Introduction

With the present appendix I intend to show that commentarial notions and ideas influenced the wording of some passages in the early discourses. The topics taken up are as follows:

• a survey of Chinese Āgama passages that differ from their Pāli parallels in ways that show close similarities to the Pāli exegetical tradition,
• an examination of Pāli discourses that exhibit comparable characteristics,
• a discussion of the relationship between discourse and commentary in general.

In this way, after first identifying the basic pattern of apparent later influences in some Chinese Āgama passages, my study proceeds by examining the same pattern observed in relation to Pāli discourses. These instances, taken together, suggest that commentarial notions could influence the discourses on which they comment during the period of oral transmission.

Such a conclusion stands in contrast to the suggestion by K.R. Norman that the commentaries were transmitted separately from the discourses. A separate transmission of commentary and discourse would of course make it improbable that the former could influence the latter. In order to clarify this point, in the third and final part of the present chapter I critically review the arguments by K.R. Norman and conclude that it seems more probable that commentary and discourse were transmitted together.

---

Study

Āgama Passages with Affinities to Later Pāli Texts

A close examination of discourses found in the Chinese Āgamas in the light of their parallel versions brings to light several instances where a particular passage reflects a form of presentation that is not found in the Pāli version of this discourse, but instead parallels later Pāli works, in particular those belonging to the commentarial and exegetical tradition.

My first example is based on the insight instructions given in the Āneñjasappāya-sutta. The instructions are to contemplate that: "this is empty of a self and what belongs to a self."[n.1] MN 106 at MN II 263,26: suññam idaṃ attena vā attaniyena vā (following Be, Ce, Se, and Horner 1959: 48 note 6 on reading suññam idaṃ instead of the Ee edition's saññam idaṃ, a reading confirmed by the occurrence of 空 in the corresponding passage in MĀ 75 at T I 542c18; cf. also Minh Chau 1964/1991: 329).

The Madhyama-āgama parallel goes into greater detail, instructing that: "this world is empty, empty of a self and of what belongs to a self, empty of being permanent, empty of being everlasting, empty of existing continuously, and empty of being unchanging."[n.2] MĀ 75 at T I 542c18: 此世空, 空於神, 神所有, 空有常, 空有恒, 空長存, 空 不變易.

Such reckoning of impermanence as a form of emptiness is not found elsewhere in the Pāli discourses. It does recur, however, in historically later Pāli works, such as the Paṭisambhidāmagga or the commentaries.[n.3] Paṭis I 109,10: suññaṃ attena vā attaniyena vā niccena vā dhuvena vā sassatena vā avipariṇāmadhammena vā; cf. also Nid II 279,12 and Vism 654,22. This case has already been noted by Baba 2004: 945.

Another example, similarly related to insight contemplation, can be found in the Ekottarika-āgama version of a set of instructions given to someone who is about to pass away. The version of this instruction found in a Majjhima-nikāya discourse and its Saṃyukta-āgama parallel simply enjoin detachment from the senses.[n.4] MN 143 at MN III 259,12 and SĀ 1032 at T II 269c16.

The corresponding passage in the Ekottarika-āgama explains that when a sense organ arises one does not know from where it comes, and when it ceases one does not know where it goes.[n.5] EĀ 51.8 at T II 819c14: "at the time of the arising of the eye, when it arises one does not know from where it comes, at the time of the cessation of the eye, when it ceases one also does not know where it goes", 若眼起時, 則起亦 不知來處, 若眼滅時, 則滅亦不知去處 (adopting a variant reading that adds 則起 after 時 in the case of arising, in analogy to the case of cessation). Among Pāli texts this type of presentation is not found in the discourses, but rather in the Visuddhimagga.[n.6] Vism 484,6 indicates that the sense-spheres "do not come from anywhere before they arise, do not go anywhere after they cease", na hi tāni pubbe udayā kutoci āgacchanti, na pi uddhaṃ vayā kuhiñci gacchanti. Baba 2004: 946 draws attention to this case as an example for a tendency of Āgama discourses to incorporate what in the Pāli tradition is found only in commentarial texts.

Such examples for the tendency in Āgama discourses to bear similarities to explanations current in the Pāli exegetical tradition are not restricted to the context of meditation. Other instances that exhibit the same pattern may involve, for example, the qualities of the Buddha or of his teaching.

The Dasabala-sutta of the Saṃyutta-nikāya proclaims that the Tathāgata, due to being endowed with ten powers and four intrepidities, claims the place of a leader, literally "the bull's place".[n.7] SN 12.22 at SN II 28,18 (to be supplemented from SN II 27,23: āsabhaṃ ṭhānaṃ paṭijānāti). This case has been noted by Wen 2006: 13. According to the Pāli commentary, the reference to a bull in this description intends former Buddhas.[n.8] Spk II 45,25: āsabhā vā pubbabuddhā. Whereas a Sanskrit fragment parallel also speaks of a bull's place,[n.9] S 472 folio 125 R1, Waldschmidt 1958: 395, reads: udāram-ārṣabhaṃ sthāna[ṃ] pratijānāti. a Saṃyukta-āgama parallel to the Dasabala-sutta indicates that the Tathāgata, due to being endowed with the ten powers and the four intrepidities, "knows the place of former Buddhas".[n.10] SĀ 348 at T II 98a15: 知先佛住處. This specification appears to be influenced by an exegetical understanding of the passage similar to the gloss now found in the Pāli commentary.

Another example, related to the teaching of the Buddha instead of to his qualities, can be identified in relation to the Gaṇakamoggallāna-sutta of the Majjhima-nikāya. The discourse highlights that the advice given by the Buddha is supreme among "things of today".[n.11] MN 107 at MN III 7,2: gotamassa ovādo paramajjadhammesu. The Pāli commentary explains that "things of today" means the six teachers (i.e., Pūraṇa Kassapa etc.).[n.12] Ps IV 70,18: ajjadhammā nāma cha satthāradhammā.

The corresponding passage in a Madhyama-āgama parallel to the Gaṇakamoggallāna-sutta proclaims that the Buddha's teaching is able to subdue all heterodox practitioners.[n.13] MĀ 144 at T I 653c7: 能伏一切外道異學; an indication made also in another parallel preserved individually, T 70 at T I 876b15: 能攝一切異學. The formulation in a way makes explicit an understanding that in the Pāli tradition only manifests in a commentarial gloss. This gives the impression that the formulation in the Madhyama-āgama discourse could have been influenced by a similar ancient Indian exegetical understanding of the implications of this passage.

Another instance can be found in the Mahāsakuludāyi-sutta. The discourse highlights that the Buddha's teaching is "with a causal [basis]", sanidāna, which the commentary explains to stand for being "with conditions", sapaccaya.[n.14] MN 77 at MN II 9,25 and Ps III 241,17. The parallel passage in the Madhyama-āgama combines both expressions by indicating that the Buddha's teaching is with causes and with conditions, not without them.[n.15] MĀ 207 at T I 783b6: 有因非無因, 有緣非無緣.

Another example involves the perennial question about the destiny of a Tathāgata after death, taken up, for example, in the Aggivacchagotta-sutta of the Majjhima-nikāya and its Saṃyukta-āgama parallel (T 99). Instead of discussing the fate of a Tathāgata, the corresponding passage in the 'other' Saṃyukta-āgama (T 100) speaks of the destiny of the "self of beings" after death.[n.16] SĀ2 196 at T II 445a18: 眾生神我; whereas MN 72 at MN I 484,27 and SĀ 962 at T II 245c5 speak of a tathāgata, 如來. This mirrors an explanation found in the Pāli commentarial tradition, which glosses occurrences of the word Tathāgata in the context of such discussions as a "living being".[n.17] Sv I 118,1: satto tathāgato ti adhippeto. This explanation seems to be standard for commenting on the tetralemma about the destiny of a Tathāgata after death and is found also in Ps III 141,21, Spk II 201,4, and Mp IV 37,22: tathāgato ti satto; on this commentarial gloss cf. also Gnanarama 1997: 236–237, Karunadasa 2007: 7–12, and Manda 2005. The formulation in the 'other' Saṃyukta-āgama gives the impression of being influenced by an exegetical understanding similar to that now found in the Pāli commentaries.

Other examples suggestive of the same type of exegetical and commentarial influence on discourses can be found in regard to circumstantial information. A case in point occurs in the two Saṃyukta-āgama parallels to the Godhika-sutta of the Saṃyutta-nikāya, which reports how Māra told the Buddha about Godhika's intention to commit suicide.[n.18] SN 4.23 at SN I 121,11: atha kho māro pāpimā āyasmato godhikassa cetasā cetoparivitakkam aññāya yena bhagavā tenupasaṅkami. This case has been noted by Wen 2006: 22. The parallel passages in the two Saṃyukta-āgama translations additionally describe Māra's reflections that motivated him to approach the Buddha.[n.19] SĀ 1091 at T II 286a17 and SĀ2 30 at T II 382c20 (for a translation of SĀ2 30 cf. Bingenheimer 2007: 65); a reflection also recorded in a discourse quotation in the Abhidharmakośopāyikāṭīkā, D 4094 nyu 32b3 or Q 5595 thu 69a6. In the Pāli tradition his reflections are described only in the commentary, not in the discourse itself.[n.20] Spk I 183,11, although its description of Māra's reflection differs from what is reported in the Saṃyukta-āgama versions.

The degree to which such additional information is reported in a discourse can vary from version to version. This can be seen with three discourses that present the famous dictum according to which the path to the end of the world can be found within this fathom-long body.[n.21] SN 2.26 at SN I 62,19; this case has been noted by Wen 2006: 18. The Rohitassa-sutta directly continues after this statement with a set of stanzas. A parallel version in the 'other' Saṃyukta-āgama (T 100) offers additional explanations of this statement before turning to its version of these stanzas, indicating that to attain the cessation of dukkha is to reach the 'end' (of the world).[n.22] SĀ2 306 at T II 477c14: 若盡苦際, 是即名為得其邊際. This explanation becomes more detailed in the parallel in the complete Saṃyukta-āgama translation (T 99), which identifies the 'world' with the five aggregates of clinging, lists each of them, and then explains that the 'end of the world' stands for the noble eightfold path, before listing its factors.[n.23] SĀ 1307 at T II 359a29: 何等為世間? 謂五受陰. 何等為五? 色受陰, 受受陰, 想受陰, 行受陰, 識受陰 … 何等為世間滅道跡? 謂八聖道, 正見, 正志, 正 語, 正業, 正命, 正方便, 正念 , 正定.

A similar line of thought can also be found in the Pāli commentary, which explains that the 'world' stands for the first noble truth, and the path to the 'end of the world' can be found in the noble eightfold path.[n.24] Spk I 117,26: lokan ti dukkhasaccaṃ … paṭipadan ti maggasaccaṃ. Spk-ṭ I 158 (Be) then explicitly refers to the five aggregates of clinging. Even though the presentations in the two Saṃyukta-āgamas and in the Pāli commentary differ to some extent from each other, all three appear to stem from a similar line of reasoning that has been made explicit to different degrees.

Another case involving the same Saṃyutta-nikāya and Saṃyukta-āgama collections occurs in relation to a cryptic injunction given in verse, according to which one should cut five, abandon five, develop five, and transcend five. Versions of this instruction in the Saṃyutta-nikāya and in a parallel discourse in the complete Saṃyukta-āgama (T 99) do not specify what these sets of five refer to.[n.25] SN 1.5 at SN I 3,15: pañca chinde pañca jahe, pañca vuttari (Be and Ce: cuttari) bhāvaye, pañca saṅgātigo (Se: saṅgātito) bhikkhu, oghatiṇṇo ti vuccati. SĀ 1312 at T II 360c26: 斷五捨於五, 五法上增修, 超五種積聚, 名比丘度流; cf. also Dhp 370 and its counterparts in the Gāndhārī Dharmapada 78, Brough 1962/2001: 129, and in T 210 at T IV 572a15. Such an explanation can be found in the Pāli commentary.[n.26] Spk I 24,12 explains that the five lower and the five higher fetters should be cut off and abandoned, the five faculties should be developed, and five types of bondage (specified to stand for passion, hatred, delusion, conceit, and views) should be transcended. In the case of another parallel in the complete Saṃyukta-āgama (T 99) as well as in two parallel discourses in the 'other' Saṃyukta-āgama (T 100), explanations regarding some of these sets of five come together with the stanzas themselves.[n.27] SĀ 1002 at T II 262c21: 斷除五捨五, 增修於五根, 超越五和合, 比丘度流淵 already has the indication regarding the need to develop the five faculties, also mentioned in the Pāli commentary. The same is also mentioned in SĀ2 140 at T II 427c19: 能斷於五蓋, 棄捨於五欲, 增上修五根, 成就五分法, 能渡駛流 水, 得名為比丘, and SĀ2 311 at T II 479a23: 除五欲受陰, 棄捨於五蓋, 增進 修五根, 成就五分身, 如是之比丘, 超渡生死海, which additionally indicate that the five hindrances and the five sense desires should be eradicated and abandoned, and the five factors of the Dharma or else of the body (whatever that may mean) should be put into operation.

Another example for the apparent influence of exegesis on a discourse can be found in the two Saṃyukta-āgama parallels to the Kassapagotta-sutta of the Saṃyutta-nikāya. The introductory narration of the Pāli discourse reports that the monk Kassapagotta admonished a hunter.[n.28] SN 9.3 at SN I 198,22: tena kho pana samayena āyasmā kassapagotto divāvihāragato aññataraṃ chetaṃ (Ce: cetaṃ) ovadati; this case has already been noted by Wen 2006: 17. The commentary then furnishes the additional information that the person in question was a deer hunter.[n.29] Spk I 289,29: cetan ti ekaṃ migaluddakaṃ (with a variant reading as chetan). Parallel passages in the two Saṃyukta-āgama translations go a step further in the same direction, since they describe that the hunter was actually setting up a trap to catch deer on that occasion.[n.30] SĀ 1339 at T II 369b19: 時有獵師名曰尺只, 去十力迦葉不遠, 張網捕鹿. 爾 時十力迦葉為彼獵師哀愍說法, and SĀ2 359 at T II 491a27: 有一獵師名連 迦, 去尊者不遠施鹿羂弶. 爾時尊者憐愍獵師為其說法.

The introductory section of a discourse is a classical instance for variations, where otherwise quite similar parallel versions can differ considerably in the degree to which they provide a detailed narrative setting for the discourse itself. A case in point is the Cūḷakammavibhaṅga-sutta, a detailed exposition on the topic of karma and its fruit given by the Buddha to the young brahmin Subha. The Pāli version simply begins with Subha's enquiry on this topic. The Pāli commentary provides an introductory account for this enquiry by relating that Subha's stingy father had been reborn as a dog in his own former household,[n.31] MN 135 at MN III 202,17 and Ps V 8,23. and it was the Buddha's revelation of the lowly rebirth of Subha's father that had motivated the young brahmin's enquiry into the working mechanism of karma. Most of the Chinese discourse parallels to the Cūḷakammavibhaṅga-sutta have incorporated a similar account into the discourse proper.[n.32] This narration can be found in MĀ 170 at T I 703c24, T 78 at T I 887b7, T 79 at T I 888b19, and T 81 at T I 895c2. Two versions do not have this introductory narration: T 80 at T I 891a21 (which does not report any enquiry by the young brahmin at all, as here the Buddha directly addresses him on the topic of karmic retribution) and T 755 at T XVII 588c22. A similar narration is also found in the Karmavibhaṅga preserved in Sanskrit; Sogdian fragments have preserved parts of the same.[n.33] Kudo 2004: 2,13, 2006a: 35,5, and 2006b: 49,8; the Sogdian fragment, Rosenberg 1920: 405, has preserved the beginning section of a version of this narration.

Needless to say, the influence of exegetical glosses and narrations is not restricted to Chinese Āgama discourses.[n.34] For a discussion of an apparent instance of commentarial influence in the case of the Dīrgha-āgama cf. Anālayo 2014b: 46f. An example of the same basic pattern affecting a discourse preserved in Tibetan translation can be seen in relation to a passage in the Mahāsuññata-sutta of the Majjhima-nikāya.[n.35] [34] Norman 1980/1991: 142 discusses another example, where the Tibetan translation of the Mūlasarvāstivāda Pravrajyāvastu version of the Śrāmaṇyaphala-sūtra shows some affinities to the Pāli commentary on the Sāmaññaphala-sutta. The Pāli version and its Madhyama-āgama parallel present contemplation of the impermanent nature of the five aggregates of clinging as the means for going beyond the conceit 'I am'.[n.36] [35] MN 122 at MN III 115,3 and MĀ 191 at T I 739b17.

In addition to referring to the conceit 'I am', the Tibetan counterpart enjoins leaving behind any desire or underlying tendency towards 'I am'.[n.37] [36] Skilling 1994: 236,7: ngaʼo snyam paʼi nga rgyal dang, ngaʼo snyam paʼi ʼdun pa dang, ngaʼo snyam paʼi bag la nyal ma spangs shing; cf. also the discourse quotation in the Abhidharmakośopāyikāṭīkā, D 4094 ju 235b6 or Q 5595 tu 269a6 which, although differently worded, has the same reference to desire and an underlying tendency. Skilling 1997: 390 notes the recurrence of this presentation in the Abhidharmakośopāyikāṭīkā and also in the Pāli commentary on the present discourse, and also highlights that the same threefold presentation recurs in SN 22.89 at SN III 130,29 (where this pattern is similarly found in the parallel version SĀ 103 at T II 30a24: 我慢, 我欲, 我使). Similar to the Tibetan version's additional reference to desire and an underlying tendency, the Pāli commentary in its gloss on the present passage speaks of conceit, desire, and underlying tendency towards 'I am'.[n.38] [37] Ps IV 163,14: asmī ti māno, asmī ti chando, asmī ti anusayo.

The same pattern can also be observed in regard to discourses of the Aṭṭhakavagga in the Sutta-nipāta, where background stories such as the tale of the murder of the female wanderer Sundarī are found only in the Pāli commentary, whereas in the case of the Chinese translation this tale has become part of the discourse itself, a status it has also acquired in the case of the Pāli Udāna collection.[n.39] [38] Pj II 518,15 commenting on Sn 780 to 787, with its Chinese counterpart in T 198 at T IV 176b13 (translated in Bapat 1945: 156–158); cf. also Ud 4.8 at Ud 43,23 and the discussion in Anālayo 2014c: 81.

At times, a whole Āgama discourse may have its parallel only in the Pāli commentarial tradition. An example is the tale of King Māndhātu who, after becoming the sole ruler in the world and eventually even being allowed to sit on the celestial throne in the Heaven of the Thirty-three together with Sakka, was still not satisfied with the range of his power and had the wish to oust Sakka. This tale occurs as a discourse in the Madhyama-āgama. Its Pāli counterpart is found in the commentarial Jātaka collection.[n.40] [39] With the difference that in MĀ 60 at T I 495c1 he merely wants to drive out the heavenly king to become the sole ruler, whereas in Jā 258 at Jā II 312,25 he wishes to kill Sakka. In the Divyāvadāna version of this tale, he only wants to drive out Śakra, Cowell and Neil 1886: 223,26, as is the case for an individual translation, T 39 at T I 823c19. According to the Chinese Udāna collection, T 212 at T IV 630a6, he had the intention of killing the ruler of the Thirty-three. Another individual translation, T 40 at T I 824c28, reports that he wished for the king of devas to be dead so that he could take over control in the Heaven of the Thirty-three.

Another case is the tale of a son of a rich family who, through negligence, eventually becomes an old pauper, even though he could have accumulated great wealth as a layman or else become an arahant if he had gone forth in his youth.

This tale, together with a stanza that compares the predicament of this man to that of an old crane, forms a discourse in the Saṃyukta-āgama. Its Pāli counterpart is found in the commentary to the Dhammapada.[n.41] [40] SĀ 1162 at T II 310a15 only records the spiritual potential of the man, 若 復剃除鬚髮, 著袈裟衣, 正信非家, 出家學道, 精勤修習者, 亦可得阿羅漢第 一上果; cf. also T 212 at T IV 707a6. Dhp-a III 131,17 also takes into account his wife, indicating that if both had gone forth, he would have become an arahant and she would have reached non-return, sace pana nikkhamitvā pabbajissa, arahattaṃ pāpuṇissa, bhariyā pi' ssa anāgāmiphale patiṭṭhahissa. This case has been discussed by Wen 2006: 26.

With several of the examples examined above, perhaps a commentary given at the time of translation influenced the way a particular discourse was translated. Zürcher (1959/1972: 31) explains that "during the work of translation, and perhaps also on other occasions, the master gave oral explanations (k'ou-chieh 口解) concerning the contents of the scriptures translated. Explanations of this kind often appear to have crept into the text."[n.42] [46] A telling case in this respect has been noted by Bapat and Hirakawa 1970: liii, where the Chinese counterpart to the Samantapāsādika has a rather unexpected passage that reads: "the Dharma teacher says: 'I do not understand the meaning of this'", T 1462 at T XXIV 706b18: 法師曰: 我未解此義, a remark absent from the corresponding Pāli passage in Sp I 179–180. This case gives the distinct impression that a comment, originally not meant for posterity, became part of the 'translation' of T 1462.

However, changes occurring at the time of translation need not be the only explanation for such instances. This possibility suggests itself from some Pāli discourses that also appear to have incorporated later passages, somewhat similarly to the pattern observed so far in relation to Āgama discourses.

Similar Patterns in Pāli Discourses

An example of the tendency to provide a wider narrative background to a discourse is the Aṅgulimāla-sutta of the Majjhima-nikāya. In agreement with a parallel version found in the Ekottarika-āgama, the Pāli discourse precedes the actual encounter between the Buddha and Aṅgulimāla with a narration of the latter's murderous deeds.[n.43] [47] MN 86 at MN II 97,22 and EĀ 38.6 at T II 719b21; an introductory narration also found in T 118 at T II 508b21 (which incorporates additional material found otherwise only in the Pāli commentaries; cf. Ps III 328,3 and Th-a III 54,31), T 119 at T II 510b18, T 202 at T IV 423b7 (with a Tibetan counterpart in Schmidt 1843: 239), and T 212 at T IV 703a25. The Pāli and Ekottarika-āgama version also relate several events that took place once Aṅgulimāla had become a monk, such as meeting with the king of the country, an asseveration of truth made by Aṅgulimāla in order to help a woman in labour, and an occasion when he was attacked while begging alms.

The two Saṃyukta-āgama versions dispense with the introductory narration as well as with Aṅgulimāla's adventures as a monk.[n.44] [48] SĀ 1077 at T II 280c18 (cf. the translation above p. 123ff) and SĀ2 16 at T II 378b17 (translated in Bingenheimer 2006: 46–49); cf. also Enomoto 1994: 22– 23. A Sanskrit fragment parallel, Hartmann 1998, also appears to have been without an extended introductory narration, similar to SĀ 1077 and SĀ2 16. Their presentation only covers his meeting with the Buddha, his going forth and attainment of liberation, and a set of stanzas spoken by him – key elements found in all versions. Hence this case gives the distinct impression that the Pāli version, just as the Ekottarika-āgama discourse, expanded on what is found only in brief in the two Saṃyukta-āgama versions, quite possibly by incorporating material that was originally part of a more commentarial narration.[n.45] [49] Notably, although SĀ 1077 does not cover any of the events that happened once Aṅgulimāla had become a monk, some of its stanzas seem to refer to the events narrated in more detail in the other versions. SĀ 1077 at T II 281b25 refers to those who feel resentment towards Aṅgulimāla, before speaking in the next line in praise of patience. These lines acquire meaning if read against the background of the occasion when Aṅgulimāla was attacked while begging for alms. The same could also underlie a line in which he indicates he has already experienced the fruits of his evil deeds, SĀ 1077 at T II 281b24. That is, the events narrated in the other versions appear to have been known to the reciters of SĀ 1077, although these are not narrated explicitly in the discourse itself; cf. above p. 137.

Another example can be found in the Bālapaṇḍita-sutta, which offers a detailed description of the seven treasures and four types of success of a wheel-turning king.[n.46] [50] MN 129 at MN III 172,14 to 177,5. Its Madhyama-āgama counterpart merely refers to them, without providing any description.[n.47] [51] MĀ 199 at T I 762b28 to b29. Elsewhere in the Madhyama-āgama, however, the same seven treasures and four types of success received a detailed treatment,[n.48] [52] MĀ 67 at T I 512a2 to 513b27. In this case, the Pāli parallel MN 83 at MN II 74,14 does not refer to a wheel-turning king at all. so their absence in the Madhyama-āgama counterpart to the Bālapaṇḍita-sutta would not be related to a wish to obliterate such descriptions. Judging from the brief enumeration in the Madhyama-āgama parallel to the Bālapaṇḍita-sutta, it seems probable that the detailed treatment in the Bālapaṇḍita-sutta itself is a case of expansion on the side of the Pāli version.

The lack of a commentary on the Madhyama-āgama makes it more difficult to find further support for such a hypothesis.[n.49] [53] Relatively few commentaries on Āgama discourses are extant, although a commentary on parts of the Saṃyukta-āgama is preserved in the Vastusaṃgrahaṇī of the Yogācārabhūmi; cf. Yìnshùn 1983a, 1983b, and 1983c for excerpts from this commentary juxtaposed with the relevant Saṃyukta-āgama discourses. A commentary on the first section of the Ekottarika-āgama can be found in T 1507 at T XXV 30a20; cf. the detailed study by Palumbo 2013. Among Sanskrit fragments, a commentary on the āryasatyas can be found in SHT III 802, Waldschmidt 1971: 2–4; a commentary on the smṛtyupasthānas in SHT V 1104, Sander and Waldschmidt 1985: 99–100; for other sūtra commentaries cf., e.g., SHT I 24, SHT I 34, and SHT I 36, Waldschmidt 1965: 16–17, 26, and 27, SHT IV 649, Sander and Waldschmidt 1980: 260–263, SHT VIII 1828, Bechert and Wille 2000: 26, and SHT IX 2013, Bechert and Wille 2004: 11–29. Nevertheless, it is noteworthy that the Pāli commentary devotes approximately four times as much space to explaining the treasures of the wheel-turning king, compared to the space it allots to commenting on the remainder of the discourse.[n.50] [54] Ps IV 214,19 to 230,2 explains the arising of the treasures of a wheel-turning king, whereas the remainder of the discourse is covered at Ps IV 210,5 to 214,18 and 230,3 to 230,10. This reflects a considerable interest in the topic of the wheel-turning king. Given this evident interest, it seems quite possible that part of a detailed commentary on a succinct statement about the possessions of a wheel-turning king became part of the Pāli discourse itself.

Yet another relevant case could be the long exposition of various aspects of the Buddhist path to liberation given in the Mahāsakuludāyi-sutta, an exposition that is missing from its Madhyama-āgama parallel.[n.51] [57] MN 77 at MN II 11,3 to 22,15 and MĀ 207 at T I 783b15 to 783b19, translated in Anālayo 2009b. Eimer 1976: 53 notes that the first part of this listing of aspects of the Buddhist path to liberation, up to the ten kasiṇas, follows a numerical ascending order, whereas the items listed after the ten kasiṇas no longer follow this order, but instead come in the same sequence as in DN 2 at DN I 73,23 to 84,12. This suggests that two originally independent listings have been combined in the present instance. Compared to its Chinese counterpart, the long exposition given in the Pāli discourse appears somewhat out of proportion: after announcing an exposition of five qualities, the first four qualities are dealt with in brief, followed by a disproportionally long exposition of the Buddhist path to liberation under the heading of the fifth quality. Due to this detailed exposition of the fifth quality, the Mahāsakuludāyi-sutta has become a rather long discourse and would find a more fitting placement in the Dīgha-nikāya, instead of being included among discourses of "middle length". This part of the Pāli version might well be a later expansion of what originally was only a shorter treatment, perhaps by incorporating material that stems from a commentarial exposition.

A similar case appears to be the Piṇḍapātapārisuddhi-sutta, where a long list of various aspects of the Buddhist path to liberation is found as part of a description of how a monk may purify the alms food he receives.[n.52] [58] MN 151 at MN III 295,13 to 297,20. This long list is absent from its Saṃyukta-āgama parallel.[n.53] [59] SĀ 236 at T II 57b21 to b25, translated in Choong 2004: 5–9; for remarks on SĀ 236 in the light of MN 151 cf. Choong 1999: 11f. Although the development of all the practices described in this part of the Piṇḍapātapārisuddhi-sutta would certainly be desirable in order to maximize the merits that will accrue to a donor of alms food, to undertake successfully the entire set described in the Pāli version the monk in question would have to become an arahant.

In contrast, the Saṃyukta-āgama version simply describes how a monk by continuously being energetic and mindful in any posture purifies the alms he receives.[n.54] [60] SĀ 236 at T II 57b24: 日夜精勤, 繫念修習, 是名比丘於行, 住, 坐, 臥淨 除乞食. The relatively brief and straightforward indication given in the Saṃyukta-āgama version seems to offer quite an adequate exposition of how a monk should act in order to become a pure recipient of alms food. Such purification could already take place at levels of development that fall short of full awakening and that have not yet done full justice to the whole range of practices mentioned in the Piṇḍapātapārisuddhi-sutta. In this case, too, it could well be that, what originally was a gloss offering additional perspectives on how a monk may purify his alms food, eventually became part of the Pāli discourse.

Other examples for what clearly appears to be an expansion of a discourse during its oral transmission by integration of exegetical material can be found in the Mahāsatipaṭṭhāna-sutta of the Dīgha-nikāya, and the Mahācattārīsaka-sutta of the Majjhima-nikāya. In both cases the content of the passage in question reflects commentarial exegesis, and in the case of the Mahācattārīsaka-sutta even terminology otherwise only found in later Pāli literature can be identified.[n.55] [56] For a detailed discussion of both cases cf. Anālayo 2014c: 91–100 and 129– 142. The intrusion of commentarial material in the case of the Mahāsatipaṭṭhāna-sutta has already been noted by various scholars. Winternitz 1920/1968: 51 considers this case to be an example of a discourse enlarged through addition of commentarial material. Bapat 1926: 11 refers to this discourse as "an amplified version of an originally small sutta … explaining, in a commentarial fashion, the details of the four noble truths"; cf. also Barua 1971/2003: 369–371. Bodhi 2005: 261 sums up that the passage in question "may have originally been an early commentary incorporated into the discourse." The case of the Mahācattārīsaka-sutta had already been discussed in detail by Meisig 1987.

A comparable combination of earlier and later textual material can also be seen in works such as the Divyāvadāna, the Lalitavistara, and the Mahāvastu. According to Oldenberg these texts at times exhibit an 'old style', similar to the Pāli discourses, as well as a 'new style', more akin in form and content to later works and the Pāli commentaries.[n.56] [67] Cf. Oldenberg 1898: 672 (northern Buddhist texts in general), Oldenberg 1912b: 156 (Divyāvadāna), Oldenberg 1882: 114 (Lalitavistara), and Oldenberg 1912a: 141 (Mahāvastu); von Simson 1985: 81 comments that this pattern is probably of general relevance and would not be limited to the texts examined by Oldenberg. It seems fair to assume that these instances reflect the same basic pattern described so far.[n.57] [68] Based on a detailed comparative study of Pāli and Sanskrit discourse material, von Simson 1965: 130 notes that often material that is found additionally in the Sanskrit version appears to stem from a commentary that has become part of the transmitted text, "zahlreiche Zusätze der Sanskritversion … scheinen geradezu aus der fortlaufenden Kommentierung des überlieferten Textes in diesen selbst eingedrungen zu sein."

In sum, the above examples from the Pāli Nikāyas suggest that commentarial passages could become part of a discourse at some point during oral transmission. In regard to the Chinese Āgamas, this would imply that the influence of commentarial exegesis on Āgama discourses need not have happened only at the time when a discourse was translated into Chinese.

The Relationship between Commentary and Discourse

The hypothesis that commentarial explanations could influence the discourse on which they comment during oral transmission would be less probable if the discourses and the commentaries should have been transmitted independently of each other. Such independent transmission has been suggested by Norman (1997: 158–160), based on his observation that the commentaries at times preserve explanations no longer meaningful due to changes in the language of the discourses.

An example provided by him is the commentarial explanation of the name of King Pasenadi as one who "defeats the army of others", parasenaṃ jināti. This explanation makes sense only with a spelling of the king's name that has a j, as in the Sanskrit form Prasenajit, but not with a spelling that has a d, as in the Pāli form Pasenadi.[n.58] [69] Lüders 1954: 100 notes that a Bhārhut inscription reads rājā pasenaji kosalo, which would fit this etymology well. Norman (1997: 158) reasons that "the fact that the commentary explains the syllable -di by jināti, shows that the text and the cty [commentary] were transmitted separately, with neither having an effect upon the other … they were in the keeping of different groups, i.e. the bhāṇakas responsible for the canonical text were not also responsible for the safe keeping of the commentary upon that text."

Yet to assume such independent transmission does not solve the problem, since the name Pasenadi occurs also in the commentarial gloss itself, which reads paccāmittaṃ parasenaṃ jinātī ti pasenadi.[n.59] [70] Ud-a 104,27. That is, the commentators would have been aware of it even if they knew only the commentary.

In other words, even if the commentators had not known the discourse at all, the very wording of the commentary would have ensured that they knew the spelling Pasenadi and thus they should have realized the inapplicability of an explanation that involved instead a j. Nevertheless, they continued to transmit an explanation that no longer suited the term it purported to explain.

Hence it seems that the old commentarial explanation parasenaṃ jināti continued to be passed on in deference to tradition, even though its etymology no longer made as much sense as it would have made originally. This in turn shows that there is no need to assume a separate transmission of discourse and commentary to explain such a phenomenon.

Besides, the Nikāyas themselves preserve popular etymologies that due to language developments are no longer meaningful. An example is a stanza in the Dhammapada, which explains that to discard evil is characteristic of a brahmin, bāhita pāpo ti brāhmaṇo.[n.60] [71a] Dhp 388. Norman (1997: 160) explains that this stanza "shows that in the dialect in which the etymology was invented the word brāhmaṇa had the form *bāhaṇa". The same popular etymology recurs in the Dīgha-nikāya, the Sutta-nipāta, and the Udāna.[n.61] [71b] DN 27 at DN III 94,1: pāpake akusale dhamme bāhentī ti kho … brāhmaṇā, Sn 519: bāhetvā (Be: bāhitvā) sabbapāpakāni … pavuccate sa brahmā, and Ud 1.5 at Ud 4,8: bāhitvā pāpake dhamme … te ve lokasmiṃ (Be and Se: lokasmi) brāhmaṇā; cf. also Schneider 1954: 578 and Mette 1973: 33 . Once even the discourses continue to pass on an explanation that no longer fits the development of the language, the fact that the commentaries did the same would simply be a sign of the fidelity of oral transmission.[n.62] [72] This degree of fidelity of the tradition may also explain another inconsistency noted by Norman 1980: 178. Some degree of inconsistency was apparently tolerable to ancient Buddhist commentators and reciters. Goonesekera 1967: 346 notes that "there are numerous instances where accounts of the same episode in the different aṭṭhakathā differ as regard details." Balbir 1991: 123, in a discussion of popular etymologies, explains that "les commentateurs sont aussi des fidèles ... il n'y a donc pas chez eux d'attitude 'critique' eu égard à la justesse ou à la fantaisie d'une explication." It does not imply that these were transmitted separately from the discourses.

In fact a large part of the commentaries, as we have them now, consists of glosses on particular expressions and words. Without knowledge of the discourse to which these refer, such glosses would hardly be intelligible.[n.63] [73] Griffiths 1999: 55 explains that a "religious commentary will take as its initial object … gobbets" and "a comment on any one of these gobbets will typically presuppose knowledge of them all, and may be incomprehensible without such knowledge." The close relationship between commentary and discourse can also be seen when sponsors of palm-leaf manuscripts have discourses and their respective commentaries written out, as can be seen, e.g., in the study of Thai manuscripts at the Lai Hin monastery by von Hinüber 2013: 33 and 37, 42 and 45, 43 and 46, 49 and 55, and 51 and 56. This makes it rather improbable that some reciters would learn only the commentary, without knowing the discourse.[n.64] [74] Although, e.g., Pj I 151,12 lists the aṭṭhakathikas alongside those who specialize in the suttas, the Vinaya, the Abhidhamma, or the Jātakas (cf. also Mori 1983: 980), such listings may only reflect some degree of specialization and need not imply that the aṭṭhakathikas knew only the commentaries and had no knowledge of the source material which the commentaries are meant to explain. Endo 2003b: 36 draws attention to references in the commentaries to akkharabhāṇakas, vyañjanabhāṇakas, vaṇṇabhāṇakas, and padabhāṇakas, which he explains to be reciters that specialized in phonetics, letters (as against meaning), praises, and syllables or sentences; cf. also Mahinda 1997: 502. Similar to the case of the aṭṭhakathikas, one would expect that these specialists were acquainted with the texts in addition to being well versed in the particular area of their specialization.

Furthermore, when presenting a discourse and its explanation to an audience, a reciter would have to alternate between portions of the discourse and commentaries on these portions. During oral performance of a discourse, the commentarial gloss would have to be given not too long after the passage on which it comments. To recite the whole discourse first in its entirety – especially if it is a longer discourse – presents the risk that the audience no longer has clearly in mind the passage to which the commentary refers.

Thus the oral performance situation of a discourse of substantial length would require the recitation of this discourse to be interspersed with explanations of a commentarial nature. This type of oral performance would combine a fixed part – be this in verse or in prose – with a (at least in its early stages) less fixed commentarial explanation. Such a basic pattern of combining fixed portions with explanations of a more commentarial nature seems to stretch all the way from Vedic ākhyāna to modern-day kavi baṇa in Sri Lanka.[n.65] [75] On the Vedic ākhyāna cf. Alsdorf 1963/1974; on the kavi baṇa Mahinda 2003. According to Kanno 2003: 302, the early Chinese commentaries were also given in an interlinear form and only later did a style of commentary develop that was more independent of the original discourse; cf. also Mayer 2004: 167–168.

Nowadays, when speaking of a "commentary", one inevitably has the aṭṭhakathās and ṭīkās in mind, which clearly differ from the source material on which they comment. Yet, such clear separation may not have been in existence in ancient times.[n.66] [76] According to Deutsch 1988: 170f, in the traditional Indian setting commentaries are seen as "not so much appendages to an otherwise fixed and complete work", but rather as a contribution "to a larger, developing work", in the sense that "from the … commentator's point of view he is not so much remarking on an already finished text", but instead "he is himself contributing to that text."

Several discourses in the four Pāli Nikāyas are in fact commentaries by disciples on a particular saying of the Buddha.[n.67] [77] A case study can be found in Anālayo 2008h. Often the concluding section of such discourses offers an endorsement by the Buddha of the exposition given by his disciple, an endorsement which turns what was originally a commentary spoken by a disciple into part of a canonical discourse.

Even without explicit approval by the Buddha, some of the explanations and comments given by his disciples have come down to us as discourses that are considered canonical.

Goonesekera (1967: 336) comments that in this way "the earliest beginnings of exegetical literature can be traced to the canon itself." According to Malalasekera (1928/1994: 89f), explanations given by the early disciples in this way "formed the nucleus of the commentaries". "When later the text of the canon came to be compiled … some of the expositions found their way into the Piṭakas." In continuation of the same tendency, as "the commentaries … were handed down … along with the texts themselves … parts of the commentaries came later to be attached to the texts which they interpreted."[n.68] [78] Mayer 2004: 166 even traces the beginnings of the commentaries to the Buddha himself, explaining that "the teachings of the dharma, from the very beginning, called for commentary … the Buddha was frequently called upon to elaborate on teachings he had given."

A case in point is the Suttavibhaṅga of the different Vinayas, where the monastic rules are found embedded in a commentary that, although being of a later nature, has itself attained canonical status. As noted by Norman (1997: 150), in this case evidently "the text and the commentary have been handed down together."

The same pattern appears to be also evident in the Pāli Udāna collection, studied in more detail in the next chapter,[n.69] Cf. below p. 463ff. whose stanzas are embedded in prose narrations that often appear to have been added at a later time. In contrast, in the case of the Dhammapada and the Jātaka, a similar narrative commentary has not attained canonical status and is still found among the aṭṭhakathās. These few examples seem to testify to the organic relationship between an original text and its commentary, which during oral transmission grew alongside the text on which it comments.

Given that a clear distinction between discourse and commentary does not seem to have been prominent during the formative stages of early Buddhist oral literature, it is perhaps not surprising if even in later times such a distinction should not always have remained a clear-cut one. In relation to the instances surveyed in the first part of the present chapter, where Āgama discourses appear to have been under the influence of notions reflecting exegetical commentary, it seems quite probable that a good part of this influence took place during their oral transmission and was already present in the Indic source text used for translation into Chinese. In this way an oral tradition to all appearances concerned with precise transmission could still result in major differences of the type that comparative study brings to light.[n.70] Cf. also Anālayo 2014b and 2015a.