Text: T1670A; 那先比丘經


Identifier T1670A [T]
Title 那先比丘經 [T]
Date 東晉 [T]
Translator 譯 Anonymous (China), 失譯, 闕譯, 未詳撰者, 未詳作者, 不載譯人 [T]

There may be translations for this text listed in the Bibliography of Translations from the Chinese Buddhist Canon into Western Languages. If translations are listed, this link will take you directly to them. However, if no translations are listed, the link will lead only to the head of the page.

There are resources for the study of this text in the SAT Daizōkyō Text Dabatase (Saṃgaṇikīkṛtaṃ Taiśotripiṭakaṃ).


Preferred? Source Pertains to Argument Details


[T]  T = CBETA [Chinese Buddhist Electronic Text Association]. Taishō shinshū daizōkyō 大正新脩大藏經. Edited by Takakusu Junjirō 高楠順次郎 and Watanabe Kaigyoku 渡邊海旭. Tokyo: Taishō shinshū daizōkyō kankōkai/Daizō shuppan, 1924-1932. CBReader v 5.0, 2014.

Entry author: Michael Radich



[CSZJJ]  Sengyou 僧祐. Chu sanzang ji ji (CSZJJ) 出三藏記集 T2145. — T2145 (LV) 22a3

In Sengyou's Chu sanzang ji ji, T1670A is regarded as an anonymous translation, that is to say, it is listed in the "Newly Compiled Continuation of the Assorted List of Anonymous Translations" 新集續撰失譯雜經錄 (juan 4):


Entry author: Michael Radich



[Fajing 594]  Fajing 法經. Zhongjing mulu 眾經目錄 T2146. — T2146 (LV) 131b26

T1670A is treated as anonymous by Fajing.

Entry author: Michael Radich



[Fei 597]  Fei Changfang 費長房. Lidai sanbao ji (LDSBJ) 歷代三寶紀 T2034. — T2034 (XLIX) 74a28, 91c11

In LDSBJ, a title corresponding to T1670 is treated as a anonymous text of the E. Jin, citing CSZJJ; but also to Guṇabhadra, Baoyun et al., citing “various catalogues”.

Entry author: Michael Radich



[Demiéville 1924]  Demiéville, Paul. "Les versions chinoises du Milindapañha." Bulletin de l'École Française d'Extrême-Orient 24 (1924) [Hanoi 1925]: 1-253. — 1-21

Demiéville critically reviews information about three Chinese versions of the Milindapañha in the traditional catalogues, with the aim of sifting out misleading or inaccurate information, and arriving at an accurate picture of the history of these texts in China.

Specht and Lévi claimed in 1893 that there were two different versions of T1670, one in the Korean edition of the canon (K) = T1670A, and one in the Song, Yuan and Ming editions (SYM) = T1670B. Pelliot subsequently showed that both versions are based on one and the same original text, a claim which Demiéville supports. K is much shorter, but this is due to a loss of some folios.

In order to determine which of these two versions is closer to the original, Demiéville compares them with evidence in three ancient glossaries, viz. the two versions of the Yiqie jing yin yi by Xuanying and Huilin, and the glossary of Kehong. He finds that these glossators must have consulted a text corresponding to the extant text, which was divided into two fascicles, like T1670A. None of the glossaries address any terms from the parts missing in K.

The two extant versions have been modified independently, and depart from the version seen by the lexicographers, whose readings substantially agree. The modifications differ between the two current versions, and older readings are retained sometimes in one of the extant versions, and sometimes in the other, so that neither could be identified as being a more ancient or exact version of the original. Comparison with Pali shows a tendency to expansion in T1670B, and to abbreviation in T1670A.

Demiéville adds that the original on which the Chinese version is based was itself probably in bad condition and characterises the original translation as stylistically clumsy and perhaps “vulgar” (vernacular?). Terminology is not translated systematically, but rather, varying translations are taken for the same terms.

In CSZJJ, T1670 is treated as anonymous, and not as an excerpt from another work, but as a complete (independent) and authoritative text 全典 (5). Sengyou records the text as extant. This version was in two fascicles, but Sengyou also mentions elsewhere a version in one fascicle, which he moreover classes as a re-translation 異出本. This second version must therefore have been incomplete. It, too, is treated as anonymous. Demiéville rejects the possibility that the text might have been known to Dao’an, on the grounds that Sengyou reproduces elsewhere lists of texts Dao’an regarded as anonymous, but T1670 is not included among them. Demiéville suggests that this may mean that T1670 did not yet exist in 374. Another record in CSZJJ records a similar title, the Naxian piyu jing 那先譬喻經, and gives the Jiu lu as the source. Sengyou treats this text as anonymous, and “missing”.

The only one of these titles treated in Fajing is the Naxian piyu jing. There, it is classified as a derivative or offshoot text 別生, i.e. an extract from another work (pp.6-7). Yet Demiéville doubts this categorization, since in the CSZJJ it was in no way described as an extract. Furthermore, one does not know if the compilers had access to this text, since CSZJJ already categorized it as missing. Fajing is also criticized in Zhisheng’s KYL for similar errors.

LDSBJ speaks again of the Naxian piyu jing. LDSBJ dates this title to the Three Kindoms period. Demiéville explains that this classification is questionable because Fei Zhangfang only based his dating on the anonymous Jiu lu and Gu lu, which he only knew through quotes from Dao’an and Sengyou. Demiéville proposes that this Naxian piyu jing should have been produced in the third century at the latest, because for later periods, Fei Zhangfang uses different earlier catalogues as his sources when discussing anonymous translations.

Demiéville then considers the question of whether or not this supposed Naxian piyu jing could have been an earlier version of the extant Milindapañha. He refuses to draw such a conclusion, partly because the term piyu is too ambiguous, as a genre label, to allow us to draw any firm inferences about the content of the text (being used for avadānas, parables, etc.).

Fei Zhangfang classifies another version of the Naxian jing among the anonymous texts of the Eastern Jing dynasty, a list that he expanded compared to the previous one by Sengyou. Demiéville argues that this classification by Fei is plausible since he possessed a diverse range of catalogues where he could have found proof for this classification. Fei also consulted, other than the anonymous versions, catalogues from the W. Jin and N. Liang (p.10).

The LDSB mentions a third version of the Naxian jing, and treats it as an alternate translation. Sengyou previously classified this version as anonymous, and yet Fei ascribes it to Guṇabhadra. While Sengyou only gives a rather sparse list of Guṇabhadra’s translations, Fei claims to complete that list by consulting “all the the catalogues” in his possession (p.10). He refers for this attribution of the Naxian jing to Zhongjing bielu of the (Liu) Song dynasty.

Demiéville asks whether Guṇabhadra could have brought the original of this *Nāgasena-sūtra from Central Asia or Ceylon, and rejects this hypothesis on the grounds that on the one hand, Guṇabhadra based his translation of the Saṃyuktāgama on an original brought by Faxian from Ceylon, and on the other hand, according to Fei and Sengyou, Guṇabhadra’s original was identical to that for the extant version from the Eastern Jin. He concludes that since Guṇabhadra would have been incapable of translating a central Asian language, this original must have been written in an Indian Prakrit (“dialecte”).

Two versions of the Naxian are again mentioned in the 14th and 15th fascicles of the LDSB, and Demiéville highlights that for the second title, the number of juan given differs from the entry for the same title elsewhere in LDSBJ. Demiéville explains this difference in suggesting that Fei Zhangfang copied information about the text from Fajing without noticing that he already mentioned the same title before, with different details: the classification methods of the table in the 14th and 15th fascicles categorize translations not by epochs and names of the translators, as in earlier parts of LDSBJ, but by vehicles; this mode of classification was first employed by Fajing. Here we confront two common methodological difficulties in the critical use of the catalogues, according to Demiéville. First, it is often unclear if information in the catalogues is based on first-hand examination of texts that the bibliographers possessed, or if they are copying previous catalogues. Second, the number of fascicles often varied in practice, but cataloguers nonetheless used length, measured in fascicles, as a criterion by which to distinguish between different versions of a text or title.

Based on a calculation of the average number of characters per folio in the traditional catalogues (p.14), Demiéville proposes that the version of Naxian biqiu jing as mentioned in Jingtai was already incomplete in 663, and contained the same gap as K.

Demiéville points out that in DZKZM, it seems as if only Guṇabhadra’s version of the Naxian biqiu jing is mentioned, and it is classified as a text without parallels (i.e. a sole exemplar, among “versions uniques”, in Demiéville’s terms; p.18). Yet according to the DTNDL, that text had been lost since 664. Demiéville further criticizes the information of DZJZM on the anonymous version, which is, against all previous catalogues, categorized under the Mahāyāna, dated to the Han dynasty, and declared independent from Guṇabhadra’s version. Demiéville counters this new information with the fact that the anonymous version was not even extant. He further argues that the numerous points of confusion in DZKZM are due to the unscientific methods of the scholars who compiled the catalogue, as described by KYL (p.18). Therefore, Demiéville concludes that none of the information given by the DZKZM can be relied upon, not even the number of folios.

In KYL, Zhisheng criticizes the sections about the anonymous versions in CSZJJ and LDSBJ: from the 53 anonymous titles dated to the Eastern Jin by LDSBJ, Zhisheng retains only two, of which one is the Naxian biqiu jing. Demiéville suggests that this judgement about the date of the text by Zhisheng must be accorded a certain value.

Demíeville concludes, on the basis of this extended analysis, that there did indeed exist three versions of the Milindapañha in China, or of some analogous work: (1) a Naxian piyu jing translated in the third century at the latest, but not necessarily any earlier, which was lost in the fifth century; (2) a Naxian biqiu jing, translated under the E. Jin, which was anonymous, and probably based upon a Prakrit original; this version appears to have still been accessible to bibliographers of the Tang and the Five Dynasties, in a witness in two fascicles which sported the same lacuna as the version transmitted to us via the Korean canon; at the same time, a complete version, in three fascicles, has been transmitted to us via the “Southern” line of transmission starting with the Song edition, and Demiéville speculates that this complete version might have been rediscovered somewhere in the South. (3) A “separate version” of (2) was executed, Demiéville believes, in present-day Nanjing or Hebei by Guṇabhadra, between 435 and 455, by Guṇabhadra, but lost by 664.

Entry author: Paulina Schilling



[Mizuno 1959]  Mizuno Kōyō 水野弘元. "Mirinda mon kyōrui ni tsuite" ミリンダ問経類について. Komazawa daigaku kenkyū kiyō 駒澤大學研究紀要 17 (1959): 17-55.

Mizuno compares Pāli and Chinese versions of the Milindapañha, along with citations in the Abhidharmakośabhāṣya, and parallels in T203. His main aim in so doing is to recontruct the development of this text or family of related texts.

Mizuno surveys external evidence for various named Chinese versions or translations of the Milindapañha (19-23) (meaning the better-known versions, 那先比丘經 T1670A and T1670B, leaving T203 for later consideration). CSZJJ lists three related titles, and treats all as anonymous. These versions, Sengyou reports, were in one, two and four juan; the supposed four-juan version presents a problem, because it is never again seen in later catalogues. Fajing and Yancong only report a single version. LDSBJ reports two versions, one of which it ascribes to Guṇabhadra. Mizuno scornfully describes Fei Changfang's information here as "nonsense" 出鱈目, and clearly regrets the fact that it was aped by such later catalogues as DTNDL. Zhisheng, in KYL, typically treats the problem in some more detail (20); Mizuno concludes from his information that only a two-juan version was actually extant under the Sui-Tang. Turning to the print canons, Mizuno notes that the N. Song and Korean canons proceed as Zhisheng would lead us to expect, and carry versions of the text in two juan. However, from the Southern Song onward, a three-juan version of the text suddenly appears, and is transmitted in the line leading through the Yuan and Ming (the so-called "three editions" 三本 of the Taishō). Nanjio based his account of the text on the Ming. In modern canons, beginning with the Shukusatsu, the two-juan version was reinstated; the Manji and Taishō versions of the canon carry both [in T, this is the T1670A vs. B distinction --- MR].

For Mizuno, this tangled history presents more than one problem. The two-juan version of Zhisheng and the northern line is not necessarily the same as the two-juan version reported by Sengyou. Moreover, no catalogue prior to the print canons reports any three-juan version, so that the three-juan version of the southern line comes out of the blue. The four-juan version of CSZJJ is also a kind of unicorn of textual history, glimpsed once at the dawn of the record, only to vanish without trace. One possible partial solution, Mizuno suggests, may be that the southern line three-juan version represents somehow the survival of a Liang four-juan version (22). Otherwise, Mizuno tends to the conclusion that there was originally only one version, in three juan; the two-juan version emerged through some defect or accident of transmission history, whereby a central part got lost (demonstrated in part through the comparative table discussed immediately below); further, the order of textual components also got jumbled at places in transmission. This means the three-juan version preserves material lost from the two-juan, and it therefore might easily appear more conservative. However, Mizuno notes that the two-juan version is subject to fewer post-translation interpolations, meaning that for parts of the text where it survives, it is more likely to preserve original readings. He also holds that the text is likely to be originally even older than the E. Jin date that it presently bears in the canon (23; he returns to the question of date later).

Mizuno tabulates the two- and three-juan versions against the Pāli, in order to show commonalities, and also differences in content and order (23-28). Perhaps the most striking lesson presented by this table, for the purposes of the textual history of the Chinese, is a large lacuna in the two-juan version (by comparison to the other two versions) (Mizuno's p. 25), equalling about 15 registers of text. He also finds that two substantial sequences of text appear in reverse order in the two versions (26-28). In Section 4 of his paper, however, Mizuno shows that even if we were to adjust for the gap in the two-juan version and the reversal of order, the two Chinese versions would not be identical: even in shared passages, details of wording also differ in numerous loci. Moreover, even the two-juan version, whose readings are more conservative, also preserves interpolated remarks which Mizuno suspects of being due to the translator—for example, an explanation of the six poṣadha days (29-30). Against the two-juan version, the three-juan version contains ample explanatory expansion of lists, etc., e.g. the thirty-seven bodhyaṅgāni (30).

Mizuno argues for a very early translation date for the base text behind T1670A and T1670B--for the original substrate, perhaps as early as the E. Han. He bases this suggestion on what he characterises as archaic translation terminology (for such lists as the eightfold path, the five skandhas, and the twelvefold chain of dependent origination), the translation of verse with prose, and reference to the "Great Qin" 大秦 (for the Greeks), and use of the plural –cao 曹 (31-34).

Mizuno suggests further that the translators of the oldest version worked without the knowledge of other translations, even those of An Shigao. He shows that the translation terminology overlaps very little with An Shigao; on this basis, he characterises the translation idiom of T1670 as archaic and "juvenile/immature" (古く、且つ幼稚なもの, 31) and "even more immature" [than that of An Shigao] (むしろ幼稚でさえある); he also speaks of a period "before Buddhist terminology settled down" (佛教用語が確定する以前, 31); and, through the comparison with An Shigao's terms, suggests that the oldest version of the text belongs to "a period when examples of translation terms" for fundamental Buddhist categories, such as the r̥ddhipāda or the five skandhas, were "completely unknown" (譯名が全くしられていない…その譯例が全く知られていなかった時代のものであることを思わせる, 31).

Against these factors that seem to him to suggest an early date, Mizuno finds it puzzling that T1670 is not mentioned in Dao'an. He proposes that this might indicate that the text was translated in a marginal location, and languished for a period outside the mainstream (33). If this were the case, he argues, it could date as late as the Three Kingdoms, and its idiosyncratic phraseology might indicate that it is a product of the fringes, outside circles where other translations circulated.

On the basis of these complex considerations, Mizuno concludes that "the translation of this scripture must be situated in the Later Han, or at the very latest, not after the Three Kingdoms"( 本経の訳出は...後漢代に置かれるべきであり、おそくとも三国時代を下るものではない). For Mizuno, then, the date of the E. Jin, still carried in the Taishō, is nothing more than "baseless freestyling" 無根拠の杜撰 on the part of Fei Changfang, which was uncritically accepted by subsequent generations (30).

Mizuno also suggests that the jumbled and opaque character of the text might have led it to have little impact, noting, for example, that it is not excerpted in the Jing lü yi xiang (he notes that a couple of passages are excerpted in the Fa yuan zhu lin and the Zhu jing yao ji).

Mizuno surveys the various editions and versions of the Pāli text, and attempts to discern relations between them (34-41). Important for our purposes is his observation that only three of seven sections in the Pāli (§§1-3, excluding §§4-7) are matched in any version in Chinese.

In the Nantuo wang yu Naqiesina gonglun yuan 難陀王與那伽斯那共論緣 of the Za baozang jing 雜寶藏經 T203(111) (ascribed to Jijiaye 吉迦夜 and Tanyao 曇曜), the protagonists are called 難陀王 (*Menander) and 那伽斯那 (Nāgasena). The content focuses on the ability of clever Nāgasena to repeatedly see through ruses set up by the king to test his intelligence. When the two finally lock horns in dialogue, the discourse revolves around the question of whether the ātman is eternal or ineternal, a question which Nāgasena smartly parries by asking the king whether the mangoes of his palace are sweet or sour (there are no such mangoes) (41-42). Few of the details in this story are matched in any other versions of the text. A similar exchange about (non-)self is found in AKBh: the question is whether the faculty of "life" (*jīva) is the same as the body. Nāgasena again resolves this thorny issue with the conceit of the mango. This exchange, and the conceit it is predicated upon, is found in neither Pāli nor Chinese versions of the Milindapañha proper (41-42). Mizuno draws the following conclusions from of this material: multiple versions of the Milindapañha circulated; T203 mentions Kaniṣka and Gandhāra, and so may have been linked to the Northwest; this may imply a link to Sarvāstivāda, and Sarvāstivāda may in turn furnish common ground with AKBh.

Mizuno suggests that material common to Pāli and T1670 is older than whatever version might lie behind T203(111) and AKBh. (Mizuno's reasoning here might be thought somewhat peculiar: he characterises T203 as childish and folksy, contrasting it with the serious, regal tone he finds in Pāli and T1670; i.e. he is assuming that the Milindapañha ultimately records a real exchange, and that the protagonists should have behaved in accordance with their station; 44.) Mizuno states that "most scholars" believe that in broad outlines, §§1-3 of the Pāli, which are shared with Chn. T1670, are older; but he claims that in the Pāli, even those sections have later additions. He believes that at least §§2-3 transmit reports of a real dialogue, but by contrast, that §§4-7 are more scholastic in tone, betraying composition in a later context informed by complex technical learning. He considers the question of where §§4-7 might have been added to the Pāli (47-52). He notes that Pāli commentaries contain only citations from §§1-3 and a little from §4, and on that basis, suggests that those sections (at least) date to before the time of the Sinhalese commentarial tradition behind Buddhaghosa (52-53).

Entry author: Michael Radich