The Transmission of the Buddhadharma from India to China: An Examination of KumÄrajÄ«vaâs Transliteration of the DhÄraá¹Ä«s of the Saddharmapuá¹á¸arÄ«kasÅ«tra

Bryan Levman

Jump to Content

Chapter 5 The Transmission of the Buddhadharma from India to China: An Examination of KumÄrajÄ«vaâs Transliteration of the DhÄraá¹Ä«s of the Saddharmapuá¹á¸arÄ«kasÅ«tra

äºBuddhist Encounters and Identities Across East Asia

èèï¼

Bryan Levman

Bryan Levman
Search for other papers by Bryan Levman in
Current site
Google Scholar
PubMed

ç±»å:: ç« è

Page Range:: 137â195

DOI:: https://doi.org/10.1163/9789004366152_007

æ¨è·å¾è®¿é®çæ¹å¼ï¼

Dar Hadith al Hassania

ä¸è½½PDF

Abstract

This article examines the transmission of the Buddhaâs teachings from India to China through the lens of the dhÄraá¹Ä«s of the Lotus Sutra. KumÄrajÄ«va was the first Chinese translator to undertake a transliteration of the dhÄraá¹Ä«s that attempted to retain their ritual efficacy for Chinese Buddhists. His source text was Prakritic in nature and shown to be centuries earlier than the Sanskrit MSS that have survived. The transmission of the buddhadharma from India to China was a highly complex process with dozens of human, temporal, spatial, dialectal, scribal, psychological and phonological variables, making it impossible to transmit the teachings error free. Although it is impossible to unravel the complex transmission history, a study of the dhÄraá¹is opens a unique window on the exchange of information between India and China in the early centuries of the common era and the interaction of two very different cultural and linguistic environments.

1 Introduction

It has long been understood that the earliest Chinese translations of the Buddhavacana were made from Prakrit (Pkt.) sources transmitted along the Silk Route caravan network from India (see references below). Indeed, the earliest translators were not indigenous Chinese but Indo-Aryan speaking missionaries and monks who brought the teachings to China and translated them, often with the help of a translation âteamâ of local Chinese scholars. Two of the earliest translators were An Shigao (fl. 148â180) and An Xuan (fl. 168â189) who spoke Parthian, a northwestern Iranian (Indo-European, IE) language, and perhaps the most influential was KumÄrajÄ«va (344â413), from the Kingdom of Kucha, an Indo-European oasis kingdom on the northern edge of the Tarim Basin in what is presently Northwestern China. Kucha was one of the stops on the Silk Route and the native language of the kingdom was West Tocharian, the easternmost branch of the IE language families. The language of the source transmission translated into Chinese has been a subject of scholarly investigation for over a century; this paper examines KumÄrajÄ«vaâs transliteration of the dhÄraá¹Ä«s of the Saddharmapuá¹á¸arÄ«kasÅ«tra (SDP) in an attempt to identify some aspects of the underlying transmissionâs phonological structure.¹ As Jan Nattier (2008: 3) has noted (referring to translations of the second and third centuries CE), these early translations âoffer a window into the Buddhist heritage of both India and China at a pivotal period in historyâ. In the case of the SDP, KumÄrajÄ«vaâs translation is earlier than any of the Indic or Tibetan witnesses by several centuries, and his was the first attempt to transliterate the dhÄraá¹Ä«s in order to retain their sonic efficacy for Chinese disciples. His translation practice provides a fascinating glimpse into the phonological structure of the source SDP text which is the subject of this paper.

2 The Nature of the Source Dialect: Previous Work

The hypothesis that source documents for the Chinese (Ch.) Ägamas were written in Middle Indic rather than Sanskrit (Skt.) is not a new one. Scholars have been investigating this issue since the early twentieth century; their primary tool has been to examine Chinese transliterations of Indic names and Buddhist terms in order to reconstruct the original transmission. In 1914, for example, Paul Pelliot (1878â1945) examined the Chinese translation of the MilindapaÃ±ha, and the transcriptions of the proper names therein. In his opinion the source document was Prakrit; a name like Skt. Kubjottara (PÄli Khujjuttara) was rendered in Chinese as Jiuchoudan é³©è®å®âaccording to Pulleyblank (PB): kuw-dÊuw-tan² âconfirming that the name in the source document did not have the Sanskrit conjunct -bj- but the Prakrit form -jh- (Pelliot 1914: 412â413). Pelliot also believed that the Chinese version preserved forms closer to the original Greek than those of the PÄli (P) text of the MilindapaÃ±ha, possibly because of a Parthian or Indo-Scythian influence. The name Menander (ÎÎÎ½Î±Î½Î´ÏÎ¿Ï), for example, appears in Chinese as Milan å½è (PB: mji/mjiÉ-lan), which Pelliot reconstructs as *Milandⁱ, maintaining it is closer to the original Greek word Menander than the P. Milinda; the change -n- > -l- is frequent in Central Asia and in the Prakrits³ and conforms to general laws of dissimilation in IE languages (ibid.: 384â385). Regardless of whether the change of two vowels between the Chinese transcription and P. is that significant, it is clear that Pelliotâs primary pointâthat the Chinese names indicate a Prakrit not a Sanskrit. source documentâis well made.

In 1915 Sylvain LÃ©vi (1863â1935) examined the Sanskrit remnants and three Chinese translations of the MahÄmÄyÅ«rÄ«sÅ«tra (âGreat Peacockâ), which contain stories and dhÄraá¹Ä«s that protect practitioners from all sorts of harm (snakebites in particular). He examines how the names of 106 yaká¹£as contained in the sÅ«tra are transliterated into Chinese in three systems of transcription: 1) in the early sixth century by Saá¹ghavarman, 2) late seventh century by Yijing and 3) early eighth century by Amoghavajra. These show changing translation practices over time and Prakritisms which are later Sanskritised. For example, Saá¹ghavarman translated the proper name of yaká¹£a PÅ«rá¹ako as Fennake åé£æ¯ (PB pun-na^h-ka),⁴ showing that in his source document, the -rá¹- conjunct had been assimilated to -á¹á¹-, as is typical of the Prakrits (LÃ©vi 1915: 41; Woolner 1928 [1996]: Â§48). In his later transcription, Yijing Sanskritises PÅ«á¹nako by adding back in an epenthetic -r-: è¯å¾⁵ æ, PulÃ¼na (PB bÉ-lwit-nraÉ¨/nÎµ:). LÃ©vi notes that the Chinese and Tibetan versions represent a state of the text prior to the surviving Sanskrit manuscripts, since they contain Prakritisms which are later Sanskritised.

In 1916 Heinrich LÃ¼ders (1869â1943) examined three fragments of Sanskrit texts found at Khadalik in Central Asia (part of the Hoernle collection). After comparing forms in the Central Asian MS like sraá¹sitavÄn, sraá¹sayati (âhe did [not] slackenâ, â[she] does not slackenâ) with corresponding forms saá¹ÅritavÄn, saá¹prakÄÅayati in the Nepalese (âhe did not clingâ, â[she] does not revealâ, and alternates janayati, saá¹mayati, praÅayati, all incorrect Sanskritisations per LÃ¼ders), he concluded that an earlier Prakrit form saá¹sitavÄ must have existed in an underlying form to account for these anomalies. He therefore concluded that both the Nepalese and the Central Asian MSS of the Saddharmapuá¹á¸arÄ«kasÅ«tra must have a common source, maintaining that the original text was written in a language that had far more Prakritisms than either of the two versions. LÃ¼ders was inclined to believe that the original was written in a pure Prakrit dialect which was afterwards gradually put into Sanskrit. This dialect was a âmixed Sanskritâ based on MÄgadhÄ«, acccording to him (1916: 161â62). While many today would disagree with LÃ¼ders about the possiblity of isolating an Urtext, his comparison of manuscripts well illustrates the complexity, ambiguities and confusions inherent in the transmission process (within just the Sanskrit texts, not to mention Chinese), especially when there is a Prakrit source involved; due to the simplified nature of the languageâwhere, for example, conjunct consonants are assimilated and intervocalic stops > -y-, -áº-, (a weakly articulated glide) or > Ãâit contains many homonyms.

In 1930 Friedrich Weller (1889â1980) examined the transliteration of Buddhist terms and proper names in the Chinese translation of the PÄá¹ikasutta from the DÄ«gha NikÄya (DN 24). Here the name of a Licchavi general Ajita (âUnconqueredâ) occurs (Ayoutuo é¿ç±é PB: Êa-juw-da or KG Ã¢-iÌá¸u-da), where -j- > -y-; -t- > -d-,⁶ along with names like Udena (Youyuan æå, PB: Êuw- wuan) where -d- > Ã or -d- > -y-, and AnupiyÄ (Anuyi é¿è¬å¤·, PB: Êa-nÉ-ji), where -p- > -y-, which prove that the source document was not composed in Sanskrit, although he did not specify which Prakrit the forms might represent (Weller 1930: 111).

At the end of his long career Ernst Waldschmidt (1897â1985) maintained: â[â¦] that the original DÄ«rghÄgama text translated into Chinese was written in some kind of archaic Prakrit and not in Sanskrit will hardly be contestedâ. He believed that it was probably translated from the Northwestern Prakrit of GÄndhÄrÄ« (G.), a hypothesis which Pulleyblank supported (Waldschmidt 1980: 137, 163; Pulleyblank 1983: 84â87; Karashima 1992). In 1932 Waldschmidt published a Central Asian Sanskrit manuscript of the MahÄsamÄjasÅ«tra (DN 20 in PÄli) together with a Chinese translation which pointed to an underlying, more Prakritic version of the sÅ«tra. In the Chinese transliterations he discovered many Prakritic forms, the most common being the ending in -u for the nom. and accus. masculine sing., which is also the case in the GDhp (Waldschmidt 1932: 230; HiÃ¤n-lin Dschi 1944: 121â44; Brough 1962 [1991]: Â§75). He also found lots of examples of intervocalic stop lenition (in the underlying source text, as transcribed in the Chinese), another feature of G: Vairocana (Bilouyena éæ¨è¶é£ PB: pji/pjiÉ-lÉw-jia-na) = Vairoyana; tejo (tiyu æè±« PB: dÎµj-jÉ¨a^h) = *teyo; vÄcÄ (poye å©è¶, PB: ba-jia)= vaya; -r- assimilation: Candra (Zhanda æ ´å¤§ PB: tÉian-daâ/daj^h) = Canda; etc., all of which are also features of GÄndhÄrÄ«,⁷ and many other Prakritic features.⁸

For some fifty years, not much work was done on this subject of Chinese transliterations. The last thirty years, however, have witnessed something of a ârenaissanceâ in this arcane sub-field of philology with studies by several important scholars: von HinÃ¼ber, Karashima and Boucher. Von HinÃ¼berâs research confirms that a GÄndhÄrÄ« version of the MadhyamÄgama existed as one of the translation stages for the Chinese text. This is the only way to account for such forms as are found in the P. UpÄlisutta like pabhinnakhilassa (Majjhima NikÄya 1, 386³), âbroken up the fallow spiritual wastelandâ, which do not correlate with parallel Sanskrit forms from the Central Asian manuscripts, prahÄ«á¹akhilasya, âhe who has abandoned the afflictionsâ. The Chinese text has duan hui æ·ç©¢ (âcut off impurityâ) which is a translation of the Central Asian text, but not the PÄli. This suggests that the source document (underlying both the PÄli and Sanskrit) must have contained the word p(r)ahiá¹a or p(r)ahina (âabandonedâ) in G., with -bh- > -h- and the -nn- > n/á¹, and no vowel quantity shown (von HinÃ¼ber 1983: 28â29). Another example von HinÃ¼ber adduces is the word pannadhaja (âwhose flag is loweredâ) from the same text, which has a Central Asian Sanskrit reflex of pará¹ajaha, translated by Saá¹ghadeva as æ§ç huÃ¬shÄng = *prajÃ±Ä-jÄta (âwisdom-bornâ); a number of strange changes and misunderstandings have taken place that are probably due to the Sanskrit and Chinese translatorsâ not recognizing the word panna (âfallenâ) as the past participle of âpad.⁹

In an important recent study, Karashima has gathered all the Prakritisms in KumÄrajÄ«vaâs and Dharmaraká¹£aâs translations, line by line, and has concluded that the Chinese translations represent an earlier stage in the transmission process when the source Indic texts were more Prakritic in nature than the current surviving Sanskrit witnesses (Karashima 1992: 13, 274â275). I will be drawing on this work further in my study of dhÄraá¹Ä«s below. Using data from Karashimaâs study, Daniel Boucher examined Dharmaraká¹£aâs translation at a lexical level, pointing out various misunderstandings due to lack of expertise, dialect (phonological) confusion, script confusion and unresolved ambiguities (e.g. the practice of âdouble translationâ, translating a word twice when it has more than one meaning) and concludes that the source text was âa very mixed and layered text [â¦] already in a hybrid languageâ which had a very complex transmission process. He does not try to identify the dialect, although features of GÄndhÄrÄ« clearly had an influence and the source document may well have been written in Kharoá¹£á¹hÄ« script. He calls for more studies that âunpack the philological clues contained in these mongrel documentsâ (Boucher 1998: 501â503; Deeg 2008: 83â118).

3 The Saddharmapuá¹á¸arÄ«kasÅ«tra

This work was originally composed in approximately the first century BCE in Prakrit or Sanskritised Prakrit, and has a complicated textual history in MI and Chinese. We possess three different MI recensions of the sÅ«tra, all fairly recent; two have been almost completely Sanskritised and the third and oldest, from Central Asia, still retains some Prakritic elements. There are six Chinese translations (completed in 255, 286, 290, 335, 406 and 601 CE) of which only the third, fifth and sixth survive. Of these, the fifth (the second of the surviving witnesses), by KumÄrajÄ«va, is considered the standard (Hurvitz 2009: xiii). So the Chinese translations are much older than the MI manuscripts that we possess, and indeed when one looks at the proper names, technical terms and dhÄraá¹Ä«s of the Chinese versions we find many Prakritic elements still preserved. As is well known, the SDP is one of the most important texts in Chinese Buddhism. A key enjoinder of the SDP is its admonishment to âreceive and keep it [the sÅ«tra], to read and recite it, to preach it, to copy it and to make offerings to itâ (Hurvitz 2009: 263). This central message of the SDP, i.e., securing its transmission to present and future generations and the accuracy of that transmission, is repeated so many times, one might argue it is the main theme of the sÅ«tra. The text starts and ends with the concept of âmastering the dhÄraá¹Ä«sâ, which does not refer solely to âmagic charmsâ as Hurvitz defines the term,¹⁰ but more germanely to memorizing and retaining the dharma. As a result, the SDP was copied thousands of times over the centuries (Stevenson 2009: 134, 145; ChÃ¼n-fang YÃ¼ 2001: 75) in a variety of Indic, Chinese, Tibetan and other languages, and a rich manuscript tradition of over sixty copies have survived in Sanskrit alone, which are available for study (Tsukamoto et al. 1986: 5). Given the importance of the text, one would expect an assiduous attention to detail in its transmission, especially with regard to the dhÄraá¹Ä«s whose accurate recitation was essential for ritual efficacy. In the Lotus samÄdhi ritual for example (Fahua sanmei chanyi æ³è¯ä¸æ§æºå), a twenty-one day repentance ritual in the Tiantai tradition involving recitation of the SDP, mistakes in the recitation were simply not permissable (bude miuwu ä¸å¾è¬¬èª¤).¹¹ But in fact, there is no uniformity amongst the Indic versions themselves, nor between these and the texts translated into Chinese. In fact, the differences are often more striking than the similiarities, especially when comparing the two earliest surviving Chinese translationsâthose of Dharmaraká¹£a in 286 CE and KumÄrajÄ«va in 406 CEâto the extant Sanskrit MSS. This lack of correspondenceâand the fact that the Sanskrit manuscripts are all fairly lateâpoints to an earlier lost manuscript tradition on which the Chinese translations relied.

3.1 Textcritical Background

Scholars recognize three major recensions for the Saddharma-puá¹á¸arÄ«kasÅ«tra Indic text: Nepalese, Gilgit and Central Asian (Bechert in Chandra 1976: 3; Karashima 1992: 12). The closest thing we have to a critical edition of any of these is Kern and Nanjioâs edition (1908â1912, hereinafter K & N), which, however, was based on only eight manuscripts (seven Nepalese and one Central Asian) and does not include all variants. None of the Nepalese texts are earlier than the eleventh century (Tsukamoto et al. 1986: 9); the Gilgit manuscripts date from the early sixth century and belong to a recension similar to the Nepalese (Watanabe 1972â1975: xi); the earliest of the Central Asian manuscripts date from the fifth or sixth century as well (Tsukamoto et al. 1986: 24; Dutt 1953: viii). These manuscripts are linguistically earlier than the Nepalese and Gilgit recensionsâcomposed before major Sanskritisation had taken placeâand contain hundreds of Prakrit forms, some of which are detailed in K & N and Duttâs later edition (K & N 1972â1975: vi f.; Dutt 1953: xvii f.; Karashima 1998: 49â68; Karashima 2001: 207â230). In his exhaustive study on Sanskritised Prakrit, which he calls Buddhist Hybrid Sanskrit (BHS), Edgerton has established a time scale which confirms that the more Prakritisms a manuscript contains, the earlier it is.¹² As stated above, there are three extant Chinese translations of the SDP: one by Dharmaraká¹£a in 286 CE, one by KumÄrajÄ«va in 406 and a third by JÃ±Änagupta and Dharmagupta in 601. There is also a partial translation, the Satanfentuolijing (T.9265), thought to predate Dharmaraká¹£a (Karashima 2003: 87). The earliest manuscript evidence we have of the Chinese translations is from Dunhuang (fifthâtenth centuries). Several Tibetan versions preserved in the various Kanjurs are presumed to date back to the early eighth century (Karashima 1992: 13).

It has long been assumed that KumÄrajÄ«vaâs translation was based on the Nepalese/Gilgit recension, while Dharmaraká¹£aâs translation was based on the earlier Central Asian manuscript (Bechert in Chandra 1976: 6; Baruch 1938: 41). However, recent studies by Karashima have problematised this view. He has shown that not only Dharmaraká¹£aâs source text, but also KumÄrajÄ«vaâs, are based on manuscripts containing a lot of Prakrit material and that in fact KumÄrajÄ«vaâs translation agrees with the earlier Central Asian recensions in 409 instances versus only 138 instances of agreement with the Nepalese/Gilgit recensions. The corresponding numbers for Dharmaraká¹£aâs translation are 622 agreements with Central Asian MSS and 230 with the Nepalese/Gilgit recensions (Karashima 1992: 256, 260). Therefore both KumÄrajÄ«vaâs and Dharmaraká¹£aâs source texts are assumed to predate the existing Sanskrit MSS, dating from a time before full Sanskritisation had taken place. In this regard, a very useful text for this studyâand one that confirms Karashimaâs findings of the Prakrit nature of the source documents at this timeâis KumÄrajÄ«vaâs translation of the arapacana syllabary in the Dazhidu lun¹³ where we find unique Prakritisms like aânoubotuo é¿è¨æ³¢é (PB: Êa-nÉw^h-pa-da) anuppÄda < Skt. anutpÄda, ânon-arisingâ, with simplification of conjunct -tp- > -pp-; zheliye é®æ¢¨å¤ (PB:tÉiaw-li-jia^h) cariya < Skt. carita, âpracticeâ, with lenition of intervocalic stop -t- > -y-; dusheta è¼éä» (PB: dÉ-dÊia-t^ha) < á¸ajÌamaá¹o¹⁴ (GDhp 75-d, 159-b) < Skt. dahyamÄna, âburningâ, with simplification of conjunct hy- > -jÌ-; hebota åæ³¢ä» (PB: É£wa-pa-t^ha) < vappatha < Skt. vÄkpatha, âfit for speechâ, with simplification of conjunct -kp- > -pp-, to name only a few (see also Appendix 1).

Given the rich and complex textual tradition, it is evident that there is no such thing as a single, monolithic SDP. There has not even been an attempt to create a critical edition (i.e. a reconstructed âoriginalââor as close as possible to the original per Maas 1958: 1âa text based on elements from all known sources); the complex tapestry of witnesses, multiple recensions and Sanskritisations and contaminatio (combination of several exemplars per Maas 1958: 3) suggests that any attempt to re-construct an Urtext would be impossible. Yet, given the importance of the SDP in East Asian religious traditions, it is a âsorry state of affairsâ that we do not make use of âall available resourcesâ when studying the text (Pye 2003: 168). Recently, this lacuna in SDP studies has been partially rectified by Karashimaâs partial publication of a trilingual edition of the SDP from all Sanskrit, Chinese and Tibetan sources (Karashima 2004: 33â104; 2006: 79â88).

From the Preface to JÃ±Änaguptaâs and Dharmaguptaâs translation Michael Pye hypothesizes that KumÄrajÄ«va worked from a Kuchean text which may well have been older than Dharmaraká¹£aâs source text. Pye arrives at this conclusion based on three factors: 1) material that was left out of KumÄrajÄ«vaâs original translation and subsequently added, 2) the separate numbering of the Devadatta chapter and 3) the arrangement of the last seven chapters (Pye 2003: 170). Dharmaraká¹£aâs translation represents a later stage of the textual tradition, but at an earlier date. Other scholars have agreed with Pye on philological grounds. Although it may be impossible to confirm the absolute chronology of the underlying source documents, one fact seems certain: based on linguistic and text historico-critical analysis, both Dharmaraká¹£aâs and KumÄrajÄ«vaâs source documents are earlier than the surviving Indic witnesses.

4 Methodological Considerations

4.1 Purpose of this Study

KumÄrajÄ«vaâs transcription of the SDPâs dhÄrÄá¹Ä«s represents a âfresh effort at transcriptionâ of MI phonemes into EMC (Pulleyblank 1983: 87).¹⁵ It therefore provides a unique opportunity to study the nature of KumÄrajÄ«vaâs transliteration practice and to identify some of the characteristics of the underlying MI source transmission dialect, while at the same time comparatively examining the main textual traditions, their differences and ambiguities. Epitomizing a central theme of the SDPâthe injunction to memorize and repeat exactly the content of the sÅ«traâthe dhÄraá¹Ä«s highlight the challenge of transmitting the Buddhadharma accurately over time, especially between a phonographic (MI) and logographic (Ch.) writing system. A tentative reconstruction of the dhÄraá¹Ä«sâ MI form is a further by-product of the study.

4.2 Methodology

This study involves the use of transcriptional data and reconstructions of the phonetic structure of EMC, based on Karlgrenâs and Pulleyblankâs work. Many scholars have questioned the validity of this approach; ZÃ¼rcher omits the use of transcribed names and Buddhist technical terms from his study on âLate Han Vernacular Elements in the Earliest Buddhist translationsâ, noting that their value is greatly reduced by a whole range of âobscuring factorsâ including 1) our ignorance of the source language, 2) distortion due to pronunciation by foreign missionaries, 3) the imperfect way the Chinese scribes may have perceived the sounds, 4) the differences in Sanskrit and Chinese phonology which make correlation of sounds problematic and 5) primitive early translation attempts which were later subsequently refined (ZÃ¼rcher 1977: 179). This last objection does not apply to KumÄrajÄ«va who had the benefit of two centuries of previous translatorsâ experience. He was also well aware of all the phonological issues involved, as his translation of the arapacana syllabary from the Dazhidu lun shows (Appendix 1); although there is certainly no exact one-to-one correspondence between Indic and Chinese languages, KumÄrajÄ«va was aware of the ambiguities and developed means of dealing with them, as will be demonstrated below. As for the source language(s), in the last thirty years there have been a lot of advances in our understanding or the underlying Prakrits, as outlined above. This is not to fully answer ZÃ¼rcherâs objectionsâespecially points 2) and 3), which are intractable; however, that we can and do know quite a bit about EMC phonology is manifest in the works of Karlgren, Pulleyblank and Coblin, and if we use the data judiciously, as Coblin recommends (1983: 7â8), using it to corroborate what we already do know from other sources (as ZÃ¼rcher recommends),âi.e. the phonological changes between Prakrit and Sanskrit,âthen the results can be very revealing.

4.2.1 DhÄraá¹Ä« Comparison

The main body of this chapter is a comparison and discussion of the linguistic form of all the SDPâs dhÄraá¹Ä«s. In what follows I list the Sanskrit texts in the three traditions, grouping the Nepalese and Tibetan together in column #1 with all variants shown in brackets;¹⁶ here I also include the Tibetan as separate items (marked âTib.â) when it differs from one of the Nepalese manuscripts, as is sometimes the case. Column #2 contains the Gilgit manuscripts (from Watanabe 1972â1975). Column #3 is the Central Asian manuscripts, some of which are shown in Kern & Nanjio and Dutt, and all of which are shown in Tsukamoto 1978.¹⁷ Column #4 shows KumÄrajÄ«vaâs Chinese translation taken from the TaishÅ (JunjirÅ et al. 1924â1934 [1974]), with assistance from CBETA and with variants shown in square brackets. The Chinese characters are transcribed in modern pinyin, and further transcribed phonetically as they sounded in Middle Chinese using Pulleyblank or Karlgren. Column #5 is a reconstruction of the source document text which KumÄrajÄ«va used, based on his transliteration. Column #6 is the meaning of the word, where known (or conjectured), and column #7 is a note on whether the word from the source document is Sanskrit or Prakritic in origin, together with any short notes that might be applicable. Longer notes follow the relevant entries. Where an entry is blank, it is missing in the appropriate document. While the dhÄraá¹Ä«s of the SDP have been transliterated four times, the earliest transliteration is KumÄrajÄ«vaâs. Dharmaraká¹£a translated the dhÄraá¹Ä«s in his Zhengfahuajing æ£æ³è¯ç¶, (T.263, 286 CE), but this is not a transcription. JÃ±Änagupta and Dharmagupta also transliterated them in their translation, TiÄnpÇn MiÃ ofÇ LiÃ¡nhuÃ¡jÄ«ng æ·»åå¦æ³è®è¯ç¶ (T.264, 601 CE); both of these are reviewed occasionally when they might be helpful in clarifying KumÄrajÄ«vaâs practice and/or intention. In addition to these testimonies, there are three other transcriptions made by JÃ±Änagupta, Xuanzang and Amoghavajra (early seventh to eighth centuries) which I have not referred to (but may be found in Karashima 2001: 380â392).

4.2.2 A Note on Vowel Notation

The transliterations are not consistent with respect to vowel notation. Sometimes the EMC phonetic sound which PB transcribes as [i] is used to represent Sanskrit/Prakrit final -e and sometimes -i. The phonetic sound [É]- is used for Sanskrit/Prakrit -e-, -a- and -i-. I assume these variations reflect dialectical variations, idiosyncrasies, allophones, etc. prevalent at the time the translations were done, which I have not tried to unravel. For derivational purposes, the consonants are much more important than the final vowel, which tends to be very variable in the dialects.¹⁸ Also, since long vowels were not notated in GÄndhÄrÄ« and even BrÄhmÄ« (GDhp 20; Norman 1997 [2006]: 107), and Chinese does not maintain the difference between long and short vowels, reconstruction of vowel length differences must be considered tentative at best. When I transcribe with long vowels in the hypothetical source document, it is usually based on Sanskrit parallels (e.g. Pkt. nÄá¸·i < Skt. nÄdi) or accent/sandhi rules (e.g. Pkt. kauÅalyÄnugada < Skt. kauÅalya-anugata or Skt. Åamita-avi = ÅamitÄvi).

5 DhÄraá¹Ä«s

5.1 DhÄraá¹Ä« #1 Spoken by Bhaiá¹£ajyarÄja è¥ç (Yaowang)¹⁹

a Tibetan is only noted when it is different than the Nepalese recension which it generally mirrors.

b The brackets a(c)chaye simply indicate that the double consonants were often not notated in G. or early BrÄhmÄ«. The double consonant represents the derivation from two consonants (-ny- > -Ã±Ã±-), which was, however, not noted in the early script.

In the Prakrits, a glide following a nasal is assimilated (-ny- > -Ã±-).²⁰ However, the sound is virtually identical, so one cannot be sure what word was in the source document. All forms ending in -e may be construed as an eastern Prakrit nominative singular.²¹ It may also be northwestern as there are lots of examples in nom. sing -e in the ShÄhbÄzgaá¹hÄ« (Sh) and MÄnsehrÄ (M) AÅokan rock edicts²² and in the Niya dialect the original nom. ending was probably in -e, although it later changed to -o.²³ The word anye could also be nom. plural in Sanskrit, PÄli and other Prakrits. While the meaning and syntax (if any) of these words and phrases is highly speculative, Tsukamoto (1978: 4f) seems to interpret the -e forms as voc. sg. fem. which is possible for some words which are feminine (anyÄ, âinexhaustibleâ; manyÄ, ânape of the neckâ), but not for nouns like citta (masc.) or ká¹£aya (masc.), nor for words like carita that appear to modify them. These would have to be loc. sing. or nom. sing. if stemming from an eastern Prakrit. The verb manye (âI thinkâ, first person sing. of âman) is a much more logical meaning than âOh! nape of the neckâ, voc. fem. sing (< Skt. manyÄ, ânape of the neckâ). In the translations that follow, I treat the -e endings as nom. sing. unless otherwise stated.

a Nom. sing. (eastern Pkt. nom. ending in -e); long -Ä- not written in G. This could also be derived from manas (P. mano, AMg maá¹a, âmindâ) or Skt. manÄ (âzeal, devotionâ) in voc. sing. as per Tsukamoto 1978: 4.

The change of -t- > -d- seems to be the first unequivocal evidence that we are dealing with a source document which is (in part at least) in a Prakrit form. Voicing of intervocalic consonants is a standard feature in GÄndhÄrÄ« (GDhp 33), and Pulleyblank (1983: 86â88) notes that intervocalic -t- is âquite consistentlyâ rendered by Ch. -d- in the Saddharmapuá¹á¸arÄ«kasÅ«tra dhÄraá¹Ä«s. We will see several other examples of this feature below, where the source document has Prakrit forms and the extant Indo-Aryan (IA) reflexes have Sanskrit words, like idime for Skt. itime, mÄdaá¹gÄ« for mÄtaá¹gÄ«, daá¹á¸avadi for daá¹á¸apati, etc. The word citte is a puzzle as it appears to be representing a source word *cile (=cire, âlong?â). The character é· is always used to represent an -l- sound or a vocalic -á¹- or consonantal -r- by KumÄrajÄ«va, but here it may be a translation rather than a transliteration (æ¨é· = lit. âcontrol oneâs intentionâ). JÃ±Änagupta and Dharmagupta have zhidi è³ªç¾ (PB: tÉit-tÉj = citte).

a The character å© is not in Pulleyblank or Karlgren. It is also not in the Guangyun shengxi å»£é»è²ç³», Song rhyming dictionary: http://ctext.org/library.pl?if=en&res=77357&by_title=%E5%BB%A3%E9%9F%BB. Accessed Nov. 2014.

b In all the Skt. texts (K & N, Wogihara & Tsuchida, Vaidya and Dutt) the word division is after -tÄË i.e., samitÄ viÅÄá¹te; other variants in the NepaleseâTibetan tradition include samitÄ visÄá¹te; samite viÅÄnte.

c A second interpretation of the ç sound is in KG: page 201, entry 2213. Coblin 1994: 246, subentry 0405, transliterates this character as *ui in Old Northwest Chinese (ONWC).

Dharmaraká¹£a translates these three words (same ÅamitÄvi ÅÄnte) as feng xiu jiran å¥ä¿®å¯ç¶ (âEsteem & cultivate quiescenceâ); the word å¥ (âesteem, revere, respectâ) is perhaps a translation for Skt. ÅÄlita (âpraisedâ). This would be the normal transliteration of shelÃ¼duo è³å±¥å¤ (PB: Éia-li-ta), i.e. with å±¥ representing the sound -li- as per PB and KG. In Soothill & Hodous (1937; S & H) we find words like tipilÃ¼ é«æ¯å±¥ (PB t^hÎµjâ-bji-l(r)i), Skt. sthavira, âelderâ, or modalÃ¼jia æ©æå±¥è¿¦ (PB: ma-tat-li-kÉ¨a), Skt. mÄtá¹kÄ âsummary, condensed statement of contentsâ where å±¥ = -r-/-á¹- and bibeilÃ¼ye èåå±¥ä¹ (PB: pjiajk-pji/pjiÉÌ-li-jiaâ), pipÄ«likÄ, âantâ where it represents an -l-. How å±¥ came to represent the -m- sound is a mystery. Pulleyblank is also puzzled and suggests that it represents an âold reading of the character that has gone unrecorded in the dictionariesâ.²⁴ In JÃ±Änaguptaâs and Dharmaguptaâs redo of the sÅ«tra two centuries later they transliterate ÅamitÄvi as æå¯å¤é¼» (PBË Éiap-mji^h-ta-bji^h), where there is no mistaking that the second syllable begins with m-. Note also for this section of the dhÄraá¹Ä« that all four versions of the Sanskrit texts transliterate ÅamitÄ viÅÄnte which is probably incorrect, as there is no such word as visÄnte in Prakrit or Sanskrit, while there is such a compound as Åamita-Ävi.

a Alternately, instead of sam-asame, this may be parsed as a distributive repetition (samasame) with a meaning of âequal in every wayâ.

b It would be noted -ká¹£- in G. and -ch- in other Prakrits. See the discussion.

The word aká¹£aye occurs later in the dhÄraá¹Ä« where KumÄrajÄ«va transliterates it as echaluo æ¡åé (PBË Êak-tÊ^haÉ¨ /tÊ^hÎµË-la= aká¹£ara), i.e., he captures the -ká¹£- conjunct perfectly, as -k- is an allowed final in EMC. Why did he not do it here? With ká¹£aye and aká¹£aye he transcribes with a retroflex affricate sound tÊhaÉ¨- and with aká¹£iá¹e he uses a velar stop with a glide -gji-. GÄndhÄrÄ« used the symbol ð¨ð¨¿ð¨® to represent Skt. -ká¹£- and it had the value of a retroflex unaspirated fricative (á¹á¹£ or aspirate á¹á¹£â),²⁵ which is how KumÄrajÄ«va transcribed it, i.e. with a sound usually represented by -(c)ch- in MI (as in chudaá¹ Gir, Rock Edict 9 < Skt. ká¹£udra, âsmall, trifling;â PÄli has both khaá¹a and chaá¹a as derived forms of Skt. ká¹£aá¹a âmomentâ). This is also the sound (å = tÊ^haÉ¨/tÊ^hÎµË) which KumÄrajÄ«va uses in his arapacana syllabary to represent Skt. -ká¹£-.²⁶ This suggests that his source document was in GÄndhÄrÄ«. The Sanskrit word aká¹£ine was not transcribed as a retroflex fricative but as a voiced aspirated stop, pronounced and/or written aghiá¹e in the source documentâwe have examples of this in PÄli where -ká¹£- > -(c)ch-, -(j)jh- as well as -(g)gh-; for example, Skt. ká¹£Äyati > P. jhÄyati and ghÄyati (âit is consumedâ). KhÄ«á¹a is the normal PÄli reflex of Skt. ká¹£Ä«á¹a, but jhÄ«á¹a also existed as a form, and possibly ghÄ«á¹a which is the same sound with [-back] > [+back].²⁷ It appears that the conjunct ká¹£- could be pronounced several ways in KumÄrajÄ«vaâs time, according to its dialectical origin. As the language of Buddhism became more and more Sanskritisedâby Xuanzangâs time, for example, in the seventh centuryâthe conjunct was always captured by a two-character sound; but the fact that KumÄrajÄ«va sometimes transcribes it with a retroflex fricative alone, and sometimes with a stop followed by a fricative suggests that he was making a deliberate distinction according to his understanding of the word and its pronunciation.²⁸

a Dharmaraká¹£a translates this as guancha guangyao è§å¯åè âobserve the splendourâ.

b Hultzsch 1925 [1969]: 117; Bloch 1950: 141. Jaugaá¸a is located in Eastern India in Kalinga. Another instance of -k- > -g- occurs in a Ch. translation of the Dhp 97 compound akataÃ±Ã±Å« (âknowing the uncreatedâ) which is translated in the Ch. version of the Abhidharma-jÃ±ÄnaprasthÄna-ÅÄstra as ä¸å¾ç¥ (bÃ¹ wÇng zhÄ«, ânot knowing what is goneâ or perhaps âknowing the not-yet frequented,â i.e. the dominion of death), indicating that the Ch. redactor had the Pkt. form agata- in front of him/her, rather than akata-. See Minoru Hara 1992: 185.

5.1.1 pratyaveká¹£aá¹i

The -ká¹£- conjunct in pratyaveká¹£aá¹i is treated the same as in aká¹£aya above, using a retroflex sibilant to express the sound. The -ty- had become palatalized and changed to -cc- as also occurred in PÄli (pa(c)cavekkhana) and GÄndhÄri and other Prakrits.²⁹ Although PÄli and all the other Prakrits lost the -r- in pr-, GÄndhÄri kept it (e.g. Skt. pratyaya > G. praceâa in GDhp 88-b), and in the NW AÅokan edicts of Sh and M it was sometimes retained and sometimes assimilated. Two and a half centuries later, when Sanskritisation was much more prevalent, this conjunct was regularly represented by two characters, e.g. in Xuanzangâs transliterated version of the PrajÃ±ÄpÄramitÄhá¹dayasÅ«tra, where the pr- in prajÃ±Ä is represented by two characters, one for p- and one for -r-: boluoeârang é¢ åèª æ (PB: pat-la-Åa/nga-É²É¨aÅ).³⁰

5.1.2 nidhiru

Prakrit (attested in Mylius 2005: 332) has á¹iviá¹á¹he; the Central Asian MS has niviá¹£á¹e. Final EMC -s had disappeared by this time in the north of China (Pulleyblank 1983: 87), so it is unclear whether KumÄrajÄ«va could have captured it with the tools at his disposal, although presumably he could have inserted a character starting with an s- to capture the sibilant sound. It does, however, appear that KumÄrajÄ«va had a Prakritic source document based on other evidence. The large number of variant forms of the Sanskrit word nidhiru shows that there was a lot of confusion concerning this form, which may be attributable in part to the alteration of -dh- >< -v- which is not uncommon in Prakrit. Norman lists several example of this in one of Buddhismâs oldest texts, the Sutta NipÄta and it is also present in the MahÄyÄna texts.³¹ The Sanskrit letter -r- cannot be pronounced by the Chinese and is automatically changed to an -l- sound which also might further mutate to a retroflex á¹ or á¸ in the Prakrits, although the change from -á¹- >-á¸- > -l- is far more common.³² The sounds -á¸a- and -la- are very similar and apparently were confused, judging by the many variants: nivita, niviá¸a, etc.

a The character å is not found in PB or KG. The åå spelling given in the TaishÅ is å¤å¯ or é½é¤, which I transliterate as âtaâ, as both å¤ and é½ have the EMC phonetic value of âtaâ or âtÉâ. However, Coblin identifies this character as a retroflex -á¸- reconstructing EMC phonetic value âá¸jeâ (1983: 164, #210), based on Xu Shenâs work (2nd century CE) in the Shuowen jiezi èªªæè§£å, an early Han Dynasty Chinese dictionary. If indeed the sound is voiced, this would be further evidence of a Prakritic influence (which tends to voice voiceless intervocalic consonants). In his later âCompendium,â Coblin (1994: 119, sub entry 0001) gives å as Qieyun tÃ¢.

Hurvitz transcribes the second word in this compound as niviá¹£á¹e, same as the immediately preceding niviá¹£á¹e, but it is not clear why KumÄrajÄ«va spells it differently this time, using å±¥ (usually signifying the sound li, but also used for vi and others)³³ for the second syllable where before he used æ¯ (-bji-; i.e. nÎµjâ-viâ-t^hÎµj^h vs^. nÎµjâ-bji-t^hÎµj^h). It certainly suggests a difference in the source text spelling, which is not immediately apparent. JÃ±Änagupta and Dharmagupta transcribe åé¼»ç (nÎµjâ-bjih-Êit)³⁴ which seems to be an attempt at transcribing niviá¹£á¹e.

There are two traditions here, abhyantarapÄriÅuddhi (âcomplete purification insideâ) and atyantapÄriÅuddhi (âperfect purificationâ). KumÄrajÄ«va has followed the second one. Coblinâs comment (see footnote 29 above) that dentals followed by a -y- were palatalized is not true in this instance, where the -y- has simply dropped off. It is clear that KumÄrajÄ«va could have represented the -ty- conjunct if he wished as -t- was a permitted final and he had characters like é¼ and å»¶ (PB: Êat, jian-), but did not use them. In GÄndhÄrÄ« -ty- usually changes to a -c- (ká¹tya > kica in GDhp 48-b), but also sometimes the -y- is just dropped as in this instance (e.g. GDhp 263-a maá¹uÅa < Skt. mÄnuá¹£ya). Alternatively, the future tense in GÄndhÄrÄ« regularly changes the -sy- > -á¹£-, as in GDhp 301-d payeá¹£idi, âhe will collectâ). This also happens in other Prakrits, for example, in AMg, where Skt. pratyeka > patteáºa (Pischel Â§80), and is seen in the AÅokan edicts, e.g. Rock Edict 5 B in GirnÄr (Gir) and ShÄhbÄzgaá¹hÄ« (Sh) where Skt. kalyÄna > Gir kalÄá¹assa and Sh kalaá¹asa.

If the m- in the Anlaut of the first mutkule is taken as the accusative singular of the previous word pÄriÅuddhim, then the utkule mutkule phrase would agree with the Chinese and the Tibetan (which has mutkule mutkule). The Chinese version accords with the Gilgit manuscript (and some of the Nepalese manuscripts) which preserve the older non-Sanskritised form (Pkt. -kk-), which was later Sanskritised to -tk-. The final -ut was permitted in EMC and KumÄrajÄ«va had access to logographs like è´ (PB: mut-) which suggests that he did not have this reading in his source document. The meaning is not clear; utkula means âan outcasteâ whereas utkÅ«la means âsloping up, highâ. Long syllables were not marked in GÄndhÄrÄ«âs Kharoá¹£á¹hi script and mukkule could simply be a euphonic -m-, often introduced in the Prakrits as a substitute for Sanskrit sandhi (euphonic junction between words); that is ukkule ukkule > ukkula ukkule (where the final -e in ukkule > -a because of the following vowel u-) would exemplify the normal connection between two nouns, one of which ended in -e and the second beginning in another vowel (u-); however, with the loss of sandhi rules this could also become ukkule-m-ukkule (Geiger Â§73.2).

A common change from Sanskrit to Pkt. is -á¸- > -á¸·- and we may be fairly certain that this is what has happened here as KumÄrajÄ«va had specified the character è¼ as the transliteration for á¸a in his translation of the arapacana syllabary, which character he could have used if his source document had araá¸e or paraá¸e, as in the Sanskrit. But he uses li é· instead, which he only uses to represent the vocalic liquids or consonants. Change of á¸- > -á¸·- is very common in P. (Geiger Â§35) and also occurs in the AÅokan edicts.³⁵ In the language of the Niya Documents (G), the letter -á¸- was either pronounced as a voiced retroflex fricative (= Ê), as an -á¹-, or as an -l-, in the case of loan-words incorporated into Khotanese Saka (Burrow 1937: Â§18) which may have been one of the languages KumÄrajÄ«va (a Kuchean) spoke, Kucha being on the north side of the Taklamakan Desert and Khotan on the south, presumably with constant interchange between the two caravan destinations. The meaning of araá¸·e/paraá¸·e is uncertain. Dharmaraká¹£a seems to associate it with turning: wuyou huixuan, suo zhouxian chu ç¡æè¿´æï¼æå¨æè, but it is not clear where he gets this derivation. JÃ±Änagupta and Dharmaguptaâs transliteration is similar to KumÄrajÄ«va: anluodi boluodi é éç¬¬ é¢éç¬¬ (PB: Êat-la^h-dÉjâ, pat-la^h-dÉjâ), preserving the -l- sound.

a Before a double consonant the vowel in Prakrit would always be short (Geiger Â§5) although it had the value of two morae. This probably means that the form KumÄrajÄ«va had in front of him was ÅukÄchi, or Åukacchi, but not ÅukÄcchi with both double -cch- and long -Ä-. In AMg this word appears as -kaá¹khÄ (P. -kankhÄ, both with short -a-) which shows the eastern change -ká¹£- > -(k)kh-; however other words like AMg kaccha (âforestâ < Skt. kaká¹£a) show the western form -ká¹£- > -(c)ch-.

Here again KumÄrajÄ«va uses a single sound (the character å·®, a retroflex fricative) to represent the conjunct -ká¹£-. It is not clear why he did not use the character å as in aká¹£aya above, but both appear to be almost identical phonetically (å = PB: tÊ^haÉ¨/tÊ^hÎµË, å·® = PB: tÊ^haÉ¨^h-tÊ^hÎµË^h). The compound ÅukÄ/Äká¹£i (ÅukÄ/Ä(c)chi) could also come from Åuka-aká¹£i (âeye of a parrotâ) which makes no sense in this context. It is much more likely that it derives from Åu-kÄá¹ká¹£Ä (âswift wishâ), where the -á¹- was omitted in the source document, as a long, open syllable was automatically nasalized in GÄndhÄrÄ« and the nasalization was often omitted in the written script (Fussman 1989: 478). Dharmaraká¹£a translates as qi mu qingjing å¶ç®æ¸æ·¨ (âtheir eyes are pureâ) taking the compound as derived from Skt. Åukra-aká¹£i which is not supported by any of the versions (which would have been spelt sukkÄká¹£i, with a double -kk- to account for the conjunct), although Tibetan (and Kernâs âKâ MS) has ÅrukÄká¹£Ä«, which might point to Åukra-aká¹£i (âpure eyeâ) by metathesis (Åruka-aká¹£i).

a So translated by Dharmaraká¹£a as deng wu suo deng çç¡æç.

The character zhi è¢ is not found in PB or KG. Hurvitz transliterates as buddhavikliá¹£á¹e, and analagous characters with the same radical (è£½) suggest a tÉiaj^h pronunciation which would almost fit (although a palatal sibilant instead of a retroflex one); however, the meaning does not seem apt for a dhÄraá¹Ä«, unless we are to take this in a Chan sense, i.e. positively (but of course this would be an anachronism). The clear Sanskrit meaning (buddhavilokite, âBuddha seenâ) is more appropriate. Dharmaraká¹£a renders jue yi yuedu è¦ºå·²è¶åº¦, âawakening to transcendenceâ so it is not clear what he was translating, but certainly not vikliá¹£á¹e. JÃ±Änagupta and Dharmagupta have bodibiluji (?) åå°é¼»ç§å,³⁶ (PB: bÉt-di^h-bji^h-lÉ-kjit-[?]) transcribing bodhi-vilokite (âenlightenment seenâ). The compound buddhavikliá¹£á¹e is an example of KumÄrajÄ«va using two characters to capture the -kl- conjunct (åå©) which suggests that the word being transcribed is either Sanskrit or Prakrit, where it would have been written with an epenthetic -i-, viz., -vikiliá¹he (G.) or vikiliá¹£á¹e (GDhp 60, with -á¹ha- = -á¹£á¹a-) and -vikiliá¹á¹he in PÄli and the Prakrits.

a Per Coblin 1983: 248, no. 173, tanmo ææ© (PBË dam/dÉm-ma) was the eastern Han transcription of dharma, which KumÄrajÄ«va inherited.

b Translated by Dharmaraká¹£a as [Ling]hezhong wuyin [ä»¤]åç¾ç¡é³ âthe silence of the Saá¹ghaâ.

The ká¹£- conjunct in Skt. dharmaparÄ«ká¹£ite is rendered as a single retroflex fricative (å·®) as in the previous compound ÅukÄ/Ä(c)chi. It appears that saá¹ghanirghoá¹£aá¹i was in the source document, as KumÄrajÄ«va has taken pains to translate the actual -rgha- conjunct, using an EMC character with final -t. This is a standard method of indicating an -r- with sarva (as he does below) and other common words (e.g. è©å©, PB sat-ba < Skt. sarva), so presumably he had a source text with the conjunct -rgh- which indicates a Skt. or Sanskritised text. The word nirghoá¹£aá¹i can either mean ânoisyâ or âwithout noiseâ (< Skt. nirghoá¹£a, ânoisyâ or âsilentâ). However as the consonant -r- > Ã in most Prakrits (but not always in GÄndhÄrÄ«),³⁷ this compound could also have a Prakrit source. Note that the word saá¹gha occurs without change in some Prakrits (e.g. P. saá¹gha, AMg saá¹gha, AÅokan Minor Rock Edict I, D, saá¹ghe), and in GDhp as saÇµÌa, where, per Brough, the letter -ÇµÌ- represents the sound of -ng- (GDhp 8, verse 102-d).

a Karashima (1992: 360) transliterates as EMC bwÃ¢-Åja-bwÃ¢-Åja-Åju di.

The Chinese spells out a word closest to the Central Asian manuscript; however the character è is usually used by KumÄrajÄ«va to represent the palatal Å, not the retroflex á¹£ as Karashima suggests (e.g. in the mantra of chapter 28 where è represents the palatal -Å- in daá¹á¸akuÅale). This would give us *bhaÅyÄbhaÅyaÅodhi which doesnât make sense; it is probably just an alternate form as we find both bhaÅadi, bhaá¹£ati and bhaá¹£adi used in GÄndhÄrÄ«.³⁸ JÃ±Änagupta and Dharmagupta transliterate bayebayeshudani è·è¶è·å¤è¼¸éæ³¥ (PB: bat-jia-bat-jia^h-ÉuÉ-dat-nÎµj^h; KG: puÃ¢-iÌa-puÃ¢-iÌa) which seems to transliterate bhaya. Dharmaraká¹£a has âWhat one states is very clear; be contentedâ.³⁹ The -á¹£y- or -Åy- conjunct suggests that this part of the source document was written in Sanskrit, as all the Prakrits would show -á¹£y- > -Å/á¹£/s- assimilation.

In this word KumÄrajÄ«va again makes use of what Xuanzang was later to name erhe yin äºåé³, or âtwo combined soundsâ to represent a conjunct consonant. This might also be the addition of an epenthetic vowel (i.e. *mantara) which is quite common in the eastern Prakrits (e.g. Pkt. ariya < Skt. Ärya, ânobleâ; Pkt. radaá¹a < Skt. ratna, âjewelâ), however the MI word *mantara or *mandara (= German Mandarin)⁴⁰ is not attested. Since there is a Prakrit form of this word (PÄli manta), we can assume that KumÄrajÄ«va was at pains to capture the full three-consonant conjunct which presumably he had before him.

The Chinese compound clearly ends in -ta, not -te like most of the Sanskrit versions and is âcorrect Sanskritâ for âmantrasâ (voc. pl.) ârule!â (2nd pers. pl. imperative). Dharmaraká¹£a has jin chu jie xian ç¡é¤ç¯é, which Karashima correlates with this section and translates as âone clears away segments and limits completelyâ. He suggests that Dharmaraká¹£aâs source document read matra or mÄtra (âmeasure, sizeâ) and -nt- > -t- in the Prakritic form that Dharmaraká¹£a had before him (1992: 236â37). This seems unlikely as nasals before stops are usually retained in Prakrit (Pischel Â§272), and the word matra occurs in the GDhp 17-b, 164-c, representing both its masculine (mÄtra) and feminine (mÄtrÄ) forms. The Prakrit for mantra would be closer to the P. manta. Dharmaraká¹£a seems to be saying that the use of mantras âeliminates limitationsâ, paraphrasing mantrÄká¹£ayata in terms of the result, which is typical of his translation approach to this dhÄraá¹Ä«. The second word in the compound (-ká¹£ayata) is treated by KumÄrajÄ«va the same as ká¹£aye above, using the retroflex fricative for the Sanskrit conjunct which is the sound it has in GÄndhÄrÄ« and other Prakrits. JÃ±Änagupta and Dharmagupta translate mandaluoqiye æ¼æéæ©å¤, (PB: muan^h-tat-la-k^hiaj^h-jia^h) which sounds like an alternate Prakrit form (-kkhaya or -khyaya) for -ká¹£aya, found in PÄli and AMg, where ká¹£- > (k)kh-.

While rute and rute kauÅalya have a clear meaning (âsoundâ and âsound and well-beingâ), the addition of the prefix u- is a puzzle, not present in any of the non-Chinese reflexes. It might be a Prakrit form of ava-,⁴¹ but avaruta is not attested either. Notice that the Central Asian reflex has a voiced intervocalic -d-, while all the other forms, including the Chinese have a voiceless dental (if indeed duo å represents such, which is not clear; see page 157, note a).

a Karashima (1992: 360) transliterates as EMC ÊÃ¢k tá¹£ha lÃ¢ (aká¹£ara).

Again KumÄrajÄ«vaâs transliteration stands apart from the Sanskrit reflexes, all of which have a different word, which has the same sense (âundecayingâ) as one of the meanings of aká¹£ara. We have seen above that KumÄrajÄ«va transcribed aká¹£aye as achayi é¿åè£, (PBË Êa- tÊhaÉ¨/tÊhÎµË-jiaj^h), omitting the -k- in the conjunct and treating it as one retroflex fricative sound; yet here he chooses to treat it as a conjunct, so it seems self-evident that he is trying to spell out the Sanskrit word aká¹£ara. The Prakrit form of this word is akkhara (P. and AMg), and there was probably a form *acchara, although not attested (as AMg accha < Skt. aká¹£a, âeyeâ is attested).

All the Sanskrit reflexes repeat the first word (aká¹£aye-) in the next compound (i.e. aká¹£aye aká¹£ayavanatÄye); however KumÄrajÄ«va changes aká¹£ara to aká¹£aya (aká¹£ayataya or aká¹£ayatÄya), while still preserving the dual consonants in the Skt. -ká¹£- conjunct. Hurvitz omits this word in his transliteration (2009: 296). The compound may be an oblique form of the Prakrit aká¹£aya-tÄ ending (âcondition of, state of imperishabilityâ).

a Karashima (1992: 360) transliterates as EMC ÊÃ¢ bwÃ¢ lwo (abalo).

b For -p-/-b- > -v- see Pischel Â§Â§199, 201; GDhp 34. See note on character å© below, under dhÄraá¹Ä« # 6.

Hurvitz transcribed é¿å©ç§ as avaru with a footnote saying that the Sanskrit has nothing to correspond to this (296, 364), but he was unaware of the Central Asian version, which the Chinese matches, with the usual Prakrit change of -b- > -v-. Dharmaraká¹£a has something similar: Yong wuli shi æ°¸ç¡åå¢, âone forever lacks strengthâ (Karashima 1992: 237). The last compound amanyanatÄya recapitulates the beginning (anye manye) in terms of sonic echo, if not in meaning. The member of the compound -nata, appears to be the past participle of ânam (âto bowâ), i.e. nata, in the dative case, which is often used as an infinitive form; if one takes manya- as Sanskrit âappearing as, thinking oneself to beâ then one may construe the meaning of the compond a-manya-natÄya, as âhomage to the non-appearance [of an I]â, but this is fanciful at best, although Dharmaraká¹£a has something similar: Wu suo sinian ç¡ææå¿µ, âlack of thoughtâ. Better to take it as a recapitulatory sonic echo of the dhÄraá¹Ä« beginning (anye manye).

In this first dhÄraá¹Ä« we have sixteen forms that could derive from either a Prakrit or Sanskrit source document, sixteen that derive from Prakrit and eleven from Sanskrit.

5.2 DharÄá¹i #2, Spoken by PradÄnaÅÅ«ra Yongshi åæ½

The difficulty here is determining what sound the character ç¤ represents, a single letter j- or a conjunct jv-. The fanqie (shi luo èªèº, PB: dÊiaj^h-lwa) suggests a single letter pronounced âdÊaâ which is similar to KumÄrajÄ«vaâs transliteration of ja in the arapacana alphabet, i.e. é (= PB: dÊia).⁴² KG transliterates ç¤ as dzâuÃ¢. Tsukamoto suggests KumÄrajÄ«vaâs transliteration = jale which is the Prakrit form of this word (e.g. AMg jala; Tsukamoto 1978: 19; Mylius 2003: 286). The character ç¤ (PB: dzwa) does suggest a slight labialization of the affricate dz-; however, since JÃ±Änagupta and Dharmagupta transliterate as shepoli æ¶ç¤ç (PB dÊip-ba-li), i.e. using two characters to capture the jv- in Sanskrit, it appears that they thought KumÄrajÄ«vaâs transliteration was Prakrit and Sanskritised it.

This is a puzzle. As Brough points out, âthe regular correspondence of the three Indian [i.e. GÄndhÄrÄ«] sibilants with the Chinese is strikingâ, yet here we have a palatal -Å- with a velar k- which never happens in Sanskrit or any of the Prakrits that I am aware (although in the AÅokan edicts, the KÄlsÄ« rock edicts use the sibilants á¹£ and Å where they are âphonetically and etymologically impossibleâ).⁴³ This of course might be a simple interchange of -Å- for -á¹£-, but KumÄrajÄ«va has not shown any âsloppinessâ in transliterating before. If he was trying to capture a ká¹£ sound in the source dialect, why didnât he use the character å (PBË tÊ^haÉ¨/tÊ^hÎµË), which he used in ká¹£aye and aká¹£aye above? Karashima (1992: 237) suggests a derivation of ukke from Sanskrit ulkÄ (âa meteor, fire-brand, torchâ). The rendition by JÃ±Änagupta and Dharmagupta is clearly Pkt. ééç®é, yujimuji (PB: Êuwk-kÉj muwk-kÉj) with a possible derivation < Skt. mukta/mukti. A possible explanation for KumÄrajÄ«va is that he was transcribing from a Prakrit where (m)ukta was pronounced (m)ukÎ´a or (m)ukza (i.e. as a fricative, as is the case in GÄndhÄrÄ«, per GDhp 43a), the sound of which he tried to capture with this character (æ³).

a Karashima 1992: 360, transliterates as EMC Êa-liei which he suggests represents *aá¸·e or *ale < aá¸e.

b Karashima 1992: 360 transliterates as EMC ÊÃ¢-lÃ¢ -bwÃ¢-diei, representing *aá¸·Ävadi or *alÄvadi.

See discussion under araá¸e parade above. The meaning is not clear. Tsukamoto suggests three possible derivations from ada (âeatingâ), Ädi (âbeginningâ) and from the root âaá¹ (âwander aboutâ), but none of these are convincing, because of lack of context (1978: 20). Dharmaraká¹£aâs âtranslationâ of this section appears to be shunlai [dang] fuzhang é ä¾[ç¶]å¯ç« meaning of which is unclear to me (âFollow, come and accept the chapterâ)? Karashima (1992: 237) correlates é ä¾ (âone comes obedientlyâ) with aá¸e < Skt. âaá¹ (âto wander aboutâ) and å¯ç« (âa piece of writing about wealthâ) with aá¸Ävati < Skt. Äá¸hya-pÄda, but the derivation of the latter is questionable).

While the preservation of the vocalic -á¹- indicates a Sanskrit derivation (as none of the Prakrits kept the vocalic -á¹-), the voicing of the voiceless dental, -t- > -d- in -vade, is a definite Prakrit feature. Dharmaraká¹£a translates as yuexi xinran æåæ¬£ç¶ (âhappy, joyfulâ), which seems like a gloss on ná¹de in its meaning âdanceâ (< Skt. ná¹tta).

a Neither of the characters æ³ or æ are in PB or KG so I have used the fanqie (å¥³æ°å) for the transliteration. According to the Guangyun, its sound is ni.

The double retroflex -á¹á¹(h)- is a common Prakrit form derived from Skt. -á¹£á¹(h)- which is how Karashima derives it;⁴⁴ there are, however lots of native Sanskrit words with the double retroflex consonants (e.g. paá¹á¹a = âclothâ; kuá¹á¹a = âbreaking, bruisingâ, etc.), so the evidence is not conclusive as to the source dialect. The meaning, as interpreted by Dharmaraká¹£a, is zhu ci li zhi yong [zhu]zuo ä½æ¤ç«å¶æ°¸[ä½]ä½ (âremains here, establishes, rules, and always actsâ); he also takes the words as derived from Skt. âsthÄ (Karashima 1992: 237).

The âdriâ sound is used by KumÄrajÄ«va for the retroflex -á¹-,⁴⁵ as was the case with the previous entry (ä¼ç·»æ = iá¹á¹ini). The last word æ¶[é·]çå¢å©åº (ná¹á¹á¹ivate) differs from the previous treatment æ¶é·å¤å©ç¬¬ (ná¹tavade) by only two characters, å¤ = ta and ç¬¬ = dÎµj^h, suggesting that KumÄrajÄ«vaâs source had a change here, as we have noted, although Hurvitz (2009: 296) transcribes them the same. The vocalic -á¹-, as mentioned above, points to a Sanskrit original, but von HinÃ¼ber suggests this is a Sanskritisation.⁴⁶ The meaning seems to be related to Skt. âná¹t (âto danceâ), however Dharmaraká¹£a translates wuhewuji ç¡åç¡é (âno joining, no gatheringâ).

In DhÄraá¹Ä« #2 most of the words have a Prakritic source. Ten are Prakritic in origin (including three âprobablyâ), two are questionable and two show elements of both Prakrit and Sanskrit.

5.3 DhÄraá¹Ä« #3 by VaiÅravaá¹a æ¯æ²é (Pishamen)

a The first character is not in PB or KG, but it is found in Coblin (1994: 264), sub-entry 0472, and he transliterates it as âprobable Qieyun nÉu,â and ONWC *nou. He notes that it is fairly common in ONWC texts.

b Karashima 1992: 238: < *anaá¸·o or *analo, EMC ÊÃ¢-nÃ¢-lwo = Dharmaraká¹£a wuliang ç¡é (âmeasurelessâ).

c Karashima (1992: 360) transliterates as ÊÃ¢ lji â¦ nÃ¢ ljiË kju nÃ¢ lji; he also reconstructs an original -á¸·- sound: *aá¸·e â¦ naá¸·i kunaá¸·i or *ale â¦ nali kunali.

All the above words show a change from retroflex dental to a retroflex -á¸·- which is typical of the Prakrits (Pischel Â§Â§238, 240; Geiger Â§35), so one may assume the Sanskrit forms have been Sanskritised at a later date and KumÄrajÄ«vaâs source document represents an earlier iteration with the Prakrit -á¸·-. This change also occurs in GÄndhÄrÄ« where, in the language of the Kharoá¹£á¹hi documents, -á¹- and -á¸- become -á¸Ì- (aspirant or fricative).⁴⁷ Brough represents this sound as [Î´] or [z] or as -r- which the Chinese translators would have heard as -l- (GDhp 42, 42a, 42b). This dhÄraá¹Ä« is translated cryptically by Dharmaraká¹£a as âwealth is tamed, game is without game, without measure (is) without wealthâhow (can there be) wealth?â⁴⁸ All six words in dhÄraá¹Ä« #3 point to a Prakrit source document.

5.4 DhÄraá¹Ä« #4 by VirÅ«á¸haka æåå¤©ç (Chiguo tianwang)

Both PB and KG transliterate ä¹¾ with a voiceless velar stop k-, suggesting kandhÄri in the source document, or at least an interpretation of the initial g- as voiceless. This may be due to the fact that the Kuchean language (KumÄrajÄ«vaâs native language) âignored the difference between voiced and voiceless consonantsâ,⁴⁹ but this fact is inconsistent with the fairly consistent practice of changing voiceless stops > voiced stops as noted above. Although Brough mentions that the -nd- conjunct usually changes to -nn- (written -n-, as in Skt. vindati > G. vinadi), in the AÅokan inscriptions we also find -ndh- > -dh- and also retained, as in the Northwestern Rock Edicts: RE 5 J MÄnsehrÄ has gadharana and ShÄhbÄzgaá¹hÄ« has gaá¹dha-ranaá¹ for the name of the GandhÄran peoples. In the Prakrits in general a nasal before a stop is usually retained (Pischel Â§272), but in the language of the Kharoá¹£á¹hi documents, loss or assimilation of the nasal before a stop is sporadic, as in the case of the AÅokan edicts.⁵⁰

a Pulleyblank gives the phonetics of è¹¬ as deÅ^h in 1983: 88. The character is not in PB.

b Reconstructed by Karashima 1992: 238.

c Karashima (ibid.: 360): EMCË zjang gjÉu lji-.

Although -r- is usually assimlated in the Prakrits (e.g. Skt. vrajati > P. vajati), it is not always assimilated in the Northwestern Prakrit GÄndhÄrÄ«, nor the language of the Niya Documents,⁵¹ so the dialect of the source document for vrusÅ«ni could be either Prakrit or Sanskrit.

Dharmaraká¹£a translates dhÄraá¹Ä« #4 as Wushu youshu, yao hei chi xiang, xiong zhou dati, yu qi shun shu, bao yan zhi you ç¡æ¸ææ¸ï¼æé»æé¦ï¼å¶åªå¤§é«ï¼äºå¨é è¿°ï¼æ´è¨è³æ (âInnumerable are the numbers. Sunshine and darkness hold perfume. A terrible curse is the main thing. By oneâs abilities, arrange and tell. Cruel words. Supreme existenceâ). Beyond the obvious meaning correlations (ç¡æ¸ = agaá¹e, ææ¸ = gaá¹e, æ = gori), the rest is obscure. Of the ten words in this dhÄraá¹Ä«, all except two could be from either a Prakrit or Sanskrit source.

5.5 DhÄraá¹Ä« #5 by the rÄká¹£asyaá¸¥ Luocha nÃ¼ ç¾åå¥³

a See note a on page 171. The character ã¹ (âhareâ) is only found in Coblin who gives the possible phonetic value ONWC *nou, which is no more explanatory than the aspirated stop value for its principal component å (t^hÉ^h), âhareâ.

Dharmaraká¹£a gives various fanciful renditions of the above words which do not correlate very well with any Sanskrit or Prakrit words: The itime sequence corresponds to yushi yusi yu er yu shi æ¼æ¯æ¼æ¯æ¼ç¾æ¼æ° (âIn this, in this place, in you, in the familyâ); the nime sequence to jishen wuwo wuwu wushen wu suo ju tong æ¥µçç¡æç¡å¾ç¡èº«ç¡æä¿±å (âno I, no self, no body, no object togetherâ);⁵² yi xing yi sheng yi cheng å·²èå·²çå·²æ, (âalready rising, already growing, already accomplishedâ) perhaps correlates with ruhe (< Skt. âruh, âto growâ); the remainder er zhu er li, yi zhu jietan, yi fei xiao tou, da ji wu de jiahai èä½èç«ï¼äº¦ä½åæï¼äº¦éæ¶é ï¼å¤§ç¾ç¡å¾å å®³ (âboth reside and stand, also to reside and sigh, also not to extinguish remnants (?), in the case of a severe illness, one should not increase itâ) presumably correlates with stuhe or haste (< âsthÄ, âto standâ or âtuh, âto painâ or âstu, âto praiseâ?) but exactly how is not clear. The voiced -d- in idime (when all the other witnesses have the voiceless -t-) suggests a Prakrit source for this word and the s- > Ã in the last three forms also confirm a Prakrit source, although the consonant is aspirated only in the last word.⁵³ Nevertheless, in the Northwest Prakrits the initial st- is generally preserved,⁵⁴ so a form like tahe (when all the other witnesses have stahe) may suggest derivation from a different Prakrit. In this last dhÄraá¹Ä« of Chapter 26 (Chapter 21 in the Sanskrit), all but two of the sequences appear to have a Prakrit source document.

In total, for this chapter we have the following:

5.6 DhÄraá¹Ä« #6 by Samantabhadra Puxian æ®è³¢

a Karashima (1992: 363) transliterates as EMC Êa-dÃ¢n-di-dÃ¢n-dÃ¢-bwÃ¢-di < *adaá¹á¸e daá¹á¸avadi < adaá¹á¸e daá¹á¸apati.

b See KumÄrajÄ«vaâs arapacana syllabary (Appendix 1) where å© = ba becomes -v- intervocalically. For -p-/-b- > -v-, see Pischel Â§Â§199, 201; GDhp 34 (âregular development -p-, -b- > vâ).

Per his syllabary (Appendix 1) KumÄrajÄ«va regularly uses the character å© for ba- and bha- initially and -va- intervocalically (e.g. SÃ pÃ³, PB: sat-ba < Skt. sarva, âallâ; huopoye ç«å©å¤ PB: xwaâ-ba-jia^h < Skt. hvaya < âhve, âto callâ), and this has been his practice in the dhÄraá¹Ä«s, e.g. aluopodi é¿ç¾å© (PB: Êa-la-ba-dÎµjâ), representing aá¸·Ävati above. For this group of words Dharmaraká¹£a has wuwo chu wo ç¡æé¤æ (âno I (anatta), eliminate the Iâ), which Karashima is suggesting be corrected > wu zhang chu zhang ç¡æé¤æ (âno staffs, removes the staffsâ). For daá¹á¸Ävartani he suggests another correction: yin wo å æ > hui zhang åæ (âswings around a staffâ; 1992: 246). JÃ±Änagupta and Dharmagupta have the same as KumÄrajÄ«va through this dhÄraá¹Ä« and appear to have copied it from the latter.

The conjunct -Åy- is usually assimilated to -Å- in GÄndhÄrÄ« and all the Prakrits.⁵⁵ It is not clear whether the -i- in the Éian transliteration is meant to represent a glide or simply a diphthong. In the examples above (ÅÄnte = PB: Éian-tÎµj^h), it certainly does not represent a glide and KG represents it phonetically as ÅiÌÃ¤n, where -iÌ- is defined as âthe subordinate vowel in a diphthongâ.⁵⁶ It is reasonable to assume, therefore, that KumÄrajÄ«vaâs source document had -paÅane, not -paÅyane. JÃ±Änagupta and Dharmagupta copy KumÄrajÄ«va. Dharmaraká¹£a has jian zhufo è¦è«¸ä½ âseeing all Buddhasâ.

KumÄrajÄ«va and previous translators use the character è© to represent the sound sat- in bodhisattva (i.e. è©è©), but KumÄrajÄ«va also seems to use it for the sounds sar- as in sarva (sapo è©å©, PB: sat-ba; see arapacana syllabary Appendix 1, s.v. sa). Since the normal GÄndhÄrÄ« reflex of this word is indeed sarva (with sava also used), this is probably the word in the source document and of course -r is not a permitted final in EMC, so he represented it this way. Later translators used three characters to capture the conjunct -rv- sound (e.g. Xuanzangâs saluofu è©åç¸ (PB. sat-la-buak, sarva; Shu-Fen Chen 2004: 129). The use of è© for the sound sar- suggests that the last word in the compound Ävatani did not have the -rt- conjunct shown in the Indic versions, as KumÄrajÄ«va does not use a character ending in -t (a permitted EMC final) but the character å© (PB: ba = MI va).

There are a few cases where KumÄrajÄ«vaâs source document agrees with the Central Asian manuscripts and not the Sanskrit (e.g. bhaá¹£yabhaá¹£ya- in dhÄraá¹Ä« #1). This is one such case, where there is also no Sanskrit reflex. Unfortunately many of the dhÄraá¹Ä«s in the Central Asian Manuscripts are missing. Dharmaraká¹£a has xing zhong zhu shuo è¡ç¾è«¸èªª (âput these many teachings into practiceâ; Karashima 1992: 247).

a See von HinÃ¼ber 2001 Â§Â§297 and 113 indicating a variation between -u- and -aá¹- in Prakrit. This is a feature of MI nasalisation.

b Karashima 1992: 247Ë saá¹ghaparÄ«ká¹£aá¹i or *saá¹ghavarÄ«ká¹£aá¹i; EMC sÉng-gja-bwÃ¢-lji tá¹£ha-á¹i.

This is another example of the character å© = Pkt. -va- intervocalically. The character å represents a retroflex fricative sound, not the Sanskrit conjunct -k- and -á¹£- sound (-ká¹£-), as discussed above in dhÄraá¹Ä« #1 (s.v. aká¹£aye).

Dharmaraká¹£a seems to translate these last two as âbring the assembly to an endâ, if we allow Karashimaâs (1992:247) proposed emendation of èè¿´è½ to åè¿´è½. The character æ¶ (PB: nÎµt) is used by KumÄrajÄ«va to represent a nir- or ná¹- sound (see saá¹ghanirghoá¹£ani in dhÄraá¹Ä« #1 or ná¹tye above in dhÄraá¹Ä« #2), as the character è© (PB: sat) is used to represent a sar- sound, which probably indicates a Sanskrit or Sanskrtised source document. Most Prakrits assimilate the -r- before a stop, but in GÄndhÄri, it is more often the case that the -r- remains, so this case is ambiguous;⁵⁷ the lenition of -t- > -d-, however, occurs only in Prakrit.

What does the character ç¥ represent? KG transcribes it as gâjiá¸ which suggests a -ghyi(a)- sound in Middle Indic (Ulving 1997: 213, entry 2607). In GÄndhÄrÄ« saá¹khy- appears as sagh- with the voicing of the stop, dropping of the anusvÄra and of the glide (GDhp 68-c saghaâi = Skt. saá¹khyÄya = P. saá¹khÄya âhaving examinedâ < Skt. saá¹+âkhyÄ, âto reckon, calculateâ, nominal form saá¹khyÄ, âcalculation, reckoningâ). So the source document word is asaá¹ghya, which shows the Prakrit change of -kh- > -gh-, but appears to preserve the glide -y- after the stop which is not a feature of the Prakrits. Dharmaraká¹£a renders wuyang shu, ji zhu ju sanshi shu deng ç¡å¤®æ¸ï¼è¨è«¸å¥ä¸ä¸æ¸ ç (âinfinite numbers; calculates phrases; is equal to the number of the three timesâ).⁵⁸

In the word tá¹adhvasaá¹ghatulya, KumÄrajÄ«va deliberately preserves the vocalic -á¹- which drops out in all Prakrits, including GÄndhÄrÄ«. Hurvitz transcribes this word as tiryÄdhasaá¹ghÄtulya (2009: 307), but a consonant + é· combination has been used above in dhÄraá¹Ä« #2 to represent the sound ná¹- sound (nielidi æ¶é·ç¬¬, for ná¹de), so it is more likely he was representing the Sanskrit word tá¹- in the Central Asian document.

a Karashima 1992: 391 leaves out é¿ç¾å¸ and transcribes å©ç¾å¸ as part of the the last compound (-prÄpte); however, the character ç¾ is always used as a separate syllable, not as part of a conjunct in all the transcriptions above. Tsukamoto (1978: 34) transcribes -pratte and has a question mark for é¿ç¾å¸.

This seems to be an echo of the pair from dhÄraá¹Ä« #1 (araá¸·e paraá¸·e), but the last character in each is different, representing a -t- sound in every other transcription (e.g. ç¾¶å¸ = ÅÄnte). Hurvitz transcribes with a retroflex araá¸e paraá¸e (2009: 307), but KumÄrajÄ«va specified the character è¼ for -á¸- in the syllabary (Appendix 1).

a See note 1 on page 171.

The -kr- conjuct is usually dropped in Prakrit (including GÄndhÄrÄ«), but it also sometimes remains as in Sh atikrataá¹ (RE 8 A).

In dhÄraá¹Ä« #6, there are eighteen items with a Prakrit source, seven with either and one with a Sanskritic source. The grand total⁵⁹ for all the dhÄraá¹Ä«s is:

5.7 Reconstruction

We may now with some confidence reconstruct the source document dhÄraá¹Ä«s which KumÄrajÄ«va had in front of him when he transliterated into Chinese:

a(Ã±)Ã±e ma(Ã±)Ã±e mane mamane cire caride Åame ÅamitÄvi ÅÄnte, mukte muktame same avisame samasame chaye a(c)chaye a(g)ghiá¹e ÅÄnte Åami dhÄraá¹Ä« ÄlogabhÄsa pac(c)ave(c)chaá¹i nivi(á¹)á¹he abhyantaranivi(á¹)á¹he a(t)tantapÄriÅu(d)dhi u(k)kule mu(k)kule araá¸·e paraá¸·e Åuka(c)chi asamasame buddhavikliá¹£á¹e dhammapari(c)chite saá¹ghanirghoá¹£aá¹i bhÄá¹£yÄbhÄá¹£yaÅodhi mantra mantrÄ(c)chÄyata uruta urutakauÅalya aká¹£ara aká¹£ayatÄya avala amanyanatÄya
jale mahÄjale ukÅe mukÅe aá¸·e aá¸·avade ná¹de ná¹tavade i(á¹)á¹ini vi(á¹)á¹ini ci(á¹)á¹ini ná¹(á¹)á¹ini ná¹(á¹)á¹ivate
aá¸·e naá¸·e nunaá¸·e anaá¸·o nÄá¸·i kunaá¸·i
agaá¹e gaá¹e gori kandhÄri caá¹á¸Äli mÄdaá¹gi jaá¹guli vrÅ«suni atte
idime idime idime idime idime, nime nime nime nime nime, ruhe ruhe ruhe ruhe, tahe tahe tahe tuhe thuhe
adaá¹á¸e daá¹á¸avati daá¹á¸avate daá¹á¸akuÅale daá¹á¸asudhÄre sudhÄre sudhÄravate buddhapaÅane sarvadhÄraá¹Ä«-Ävatani sarvabhÄá¹£Ävatani su-Ävatani saá¹ghavari(c)chani saá¹ghanirghÄdani asaá¹ghya saá¹ghÄvagadi tá¹adhvasaá¹ghatulya arate parate sarvasaá¹gha samadigrandi sarvadharmasupari(c)chite sarvasatvarudakauÅalyÄnu-gada siá¹havikrÄ«á¸ite.

6 Discussion

The numbers show that among the 105 items (words and compounds) analyzed, the dhÄraá¹Ä«s had a Prakrit item in the source document in about 53% of the cases with a Sanskrit one in 12% of the cases (with the remainder being either or indeterminable). The Prakrit:Sanskrit ratio is approximately 4.38:1 (57:13), which is higher than the ratios Karashima found in his study of agreement/disagreement with Central Asian manuscripts (2.2:1 for Dharmaraká¹£a and 2.7:1 for KumÄrajÄ«va; 1992: 254, 257). While these ratios are comparing different things, they do show that KumÄrajÄ«vaâs source document had much more in common with a Prakrit source document, than a Sanskrit one and Karashimaâs conclusionsâthat KumÄrajÄ«vaâs translation is closer to the Central Asian manuscripts which are known to be more Prakritic in nature⁶⁰ âpoint in the same direction. The high Prakrit:Sanskrit ratio may also indicate that the dhÄraá¹Ä«s received special attention in their transmission, in an attempt to guarantee their accuracy and efficacy. Since we know that the more Prakritisms a manuscript contains, the earlier it is, we may safely conclude that KumÄrajÄ«vaâs source document was earlier than the manuscripts of the Nepalese and Gilgit traditions, which are almost wholly Sanskritised. Heinrich LÃ¼ders (1916: 161), as mentioned above, believed that the âoriginalâ text of the Saddharmapuá¹á¸arÄ«kasÅ«tra was written in a pure Prakrit dialect which was afterwards gradually put into Sanskrit. We have argued that an original text of the Saddharmapuá¹á¸arÄ«kasÅ«tra is unrecoverable, because of the complexity of the transmission; however, that KumÄrajÄ«va had an earlier, more Prakritic text in front of him than the surviving Sanskrit witnesses is certain.

In addition to the large number of Prakritisms discernible in KumÄrajÄ«vaâs transliteration, the most striking phenomenon is the number of divergences from the existing manuscript traditions. There are several instances in which there are noticeable disagreements with the Nepalese/Gilgit recensions: in dhÄraá¹Ä« #6, for example, there are several words which only correspond to the Central Asian recension and are lacking in the Nepalese/Gilgit, i.e. sarvabhÄá¹£Ävatani, asaá¹gha, saá¹ghÄvagadi, tá¹adhvasaá¹ghatulya, sarvasaá¹gha, sarvadharma-suparÄ«ká¹£ate. There are also a number of words which correspond more closely with the Central Asian recension than the Sanskrit one:⁶¹

su-Ävatani su-Ävartane â saá¹vartane

sarvadhÄraá¹i sarvadhÄrani â dhÄraá¹i (dhÄraá¹Ä« #6)

atte agasti â sisi

jaá¹gali jÄá¹guli â saá¹kule (dhÄraá¹Ä« #4)

avala abale â valoá¸a

bhÄá¹£yÄbhÄá¹£yasodhi/bhÄá¹£yabhÄá¹£yasoddhÄ« â bhayÄbhayÄviÅodhani

niviá¹£á¹e niviá¹£á¹e â nidhiru (dhÄraá¹Ä« #1)

However, since parts of the dhÄraá¹Ä«s are missing in the Central Asian manuscripts, it is impossible to arrive at any general conclusions on the matter, except as already stated by Karashima, that the correspondence between KumÄrajÄ«vaâs translation and the Central Asian MSS is significantly higher than the correspondence with the Sanskrit versions.

Divergences are manifold in almost every entry. Sometimes these are minor, with a change in only one syllable or vowel (e.g. Skt. Åame or same = KumÄrajÄ«va Åami) and sometimes the words are barely recognizable (Skt. buddha-vilokite â KumÄrajÄ«va buddha-vikliá¹£á¹e) and clearly point to different manuscript traditions. Often within the two Sanskrit recensions (Nepalese and Gilgit) there are multiple versions of a word or compound (e.g. the second member of the compound abhyantara-niviá¹£á¹e in dhÄraá¹Ä« #1 where we find the variants -niviá¹he, -niviá¹£á¹he, -niviá¹£á¹a, -nirviá¹£á¹a, -visiá¹£á¹a, -viciá¹£á¹a, -piviá¹£á¹e, -praviá¹£te, -vivaá¹£á¹e). Many of the differences between KumÄrajÄ«va and the Sanskrit versions are because KumÄrajÄ«va was working with an earlier, more Prakritic version of the text, as the discussion above has tried to show.

Dharmaraká¹£aâs translation (vs. KumÄrajÄ«vaâs transliteration) of the dhÄraá¹Ä«s allows a fascinating glimpse into the Indian nirukti mind at work attempting to find meaning in the dhÄraá¹Ä« sonic formulae.⁶² Sometimes this is a simple one-to-one tracking: ÄlogabhÄsa = è§å¯åè âobserve the splendourâ = Skt. Äloka bhÄsam, or; çç¡æç âequal to the unequalledâ for asamasame, idem; sometimes it seems to be a âmistranslationâ based on phonologically similar words: like å¶ç®æ¸æ·¨ âtheir eyes are pureâ, taking the compound as derived from Skt. Åukra-aká¹£i, while the Sanskrit suggests it is from ÅukÄá¹ká¹£i, âswift wishâ. Sometimes multiple nuances are expressed for a repeated word, as in æ¥µçç¡æç¡å¾ç¡èº«ç¡æä¿±å âno I, no self, no body, no object togetherâ for the nime, nime, nime, nime, nime sequence of dhÄraá¹Ä« #5, perhaps related to Skt. nir me; and sometimes the explanation seems to be invented to explain what on the surface appears to unexplainable, as in most of the explanations related to dhÄraá¹Ä« #4 above. In Zhiyiâs commentary on the dhÄraá¹Ä« chapter he says that it is not necessary to understand the meaning of a dhÄraá¹Ä« in order for it to work; but since it is the âsecret word of the Buddhasâ (æ¯è«¸ä½å¯èª),⁶³ exegetes must have felt compelled to delve into the significance of the sonic formulae, and indeed most of them do have an OI/MI phonotactical structure which suggests a meaningful derivation. Nevertheless, without the contextual âsemantic walking stickâ a translation of the dhÄraá¹Ä«s does not appear to be very tenable.⁶⁴

6.1 An Urtext?

If it were possible to establish an Urtext, we would have to fully account for all the variants in the existing witnesses by understanding:

The very complex transmission process involving multiple recensions, each with perhaps hundreds or even thousands of manuscripts.
The scribal errors that have entered into the text because of the ânormalâ copying process of omission, incorrect word division, parablepsis (omission of words caused by repetition of one or more words in the same context), interchange of letters (metathesis), etc.
Errors that have entered into the text because of epigraphical considerations (misreading of scripts), due to unfamiliarty, similarity of letters, etc.
Errors that have entered into the text because of inaccurate translation practices, either between Prakrit and Sanskrit or Prakrit and Chinese. In the latter case especially, there are many phonetic forms in MI which can not be easily represented in EMC, as we have seen above.
The impact of the native dialect of the translators. KumÄrajÄ«va was a Kuchean who spoke a Tocharian language; how did this impact his perception and understanding of MI and EMC?

This complex transmission tapestry becomes even more intractable when one adds in the component of time. The SDP is one of the oldest of the MahÄyana sÅ«tras, possibly dating from the first century BCE (Nakamura 1980: 186â87; Keown 2005: 158), which means that as much as four or five centuries had elapsed between its composition and its translation by KumÄrajÄ«va in 406 CE, probably with numerous other (now lost) copies and translations being made in between. If indeed the original version was composed in Prakrit, as LÃ¼ders has suggested, the source document that KumÄrajÄ«va had in front of him was at least in part already Sanskritised. But the process and timescale in which this took place is impossible to reconstruct. The complexity of this transmission scenario suggests that the establishment of an Urtext for the SDP is not a valid endeavour. However, we can learn quite a bit about the nature of the text that KumÄrajÄ«va and his translation team had in front of him in the early fifth century, namely, the dhÄraá¹Ä«s reveal a lot about the underlying transmission dialect of the source document which KumÄrajÄ«va used. Here is a list of the principal Prakritisms found in KumÄrajÄ«vaâs source document, as reflected in his Chinese transliterations:

The -e ending to most of the nouns and adjectives in the dhÄraá¹Ä«s is most likely a Prakritism. As is well known, it is the nom. sing. ending for the eastern AÅokan Prakrits (REs of Jaugaá¸a and KÄlsÄ«) and MÄgadhÄ«. It also appears in the northwest edicts of Sh and M and historically in the Niya Documents of the Northwest China kingdom of Shan Shan. It can be interpreted as the fem. sing. vocative (where there are fem. nouns) or loc. sing. of masc. nouns, but this does not harmonize with the context or the meanings, nor is it consistent with the sÅ«traâs Prakrit heritage.
Intervocalic lenition. I have isolated all the instances where this has taken place (usually -t- > -d-, but also -k- > -g- and -khy- > -ghy-). This occurs quite a lot throughout the dhÄraá¹Ä«s, but not universally, as in the case of KumÄrajÄ«va su-Ävatani. Although intervocalic lenition is a standard feature in GÄndhÄrÄ« and most Prakrits, it is not a consistent occurrence in all the dialects. In P., for example, voiceless intervocalics often remain, and sometimes voiced stops are subject to fortition (voiced > voiceless), which also happens in the case of the word gÄndhÄri which KumÄrajÄ«va represents as kandhÄri; this, however, may simply be due to orthography in G.⁶⁵
In all but two cases (aká¹£ara, aká¹£ayatÄya), KumÄrajÄ«va transliterates Skt. -ká¹£- as a single retroflex fricative sound [Ê], not as a conjunct. We know that this is close to how it was pronounced in G. In most other Prakrits the sound was notated by -(c)cha- or -(k)kha-.⁶⁶
The -ty- conjunct is palatalized and changed to -c(c)- (as in pac(c)ave(c)chaá¹i) or assimilated to -t(t)- (as in at(t)antapÄriÅuddhiâ).
Conjunct assimilations: The -á¹£á¹- conjunct is assimilated to -(á¹)á¹-.
The conjunct -jv- has changed to -j-.
Conjunct -Åy- > -Å-; -á¹£y- > -á¹£-.
Conjunct -rt- is assimilated to > -t-.
Retroflex -á¹- and -á¸- have changed to -á¸·-.
Labials -p- and -b- changes to -v- intervocalically.
The letter s- > Ã when before a consonant and the following consonant is usually aspirated.

None of these phonological changes are inconsistent with a GÄndhÄrÄ« source document. But since they are also not inconsistent with many other Prakrits (except for the retention of the distinction between the sibilants: dental -s-, retroflex -á¹£-, and palatal -Å-, which is only preserved in GÄndhÄrÄ«), we cannot make any final conclusions about the provenance of the source dialect, only noting the âprobabilityâ of GÄndhÄrÄ« as the transmission dialect, along with Waldschmidt, Pulleyblank and other researchers.⁶⁷ However, recent discoveries in Pakistan of GÄndhÄrÄ« MSS of a PrajÃ±ÄpÄramitÄ and an Aká¹£obhyavyÅ«ha type text (in Kharoá¹£á¹hÄ« script) dated to the first or second centuries CE certainly make the âGÄndhÄrÄ« hypothesisâ even more plausible. It is quite possible that a GÄndhÄrÄ« version of the Lotus SÅ«traâor fragments thereofâwill eventually be uncovered in the monastic ruins of ancient GandhÄra.⁶⁸

7 Conclusions

From the above data, we can draw the following conclusions about the dhÄraá¹Ä«s in KumÄrajÄ«vaâs source:

The source document was a Prakrit one with limited Sanskritisations: only 12% of the items (words and compounds) in the dhÄraá¹Ä«s can be shown to have had a Sanskrit source.
KumÄrajÄ«vaâs source document cannot be said to match any of the three recensions, although it appears to be closest to the Central Asian recension in the examples shown above. Due to the absence of data, this is not fully conclusive.
KumÄrajÄ«vaâs Prakrit source document pre-dates the Nepalese and Gilgit recensions, probably by centuries, based on Edgertonâs Sanskritisation â (varies as) time rule (footnote 12). Whether it goes back to an âoriginalâ source is impossible to tell, but considering the vagaries of the transmission process, probably not.
The abundant variant forms in the different recensions point to divergent source texts. In addition, there appear to be numerous intra- and inter-recensional scribal errors or confusions, taking the form of incorrect word division (e.g. Skt. samitÄ-viÅÄnte versus KumÄrajÄ«va samitÄvi-ÅÄnte) and confusion of -i and -e endings throughout;⁶⁹ misspellings (e.g. Skt. nÄá¸i vs. Tib. nÄá¹i); metathesis (e.g. Skt. kunaá¸i/kuá¹ani); intervocalic consonantal confusion (Skt./KumÄrajÄ«va citte/cire; aká¹£aya/aká¹£ara; vrÅ«sali/ vrÅ«suni); omissions and additions of whole words (as in dhÄraá¹Ä« #6 for the Central Asian recension above) or syllables (e.g. Skt./KumÄrajÄ«va aká¹£ayavanatÄye/aká¹£ayatÄya or bhayÄbha-yÄÅviÅodhani/bhÄá¹£yÄbhÄá¹£yaÅodhÄ«); different words (Skt. buddha-vilokite vs. KumÄrajÄ«va buddhavikliá¹£á¹e or Skt. -nirghoá¹£ani vs. Skt. -nisaá¹ghani); etc. This may also be due to oral/aural problems in the transmission process.
Many of the MSS show a scribal misunderstanding or confusion re: Prakrit dialects: e.g. -dh- >< -v-, Skt. nidhiru > KumÄrajÄ«va nivi(á¹)á¹e; interchange of -Å- and -s- in words like KumÄrajÄ«va Åame â Skt. same; interchange of -á¹- and -i- (in Skt.: ná¹tyÄvati vs. nityÄvati), omission of anusvÄra (KumÄrajÄ«va Åuka(c)chi vs. Skt. sukÄá¹ksi), confusion on voicing (Skt. gandhÄri vs. KumÄrajÄ«va kandhÄri), etc.
Sanskritisation of the Nepalese & Gilgit MSS is almost one hundred per cent. Very few Prakrit forms survive (e.g. iá¹á¹ini, nityÄvati are two surviving Prakrit forms.)
Dharma transmission from MI to EMC is a highly complex process, with dozens of human, temporal, spatial, dialectal, scribal, perceptual, accentual, psychological, etc. variables, making it impossible to transmit something accurately and error free. The complicated dharma transmission process has several imporant cultural and religious ramifications, not the least of which is the impossibility of establishing an âoriginalâ text when the transmission takes place over centuries between phonologically disparate languages.

The reason why KumÄrajÄ«vaâs dhÄraá¹Ä« transcriptions are so different from the Sanskrit versions should now be clear: KumÄrajÄ«vaâs source document was quite unlike the surviving Sanskrit exemplars, and based on an earlier MS tradition which was much more Prakritic. In addition, there are numerous transmission errors and confusions present, both within the MI recensions themselves and between the MI and KumÄrajÄ«vaâs Chinese transliteration. Given the long, almost two-millenium timescale involved, it is impossible to unravel the complex transmission tapestry. All the MI versions have undergone significant Sanskritisation (Gilgit & Nepalese the most), and while the Central Asian recension preserves many more Prakritisms and correlates better with KumÄrajÄ«vaâs translation overall, much of the dhÄraá¹Ä« material is missing.

As well as uncovering the nature of KumÄrajÄ«vaâs underlying source, this study has also tried to demonstrate the complexity of the transmission and translation process, whether Indic to Indic, that is Prakrit > Sanskrit, or Indic to Chinese, and the many different temporal strata, linguistic and human factors involved. It also provides a unique perspective on the interaction and exchange of Buddhist teachings in the early centuries of the common era. These teachings were all mediated by Indo-Aryan translators working from Prakrit sources, which like PÄli were themselves translations of an earlier, underlying transmission variously styled une langue prÃ©canonique (LÃ©vi 1912), a lingua franca (Geiger 1916: 3â4), a koine gangÃ©tique (Smith 1952), a Kanzleisprache (the administrative language of the ruling government in PÄá¹aliputra; LÃ¼ders 1954: 8), or Buddhist Middle Indic (von HinÃ¼ber 1983: 192â193). I have discussed this âlanguage of early Buddhismâ elsewhere (Levman 2014; Levman 2016) and it goes well beyond the scope of this article, except insofar as it illuminates the framework within which the Indian and Chinese cultures interacted, a framework which was at least in part determined by an ambiguous linguistic environment where various Prakritic homynyms could result in two meanings (Boucher 1998: 489â493). While the dhÄraá¹Ä« meanings are not always clear, KumÄrajÄ«vaâs transliterations provide a clear snapshot of the phonological state of the Prakrit in the early fourth century CE.

Appendix 1: KumÄrajÄ«vaâs Syllabary

From Dazhidu lun, TaishÅ Volume 25, SÅ«tra 1509 (å¤§æºåº¦è«), p.408b15 and following.

These correspondences are phrased in the following form: è¥èç¾åï¼å³é¨ç¾©ç¥ä¸åæ³é¢å¢ç¸ãç¾éï¼ç§¦è¨å¢ Ruo wen luo zi, ji sui yi zhi yiqie fa li gou xiang. Luo she, Qin yan gou; âIf one hears the character ç¾, the meaning immediately follows that all dharma are apart from the characteristic of filth, rajas [the Sanskrit word], which is âfilth å¢â in the language of the Qin dynastyâ.

a Letter -jÌ- has a macron over it indicating -jh- per Brough GDHp Â§6.

b Brough 1977: 89.

c There seems to be some confusion about the headword. Here å© is given, not è©, but since å© = ba above, it must be a mistake.

d å is not in Pulleyblank or Karlgren; here å¤ is the closest parallel. See page 157, note a, where Coblin gives it the value tÈ.

e Here both Prakrit forms of Skt. artha are given: aritha with epenthetic vowel and attha with conjuncts assimilated.

f å is not in Pulleyblank; here ä» is the closest parallel.

h Why does KumÄrajÄ«va leave out the -n- in khanda? Available to him were characters like æ ¹ (gen = PB kÉn) if he wanted to capture this -n- + consonant sound. See Brough 1962 Â§48: âsporadic weakening or loss of the nasal before voiced consonantsâ in G. Also see Geiger Â§6.3 where short nasalized vowels are not infrequently replaced by a pure long vowel (so khadha = khÄdha above) and Fussman 1989 Â§33.5 where an open long syllable was automatically nasalized in G. Of course sometimes the anusvÄra was simply omitted.

i å¤ alternate form.

j Brough 1977: 94.

Appendix 2: Phonetic Abbreviations

â (apostrophe)	rising tone
^h (superscript h)	sign of aspiration, including aspiration in the departing tone
[Î´]	a dental voiced fricative
eÌ¯ iÌ¯ ÉÌ¯, etc.	subordinate vowels in diphthong per Karlgren (in Ulving 1997, 13)
level tone	unmarked
entering tone	syllables ending in -p, -t, -k
É²	palatal nasal
Å	velar nasal
É	schwa
ÉÌ	a schwa like off-glide found in combinations like iÉÌ; see Pulleyblank 1991, 5
É	lower mid back rounded vowel like âlongâ in English
É¤	closed mid back unrounded vowel
É	a voiceless palatal fricative, Å in Sanskrit. Also found as tÉ which is an affricate form
Ê	a voiceless retroflex fricative, á¹£ in Sanskrit. Also found as tÊ which is an affrciate form
É	lower mid-front vowel
j	high front glide like the consonant y in English
É¨	high, central unrounded vowel
Ë	long vowel
[ ]	alternate reading in the different Taisho editions or alternate phonetic spelling (depending on context)
vÌ	(vowel with subordinate marker in a diphthong, e.g. diphthong -iÌÃ¤-)
É£	voiced velar fricative
Ê	glottal stop
Ê	voiced retroflex fricative

The word dhÄraá¹Ä« is a multi-faceted term with many meanings, common to all of which is the notion of retention of the Buddhaâs teachings. The dhÄraá¹Ä« formulas were expected to be memorized exactly in order to preserve their ritual efficacy; therefore they are particularly apt for the study of the transmission of the dharma, as special care was taken with their memorization and transmission. For recent studies on the dhÄraá¹Ä«s, see Braarvig 1985, McBride 2005, Copp 2008, and Davidson 2009.

PB = Pulleyblank 1991, which reconstructs the pronunciation of Early Middle Chinese (EMC). The abbreviation PB will be used before all transcriptions (phonetic reconstructions) which use this system. Transcriptions are given in regular type, with italic type reserved for Middle Indic words and Chinese pinyin. When I consult Bernhard Karlgrenâs transcriptions, I use Ulving 1997, abbreviation KG. âEarly Middle Chineseâ is the Chinese codified in the Qieyun rhyme dictionary (601 CE), representing the Chinese language of the fourth to seventh centuries CE, per Pulleyblank 1984: 3.

For example, see Geiger 1916 [2005]: Â§43.2, e.g. Skt. enas > P. ela (âfaultâ); hereinafter âGeiger.â

LÃ©vi 1915: 41. LÃ©vi notes that Saá¹ghavarman was a âdemi-hindou, originaire de lâIndo-Chineâ and Yijing was âun pur styliste chinois instruit par un long sÃ©jour dans lâIndeâ (ibid.: 122).

See Pischel 1900 [1981] Â§236 (hereinafter Pischel); the change of an intervocalic stop to a glide (-y-) or a weakly articulated glide (-áº-) or even its disappearance is quite common in the Prakrits. For an example from one of the earliest Buddhist suttas, see Norman 1980: 175 (also in Norman 1991, vol.2: 151), where, for example, the earlier word virayo can be confidently derived from two MI reflexes: in the P. Sutta NipÄta, virato, and in the MahÄvastu, virajo. The voicing of -t- > -d- is also a common Prakrit phenomenon, especially in GÄndhÄrÄ«. See Brough 1962 [2001] Â§33 (hereinafter Brough 1962 = GDhp = GÄndhÄrÄ« Dhammapada).

Waldschmidt 1932: 231. For intervocalic stop lenition in GÄndhÄrÄ« (G.), see GDhp 28. For assimilation of -r- (which is sporadic per Waldschmidt, 232), see Burrow 1937 Â§37 and 38. Note that candra occurs in the GDhp as cadra and usually. appears in P. as canda and in AMg as caá¹da. In G., -r- is usually assimilated after a stop (kodha < krodha in GDhp 280), but sometimes it is kept (pridi < Skt. prÄ«ti in GDhp 56, 224). For discussion, see GDhp 51; vÄcÄ as vaya occurs in several GDhp gÄthÄs (53, 290, 291, etc.).

Waldschmidt 1932: 231â234. To name a few: change of aspirated stops to -h- (abhikrÄntÄ > transcribed in Ch. as ahikanta); same, plus assimilation of -r- before -á¹£- (abhivará¹£a > transcribed in Ch. as ahivaá¹£a); change of -á¹£á¹- > -á¹á¹- (Åreá¹£á¹ha > transcribed in Ch. as Åeá¹ha).

Ibid.: 29â32. Some of this confusion is understandable as panna = prajÃ±Ä in AMg (Pischel Â§226). See also Mylius 2003: 413, s.v. paá¹á¹a = prajÃ±Ä. Brough 1962, GDhp, also discusses this confusion in Â§45.

Ibid.: 3. Compare: jie de tuoluoni çå¾éç¾å°¼ âAll had attained the dhÄraá¹Ä«sâ (T.9.262: 2a3). See also Hurvitz 2009: 309: âincalculable, limitless bodhisattvas [â¦] attained the dhÄraá¹Ä« [â¦].â

Bude miuwu (T.46.1941: 954a5), translated in Stevenson 1986: 69. See footnote 63.

Edgerton 1953 [1998], vol.1: xxv. Dutt (1953: xvii, citing LÃ¼ders, Hoernle and Mironov, with no reference) agrees and gives the example of Central Asian MSS written in Upright Gupta script in the early 5th or 6th centuries containing more Prakritisms than those written in the calligraphic script of the 7th century. Hoernle (1916: xxxi) discusses the Northern and Southern canon and maintains that they were originally written in the âvernacular language of Magadhaâ which is of course the essence of Heinrich LÃ¼dersâ thesis as documented in his 1954 posthumous opus Beobachtungen Ã¼ber die Sprache des Buddhistischen Urkanons. Another scholar to make this point, specifically about the SDP, is HiÃ¤n-lin Dschi (1944: 139: âIch glaube frÃ¼her gezeigt zu haben, daÃ das Werk [Saddharmapuá¹á¸arÄ«kasÅ«tra] ursprÃ¼nglich in der Alt-ArdhamÄgadhÄ« abgefasst worden war, daÃ die Kashgar-Rezension [Central Asian] dem Original viel nÃ¤her steht also die nepalesische, und schliesslich, daÃ die Alt-ArdamÄgadhÄ«-Formen der Kashgar-Rezension in der nepalesischen Rezension mit der Sanskritisierung nach und nach beseitigt wurdenâ (âI believe to have shown earlier that the text was originally composed in Old ArdhamÄgadhÄ«, that the Kashgar recension was nearer to the original than the Nepalese and finally that the Old ArdhamÄgadhÄ« forms of the Kashgar recensions were gradually removed in the Nepalese recension with Sanskritisationâ).

This rendering was done between CE 404 and 406, just before KumÄrajÄ«va began the SDP translation. Lamotte (1944) called it the MahÄprajÃ±ÄpÄramitÄÅÄstra but it was known by various names in medieval China. For a discussion on the name, see Benn 2009: 12â13, fn. 1. For the arapacana syllabary, see Brough 1977: 86, who suggests that the list of headwords which the syllabary represents âmight have been in origin a mnemonic device to fix the order of the verses or paragraphs of some important text, by taking the first word of each. Thereafter, the mnemonic would have been further reduced to initial syllables where possibleâ (ibid: 94). For more information on the arapacana syllabary, see Salomon 1990: 255â273. Salomon argues (convincingly) that the arapacana syllabary was formulated in GandhÄra, in the Kharoá¹£á¹hÄ« script, based on epigraphical and internal evidence.

The letter -jÌ- = -jh- per Brough 1962, GDhp 6a.

KumÄrajÄ«va was of course not the first translator to transcribe MI. The Kucha monk Lokaká¹£ema (latter part of the second century CE), for example, transcribed Buddhist names and technical terms, while An Shigao transcribed personal and place names and translated technical terms (Nattier 2008: 4). KumÄrajÄ«va preferred to transcribe technical terms (or use existing transcriptions, like è©è© = bodhisattva, æ¶æ§ = nirvÄá¹a, æ³¢ç¾è = pÄramitÄ; éç¾å°¼ = dhÄraá¹Ä«), while he generally translated proper names (bhaiá¹£ajyarÄja = Medicine King = Yaowang è¥ç; aká¹£ayamati = Inexhaustible Mind = Wujin yi ç¡ç¡æ), but sometimes used existing transcriptions (e.g. ææ®å¸«å© for MaÃ±juÅrÄ«). He was the first translator to attempt a transcription of the SDPâs dhÄraá¹Ä«s.

Some of these variants are found in K & N; all are shown in Tsukamoko 1978: 1â35. The various manuscripts in the Nepalese tradition are listed and described in Tsukamoto et al. 1986.

These are, except for the LÃ¼shun Museum fragment, also available in Toda 1981. The Central Asian facsimile manuscripts are available in Institute for the Comprehensive Study of Lotus Sutra 1977: vol. 11, 12 and the Kashgar manuscript in a facsimile edition in Chandra 1976, which contains Bechertâs Foreword.

See for example the different endings in the nom. sing. in G. which can be either -e, -a, -o or -u, per GDhp: 75, 76. AÅokaâs Rock Edicts from ShÄbÄzgaá¹hÄ« (Northwest MI) show -a, -o and -e (Hultzsch 1925 [1969]: xc). In GDhp 21, Brough (1962) notes that in G., -e in final position regularly appears as either -e, or -i. See also Fussman 1989: 459, which notes âlâequivalence phonÃ©tique en finale de -e, -o et -aâ.

The dhÄraá¹Ä«s are found as follows: #1: K & N pp. 396¹â397²; #2 398^4â5; #3 399^1â2; #4 399⁹â400¹; #5 401^2â3; #6 477^1â4. DhÄraá¹Ä«s 1â5 are in Chapter 21, dhÄraá¹Ä« #6 in Chapter 26. Chinese versions may be found at dhÄraá¹Ä« #1 in T.9.262: 58b19 to T.9.262: 58c03; dhÄraá¹Ä« #2 in T.9.262: 58c14 to T.9.262: 59a03; dhÄraá¹Ä« #3 in T.9.262: 59a10 to T.9.262: 59a11; dhÄraá¹Ä« #4 in T.9.262: 59a18 to T.9.262: 59a19; dhÄraá¹Ä« #5 in T.9.262: 59b01 to T.9.262: 59b04; dhÄraá¹Ä« #6 in T.9.262: 61b19 to T.9.262: 61b2.

Pischel Â§282; GDhp 260 aÃ±a < Skt. anya; P. aÃ±Ã±a < Skt. anya; AMg aá¹á¹a < Skt. anya.

LÃ¼ders 1954: 10 and Â§Â§1â11.

Hultzsch 1925 [1969]: xc. For instance, jane in Sh Rock Edict (RE) 10 A vivade in RE 6 F; devanapriye in RE 10, A, etc. Capital letters refer to location reference used by Hultzsch. See also Brough 1962, GDhp 76. See also HiÃ¤n-lin Dschi 1944: 143, quoting Konow, who associates the -e dialect with the MÄnsehrÄ dialects and the Niya Prakrit.

Burrow 1937: Â§53. The Niya documents represent the administrative language of Shan-Shan (Northwest China) in the third century CE (ibid.: v, Introduction).

Pulleyblank 1983: 100, footnote 13. The character is used for Skt. syllables mi, me and vi as well as the usual ri and á¸i. See also Coblin 1983: 155, # 43 where å±¥ is transcribed as lji from the Baihu tongyi ç½èéç¾© paranomastic glosses (first century CE).

For a full discussion see H. W. Bailey 1946: 770â775. See also GDhp Â§16. Most Prakrits used the notation -(c)ch- or -(k)kh- to represent Skt. -ká¹£- (brackets indicate that the doubled consonants were often not shown in Pkt.). See also HiÃ¤n-lin Dschi 1944: 143, who makes the same point that Skt. ká¹£a changed in the west and northwest to cha and was represented in Ch. by tscha. See footnote 66 for further references. In G., -(c)ch- could also apparently be mistaken for a palatal fricative, as in GDhp 12-b which has Åotria (âlearned in the Vedaâ) paralleling Dhp 294-b and P. Dhp 47-b khattiye (âwarrior casteâ < Skt. ká¹£atriya); here the western ch- sound (< ká¹£-) has apparently been heard or interpreted as a Å- sound.

T.25.1509: 408c17 (Dazhi du lun å¤§æºåº¦è«). Here he uses the same word as an example: chaye åè¶ (PB: tÊ^haÉ¨/tÊ^hÎµ:-jia) < Skt. ká¹£aya. See also Appendix 1.

Norman 1995: 283. See also Sheth 1963: 308, where Pkt. ghitta for Skt. ká¹£ipta is found, so presumably ghÄ«á¹a < Skt. ká¹£Ä«á¹a is possible, if not attested.

For transliteration of -ká¹£ by Xuanzang, see Shu-Fen Chen 2004: 123 (caká¹£uá¸¥), 144 (laká¹£aá¹a), and 146 (ká¹£ayo). The Digital Dictionary of Buddhism (DDB) and PB (p. 47) give the character cha å¹ (PB: tÊ^ha É¨t/tÊhÎµ:t) as the transcription character for Skt. ká¹£a(t) and ashaluo é¿å¹ç¾ (PB: Êa-tÊ^haÉ¨t/tÊhÎµ:t-la) is an additional transcription possibility for aká¹£ara (S & H), but one which KumÄrajÄ«va did not use, as there was evidently no standard for him to follow. DDB is found at http://www.buddhism-dict.net/ddb/accessed Nov. 2014.

See Pischel Â§280 and Coblin 1983: 35Ë âIt therefore seems safe to conclude that earlier dentals followed by y had become palatalized in the underlying language(s) of the BTD textsâ (BTD = Buddhist Transcription Dialect).

Shu-Fen Chen 2004: 115. The character èª is neither in Pulleylank or Karlgren, so nga represents the authorâs (Chenâs) transliteration, which is taken from William H. Baxter, âAn Etymological Dictionary of Common Chinese Charactersâ (manuscript, 2000). Coblin (1994: 123, entry 0011) gives it the same Qieyun value (ngÃ¢, based on the fanqie spellings in the Guangyun).

K. R. Norman 2006: 157. vÄ«ra/dhÄ«ra, vaá¹kaá¹/dhaá¹kaá¹, avibhÅ«/adhibhÅ«, etc. This also occurs in the VimalakÄ«rti SÅ«tra, where in Chapter 9, 56a1, page 92, the manuscript reads avodigbhÄga, and it has been changed in critical edition to adhodigbhÄga. See Study Group on Buddhist Sanskrit Literature 2006.

Pischel Â§Â§238, 240. In the AÅokan edicts, for example, we find Skt. á¸uli written as duá¸Ä« (âturtleâ) in Pillar Edict 5 B (AllÄhÄbÄd-Kosam) and Skt. mahilÄ written as mahiá¸Ä (âwomanâ) in Gir RE 9 C. For other examples see Levman 2010: 66.

See footnote 30.

The character å is neither in PB or KG; Coblin 1994: 219, entry 0307a, gives it a Qieyun value of nÃ¯ and an ONWC value of *nii.

Skt. eá¸aka > eá¸·akÄ (âramâ) in Pillar Edict 5 C.

This is the closest character I could find to what is shown in the TaishÅ, notated as [ç¾*(å¥-å£+ç¦)]. However this character is missing the å¥-å£, and I donât know what the phonetic value might be (or the pinyin).

See discussion below in dhÄraá¹Ä« #6, s.v. saá¹ghanirghÄtani, p. 179.

See Dictionary of GÄndhÄri s.v. bhaÅadi (British Library Kharoá¹£á¹hÄ« Fragment 18 available at gandhari.org/a_manuscript.php?catid=CKM0020, accessed Nov. 2014), bhaá¹£ati (Bhajur Fragment 2 available at gandhari.org/a_manuscript.php?catid =CKM0265, accessed Nov. 2014) and bhaá¹£adi (GDhp 114-b, 201-d, 202-d).

T.9.263: 130a18â19: suoshuo [xian] jieming er huai zhi zu æèªª[é®®]è§£æèæ·æ¢è¶³.

Mayrhofer 1963: vol. 3, 578. The word Indara for Indra appears in the Mitanni-Hittite treaty, c. 1350 BCE, in Norman 1995: 1, but this is probably due to the cuneiform writing system.

See von HinÃ¼ber 2001 Â§139. The character å¸ may simply represent a strong initial r-, per Prof. Max Deeg (private communication).

T.25.1509: 408c10 (Dazhi du lun å¤§æºåº¦è«).

Hultzsch 1925 [1969]: lxxii. Could mukÅe represent a Tocharian influence (i.e. from KumÄrajÄ«vaâs native language), where velars > palatals before i or e (Adams 1988: 40â43)? Here muk-ke > muk-Åe? I thank Prof. Alexei Kochetov for this suggestion.

See Pischel Â§Â§303â304. Karashima 1992: 237 derives -á¹£á¹h- > -á¹á¹h- > -á¹á¹-

The character ç·» may also designate a retroflex -á¸-, but I have been unable to find another example where KumÄrajÄ«va uses it as such; in his arapacana syllabary he uses è¼ (PB: dÉ) for retroflex -á¸- and å (PB: traÉ¨^h/trÉ:^h) for retroflex -á¹-

Quoted in Pulleyblank (1983: 101): â[â¦] they should be derived from an original text having naá¹-, the -á¹- being due to a part-Sanskritisation [â¦].â

Burrow 1937 Â§18. At present in the Northwest intervocalic á¸ is represented by á¹ which may have been the ancient pronunciation (which the Ch. would have heard as á¸· ). Also, in loanwords from Khotan, the á¹ or á¸ usually appear with l.

T.9.263: 130b09: å¯æèª¿æ²ç¡æ²ï¼ç¡éç¡å¯ä½å¯; Karashima (1992: 238) translates: âOne richly has (Ridicule. No ridicule) [â¦] No riches. What is richness?â However, this does not seem to be a sentence, but simply a group of words mirroring the dhÄraá¹Ä«. I thank Prof. Max Deeg for the suggested translation above.

Pelliot 1914: 402, footnote 1, and Burrow (1937: viii), who says the same thing about the language of the Niya Documents (which he terms âKrorainicâ, named after the capital of the kingdom; âit was devoid of voiced stopsâ). Shan Shan was on the south side of the Karim basin in NW China and Kucha on the north side (within 200 kms of each other).

Burrow 1937 Â§45. In P. as well this phenomenon occurs as in abaddho (âunboundâ) with variant abandho (idem) in Sn v. 39-a.

Burrow 1937 Â§36. See also GDhp words like bramaá¹a, praá¹a, etc.

Karashima (1992: 239) derives this from nir me (âwithout meâ).

The general rule in the Prakrits is that when a sibilant occurs before a stop, the sibilant is assimilated and the stop is aspirated (e.g. Skt. stana > P. thana). See Woolner 1928 [1996] Â§38.

Burrow 1937 Â§49, except in cases of words having the root âsthÄ, of which tahe may be an example. See GDhp 209-f stuka-stoka.

See GDhp 5-b, 106-b, 108-b, etc., paÅadi < Skt. paÅyati; see also Burrow 1937 Â§41 and Woolner 1928 [1996] Â§49.

Ulving 1997: 340, entry #6688 and p. 13 for the definition of -iÌ-.

Burrow, 1937 Â§37. See also GDhp 24-c, 254-c, 255-a, artha = Skt. artha where other Prakrits (e.g. P., AMg) have attha < Skt. artha.

Karashima 1992: 247. His notes: saá¹ga, *saá¹ghÄ < Pkt. saá¹khÄ < saá¹khyÄ (ânumerationâ); -á¹g- /-á¹gh- < -á¹kh- < -á¹khy-.

These numbers count chart entries, not words, except in cases where one word of a compound can be demonstrated to show a different derivation than another as in sarvadhÄraá¹Ä« Ävatani where the first karmadhÄraya (descriptive compound) could derive from either Skt. or Pkt., but the second derives from a Pkt. source.

For a partial list of Prakritisms in the Central Asian manuscripts, see K & N, vi f.; Dutt, xix f.

In the following groups of three words, the first word is the transliterated Ch., the second the Central Asian manuscript, and the third the Skt. from K & N.

For an excellent introduction to the Indian love of etymologizing and finding multiple meanings in words, see M. Deeg 1995: 33â73. For an example of âDie sprachwissenschaftliche Etymologie,â practised by vaidikas and also by the Buddhist commentators, see ParamatthajotikÄ 2, 208^12â13, where Buddhaghosa tries to explain why the Buddha is called nÄga, âsnake,â evidently an embarrassing epithet: nÄgan ti punabbhavaá¹ nâ eva gantÄraá¹, atha vÄ Ägun na karotÄ« ti pi nÄgo, balavÄ ti pi nÄgo, taá¹ nÄgaá¹, âhe is called ânÄgaâ since he does not go to a new birth [taking the ga- in nÄga as derived from the MI verb gam, âto goâ with na- as the negative adverb], or he does not commit a fault [na- -Ägu, âno, faultâ] and also since he is strong.â

T.34.1718: 146c21; for Zhi Yiâs commentary, see ééç¾å°¼å in å¦æ³è®è¯ç¶æå¥, T.34.1718: 146b29â146c26.

I thank Prof. Max Deeg for this useful expression (private communication).

For intervocalic lenition, see Pischel Â§186f. For P., see Geiger Â§Â§35, 38, 39. For the use of -k- for -g- in G., see GDhp 30.

See Pischel Â§317f. Generally -ccha- was used in the west and -kkha- in the east per Woolner 1924 [1996] Â§40; also Geiger Â§56. PÄli shows both notations. The sound -(k)kha- represents a voiceless velar stop + a velar aspirated stop; the sound -(c)cha- represents a voiceless palatal stop + an aspirated palatal stop. Also, -ká¹£a- can become -(j)jha- in Prakrit (Pischel Â§326). An interesting example of this occurs at DN 2, 161^21â22, where jhÄpenti (< Skt. âká¹£ai, to âburnâ) is used: âthey burn the body of the universal monarchâ (raÃ±Ã±o cakkavattissa sarÄ«raá¹ jhÄpenti); in the corresponding MPS version (Waldschmidt 1950â1951 Â§46.7, p. 410 we find dhyÄpyate (âit was burntâ), which is a hyperform, as Edgerton points out (BHSD, 288 s.v. *dhyÄyati)âthe translator misunderstood P. jhÄpenti as being dervied from dhyÄpenti (< Skt. âdhyai, âto meditate,â caus. dhyÄpayati), when it was actually derived from Skt. âká¹£ai, âto burn,â caus. ká¹£Äpayati. He/she therefore wrongly Sanskritised the jh- > dhy-.

For example, Bernhard (1970: 57) argues that G. was the âmedium 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.â See also HiÃ¤n-lin Dschi (1944: 141â142), who establishes the translation sequence from Alt-ArdhamÄgadhÄ« > northwestern dialects > Sanskritisation, a sequence he says applies not only to the Lalitavistara and the SDP but for all old Buddhist writings where the ending -u appears for -aá¹ (in the nominative and accusative singular, which is also prevalent in the SDP Skt. recensions). âMagadha was the homeland and GandhÄra, âthe second holy land of Buddhismâ [here Waldschmidt 1925: 12 is quoted]. Numerous old Buddhist texts wandered through both lands and carried the traces of themâ (authorâs translation, pp. 141â142). Norman (1976: 117â127) suggests that certain anomalous forms in P. (nom. sing. ending in -e and the gen. pl. ending in -uno) were taken over from a Northwestern Pkt., i.e. G. (pp. 125â126).

For the Aká¹£obhya-type text, see Strauch 2008: 47â60. For the PrajÃ±ÄpÄramitÄ, see Falk & Karashima 2012: 19â62, and Falk & Karashima 2013: 97â169.

This may simply reflect the fact that in G. an -e at the end of a word can be writtten either as -e or -i, as per Brough, GDhp 21.

å¼ç¨ä¿¡æ¯

Save
Cite
Email this content

Share Link

Copy this link, or click below to email it to a friend
Email this content
or copy the link directly:

https://brill.edhh.ma/edcollchap-oa/book/9789004366152/BP000016.xml
The link was not copied. Your current browser may not support copying via this button.

Link copied successfully

Buddhist Encounters and Identities Across East Asia

ä¸ç¼ï¼ Dynamics in the History of Religions, å·ï¼ 10

ISBN:: 9789004366152

åºçç¤¾:: Brill

å°å·åºçæ¥æ:: 23 Apr 2018

Subjects
- Asian Studies
- Religious Studies
  - Religion in Asia

Front Matter

Back Matter

ç»è®¡æ°æ®

	å¨é¨æé´	è¿å»ä¸å¹´	è¿å»30å¤©
æè¦æµè§æ¬¡æ°	0	0	0
å¨ææµè§æ¬¡æ°	1448	1043	463
PDFä¸è½½æ¬¡æ°	1065	702	138

African Studies	Education	Media Studies
American Studies	History	Middle East and Islamic Studies
Ancient Near East and Egypt	Human Rights and Humanitarian Law	Musicology
Art History	International Law	Philosophy
Asian Studies	International Relations	Religious Studies
Biblical Studies	Jewish Studies	Slavic and Eurasian Studies
Biology	Languages and Linguistics	Social Sciences
Book History and Cartography	Life Sciences	Theology and World Christianity
Classical Studies	Literature and Cultural Studies

Subjects

AppBar-Authors 2025

AppBar-Open Access 2025

AppBar-Product Information 2025

AppBar-Company 2025

AppBar-Contact 2025

Chapter 5 The Transmission of the Buddhadharma from India to China: An Examination of KumÄrajÄ«vaâs Transliteration of the DhÄraá¹Ä«s of the Saddharmapuá¹á¸arÄ«kasÅ«tra

Abstract

1 Introduction

2 The Nature of the Source Dialect: Previous Work

3 The Saddharmapuá¹á¸arÄ«kasÅ«tra

3.1 Textcritical Background

4 Methodological Considerations

4.1 Purpose of this Study

4.2 Methodology

4.2.1 DhÄraá¹Ä« Comparison

4.2.2 A Note on Vowel Notation

5 DhÄraá¹Ä«s

5.1 DhÄraá¹Ä« #1 Spoken by Bhaiá¹£ajyarÄja è¥ç (Yaowang)19

5.1.1 pratyaveká¹£aá¹i

5.1.2 nidhiru

5.2 DharÄá¹i #2, Spoken by PradÄnaÅÅ«ra Yongshi åæ½

5.3 DhÄraá¹Ä« #3 by VaiÅravaá¹a æ¯æ²é (Pishamen)

5.4 DhÄraá¹Ä« #4 by VirÅ«á¸haka æåå¤©ç (Chiguo tianwang)

5.5 DhÄraá¹Ä« #5 by the rÄká¹£asyaá¸¥ Luocha nÃ¼ ç¾ åå¥³

5.6 DhÄraá¹Ä« #6 by Samantabhadra Puxian æ®è³¢