norman has shown that these "translations" 
            were often carried out by scribes who applied certain phonetic rules 
            mechanically.(125) Nevertheless, some of these transpositions led to 
            hypercorrections and mistaken interpretations, suggesting that the 
            movement between these dialects was not always clear even to learned 
            scribes.(126) This problem was especially acute in Gandhari, as 
            Gerard Fussman has recently indicated: 
            Il ne faut pas surestimer la gene qu'apporte h l'usager l'existence 
            d'une orthographe vieillie assez eloignee de la prononciation 
            reelle. . . . Dans ces conditions les textes bouddhiques gandh. 
            s'ecartaient tellement de la norme parlee qu'ils n'etaient parfois 
            plus comprehensibles, meme a leur redacteur.(127) 
            For our purposes then it is important to realize that before an 
            Indian sutra arrived in China, it may have undergone one or more 
            stages of transference between Middle Indic languages. This process 
            almost certainly would have resulted in a very mixed and layered 
            Moreover, it is precisely this predicament, Fussman suggests, that 
            led Buddhists in the northwest to adopt the use of Sanskrit as their 
            linguistic norm: 
            Surtout il n'existait a ma connaissance aucun texte gandh. dont le 
            prestige fut tel qu'il put servir de norme: on sait bien que le 
            bouddhisme n'est pas originaire de Gandhara et les grands sutra 
            bouddhiques, s'ils existaient en gandh., n'y existaient qu'an 
            traduction faite ou refondue sur un original en m-i gangetique. La 
            seule norme possible etait le skt., dont le prestige est bien 
            atteste aux environs de n.e. . . .(129) 
            We would expect then that the Indic text of the SP was shaped by the 
            burgeoning role of Sanskrit in north India beginning from 
            approximately the first century B.C.E.(130) Edgerton has in fact 
            already shown that the idiom he called Buddhist Hybrid Sanskrit "was 
            not a pure Prakrit but a hybrid dialect, based on a Prakrit, but 
            partially Sanskritized from the start."(131) Though the Indic text 
            underlying Dharmaraksa's translation would have certainly contained 
            many more Prakritic forms that were increasingly disguised with an 
            orthographically Sanskritic veneer, there can be little doubt that 
            the original composition was already in the hybrid language. And 
            this language, as Edgerton repeatedly emphasized, was an artificial 
            language, in no way identical to any living vernacular or otherwise 
            literary Prakrit. Given the debate that has surrounded the 
            linguistic status of BHS since Edgerton's monumental study, as well 
            as the continued uncertainty as to the location(s) of the early 
            Mahayana, these philological discussions are likely to have 
            ramifications beyond any particular text. 
            It must be emphasized at this point that I have not proven - nor 
            have I attempted to prove - that Dharmaraksa's underlying Indic 
            manuscript was not written under the influence of Gandhari Prakrit. 
            If, despite some qualifications, there is sufficient evidence that 
            points to this manuscript as having been written in kharosthi 
            script, we would expect a fair number of Gandhari features to be 
            represented even in a Buddhist Hybrid Sanskrit text.(132) But what I 
            have attempted to show is that these early Chinese translations are 
            imperfect testimonies to the Indian source texts. There is much that 
            is not well understood about these early translations and much of 
            this will have to be solved within sinology. As should be abundantly 
            clear by now, the Chinese-ness of these texts intrudes throughout 
            and must be taken seriously in any assessment of the source 
            The gist of this long digression is that any proposal that a Chinese 
            Buddhist translation derives from Gandhari must also take into 
            account the complex history of Indian Buddhist texts, generally, and 
            the process of their translation into Chinese, specifically. Given 
            the importance of such philological discussions for Buddhist textual 
            history, we obviously must proceed carefully.(133) 
            Despite ali the uncertainties, I hope to have shown that these early 
            Chinese translations hold tremendous potential for advancing our 
            knowledge about the language of the Buddhist texts transmitted from 
            India in the first half of the first millennium. Above all else it 
            should be evident that we need fewer generic statements that merely 
            repeat the scholarly assumptions of our predecessors and more 
            focused studies - one text at a time - that unpack the philological 
            clues contained in these mongrel documents. Karashima's study is but 
            the first serious attempt in this regard. Obviously we are in need 
            of many more. 
            I have been fortunate to receive the kind advice and suggestions of 
            several scholars who read an earlier version of this paper. I would 
            like at this point to extend my profound gratitude to Victor H. Mair 
            and Seishi Karashima for comments on things Chinese; to Klaus Wille 
            and Jens-Uwe Hartmann on various Indian matters; to Richard Salomon 
            and Gerard Fussman for very useful suggestions on Gandhari matters; 
            and to Jan Nattier and Paul Harrsion for miscellaneous suggestions 
            throughout. All of these scholars contributed greatly in helping me 
            to avoid a number of mistakes; those that remain are where I strayed 
            1 See Zurcher 1977 and 1991. 
            2 Pelliot 1914. 
            3 Weller 1930 and Waldschmidt 1932, esp. pp. 226-49. 
            4 A revised edition of this text and a discussion of its language in 
            light of fifty more years of research can be found in Waldschmidt 
            5 Waldschmidt 1932, esp. pp. 231 ff. 
            6 Bailey 1946. 
            7 Prior to Bailey's article, the language of the Asokan edicts had 
            received extensive analysis by such scholars as Johansson, Senart, 
            Buhler, and Woolner. For a systematic description of the language of 
            the kharosthi edicts, see Hultzsch 1925, lxxxiv-xcix. The corpus of 
            Asokan studies that has since accumulated is now quite large, 
            constituting something of a sub-field in its own right. 
            8 On the language of these inscriptions, see Konow 1929, xcv-cxv. 
            Many important contributions have since been made toward clarifying 
            some of the problems posed by these epigraphs, particularly by H. W. 
            Bailey, Gerard Fussman, and Richard Salomon; see the bibliography in 
            Fussman 1989, 488-98. 
            9 For a list of the early studies on the linguistic problems of this 
            text, see Brough 1962, viii-x. 
            10 Boyer et al. 1920-29 and Burrow 1937. See also the rather 
            comprehensive list of kharosthi text/Gandhari Prakrit related 
            publications focusing on finds from Chinese Turkestan (Xinjiang) in 
            Lin 1996. 
            11 Brough 1962, 50-54. 
            12 Brough 1965, 587. 
            13 Bernhard 1970, 57. 
            14 On the problem of categories of language in Central Asia, see 
            Nattier 1990. 
            15 Bailey 1946 discussed this influence on Khotanese and Tokharian 
            among other Central Asian languages; for a survey of the impact of 
            Gandhari on Parthian and Sogdian, see Sims-Williams 1983. 
            16 Douglas Hitch has attempted to pinpoint this influence more 
            precisely (Hitch 1988). He argues that Kushan control of the 
            southern silk route and the northwest Tarim Basin coincided with the 
            rise of Kaniska - taking his ascension as the traditional 78 C.E. On 
            the basis of Chinese historical accounts and numismatic evidence, 
            Hitch hypothesized that this domination probably lasted only until 
            midway through the reign of Huviska, or approximately thirty-five 
            years, when Chinese campaigns reasserted themselves in the western 
            regions (Hitch 1988, 185-86). Hitch's thesis, however, depends upon 
            the often repeated but never substantiated supposition that Kushan 
            expansion beyond the Pamirs could only have occurred under Kaniska. 
            Moreover, the evidence of Kushan control of the Tarim Basin has 
            consisted of little hard data: "The paucity of Kushan coins in the 
            area and the absense of other substantial evidence, literary or 
            archeological, make it likely that Kushan interests were strategic 
            or commercial and that they did not rule directly over much of the 
            region for any considerable time" (Rosenfield 1967, 43). 
            17 The only clear case to draw from on this issue is the corpus of 
            texts from the ancient kingdom of Shanshan. While these 
            administrative documents are written in a kind of Gandhari Prakrit, 
            it is also clear from internal linguistic evidence that the local 
            spoken language of this region was a Tokharian dialect, albeit one 
            that differs from that of either Agni or Kucha; see Burrow 1935. In 
            addition, we know that Chinese became used at least for business 
            purposes from the time of Emperor Wu's conquest of Kroraina (Chin. 
            Loulan) in 263. Like much of Central Asia, Shanshan was clearly a 
            multilingual society. For a recent description of what these 
            documents reveal about social, political, and religious life in this 
            region, see Atwood 1991. 
            18 Bernhard 1970, 59-61. 
            19 For example, Bernhard claimed that an early Chinese translation 
            of the Karmavacana belongs to the Dharmaguptaka school. However, as 
            Hisashi Matsumura has recently pointed out, the text in question is 
            a mere extract from the Dharmaguptakavinaya: "Once it has become 
            clear that the extant two Chinese Karmavacana texts of the 
            Dharmaguptakas were compiled in China, it is entirely meaningless to 
            discuss what the original language of the Karmavacana of this school 
            was" (Matsumura 1990, 69). 
            20 On the problem of school affiliation in relation to the preserved 
            Sanskrit remains from Central Asia, see von Simson 1985, esp. pp. 
            84-85 on the evidence of the Dharmaguptakas. As von Simson points 
            out, the only extant vinaya fragment of the Dharmaguptakas is in 
            hybrid Sanskrit and a sutra fragment attributed to this sect is in 
            pure Sanskrit. Dr. Klaus Wille has informed me (personal 
            communication, June 1995) that there may be some additional 
            fragments of the Dharmaguptakavinaya in the Pelliot Sanskrit 
            collection; they too are written in Sanskrit. 
            21 Levi 1915, 440. 
            22 Salomon 1990. 
            23 Weller and Waldschmidt examined only a relatively small portion 
            of the entire text in their early studies. The underlying language 
            of the Chinese Dirghagama will now have to be reconsidered in light 
            of the thorough study by Karashima (1994). Karashima makes it clear 
            that the situation is more complicated than generally stated: "As we 
            have seen above, the original language of the Chang ahan jing is not 
            something that can be simply decided upon as Gandhari. When one 
            looks at the particulars, complex aspects emerge in which elements 
            of Sanskritization, Prakrits, and local dialects were harmonized in 
            addition to specific features of the Northwest dialect. We may still 
            be able to call this dialect Gandhari in a broad sense, with the 
            necessary proviso that it differs considerably from the Gandhari 
            language as reflected in the Northwest inscriptions" (Karashima 
            1994, 51-52). 
            24 See, among others, von Hinuber 1982, esp. p. 250: "If there has 
            been a Gandhari text of the Upaligathas, it does not seem to be too 
            far fetched an assumption that the whole text of the Madhyamagama 
            passed through a stage of development when it was written in this 
            language once widely used in Central Asia" (von Hinuber follows this 
            remark by citing Brough 1965). See also von Hinuber 1983 and 
            Nishimura 1987. 
            25 See Pulleyblank 1983. Pulleyblank's adherence to the Gandhari 
            hypothesis is clear: "The hypothesis that the texts brought by the 
            first Buddhist missionaries to China were written in Gandhari . . . 
            seems to make good sense in terms of the historical situation and 
            has been supported by linguistic arguments by Bailey and Brough" 
            (Pulleyblank 1983, 84). 
            26 Coblin 1983, 34-35. Coblin's study does in fact add a 
            considerable amount of data to the transcriptional corpus from some 
            of the earliest translations of Buddhist texts into Chinese, though 
            much more work remains to be done. Moreover, Coblin has suggested a 
            more cautious approach to the underlying Indic languages vis-a-vis 
            the Chinese transcriptions in his more recent study, Coblin 1993, 
            27 Bechert 1991, 17. 
            28 We call these early translators "missionaries" by convention; 
            while it is likely that their endeavors included activities that we 
            would typically label as missionizing, there is increasing evidence 
            that suggests some of them may have been more what we should call 
            refugees than proselytizers. See Forte 1995, 65-70 for some 
            tentative suggestions regarding the motives of An Shigao, the first 
            translator in China. This issue is tangential to this paper, though 
            a more careful consideration of the possible motives of these first 
            Buddhist teachers in China may reveal some interesting facts about 
            the homelands they left. 
            29 Karashima 1992. 
            30 Dharmaraksa, Chin. Zhu Fahu (ca. 233-311), was born at Dunhuang 
            and studied under an Indian teacher there. He was the most prolific 
            of the early translators; his career spans over forty years and the 
            earliest bibliography of Chinese Buddhist translations credits him 
            with 154 translations, approximately half of which are extant. The 
            best overview of his life and translation career can be found in 
            Tsukamoto and Hurvitz 1985, 193-230. 
            31 A survey of the extant manuscripts can be found in Karashima 
            1992, 16-19. 
            32 The following abbreviations are used throughout the rest of the 
            Dh: Dharmaraksa's translation of SP (references to the translation 
            are to Taisho 263, vol. 9, by page, register, and line number). 
            KN: Kern/Nanjio 1908-12 (references are to page and line numbers). 
            Kash: Chandra 1976; unless otherwise stated, this manuscript has 
            been cited from the transcription of Toda 1981 by folio, side, and 
            line number. 
            BHS G and D: Edgerton 1953. 
            Krsh: Karashima 1992. 
            Translations throughout are mine unless otherwise indicated. 
            33 For a recent explanation of this convention in kharosthi script 
            and its implications for understanding the phonology of Gandhari, 
            see Fussman 1989, [sections]33-34. 
            34 Note also that Dharmaraksa has confused etadrsa and indriya 
            (faculties), a confusion that is not easily explained in 
            phonological terms. 
            35 We should note that this example is a bit ambiguous. While it is 
            likely that Dharmaraksa mistook balavantu as from balavat as 
            Karashima has suggested, he also translated balavantu in his 
            rendering da . We will see other instances of this kind of "double 
            translation" below. 
            36 Karashima has suggested an alternative rendering: "(The Buddha) 
            burns [ran] a great candle." 
            37 Furthermore, sexiang occurs elsewhere in Dharmaraksa's SP for 
            rupa: KN 76.3/Dh 75c.2, KN 290.12/Dh 109c.10, KN 295.10/Dh 110b.11, 
            etc. I would like to thank Prof. Karashima for calling these 
            additional examples to my attention. 
            38 Thus Chinese renderings (translations or transcriptions) 
            established in the early period under possible Gandhari influence 
            cannot be cited from later texts as evincing the continued influence 
            of the Northwest Prakrit. Once these terms became part of the 
            indigenous Chinese Buddhist vocabulary, translators often defaulted 
            to them even if their Indic text may not have reflected the same 
            phonology or exact meaning. A common example that could be cited is 
            shamen (Early Middle Chinese: sa-men), which transcribes sramana but 
            appears to reflect the particular Gandhari development of sr > s, 
            (samana). On this issue, see de Jong 1981, 111-12 and Nishimura 
            1987, 51-52. 
            39 Note that Kash 21a.6-7 reads: ghanthasamudgebhi. 
            40 Cf. Kash 62a. 1: . . . daruni utpamna satvesu kasatthamadye. With 
            regard to our examination of the confusions related to aspiration in 
            Dharmaraksa's translation, we should also note that in this one line 
            the Kashgar MS itself has made two errors of this kind: kasattha 
            presumably stands for kasatta (cf. BHSD 174; note also Kash 53a.2: 
            kasatrra) and madye here is a mistake for madhye. The manuscript is 
            quite clear in both cases. 
            41 Karashima proposes that biqiu here represents an instance of 
            metathesis in Dharmaraksa's translation: (sam)-ksobhi/bhiksu (Krsh, 
            42 variant: shun. 
            43 The situation in Gandhari is actually more complicated than this. 
            Brough has astutely hypothesized (1962, [section]43a) that the 
            appearance in the Gandhari Dharmapada of -dh- in place of -t- 
            results from a further weakening of the intervocalic stop to the 
            point at which it would have sounded like -dh- to at least some 
            scribes. Such a shift would have been facilitated by the fact that 
            -dh- had already taken on the value of a fricative [[Delta]]; as -t- 
            and -d- both weakened over time towards the spirant, a difference 
            between the aspirated and unaspirated stop was no longer felt. But 
            this confusion could also have occurred in Dharmaraksa's translation 
            under the influence of an Iranian pronunciation, without specific 
            connection to Gandhari Prakrit. Cf. also note 47 below. 
            44 Other examples of a confusion between bhoti and bodhi include the 
            following: KN 63.2/Dh 73c.25-26 (Krsh, 63); KN 117.4/Dh 82c.2 (Krsh, 
            89); KN 177.6/Dh 91b.10 (Krsh 113); KN 287.8/Dh 109b.5 (Krsh, 170); 
            KN 287.10/Dh 109b.8 (Krsh, 170); Kash 342a.5 [KN 355.10: bhavet]/Dh 
            119b.10 (Krsh, 198); KN 394.3/Dh 125a.2 (Krsh, 215). 
            45 Further examples of this confusion between bhonti/bho(n)di and 
            bodhi can be found: Kash 54a.1[KN 43.3: bhavanti]/Dh 70b.4 (Krsh, 
            51); Kash 224b.5/Dh 102a.21 (Krsh, 142); KN 236.5/Dh 102a.24 (Krsh, 
            143); KN 296.1/Dh 110b.13 (Krsh, 174); KN 326.10/Dh 115b.3 (Krsh, 
            188); KN 355.1/Dh 119b.2 (Krsh, 198). 
            46 KN 40.11 reads: tathagatajnanadarsanasamdarsaka evaham sariputra. 
            47 As noted above, Brough has already well explained the use of -dh- 
            [= [Delta]] in place of original -t- or -d- in the Dharmapada. This 
            convention could have been known to Dharmaraksa and/or his 
            assistants. We could also speculate that if Dharmaraksa's 
            pronunciation habits were influenced by a Tokharian idiom as, for 
            example, the Gandhari texts from Niya were (cf. Burrow 1937, 
            [sections]14-15, 19), intervocalic consonants could have been orally 
            represented by Dharmaraksa as devoiced. In such circumstances his 
            assistants would still have had to deduce the derivation of the word 
            from a pronunciation in which voiced and unvoiced consonants 
            collapsed together, but they might have been more likely to choose 
            the devoiced equivalent under such conditions. 
            48 For a survey of scholarly opinions up to 1948, see Mallmann 1948, 
            59-82; the few studies that have appeared since this work have 
            contributed little to the discussion. 
            49 See Mironov 1927, 243. The rendering guanshiyin has been said to 
            originate with the translator Kang Sengkai (Sanghavarman?) 
            (mid-third cent.) in a translation of the Sukhavativyuha (see von 
            Stael-Holstein 1936, 352, n. 3), but the attribution of this 
            translation to him is highly questionable. Before Kang Sengkai, the 
            Parthian translator An Xuan (ca. 180) rendered the name as kuiyin 
            ("[the one who] watches over [i.e., hears] the sounds"?) in his 
            translation of the Ugrapariprccha (Taisho 322, vol. 12, 15b.7) as 
            did Zhi Qian (ca. 220-52) in his translation of the 
            Vimalakirtinirdesasutra (Taisho 474, vol. 14, 519b.16). The 
            translation of the Sukhavativyuha attributed to Lokaksema (ca. 
            168-88) transcribed the name: helougeng (Taisho 361, vol. 12, 
            290a.27); see von Stael-Holstein 1936, 351-52, n. 3 and Brough 1970, 
            83 and nn. 13-16. Once again, however, this attribution is quite 
            improbable. Brough's attempt to link this transcription with a name 
            that appears in a second-century Gandharan inscription is also not 
            without problems (cf. Brough 1982). 
            50 KN 120.8 reads: bahuprakaram hi braviti dharmam. 
            51 See Konow 1936, 610; Bailey 1946, [section]4; Brough 1962, 
            [section]16; for a fuller discussion of the problems related to this 
            conjunct in Middle and New Indo-Aryan, see Turner 1936. 
            52 I have read here with the variant. 
            53 Konow remarks with regard to these two nasals: "Here there is an 
            apparent difference between the system of Dhp. and that of Doc. 
            [Niya Documents] and, so far as we can see, Indian Kharosthi 
            inscriptions. It is, however, remarkable that the Kurram casket 
            inscription, which contains a quotation of a canonical passage 
            written in practically the same language as Dhp., has no trace of 
            the Dhp. distinction between n and n. We are left with the 
            impression that Dhp. in this respect represents a normalization 
            which may be due to the influence of another literary Prakrit, or 
            belongs to a limited territory within the area covered by this 
            dialect, where the treatment of n was different" (Konow 1936, 607). 
            54 Variant reads fu. 
            55 On these BHS forms, see BHSG, [sections]19.29-30. 
            56 See Yuyama 1992. 
            57 In the case of 200, we could hypothesize that it was read as 
            duvisati and that the aksara du- was perhaps mistaken as a particle 
            (= tu). 
            58 Fussman 1989, [section]33 ff. 
            59 Fussman 1989, 478; for remarks on nasalization in the Niya 
            documents, see Burrow 1937, [section]47. 
            60 Geiger 1994, [section]6. 
            61 BHSG, [section]3.1-4. 
            62 Brough 1954, 355. 
            63 Zurcher 1959, 31. For other scholarly discussions of the 
            translation process in China, see Fuchs 1930, van Gulik 1956, 
            Hrdlickova 1958, Ch'en 1960, Tso 1963. 
            64 Taisho 2145, vol. 55, 56c.16-24. 
            65 kouxuan (var. adds chuan)chu. 
            66 bishou, literally "received with the brush." 
            67 The question, of course, is first year of which reign period. 
            Tsukamoto and Hurvitz 1985, 551, note 3 assume the reign period to 
            be Yongkang [= 291], but that is unlikely given the fact that that 
            reign period only begins in the third month. The first new year 
            after the Taikang period is Taixi, which would make this date 
            equivalent to March 3, 290. This problem is exacerbated by the fact 
            that there are four rapidly succeeding changes of reign titles in 
            the years 290-291; whether the anonymous colophon writer was in 
            touch with such changes at court is impossible to determine. Tang 
            1938, 112 and Okabe 1983, 21 read yuan nian here as a mistake for 
            [Taikang] jiu nian [= March 25, 288]. This reading has the advantage 
            of explaining why a new reign title was not specifically mentioned 
            in the notice. 
            68 Exactly what the Chang'an devotee Sun Bohu did is not entirely 
            clear. The colophon states that he xie sujie, "copied [the 
            translation, making] a simple exegesis." Okabe 1983, 21 proposes to 
            read xie sujuan, "copied it onto pure silk." Though perhaps a 
            clearer reading, there is no obvious reason to adopt such a 
            emendation. Interestingly, Sun Bohu is mentioned in Dharmaraksa's 
            biography in the Gaoseng zhuan (Taisho 2059, vol. 50, 327a. 6-7) as 
            one of the several people who regularly "held the brush and collated 
            [the translation] in detail at the request of Dharmaraksa." It is 
            not unreasonable to hypothesize that if Sun Bohu did in fact play a 
            significant role on Dharmaraksa's translation committees as the 
            Gaoseng zhuan suggests, then he very well may have produced a series 
            of exegetical notes to the Saddharmapundarikasutra for the faithful 
            in Chang'an as he copied down the text, perhaps even at the request 
            of Dharmaraksa himself. 
            69 Among the Chinese on this translation committee are three members 
            of the Zhang clan: two scribes and one of the patrons. Wolfram 
            Eberhard (1956, 213-14) has listed this clan name among the 
            prominent families at Dunhuang from early times, and members of this 
            clan are known to have been particularly active in the production of 
            Buddhist texts at Dunhuang in later periods (see Teiser 1994, 146, 
            n. 26). With regard to Zhu Decheng and Zhu Wensheng, who "took 
            pleasure in encouraging and assisting" the work on the 
            Saddharmapundarikasutra, Hurvitz states: "These two Chinese lay 
            brethren with the surname Chu [Zhu] must have been very devout 
            indeed, since, although still laymen, they had left the secular 
            community, an act symbolized by abandoning their clan name and 
            taking instead the name Chu, which, as indicated above, is short for 
            'T'ienchu,' i.e., India" (Tsukamoto and Hurvitz 1985, 486, note 
            "ad"). Hurvitz's speculation - and that is all this is - is dubious 
            for two reasons. For one, despite the Chinese-looking personal 
            names, it is not impossible that they were both naturalized Indians 
            living in China. Secondly, if they were Chinese, it is likely that 
            they were monks, given that they had adopted the ethnikon of a 
            foreign master, perhaps even Dharmaraksa himself (cf. Zurcher 1959, 
            68). Among the assistants on Dharmaraksa's various translation 
            committees with the ethnikon zhu, only two, Zhu Li and Zhu Fashou, 
            are clearly of Indian descent and both are described as sramanas. 
            70 The ethnic identity and linguistic affiliation of the Yuezhi is 
            one of the most vexed subjects in Central Asian history. Despite 
            decades of studies drawing upon Greco-Roman, Chinese, Tibetan, and 
            Central Asian sources, there has yet to be a consensus on many of 
            the most fundamental issues. Much of the problem lies in the great 
            difficulty - and probable impossibility - of pinpointing the 
            identity of the Yuezhi before their expulsion by the Xiongnu out of 
            Gansu in the second century B.C.E. Maenchen-Helfen 1945 is almost 
            certainly correct in suggesting that the ethnikon Yuezhi in Chinese 
            sources ceased as a sociological-ethnic term after the migration of 
            the Great Yuezhi to the west. From that point, this designation 
            represented a composite people: one group (the Dayuezhi) settled in 
            the western Tarim Basin and eventually conquered Bactria, where they 
            adopted an Iranian language and culture; others (the Xiaoyuezhi) 
            remained in the Nanshan region (in modern Gansu) among the Qiang 
            tribes and probably spoke a Tokharian language. The problem of 
            Dharmaraksa's ethnic identity is not without significance for this 
            investigation. As noted several times already, Dharmaraksa's own 
            pronunciation habits could have been responsible for some of the 
            translation confusions we have considered and will consider below. 
            It would be of some interest then to know if his native 
            pronunciation was affected by Iranian habits, perhaps to a greater 
            degree than Gandhari speakers in northwest India, or by a Tokharian 
            dialect, as the inhabitants of the Shanshan kingdom appear to have 
            71 The year Taikang 7 was an especially active period for 
            Dharmaraksa. Besides the SP, he also translated the 
            Pancavimsatisahasrikaprajnaparamita, the 
            Visesacintibrahmapariprccha, and the Ajatasatrukaukrtyavinodana, all 
            of which are sizable texts. 
            72 Nie Chengyuan was without a doubt Dharmaraksa's closest disciple. 
            He is mentioned in a number of colophons to Dharmaraksa's 
            translations, including the earliest, the 
            Suvikrantacintidevaputrapariprccha, translated in 267 C.E. Thus he 
            had over twenty years of experience working on Dharmaraksa's 
            translation teams by the time of the rendering of the SP. 
            Furthermore, he is eulogized in Dharmaraksa's biography as follows: 
            "[Nie] Chengyuan was wise and experienced, talented and principled - 
            devout in the work of the dharma. When Master Hu [Dharmaraksa] 
            issued scriptures, he [Nie Chengyuan] would frequently examine and 
            revise them" (CSZJJ, vol. 55, 98a.2627). I will return to Nie 
            Chengyuan and his possible influence on the translation of the SP 
            73 The crucial word here is chu, a very common verb, yet difficult 
            to pin down, describing translation procedures in China. It is often 
            translated as "to publish," but that does nothing to clarify the 
            designated activity. Arthur Waley (1957, 196) has argued that chu 
            refers to an oral translation as opposed to yi, a written one. Since 
            all translations by Indian and Central Asian missionaries were 
            carried out orally, there appears little point to such a contrast. 
            Richard Robinson (1967, 298, n. 28) contends that chu at least 
            sometimes refers to the recitation of the Indic text, not its 
            translation into Chinese; he cites several examples. Robert Shih 
            seems, in part, to support this position: "Dans les prefaces, la 
            difference entre 'publier' et 'traduire' apparait clairement. Celui 
            qui tient en mains le texte indien joue un role plus important que 
            celui qui traduit l'indien en chinois" (Shih 1968, 168). While the 
            greater importance of the foreign master was certainly acknowledged 
            by the Chinese bibliographers, we have looked at data that calls 
            this into question - at least without substantial qualification. 
            Arthur Link has gone further to suggest that chu is "an abbreviation 
            for the technical Buddhist compound i-ch'u. . . . That is i-ch'u 
            means 'translated [with the result that a book] is issued,' or more 
            simply, 'translate'" (Link 1960, 30). None of these positions is 
            fully satisfying. To "issue" an Indian text is to bring it out of 
            its native guise, to make it available. That process, however, 
            required at least two steps that were not necessarily performed by 
            the same person. The Indian text had to be recited aloud, its 
            esoteric script being otherwise impenetrable to native assistants. 
            It also had to be glossed in Chinese, since the Indic sounds were no 
            less befuddling than the manuscript. While we can reasonably 
            hypothesize that Dharmaraksa both recited the Indian text of the SP 
            and explained it in at least general terms for his Chinese 
            assistants, it is unlikely that chu can be thought of as "to 
            translate" in the way that we now use the word. 
            74 CSZJJ, vol. 55, 48b.22-26. 
            75 See, for example, the colophon to his Lalitavistara, translated 
            in 308 C.E. (CSZJJ, vol. 55, 48b.27-c.1). 
            76 Aside from the pilgrims who studied extensively in India, it is 
            unlikely that any Chinese in traditional times truly commanded any 
            Indian literary language. Cf. the remarks by R. H. van Gulik: ". . . 
            [T]he average Chinese scholar considered a knowledge of the Indian 
            script alone tantamount to a knowledge of the Sanskrit language. 
            Chinese terms like fan-hsueh-seng 'a monk who has studied Sanskrit' 
            as a rule means nothing more than 'a monk who has mastered the 
            Indian script'" (van Gulik 1956, 13). For a fascinating discussion 
            of how even a very learned Chinese Buddhist scholastic fundamentally 
            misunderstood the nature of Indian languages, see Link 1961, 281-99. 
            77 Read si with the variant. 
            78 I read benzhai (lit., "original fast") as referring to the 
            monastic holy day, the uposadha, at which time monks often recited 
            the pratimoksa and laymen took special vows. This designation occurs 
            again in the colophon to Dharmaraksa's translation of the 
            Lalitavistara (CSZJJ, vol. 55, 48b.28). 
            79 Shitan appears to be a translation-transcription of dana, 
            80 CSZJJ, vol. 55, 56c.25-57a.2. 
            81 The ethnikon kang is generally taken to represent Sogdian, but 
            Wolfram Eberhard has shown that there is some reason to believe that 
            early use of this ethnikon may have designated two different clans: 
            one that was native to Kangguo (present-day Samarkand) and another, 
            the old Kangju, who were native to Gansu before being forced to 
            emigrate to Transoxiana; these latter may have been Yuezhi (Eberhard 
            1955, 150). It is also possible that this ethnikon was adopted by a 
            Chinese monk after ordination by a Sogdian preceptor, a practice 
            which became common among Chinese clerics in the third and fourth 
            82 This appears to be one of the few recorded instances of 
            Dharmaraksa travelling this far east. The vast majority of his 
            translations were carried out in Dunhuang and Chang'an. Though not 
            explicitly stated, there are several indications that Dharmaraksa's 
            translation of the SP was carried out at Chang'an. 
            83 The dubiousness of a Gandhari influence in this example has been 
            further emphasized by Gerard Fussman (personal communication, June 
            1995): bhavati is almost always attested as hoti (pronounced hoti, 
            hodi, or ho'i) in northwest kharosthi inscriptions; bodhi is often 
            written bosi (pronounced [bozi]), at least from the first century of 
            the common era. A phonemic overlap between the two words is thus 
            highly unlikely in a Gandhari text dating from the third century. 
            84 For an interesting parallel to this phenomenon in the Uighur 
            translations of Chinese Buddhist texts, see Zieme 1992. 
            85 We might also hypothesize that such a confusion could have 
            resulted from a kharosthi manuscript in which the notation -k- could 
            stand for -g-, as in the Gandhari Dharmapada (cf. Brough 1962, 
            [sections]30-31) or, conversely, the notation -g- [= [Gamma]] could 
            stand for either -k- or -g- as in the Niya documents (cf. Burrow 
            1937, [section]16). If this were the case, it is possible that 
            Dharmaraksa himself would have been unclear as to the actual word 
            intended by the Indic manuscript. At the very least we are reminded 
            of the complexity of deciding among multiple indeterminable factors 
            in the transmission and reception of these texts. 
            86 There have been two rather unsatisfactory monographs on the 
            pratyekabuddha figure: Kloppenborg 1974 and, more recently, 
            Wiltshire 1990. On the latter see the review by Collins (1992). 
            87 Norman 1983. 
            88 I will in this section draw upon an article that Karashima 
            published in 1993. He there makes the provocative claim that the 
            very conception of "vehicle" as a central motif of identification 
            for the Mahayana may very well be founded on an incorrect 
            back-formation of the Middle Indic word for "knowledge" in the 
            process of Sanskritization. It is my intention to produce an English 
            translation of this very interesting article in the near future. 
            89 Karashima (1993, 139) notes that one Sanskrit MS (Add 1682 housed 
            at the Cambridge University Library) reads jnana here. 
            90 Kash 183a.3 reads: buddhayanam. 
            91 Kash 183a.4 reads: buddhayanam. 
            92 For example, in the Sanskrit kharosthi document no. 511 from Niya 
            we find dhyana represented as jana: te jana parami gata (they attain 
            mastery in meditation); see Boyer et al. 1927, 186 (reverse, 1.6). 
            93 KN 23.6 reads: samadapeti bahubodhisattvan acintiyan uttami 
            94 There are other instances in which jnana is confused with jana, 
            prajana, and jina; see Karashima 1993, 147-48. 
            95 But note that in the first pada of this same line Dharmaraksa 
            renders the word dhyana correctly: dhyayanta varsana sahasrakotya 
            ("being in concentration for thousands of kotis of years . . ."); 
            ("meditating for hundreds of thousands of kotis of years . . ."). 
            96 Besides the dhyana/dana confusion, Karashima also proposes that 
            Dharmaraksa mistook parityajantah as pratyaya [-square root of jan], 
            leading to his rather bizarre rendering. 
            97 Watanabe 1975, 2: 10.8. 
            98 In a Taxila seal inscription (Konow 1929, 100), for example, we 
            find a case in which mahajana almost certainly stands for mahadhana, 
            where -j- = [z] was interchanged with -dh- = [[Delta]]; cf. Brough 
            1962, [section]6b. 
            99 See Edgerton 1935 and 1946. 
            100 It is important to recognize, however, that these translations 
            of jnana and dhyana do not require a text written in Gandhari 
            Prakrit, but only a Buddhist Hybrid Sanskrit manuscript read aloud 
            under the influence of a Prakrit dialect in which both jr- and dhy- 
            were assimilated to j- (jh-). We cannot assume that Indian texts 
            were pronounced by Central Asians in Indian fashion. Even in the 
            Buddhist Sanskrit texts preserved at Gilgit, we find evidence for 
            pronunciations differentiated under the influence of Iranian habits; 
            see von Hinuber 1989, 357-58. Given the overwhelming importance of 
            oral/aural interaction to the Chinese translation process, such a 
            consideration must always be central to our examination of data for 
            the underlying Indic text. 
            101 With regard to Gandhari, Fussman (1989, [section]18 and n. 32) 
            gives an early example of the development of j  ya-sruti shift is exhibited 
            in one of Dharmaraksa's few transcriptions: the name Ajita is 
            rendered as ayi (66a.17), Early Middle Chinese ?a jit (j here is IPA 
            high front glide). Elsewhere he translated this name as rnoneng 
            sheng ("cannot be surpassed"). 
            102 See Konow 1929, 48 (Mathura Lion Capital) and 87 (Taxila Vase 
            103 In all fairness to Dharmaraksa, it would appear that Kern also 
            mistranslated this verse in his English rendering (Kern 1884, 242, 
            v. 32); Iwamoto's Japanese translation is to be preferred (Iwamoto 
            1964, 2: 199, v. 32). 
            104 Karashima would also like to see a mistake here between braviti 
            ("speaks, teaches") and bhaveti (