Source: Harrison 1990

Harrison argues that the Banzhou sanmei jing 般舟三昧經 (in three juan) T418 (BZSMJ) can be tentatively attributed to Lokakṣema, with some reservations. According to Harrison, the text is preserved both in the Korean edition of the canon ("K") and the “printed editions” of the Song, Yuan and Ming Dynasties ("SYM"). Harrison summarises the differences between K and SYM as follows [“chapter” refers to divisions based upon the Tibetan, rather than pin 品 divisions in the Chinese]:

1. The opening paragraph of the nidāna demonstrates a “redactional difference” between K and SYM. Harrison suggests that this may be the result of an insertion into SYM of a “well known śrāvaka-guṇa formula,” without reference to the Indic manuscript of the Pratyutpannabuddhasammukhāvastitasamādhi-sūtra (PraS). However, because the nidāna of SYM corresponds with that of T416 and the Tibetan version, Harrison considers it likely that the nidāna does go back to the Indic, and is not a “formulaic application by a Chinese writer of the nidāna as it appears in K.”

2. The gāthās at the end of Chapters 3, 4 and 5 are “clearly independent translations of an original text”. Harrison bases this conclusion on comparison to T416, T419 (which is closer to K than SYM), and the Tibetan version.

3. There are roughly 235 variant readings between the two versions, which Harrison regards as “of minor importance”. They usually involve no more than a single character.

Approximately half of these variants are to be found in the “first third of the text.” Chapters 1-6 of K show significantly more deviation in transmission from SYM than Chapters 7-26; and the opening section of the nidāna and the gāthās of these chapters are “clear proof of an independent textual tradition”. From Chapter 7 onwards, K and SYM are “virtually identical”. We can therefore conclude that “all the prose of the BZSMJ (excluding 1A) and all the gāthās from Chap. 7 onwards in both K and SYM go back to the same original text. Because the variant readings contained therein are almost totally artefacts of transmission, we can also conclude that the only significant redactional differences between K and SYM lie, in fact, in the gāthās of Chapters 4-6 and in the nidāna of Chapter 1 (i.e. in Chapter 1-6 of K). In view of this, “we are justified in designating K 1-6 as a separate redaction of the BZSMJ.”

Thus Harrison divides the texts into Reaction A (K Chapters 1-6), and Redaction B (all of SYM and Chapters 7-26 of K). He offers the following hypotheses to explain the existence of these two separate redactions:

1. Lokakṣema translates a complete version of the PraS with his customary “shortened nidāna” and “prose rendition of the gāthā”. This translation is Redaction A.

2. (a) “A later translator”, with a different Indic manuscript, redoes the nidāna and the gāthās, but “leaves the prose of Redaction A virtually unchanged”, thus producing Redaction B.

Or:

3. (b) “A later translator”, with a different Indic manuscript, creates an original translation of the PraS (which Harrison terms “X”). Subsequently the gāthās and nidāna of this translation are conflated with the prose of Redaction A”, thus producing Redaction B.

4. Redaction B becomes the “standard edition of the large BZSMJ in China and is printed in the standard editions of the Chinese Canon (=SYM)”.

5. At some point, “either at the redaction of the Korean edition of the Canon or – more probably – on the occasion of some earlier edition, a version of the BZSMJ is produced which combines Chapters 1-6 of A with Chapters 7-26 of B”, yielding the present K.

Harrison notes that T419 (which consists of six chapters) and Redaction A end at the same point. This would appear to suggest that Lokakṣema’s original translation consisted only of Chapters 1-6. However, Harrison argues that the “homogeneity of prose” in both versions of the BZSMJ suggests that the prose portions of Chapters 1-6 and 7-26 are the work of the same person or persons.

In what follows, Harrison attempts to establish this homogeneity, and to determine the authorship of the text by comparison to T224 ("DXJ" = Daoxing jing), as DXJ is “the only text which can be regarded with certainty as a genuine product of his [Lokakṣema’s] translation work.”

Harrison concludes that features common to BZSMJ and DXJ prove that both are the work of the same person or school. Thus, in the two extant versions of the BZSMJ in three juan, Chapters 1-6 of K go back to Lokakṣema’s original translation, as does, “with some hesitation”, the prose of Chapters 7-26 in both K and SYM. However, the gāthās of SYM 1-26 and K 7-26 “are the work of another translator.”

To illustrate his point, Harrison presents a list of terms which are common to both the BZSMJ and DXJ. He finds that the majority of translation terms in the two texts exhibit a close affinity, especially those which are considered “particularly characteristic of Lokakṣema … [e.g. terms for] tathatā, Buddha, bhūta-koṭi etc.” Harrison considers small discrepancies in the treatment of these terms to be “minor alterations” made “during the course of a later revision.”

While the prose portions of both Redactions agree with Lokakṣema’s style, the verse gāthās of Redaction B are not in the style of Lokakṣema, “who is known to render gāthās in prose.” Therefore, Harrison considers these verses to be the work of another hand. He thinks it most likely that at some point in the third century, someone with access to an Indic manuscript revised Lokakṣema’s translation and replaced its gāthās with verse, while also making certain other adjustments. While he does not himself investigate the author of these adjustments, Harrison suggests Zhi Qian and Dharmarakṣa as possible candidates.

Harrison then examines external evidence. Sengyou’s catalogue indicates that Dao’an attributed a text entitled Banzhou sanmei jing 般舟三昧經 to Lokakṣema, as did the Jiu lu 舊錄, and both indicated that Lokakṣema either translated or published the translation on the “eighth day of the tenth month, 179 C.E.” (which, as Harrison points out, is the same date on which he released DXJ!). In addition, a colophon to the BZSMJ (by an unknown author) tells us that Lokakṣema translated the text along with Zhu Foshuo 竺佛朔, an Indian who was also involved in the translation of the Aṣṭa (which, according to Harrison, raises the possibility that the PraS was brought from India at the same time). Zhu Foshuo is said to have “recited the text in the original language, while Lokakṣema translated it orally into Chinese for his Chinese assistants to take down in writing.” Fei Changfang attributed an additional translation of the PraS to Zhu Foshuo, but Harrison considers this to be improbable. Additional evidence for the text’s attribution to Lokakṣema can be found in another colophon (written by Zhi Mindu 支愍度). Thus, Harrison concludes, “The attribution of a work entitled BZSMJ to Lokakṣema is well established in the earliest sources.”

Harrison adds that “since Sengyou does not mention any other extant works with this title, we can assume that there were only two works of that name;” the longer version popularly attributed to Dharmarakṣa, and the shorter to Lokakṣema (T417). Fajing attributed a two juan version to Dharmarakṣa, and lists a separate and partial translation in one juan by Lokakṣema. However, after its appearance in the Fajing lu, Lokakṣema’s version in one juan disappears. It is not mentioned by LDSBJ, and the Renshou lu (Yancong), Jingtai and DZKZM list it among lost texts. The text did not appear in the Song, Yuan or Ming editions of the Canon, nor the Qisha 磧砂 edition. Yet it surfaced in the Korean edition, and has since found its way into the Taishō as T417. How it did so, Harrison writes, remains a mystery.

Returning to the larger BZSMJ, Harrison makes two main points: the extant two or three juan version (most likely Redaction B) “was ascribed to Dharmarakṣa until the eighth century”; “Lokakṣema came to be credited for his own two juan version, along with the temporarily lost one juan text.” The first extant catalogue after Sengyou (CSZJJ) to attribute a two juan version to Lokakṣema was Fei Changfang’s “notoriously unreliable” LDSBJ, in which he takes the alternative date and title from Sengyou's CSZJJ. Harrison is unsure where Fei got the variant juan count; it possibly derives from another source.

In KYL, Zhisheng “overturned the attribution of the complete version of the PraS” to Dharmarakṣa. Zhisheng lists Dharmarakṣa’s version as lost, and attributes the three juan version to Lokakṣema. According to Harrison, this reattribution was most likely based on internal evidence, the “testimony of Nie Daozhen’s catalogue,” and the “Wu lu 吳錄 recorded by Fei Changfang”. The text is listed as fifty pages long, “which tallies exactly with the number of columns in the Taishō edition.”

Zhisheng’s assertion has been regarded as final “down to the present day”, and Harrison believes this to have been “borne out by an examination of [the text’s] style”. However, he also notes that for the existence of Dharmarakṣa’s translation, we only have Sengyou’s word (based on Dao’an’s), which all subsequent catalogues followed. Although Dao’an is considered to be reliable, Harrison wonders if such a translation existed in the first place. He writes that it is possible that Dao’an was referring to Dharmarakṣa’s revision of a previous translation, rather than an original. As discussed above, based on his examination of stylistic cues, Harrison does not consider Dharmarakṣa to have been involved, because “the relevant portions of B are closer to Lokakṣema’s A than Dharmarakṣa’s translation style.”

In view of these facts, Harrison argues that Redaction B was created by a revision which (according to the colophon to the BZSMJ) took place in 208, “in the Han capital of Xu(chang).” He reasons that if the original translation had taken place in 170, a revision so long afterwards "would not have been the stylistic touching up which usually occurred after a translation”. Therefore the revision in 208 appears to have been a “major overhaul”, which Harrison suggests was undertaken by students of Lokakṣema’s school with a slightly different manuscript. Those students would have been more familiar with Chinese and thus able to “convert gāthās into unrhymed Chinese verse” and utilise a more “sinicised vocabulary”.

Harrison concludes that Dharmarakṣa’s part, if he ever was involved, “remains unknown”, but “as far as the origins of Redaction B of the BZSMJ is concerned, I doubt we ever need to look further than the revision of 208 C.E.”

Edit

223-249

Harrison records three texts listed in the catalogues “which may or may not be related to the [Pratyutpannabuddhasammukhāvasthitasamādhi-sūtra],” none of which have survived. Among these is the Fatuo pusa bai’ershi nan jing 颰陀菩薩百二十難經 (in one juan) [sic for Fatuo---dictionaries give bá or fú for 颰, 颰陀 Skt *Bhadra---MR]. The text first appears in CSZJJ, listed as a lost, anonymous text. Harrison suggests that this could be an “alternative title” of one of the earliest translations of the Pratyutpannabuddhasammukhāvasthitasamādhi-sūtra. Harrison adds that any three of these texts may be the source for the passages in T417 which are not based on the BZSMJ in three juan.

Edit

269

Harrison writes that the 大方等大集經賢護分 T416 is “the only Chinese version of the PraS concerning which we know definitely when, where, and by whom it was translated.” The catalogue in which it first appears was published two years after the text, and Fei Changfang’s association with the translator (Jñānagupta) leaves little room for error. Changfang lists T416 as Jñānagupta’s sixth translation among thirty-one (at the time of the compilation of the catalogue), and his attribution has been accepted by all subsequent catalogues. Harrison considers the strength of the external evidence to outweigh the need for internal evidence, but for good measure adds that T416 “exhibits all the characteristics of the later Chinese Buddhist translations:” standard terminology, fidelity to the text, consistency in the use of formulae. He concludes that T416 was translated in the “early months of the year 595 at the Daxingshan-si in Chang’an.”

Edit

212-215

Harrison argues that the Banzhou sanmei jing 般舟三昧經 (in three juan) T418 (BZSMJ) can be tentatively attributed to Lokakṣema, with some reservations. According to Harrison, the text is preserved both in the Korean edition of the canon ("K") and the “printed editions” of the Song, Yuan and Ming Dynasties ("SYM"). Harrison summarises the differences between K and SYM as follows [“chapter” refers to divisions based upon the Tibetan, rather than pin 品 divisions in the Chinese]:

1. The opening paragraph of the nidāna demonstrates a “redactional difference” between K and SYM. Harrison suggests that this may be the result of an insertion into SYM of a “well known śrāvaka-guṇa formula,” without reference to the Indic manuscript of the Pratyutpannabuddhasammukhāvastitasamādhi-sūtra (PraS). However, because the nidāna of SYM corresponds with that of T416 and the Tibetan version, Harrison considers it likely that the nidāna does go back to the Indic, and is not a “formulaic application by a Chinese writer of the nidāna as it appears in K.”

2. The gāthās at the end of Chapters 3, 4 and 5 are “clearly independent translations of an original text”. Harrison bases this conclusion on comparison to T416, T419 (which is closer to K than SYM), and the Tibetan version.

3. There are roughly 235 variant readings between the two versions, which Harrison regards as “of minor importance”. They usually involve no more than a single character.

Approximately half of these variants are to be found in the “first third of the text.” Chapters 1-6 of K show significantly more deviation in transmission from SYM than Chapters 7-26; and the opening section of the nidāna and the gāthās of these chapters are “clear proof of an independent textual tradition”. From Chapter 7 onwards, K and SYM are “virtually identical”. We can therefore conclude that “all the prose of the BZSMJ (excluding 1A) and all the gāthās from Chap. 7 onwards in both K and SYM go back to the same original text. Because the variant readings contained therein are almost totally artefacts of transmission, we can also conclude that the only significant redactional differences between K and SYM lie, in fact, in the gāthās of Chapters 4-6 and in the nidāna of Chapter 1 (i.e. in Chapter 1-6 of K). In view of this, “we are justified in designating K 1-6 as a separate redaction of the BZSMJ.”

Thus Harrison divides the texts into Reaction A (K Chapters 1-6), and Redaction B (all of SYM and Chapters 7-26 of K). He offers the following hypotheses to explain the existence of these two separate redactions:

1. Lokakṣema translates a complete version of the PraS with his customary “shortened nidāna” and “prose rendition of the gāthā”. This translation is Redaction A.

2. (a) “A later translator”, with a different Indic manuscript, redoes the nidāna and the gāthās, but “leaves the prose of Redaction A virtually unchanged”, thus producing Redaction B.

Or:

3. (b) “A later translator”, with a different Indic manuscript, creates an original translation of the PraS (which Harrison terms “X”). Subsequently the gāthās and nidāna of this translation are conflated with the prose of Redaction A”, thus producing Redaction B.

4. Redaction B becomes the “standard edition of the large BZSMJ in China and is printed in the standard editions of the Chinese Canon (=SYM)”.

5. At some point, “either at the redaction of the Korean edition of the Canon or – more probably – on the occasion of some earlier edition, a version of the BZSMJ is produced which combines Chapters 1-6 of A with Chapters 7-26 of B”, yielding the present K.

Harrison notes that T419 (which consists of six chapters) and Redaction A end at the same point. This would appear to suggest that Lokakṣema’s original translation consisted only of Chapters 1-6. However, Harrison argues that the “homogeneity of prose” in both versions of the BZSMJ suggests that the prose portions of Chapters 1-6 and 7-26 are the work of the same person or persons.

In what follows, Harrison attempts to establish this homogeneity, and to determine the authorship of the text by comparison to T224 ("DXJ" = Daoxing jing), as DXJ is “the only text which can be regarded with certainty as a genuine product of his [Lokakṣema’s] translation work.”

Harrison concludes that features common to BZSMJ and DXJ prove that both are the work of the same person or school. Thus, in the two extant versions of the BZSMJ in three juan, Chapters 1-6 of K go back to Lokakṣema’s original translation, as does, “with some hesitation”, the prose of Chapters 7-26 in both K and SYM. However, the gāthās of SYM 1-26 and K 7-26 “are the work of another translator.”

To illustrate his point, Harrison presents a list of terms which are common to both the BZSMJ and DXJ. He finds that the majority of translation terms in the two texts exhibit a close affinity, especially those which are considered “particularly characteristic of Lokakṣema … [e.g. terms for] tathatā, Buddha, bhūta-koṭi etc.” Harrison considers small discrepancies in the treatment of these terms to be “minor alterations” made “during the course of a later revision.”

While the prose portions of both Redactions agree with Lokakṣema’s style, the verse gāthās of Redaction B are not in the style of Lokakṣema, “who is known to render gāthās in prose.” Therefore, Harrison considers these verses to be the work of another hand. He thinks it most likely that at some point in the third century, someone with access to an Indic manuscript revised Lokakṣema’s translation and replaced its gāthās with verse, while also making certain other adjustments. While he does not himself investigate the author of these adjustments, Harrison suggests Zhi Qian and Dharmarakṣa as possible candidates.

Harrison then examines external evidence. Sengyou’s catalogue indicates that Dao’an attributed a text entitled Banzhou sanmei jing 般舟三昧經 to Lokakṣema, as did the Jiu lu 舊錄, and both indicated that Lokakṣema either translated or published the translation on the “eighth day of the tenth month, 179 C.E.” (which, as Harrison points out, is the same date on which he released DXJ!). In addition, a colophon to the BZSMJ (by an unknown author) tells us that Lokakṣema translated the text along with Zhu Foshuo 竺佛朔, an Indian who was also involved in the translation of the Aṣṭa (which, according to Harrison, raises the possibility that the PraS was brought from India at the same time). Zhu Foshuo is said to have “recited the text in the original language, while Lokakṣema translated it orally into Chinese for his Chinese assistants to take down in writing.” Fei Changfang attributed an additional translation of the PraS to Zhu Foshuo, but Harrison considers this to be improbable. Additional evidence for the text’s attribution to Lokakṣema can be found in another colophon (written by Zhi Mindu 支愍度). Thus, Harrison concludes, “The attribution of a work entitled BZSMJ to Lokakṣema is well established in the earliest sources.”

Harrison adds that “since Sengyou does not mention any other extant works with this title, we can assume that there were only two works of that name;” the longer version popularly attributed to Dharmarakṣa, and the shorter to Lokakṣema (T417). Fajing attributed a two juan version to Dharmarakṣa, and lists a separate and partial translation in one juan by Lokakṣema. However, after its appearance in the Fajing lu, Lokakṣema’s version in one juan disappears. It is not mentioned by LDSBJ, and the Renshou lu (Yancong), Jingtai and DZKZM list it among lost texts. The text did not appear in the Song, Yuan or Ming editions of the Canon, nor the Qisha 磧砂 edition. Yet it surfaced in the Korean edition, and has since found its way into the Taishō as T417. How it did so, Harrison writes, remains a mystery.

Returning to the larger BZSMJ, Harrison makes two main points: the extant two or three juan version (most likely Redaction B) “was ascribed to Dharmarakṣa until the eighth century”; “Lokakṣema came to be credited for his own two juan version, along with the temporarily lost one juan text.” The first extant catalogue after Sengyou (CSZJJ) to attribute a two juan version to Lokakṣema was Fei Changfang’s “notoriously unreliable” LDSBJ, in which he takes the alternative date and title from Sengyou's CSZJJ. Harrison is unsure where Fei got the variant juan count; it possibly derives from another source.

In KYL, Zhisheng “overturned the attribution of the complete version of the PraS” to Dharmarakṣa. Zhisheng lists Dharmarakṣa’s version as lost, and attributes the three juan version to Lokakṣema. According to Harrison, this reattribution was most likely based on internal evidence, the “testimony of Nie Daozhen’s catalogue,” and the “Wu lu 吳錄 recorded by Fei Changfang”. The text is listed as fifty pages long, “which tallies exactly with the number of columns in the Taishō edition.”

Zhisheng’s assertion has been regarded as final “down to the present day”, and Harrison believes this to have been “borne out by an examination of [the text’s] style”. However, he also notes that for the existence of Dharmarakṣa’s translation, we only have Sengyou’s word (based on Dao’an’s), which all subsequent catalogues followed. Although Dao’an is considered to be reliable, Harrison wonders if such a translation existed in the first place. He writes that it is possible that Dao’an was referring to Dharmarakṣa’s revision of a previous translation, rather than an original. As discussed above, based on his examination of stylistic cues, Harrison does not consider Dharmarakṣa to have been involved, because “the relevant portions of B are closer to Lokakṣema’s A than Dharmarakṣa’s translation style.”

In view of these facts, Harrison argues that Redaction B was created by a revision which (according to the colophon to the BZSMJ) took place in 208, “in the Han capital of Xu(chang).” He reasons that if the original translation had taken place in 179, a revision so long afterwards "would not have been the stylistic touching up which usually occurred after a translation”. Therefore the revision in 208 appears to have been a “major overhaul”, which Harrison suggests was undertaken by students of Lokakṣema’s school with a slightly different manuscript. Those students would have been more familiar with Chinese and thus able to “convert gāthās into unrhymed Chinese verse” and utilise a more “sinicised vocabulary”.

Harrison concludes that Dharmarakṣa’s part, if he ever was involved, “remains unknown”, but “as far as the origins of Redaction B of the BZSMJ is concerned, I doubt we ever need to look further than the revision of 208 C.E.”

Edit

223-249

Harrison records three texts listed in the catalogues “which may or may not be related to the [Pratyutpannabuddhasammukhāvasthitasamādhi-sūtra],” none of which have survived. Among these is the Xiao-an banzhou sanmei jing 小安般舟三昧經 (in one juan). This text first appears in CSZJJ among the “lost works,” and is accompanied by the notice “Listed in the Jiu lu.” Harrison claims that the “Jiu lu” may refer to Zhu Daozu’s catalogue, which would date the text “to the fourth century or earlier”. Subsequent catalogues assign the text to the Wei-Wu period (220-277). It is sometimes listed as “a lost Mahāyāna work” and sometimes as “a lost ‘Hīnayāna’ work”. Harrison does not know of any catalogue which has attempted to relate this text to the Banzhou sanmei jing 般舟三昧經.

However, Hayashiya argued that this text is related to T417. Hayashiya reasoned that “if the Jiu lu had a Da banzhou sanmei jing 大般舟三昧經… it ought by rights to have had a Xiao banzhou sanmei jing (the an 安 being a scribal blunder)” (Hayashiya 1945: 564-567). Harrison considers this “plausible” but is not convinced due to “insignificant evidence”. He adds, “We would do well to view with suspicion any conclusion to which even cataloguers like Fei Changfang refused to jump.” Harrison adds that any three of these texts may be the source for the passages in T417 which are not based on the Banzhou sanmei jing in three juan.

Edit

268-269