Introduction

By way of introduction to the studies collected in this volume, in this chapter I attempt an evaluation of the historical perspective that forms the basis for comparative studies of the 'early' discourses. In what follows I intend to arrive at an estimate of the time period the Pāli discourses,[n.1] By the expression "Pāli discourses" I mean the four Nikāyas (DN, MN, SN, and AN), Sutta-nipāta, Udāna, and Itivuttaka. and by implication their Āgama parallels, can be considered as representing.

The questions that occupy me in the following pages are the following: From what time period onwards would the Pāli discourses have been in existence in a form that roughly resembles what we currently have at our disposal (leaving aside minor changes or errors by copyists, etc., such as are evident in variant readings)? What is therefore their historical value? What period in the development of Buddhist thought can they be taken to reflect?

My presentation covers the following topics: additions recorded in the Pāli commentaries, the question of later levelling, additions made in Sri Lanka, the finalization of the Pāli commentaries, and the political and geographical setting of the Pāli discourses.

---

Additions Recorded in the Commentaries

The Pāli commentarial tradition appears to have been well aware of the fact that the discourses in the four Pāli Nikāyas could not have been recited in their present form at the first so-called "council" or better "communal recitation", saṅgīti. Thus, for example, the commentaries report that the Bakkula-sutta was only included in the Majjhima-nikāya at the second saṅgīti.[n.2] Ps IV 197,2: idaṃ pana suttaṃ dutiyasaṅgahe saṅgahītan ti; cf. also Anālayo 2010g: 45. Another discourse that would fall into the same category is AN 5.50 at AN III 57–62, which has King Muṇḍa as one of its protagonists, presumably the great-grandson of King Ajātasattu (cf. Mahāvaṃsa 4.2); for a translation of EĀ 32.7, the parallel to AN 5.50, cf. Pāsādika 2006. As already noted by Pande 1957: 47, the doctrinal contents of this discourse do not involve any innovation, as they just repeat what according to AN 5.48 and AN 5.49 at AN III 54,7 the Buddha had taught on earlier occasions to the monks and to King Pasenadi. In regard to the Mahāparinibbāna-sutta, the commentaries go further. Not only do they indicate that a remark was added to its concluding section at the time of what Theravāda tradition considers to be the third council, but they also report that some stanzas were added at a still later date by the elders from the island of Tambapaṇṇi, i.e., Sri Lanka.[n.3] Sv II 615,16: aṭṭhadoṇaṃ cakkhumato sarīran ti ādi gāthāyo pana tambapaṇṇitherehi vuttā. In his detailed study of these stanzas, Endo 2010: 117 comments that, "as far as the Pāli tradition goes, there is no evidence to deny that the verses were added in Sri Lanka."

The fact that the commentaries openly acknowledge such additions clearly shows that the Pāli discourse collections were not considered completely closed at the time they arrived in Sri Lanka. That such an acknowledgement is made, however, also gives the impression that additions and changes in Sri Lanka would not have been the rule, otherwise there would have been little reason for the commentators to record explicitly the addition of a few stanzas regarding the way the Buddha's relics were believed to have been distributed.

Similar stanzas also occur in the Sanskrit fragments of the Mahāparinirvāṇa-sūtra,[n.4] Fragment 376 R5–6 (no. 251), Waldschmidt 1950: 83f. and in Chinese and Tibetan versions of this discourse preserved in the Mūlasarvāstivāda Vinaya.[n.5] T 1451 at T XXIV 402b24 and Waldschmidt 1951: 451,24 (§51.23–24). This makes it fairly probable that these stanzas were already known within the Theravāda and Mūlasarvāstivāda traditions,[n.6] Waldschmidt 1948: 331 comments that the stanzas common to the final parts of the Mahāparinibbāna-sutta and the Mahāparinirvāṇa-sūtra must stem from an earlier tradition ("auf älterer Tradition beruhen müssen"). Thus it seems probable that a tradition on the distribution of the relics, at first handed down independently from the discourse (perhaps in the form of a commentary), eventually made its way into the discourse itself in different Buddhist schools. There is in fact evidence indicating that, in the course of transmission, the dividing line between commentary and discourse could become blurred; cf. Anālayo 2010c. Granting the possibility that an originally more commentarial type of information became part of the canonical discourses would provide a considerably more plausible explanation than assuming that the Sri Lankan elders invented the stanzas without any input from India (and without adding any reference to a portion of the relics having reached Sri Lanka, so as to boost the local cult), and that this Sri Lankan invention then somehow managed to be incorporated in the Mūlasarvāstivāda discourse version(s) transmitted in India. but presumably had not yet become part of the Mahāparinibbāna-sutta of the Dīgha-nikāya at the time of its arrival in Sri Lanka. That is, the stanzas that according to the commentaries were added to the Mahāparinibbāna-sutta do not appear to be a Sri Lankan innovation, but would have originated in India.

---

The Question of Later Levelling

My assessment of the significance of the stanzas found at the end of the Mahāparinibbāna-sutta and its parallels is based on a tacit assumption, which has been questioned in an influential paper by Schopen (1985). Therefore in what follows I examine his arguments in this respect. Criticizing the notion that similarities between parallel versions of a text point to a common early core, Schopen (1985: 17) instead proposes that

if all known versions of a text or passage agree, that text or passage is probably late, i.e. it probably represents the results of the conflation and gradual leveling and harmonization of earlier existing traditions.

To prove his point, he takes up the story of the stūpa of the former Buddha Kassapa/Kāśyapa, basing himself on a study of this story in five Vinayas preserved in Chinese by Bareau (1960: 257–261). Schopen (1985: 19) summarizes the narrative found in three of these Vinayas – belonging to the Mahāsāṅghika, Mahīśāsaka, and Dharmaguptaka traditions – as well as in a Theravāda version of this tale, preserved in Pāli, as follows:

the Buddha is travelling in Kosala; he reaches a spot near a village called Tou-i, Tou-tseu, Todeyya (all = Skt. Toyikā); he has an encounter with a man working in a nearby field as a result of which it becomes known that the stūpa of the Buddha Kāśyapa is buried under this spot; the Buddha then makes the stūpa momentarily appear and after it disappears he and/or the monks construct a stūpa on that spot from mud.

Schopen notes that the Mahāsāṅghika, Mahīśāsaka, Dharmaguptaka, and Theravāda versions of this tale

agree on these basic elements of the 'text', and since each version belongs to a separate school, we can conclude – according to the accepted principle – that the essential elements of this account must go back to a very old or pre-sectarian stage of the tradition.

Schopen (1985: 21) then contrasts this agreement with a version of this tale found in the Mūlasarvāstivāda Vinaya preserved in Sanskrit and in the Divyāvadāna, noting that these "only knew a form of the relic cult in which the stūpa did not yet have a part". Schopen reasons that this must be

the earliest known account which knows nothing of a stūpa at Toyikā – and this is an important stūpa – set over against the versions of the Mahāsāṅghika, Mahīśāsaka, Dharmaguptaka and the Theravāda, all of which must be later and all of which agree that there was a stūpa of Kāśyapa at Toyikā.

Schopen (1985: 15 and 22) concludes that this case serves to show that the tenet "if all known 'sectarian' versions of a text or passage agree, that text or passage must be very old" should be discarded, as in the present case

the Mahāsāṅghika, Mahīśāsaka, Dharmaguptaka and Theravāda accounts agree not because they represent the old presectarian version, but because they almost certainly represent later, conflated, and fundamentally altered versions.

Since Schopen has chosen this tale as a "fine example of how the accepted principle … is supposed to work, and … when we can actually check it, it is also clear that it does not seem to work very well at all" (p. 19), it seems fair to examine this particular story in detail as the best way of assessing if the position taken by Schopen is indeed to be preferred over the notion that similarities between parallel versions point to earliness rather than reflecting lateness.

Closer inspection of the relevant primary sources brings to light that the difference between the Mūlasarvāstivāda account and the other versions is not as pronounced as Schopen suggests in his presentation. The Sanskrit and Tibetan Mūlasarvāstivāda versions and the Divyāvadāna indeed do not report the display of an ancient stūpa that then disappeared again. However, versions of the tale under discussion found in the Chinese Mūlasarvāstivāda Vinaya, which has been summarized by Bareau (1960: 260) in the study on which Schopen relied in his paper, do report that such a stūpa had been constructed at the time of the Buddha Kassapa/ Kāśyapa.[n.7] T 1451 at T XXIV 249b4 indicates that the king at the time of the Buddha Kāśyapa "raised a stūpa [made of the] four gems", 起四寶塔, and in another reference to the present event, T XXIV 261c26, reports that: "he constructed a great stūpa made completely out of the four gems", 造大窣覩波皆用四寶 (adopting the variant 窣覩波 instead of 宰堵波); cf. also the summary in Bareau 1960: 260: "[il] éleva un stūpa" and "il éleva un grand stūpa fait des quatre joyaux."

Moreover, once the display of the former Buddha's remains is over, according to the Sanskrit and Tibetan Mūlasarvāstivāda Vinaya as well as the Divyāvadāna people piled up balls of clay on the spot where the bodily remains of Kassapa/Kāśyapa had disappeared.[n.8] The Bhaiṣajyavastu reports that a layman offered a ball of clay to the spot where the bodily remains of Kāśyapa had disappeared. When the Buddha eulogized the merit to be gained from such an action, others followed suit; cf. Dutt 1984a: 76,11: tac chrutvānekaiḥ prāṇiśatasahasrair mṛttikāpiṇḍasamāropaṇaṃ kṛtaṃ, with the Tibetan parallel in D 1 kha 161a6 or Q 1030 ge 149b6: de thos nas srog chags brgya stong phrag du ma dag gis kyang ’jim pa’i gong bu bcir to, and a similar passage found in the Divyāvadāna, Cowell and Neil 1886: 78,17 (cf. also 467,17): tataḥ śrutvānekaiḥ prāṇiśatasahasrair mṛttikāpiṇḍasamāropaṇaṃ kṛtaṃ. Thus in this respect the Sanskrit and Tibetan Mūlasarvāstivāda versions resemble the narrations found in the Dharmaguptaka, Mahīśāsaka, and Mahāsāṅghika Vinayas, which also describe a piling up of balls of clay that results in the construction of a stūpa.[n.9] The Dharmaguptaka Vinaya, T 1428 at T XXII 958b7, records that a farmer was tilling the land, whereon "the Buddha went to that furrow, took hold of a ball of clay, came, and placed it in this location", 佛往彼間, 取一摶泥, 來置 此處, followed by indicating, T XXII 958b23, that "then, the monks, nuns, male lay followers, and female lay followers all took a ball of clay and placed it in this location, soon completing a great stūpa", 時諸比丘, 比丘尼, 優婆塞 優婆夷, 皆以一摶泥著此處, 即成大塔. The Mahāsāṅghika Vinaya, T 1425 at T XXII 497b28, reports that "at that time, the Blessed One soon [began] to make a stūpa for the Buddha Kāśyapa in that location. The monks said to the Buddha: 'Blessed One, are we permitted to contribute with clay to it?' The Buddha said: 'I permit you to contribute to it'", 爾時世尊即於彼處作迦葉佛 塔, 諸比丘白佛言, 世尊我得授泥不, 佛言得授. The Mahīśāsaka Vinaya, T 1421 at T XXII 173a2, indicates that "the Buddha in turn put four balls, each of clay, on the location where the stūpa had disappeared. One thousand two hundred and fifty monks each also placed four clay balls on top of it. Thereupon the monks wished to raise the clay [balls] in this location into a stūpa for the Buddha Kāśyapa. The Buddha said: 'I permit you to raise it,' They in turn soon together raised it. At this time, this was the first stūpa raised on the grounds of Jamudvīpa", 佛便以四摶泥泥塔沒處, 千二百五十比丘亦各上泥四摶, 於 是諸比丘欲於所泥處為迦葉佛起塔, 佛言聽起, 即便共起, 是時於閻浮提地 上最初起塔. In other words, in the Mūlasarvāstivāda version "lay people of the area do end up building a stūpa there out of balls of mud or clay (mṛttikāpiṇḍa)", as pointed out by Strong (2004: 36).

The Mūlasarvāstivāda versions continue by reporting a stanza spoken by the Buddha in praise of the merits of placing a single ball of clay on the caityas of a Buddha.[n.10] Dutt 1984a: 76,10 and Cowell and Neil 1886: 78,16 (cf. also 467,15): yo buddhacaityeṣu prasannacitta āropayen mṛttikapiṇḍam ekam; cf. also D 1 kha 161a6 or Q 1030 ge 149b5: gang zhig rab dang sems kyis sangs rgyas kyi mchod rten la ni ’jim gong ’jog byed na. It is not clear to me on what Schopen (1985: 21) bases his assessment that this reference "quite clearly has nothing to do with stūpas".[n.11] Judging from Schopen 1975/2005: 28, the distinction between caitya and stūpa is not necessarily self-evident, at least when it comes to references to the construction of memorials of the type similar to the present case. In relation to the Pāli term cetiya, Schopen 1989: 91ff even argues its equivalence to stūpa.

Nevertheless, granting for the moment that the reference to caityas in the Mūlasarvāstivāda version should indeed be considered as having "nothing to do with stūpas", the term used in the Dharmaguptaka, Mahāsāṅghika, and Mahīśāsaka Vinayas to describe the construction raised in memory of Kassapa/Kāśyapa, 塔, can render either stūpa or caitya.[n.12] Hirakawa 1997: 301. This case is thus similar to the Tibetan mchod rten, found in the above quoted passage from the Tibetan Mūlasarvāstivāda Vinaya. Schopen 1975/2005: 55 note 15 explains: "Tibetan has only a single term, mchod rten, that serves to translate both caitya and stūpa." The second of the two passages from the Chinese Mūlasarvāstivāda Vinaya, cited above p. 20 note 7, however, uses a transcription of the term stūpa. Thus if the construction of a memorial with the help of mud balls should be termed a caitya and not a stūpa, then Bareau's translation needs to be revised by replacing stūpa with caitya as the more appropriate rendering for 塔 in the other Vinaya versions.

Whether the term adopted is caitya or stūpa, it seems clear that the form of the relic cult envisaged in the Mūlasarvāstivāda versions is similar to what the other Vinaya versions report, and that they all knew of a memorial constructed at Toyikā in honour of the former Buddha Kassapa/Kāśyapa.

The Theravāda version consulted by Schopen differs in so far as it does not report a collecting of mud or clay balls. That is, this version differs in this respect from the summarized narrative given by Schopen (1985: 19) for the four versions – three Chinese Vinayas and the Theravāda version – which he sees as being in general agreement, in contrast to the Sanskrit Mūlasarvāstivāda Vinaya version. Instead of mud or clay balls, the Theravāda version only indicates that, once the magical display was over, a stone cetiya remained in that place.[n.13] Dhp-a III 253,7: tatth' eva taṃkhaṇe mahantaṃ pāsāṇacetiyaṃ ahosi.

Another Vinaya version of the present tale, summarized by Bareau (1960: 260) in the study on which Schopen based his presentation, also does not mention any collecting of clay balls. In fact, it does not mention the coming into being of a new stūpa at the time of the Buddha Gautama at all. According to Demié- ville (1951: 243 note 6), this text probably belongs to the Sarvāstivāda tradition, an identification confirmed by Yuyama (1979: 7).[n.14] This is the 鼻奈耶, T 1464 at T XXIV 897c20, on the school affiliation of which cf. also, e.g., Willemen et al. 1998: 85, Karashima 2006: 358, and Clarke 2015: 72. Bareau 1960: 230 just introduces this text as a "Vinaya anonyme".

Now this Sarvāstivāda version differs in nearly all aspects from the other Vinayas. It has nothing about the Buddha reaching a certain spot while travelling, about him encountering a man working in the fields, etc. Instead, it reports a visit by King Pasenadi, whom the Buddha tells about a stūpa constructed at the time of the passing away of the former Buddha Kassapa/Kāśyapa. The king living at that time in the past worshiped the Buddha's bodily remains with flowers, etc., and decided to construct a stūpa, for which he received help from four nāga kings. Once the stūpa was completed, a young man stole some of the flowers offered to it. This event then leads on to a tale about the karmic retribution he experienced for that theft.[n.15] For a summary of this tale and a discussion of some of its features cf. also Strong 2004: 35.

Thus already consulting this Sarvāstivāda version shows that the extant versions do not "agree on the basic elements of the text". If Schopen had taken this version into consideration, it would have become clear to him that the various Vinaya versions of this tale actually do not go back to a common early core.[n.16] Schopen 1985: 18 indicates that "here, unfortunately, I can give only a very condensed summary of this important case." I find it nevertheless somewhat puzzling that, in a case study aimed at drawing conclusions of considerable methodological significance, Schopen does not mention the existence of this Vinaya version at all, even though its relevant aspects were available in Bareau's study. The circumstance that Bareau does not identify the school of this Vinaya does not change the fact that we have here a Vinaya version that substantially differs from the other accounts.

It seems to me in fact improbable that serious scholars would, just because three Vinayas agree, necessarily "conclude – according to the accepted principle – that the essential elements of this account must go back to a very old or pre-sectarian stage of the tradition", this being the position that Schopen (1985: 20) depicts in order to then refute it.[n.17] The present case conforms to what to me appears to be a recurrent trope in Schopen's writings, where he depicts naïve conclusions drawn by often unnamed scholars that he then sets out to refute. I sometimes find it difficult to think of reputable scholars who indeed hold such opinions.

Another problem is that the version referred to by Schopen as the "Pāli version" or the "Theravāda version" is actually found in the commentary to the Dhammapada. This reflects a methodological problem, since Schopen compares Vinaya versions of other schools with a commentary of the Theravāda tradition. Bareau differs in this respect, as he does not consider the Dhammapada commentary on a par with the Vinaya accounts, but only compares canonical versions with each other.

Staying within the scope of Vinaya material, it is quite significant that in the Theravāda Vinaya no reference to a stūpa of Kassapa is found at all. This suggests the present case to be similar to the topic of monastic regulations for appropriate behaviour in regard to a stūpa. In contrast to other Vinayas, the Theravāda Vinaya has no reference to such behaviour at all and thus appears to testify to an early period when such regulations were not yet in existence.[n.18] The suggestion by Schopen 1989: 95 that "the total absence of rules regarding stūpas in the Pāli Vinaya would seem to make sense only if they had been systematically removed" has met with immediate criticism by Gombrich 1990b, Hallisey 1990, and von Hinüber 1990. Instead, the absence of any reference to stūpa worship indicates that the Pāli Vinaya was closed by the time stūpa worship came to be generally practised; cf., e.g., Bareau 1960: 230, Sasaki 1979: 196, Roth 1980: 186, Hirakawa 1993/1998: 272, Vetter 1994a: 1248 note 17, and Enomoto 2007: 194. Wynne 2005: 45 notes that "Schopen seems to have proved, inadvertently, that the Pāli canon was relatively closed … at an early date." On another unconvincing argument by Schopen on the absence of references to stūpa construction for nuns cf. Dhammadinnā 2016a: 44–53; and on the absence of the śikṣādattaka observance in the Theravāda Vinaya as another pointer to a comparatively early closure of this text cf. Anālayo 2016e.

The same would hold true for the Kassapa stūpa, in that the absence of any record of this stūpa in the Theravāda Vinaya suggests that, by the time this story came into vogue, the Theravāda Vinaya was already considered closed, so that this tale only found a placing in a Theravāda commentary. In other words, the very fact that in the Theravāda tradition a tale related to the Kassapa stūpa is only extant in the commentaries is yet another indicator that the various Vinaya versions of this tale do not go back to a common early core.

For a better assessment of the growth of the Kassapa tale, still another Vinaya text can be taken into account. This is the Mahā-vastu, which reports that the Buddha, who is travelling in Kośala, displays a smile. Asked by Ānanda why he smiled, the Buddha explains that on the same spot where they are now, the Buddha Kāśyapa had been seated, as well as two Buddhas previous to Kāśyapa.[n.19] Senart 1882: 318,12. Schopen 1985: 19 note 31 refers to this tale as one of various instances of "the smile of the Buddha". Ānanda then invites the Buddha to sit on the same place.

A similar tale can be found in the Ghaṭikāra-sutta and its Madhyama-āgama parallel, differing in so far as these only mention that the Buddha Kassapa had been seated on this spot, without any reference to other former Buddhas.[n.20] MN 81 at MN II 45,17 and MĀ 63 at T I 499a16. Other versions of this tale can be found in the Saṅghabhedavastu, Gnoli 1978: 22,1, and in an avadāna collection, T 197 at T IV 172c5. These two versions have a different setting, as they attempt to explain the karmic precedents for Gautama's practice of asceticism. Hence they do not begin with a reference to the spot on which the former Buddha had been seated. For a comparative study of MN 81 cf. Anālayo 2011b: 441–451. Oldenberg (1912b: 138) sees the Mahāvastu reference to three former Buddhas as an attempt to improve on the narrative, since the sudden reference to two additional former Buddhas contrasts with the preceding account in the Mahāvastu, which only mentions the Buddha Kāśyapa. In this respect the Mahāvastu appears to show a more developed stage than the Ghaṭikāra-sutta and its Madhyama-āgama parallel.

The motif of the smile displayed by the Buddha as a marker of the special nature of the spot related to the Buddha Kassapa recurs in the Dharmaguptaka, Mahāsāṅghika, and Mahīśāsaka Vinayas.[n.21] The Buddha's smile is reported in the Dharmaguptaka Vinaya, T 1428 at T XXII 958a27, in the Mahāsāṅghika Vinaya, T 1425 at T XXII 497b20, and in the Mahīśāsaka Vinaya, T 1421 at T XXII 172a5. Ānanda's invitation to sit on the same spot so that this place has been used by two Buddhas is recorded in the Mūlasarvāstivāda versions.[n.22] Ānanda's invitation to sit is reported in the Bhaiṣajyavastu, Dutt 1984a: 74,17 and D 1 kha 160a6 or Q 1030 ge 148b6, as well as in the Divyāvadāna, Cowell and Neil 1886: 77,4 (cf. also 466,1). This and other similarities point to the continuity between the Ghaṭikāra-sutta episode and the stūpa narrations in the different Vinayas. The version in the commentary on the Dhammapada in fact explicitly refers to the Ghaṭikāra-sutta.[n.23] Dhp-a III 251,9 reports that the Buddha narrated the Ghaṭikāra-sutta of the Majjhima-nikāya before revealing the stūpa of the Buddha Kassapa. Strong (2004: 39) must be correct when he concludes that the Ghaṭikāra-sutta tale stands at the beginning point of a textual development that eventually led to the narrations regarding Kassapa's stūpa.[n.24] As already noted by Strong 2004: 33, the existence of a stūpa of the former Buddha Kāśyapa was also known to Fǎxiǎn (法顯) and Xuánzàng (玄奘); cf. T 2085 at T LI 861a17 and T 2087 at T LI 900c18.

According to Mus (1935: 481), the reference in the Mahāvastu to a location used by former Buddhas points to a simple form of cult, before sacred architecture became important. On this reasoning the Ghaṭikāra-sutta and its Madhyama-āgama parallel, together with the Mahāvastu, would testify to an early stage when a relationship between the former Buddha Kassapa/Kāśyapa and a particular spot in the Kośala country was established without a reference to his bodily remains or to a stūpa. The tales in the Vinaya versions then would be subsequent developments of this motif, reflecting the coming into vogue of stūpa worship.

Even if one were to doubt the continuity between the reference to a mere spot where the Buddha Kassapa/Kāśyapa once sat and the story of his stūpa, the absence of any tale related to his stūpa in the Theravāda Vinaya makes it clear that this motif must indeed be a later development. Thus the case chosen by Schopen does not conflict with the tenet that regards agreement between parallel versions as a criterion for earliness, simply because in the present case the parallel versions do not agree. Close inspection shows that they in fact disagree in significant details, and paying attention to these disagreements and to all the available textual evidence allows us to trace the relative chronology of the different versions.

Instead of disproving this tenet, the present case shows that, although versions of the Kassapa/Kāśyapa stūpa tale were known in other Vinayas – in fact even a commentary belonging to the Theravāda tradition shows knowledge of this stūpa – nevertheless a story related to this stūpa was not incorporated in the Ghaṭikāra-sutta or in the Theravāda Vinaya. The present case thereby exemplifies a tendency noted by Norman (1997: 140), where popular Buddhist tales

are known in the Theravādin tradition only in the commentaries, although they are found in texts which are regarded as canonical in other traditions … we have, therefore, to assume that at least the Vinaya- and the Sutta-piṭaka had been closed at an earlier date.

Von Hinüber (2009: 47 note 43) explains that

during the past decades the age of the Theravāda tradition has been thrown into doubt occasionally, not rarely in a rather general and sweeping way … in this context it is remarkable that new concepts sometimes found their way only into the Theravāda commentaries, while they still could be included in the canonical scriptures of other traditions, which points to a rather early closure of the Theravāda canon.

Endo (2012: 41) notes, in regard to the Pāli commentarial tradition, that although

the Sri Lanka bhāṇakas had opportunities to homogenize the texts, if they so wished … nevertheless, what we find in the Pāli commentaries is contrary to our expectation … in other words, the bhāṇakas did not … homogenize the texts.

If even at the commentarial level a general trend towards homogenization is not evident, it seems fair to assume that in the case of the discourses, perceived by tradition as records of the Buddha's own word and therefore invested with much higher authority, it is rather improbable that they should be the product of wholesale later levelling.

The conclusions that can be drawn from the present case study are applicable to the case of other Pāli discourses in general. Once these are placed side by side with their parallel versions and subjected to a detailed comparative study, in the majority of cases a host of minor variations emerge alongside an impressive degree of similarity. This gives a strong impression that we are indeed dealing with the authentic remains of an oral transmission, not with material that is for the most part the product of later harmonization and levelling.[n.25] My assessment in this respect is based on my habilitation research in which I compared the one hundred and fifty-two discourses of the Majjhima-nikāya with over four hundred parallels preserved in Chinese, Sanskrit, Tibetan, etc., cf. Anālayo 2011b.

In relation to the early discourses in general, Salomon (2017: 261f) points out that

some skeptically inclined modern scholars have doubted the antiquity of the Buddhist canons as we know them, on the grounds that we have no written record of them until many centuries after the time of the Buddha. But I feel that there are two major arguments against what I am inclined to see as a hyperskeptical point of view.

First, there is the overall agreement, in broad terms, between the different Sūtra-āgamas with respect to their contents and arrangement, or at least their principles of arrangement. Secondly, it is striking that even the earliest representations of Āgama collections in written form, in the form of selective anthologies, clearly indicate the prior existence of complete Āgamas similar to those which we know from later times.

In sum, contrary to the position argued by Schopen, it seems to me quite sensible to assume that, if all known versions of a text or passage agree, that text or passage is probably old.[n.26] I have slightly reworded the maxim discussed by Schopen, especially replacing "must be very old" with "is probably old", since I would not wish to exclude categorically the possibility of levelling, which I think is a possible, but in most cases the less probable, explanation.

---

Additions Made in Sri Lanka

Returning to the topic of later additions to the Pāli discourses, besides the indications given in the Pāli commentarial tradition itself, modern scholarship has identified several cases that appear to be additions made in Sri Lanka to the canonical discourses.

Adikaram (1946/1994: 11) argues that, since the Paramatthajotikā reports that the Mahā-aṭṭhakathā did not comment on the last two stanzas of the Kokālika-sutta of the Sutta-nipāta, it follows that these stanzas must have been added in Sri Lanka.[n.27] Pj II 477,13: avasāne gāthādvayaṃ eva pana mahā-aṭṭhakathāyaṃ vinicchitapāṭhe n' atthi, made in relation to Sn 677f (Ee title: Kokāliya-sutta). Another two cases are noted by Abeynayake 2000: 172f, who quotes the discussion in Prajñārāma 1946: 19 and 31ff to the effect that additions to Pāli discourses made in Sri Lanka took the form of eight sentences added to the Mūlapariyāya-sutta and four stanzas added to the Sammādiṭṭhi-sutta (MN 1 at MN I 6 and MN 9 at MN I 55), each case betraying the influence of Sinhala Prākrit. In the case of the Mūlapariyāya-sutta, these sentences are uddānas placed after the different subsections of the discourse, e.g., after the exposition on the worldling: puthujjanavasena paṭhamanayabhūmi paricchedo, where the readings found in the Ceylonese edition differ from the Burmese and Siamese editions, perhaps having suffered some textual corruption. In the case of the Sammādiṭṭhi-sutta, a set of four uddānas summing up the contents of the whole discourse is found at the end of the Ceylonese and Siamese editions of the Sammādiṭṭhi-sutta, but is absent from the Burmese edition. Unlike the case of the Mahāparinibbāna-sutta or the Kokalika-sutta, however, these two instances do not add anything new to the content of the discourses itself, since they are only concerned with uddānas. Norman (1992: 275), however, points out that

their absence from the aṭṭhakathā may simply mean that no cty [commentary] on the verses was brought from India by the early missionaries. It seems unlikely that the two verses were added in Ceylon, for their metre seems to be unique and old. Since the verses give the answer to the question asked in the prose introduction, it is likely that they were composed at the same time as the introduction.

Another suggestion of Sri Lankan influence has been made by Bechert (1957), in that -e terminations for nominative singular found in the description of Makkhali Gosala's view in the Sāmaññaphala-sutta reflect Sinhalese influence. [n.28] DN 2 at DN I 53,29; cf. also Anālayo 2011b: 414 note 119. However, Norman (1976: 120) has drawn attention to similar -e terminations in a Jain text also reporting the views of heretical teachers. This suggests a common source for the form in which the views of these teachers were reported, independent of Sri Lankan influence.[n.29] A few cases of Sinhalisms in the Pāli canon have been noted by von Hinüber 1985/2001: 172 (§209) and 193 (§248). These concern minor matters related to pronunciation and do not affect the content in any way.

Substantial changes taking place after the canonical material had reached Sri Lanka have been suggested by Vetter (1994b), based on a comparative study of the Mahānidāna-sutta in the light of four Chinese parallels. Since this case exemplifies the type of conclusions that can be drawn based on comparative studies and the dates of the translation of Chinese parallels, in what follows I examine the arguments made by Vetter in some detail.

In an English summary at the end of his paper, Vetter (1994b: 159) explains that the Mahānidāna-sutta

appears, in a considerable number of instances, to be more 'developed' than the translations ascribed to An Shigao (late 2 c. C.E.) and in the Madhyamāgama (late 4 c.), but less than the versions in the Dīrghāgama (early 5 c.) and by Shi Hu (late 10 c.).

So it seems that even in the first centuries C.E. the Pāli text has been enlarged. Maybe this process came to a standstill as soon as the Sumaṅgalavilāsinī (compiled around 400 C.E.?) was accepted as an authoritative commentary and thereby encapsulated the Sutta.

In a subsequent paper, Vetter then refers back to his research on the Mahānidāna-sutta as the basis for his conclusion that the whole of the Pāli canonical texts are of less use for historical purposes than the early Chinese translations.[n.30] Vetter 1996: 49: "der volle Textbestand des Pāli-Kanon ist für historische Zwecke weniger brauchbar als frühe chinesische Übersetzungen." The conclusions by Vetter 1994b are quoted approvingly by Bronkhorst 2000: 77 note 2.

If I understand him correctly, the approach Vetter adopts is based on the notion that the date of translation of the Chinese versions reflects their respective degree of development. By assessing the Pāli version's degree of development in comparison, we can then determine the date of its final redaction.

As far as I can see, on seven occasions Vetter explicitly identifies a particular passage in the Mahānidāna-sutta as later than one or more of its parallels. In what follows, I briefly survey these seven passages from the viewpoint of how far the suggested stage of development reflects the date of translation or assumed closure.

1) The first passage identified by Vetter as late in the Pāli version has a shorter and thus presumably earlier counterpart in the translation attributed to Ān Shìgāo (安世高, late second century),[n.31] Vetter 1994b: 139 note 13 refers to Zürcher 1959/1972: 33 and page 331 note 83, who includes T 14 among the translations that can safely be attributed to Ān Shìgāo (安世高); cf. also Nattier 2008: 45, Zacchetti 2010a: 252, and Vetter 2012: 7. in the Madhyama-āgama version (late fourth century), and in the translation attributed to Shīhù (施護, late tenth century).[n.32] The passage in question has an introductory statement regarding the profundity of dependent arising. DN 15 at DN II 55,14 stands alone in indicating that lack of understanding this principle leads beings to rebirth in lower realms, pajā … apāyaṃ duggatiṃ vinipātaṃ saṃsāraṃ nātivattati; Vetter 1994b: 143 comments: "Der Abweg, der sich nur im Pāli-Text (mit den Worten apāya-, duggati- und vinipāta-) findet … dürfte ein später Zusatz sein." The problem of being subject to saṃsāra, without a reference to lower realms, is also mentioned in T 14 at T I 242a7, MĀ 97 at T I 578b20, and T 52 at T I 844b24. The Dīrgha-āgama version (early fifth century) does not have this passage at all, which Vetter interprets as a sign of loss or intentional removal.[n.33] DĀ 13 at T I 60b9. Vetter 1994b: 143 comments that in this case such a passage "ist also eher weggefallen oder weggelassen"; cf. also the general reflections in Vetter 1994b: 140 note 14. It is not clear to me why the short version found in DĀ 13 must be either the result of a textual loss or else an intentional omission. Since the discourse would be coherent without the passage in question, it could just as well be that DĀ 13 testifies to an early stage, with the remaining versions being subsequent expansions.

2) Regarding the second instance of an alleged lateness in the Pāli version,[n.34] This case concerns the initial exposition of the links of dependent arising in reverse order, which begins with old age and death and leads up to the reciprocal conditioning of consciousness and name-and-form, DN 15 at DN II 56,28. Vetter 1994b: 144 then suggests this to be a later addition: "Daß die 'Katechese' schon hier eine Reihe von Alter und Tod bis zu nāmarūpa- und viññāṇa-, welch letztere sich gegenseitig bedingen, zeigt, dürfte spät sein." the situation in the parallels is as follows: the translation attributed to Ān Shìgāo (late second entury), the Madhyama-agama version (late fourth century) and the translation attributed to Shīhù (late tenth century) have a shorter and thus presumably earlier counterpart,[n.35] T 14 at T I 242b12, MĀ 97 at T I 578c2, and T 52 at T I 844c15 only go up to craving. while the Dīrgha-āgama version (early fifth century) goes further than the Pāli discourse.[n.36] DĀ 13 at T I 60b23 continues beyond consciousness by mentioning formations and ignorance.

3) In the case of the third identification by Vetter of what he believes to be a late element in the Mahānidāna-sutta,[n.37] This instance concerns the beginning of a series of conditions that in all versions lead from questing to quarrelling and fighting. DN 15 at DN II 58,31 starts with feeling, where Vetter 1994b: 145 comments that this would be a later addition, "vedanā- hat hier aber, wie die anderen Versionen nahelegen, früher wohl nicht gestanden." all of the four Chinese parallels (late second to late tenth century) show a shorter and thus presumably earlier presentation.[n.38] In DĀ 13 at T I 60c19, T 14 at T I 242b18, MĀ 97 at T I 579a1, and T 52 at T I 844c16 the starting point is craving.

4) Another passage in the Mahānidāna-sutta is according to Vetter a later development of an originally more simple presentation,[n.39] The passage begins in DN 15 at DN II 62,10, with Vetter 1994b: 150 proposing that only three of the four extant enquiries are original, which, however, would have been simpler, "anfänglich nur drei negative Wenn-dann-Sätze gab, die nicht die Verwicklungen der Pāli-Version besaßen". such as found in three of the Chinese parallels (late second to early fifth century),[n.40] DĀ 13 at T I 61b3, T 14 at T I 243b8, and MĀ 97 at T I 579c7. whereas a counterpart to this exposition is not found at all in the latest Chinese translation (tenth century).[n.41] In relation to T 52, Vetter 1994b: 147 assumes an intentional omission motivated by lack of understanding, "Verzicht auf eine schwierige und seinerzeit wohl kaum mehr verständliche Diskussion".

5) Another supposedly late part of the Mahānidāna-sutta is absent from all of the Chinese parallels (late second to late tenth century).[n.42] The passage identified as late in DN 15 at DN II 63,27 indicates, in relation to the reciprocal conditioning of consciousness and name-and-form, that it is to this extent that one is born, becomes old, dies, passes away, and appears again. Vetter 1994b: 151 comments on this passage that "in den anderen chinesischen Versionen ist dazu nichts zu finden. Es ist daher möglich, daß es in später Zeit … aus dem Mahāpadāna-sutta (D II/32) hierher verschleppt wurde", and again Vetter 1994b: 152: "es ist nicht wahrscheinlich, daß der Satz ettāvatā kho … uppajjetha (v.l. upapajjetha) vā zum alten Kern … gehörte."

6) The next identification of a late passage in the Mahānidāna-sutta has counterparts in three of the Chinese parallels (late second to early fifth century),[n.43] The passage in question concerns an examination of four types of self, DN 15 at DN II 64,3, which according to Vetter 1994b: 154 "gehört kaum zum frü- hesten Bestand". Counterparts to this passage can be found in DĀ 13 at T I 62a17, T 14 at T I 244a29 and MĀ 97 at T I 580c3. but is absent from the latest Chinese translation (late tenth century).[n.44] Vetter 1994b: 153 note 40 comments that this passage appears to be missing from T 52.

7) Another Mahānidāna-sutta passage, identified as a later element,[n.45] This concerns the path for designation, adhivacana, DN 15 at DN II 68,18, on which Vetter 1994b: 156 comments that "auch hier weist MA also auf ein Stadium hin, in dem weniger Begriffe vorkamen als heutzutage im Pāli." is absent from the earliest and latest translations (second and tenth century), whereas a shorter counterpart can be found in the Madhyama-āgama version (late fourth century) and another corresponding passage in the Dīrgha-āgama (early fifth century).[n.46] DĀ 13 at T I 62a8 and MĀ 97 at T I 580b22.

In sum, even if all these seven cases should indeed be later additions to the Mahānidāna-sutta, which I find rather doubtful,[n.47] Regarding Vetter's assessment of the implications of the above differences, I am under the impression that a central principle behind his analysis is the assumption that the majority of sources must have the correct reading. Although this may often be the case, it seems to me that at times the correct reading could also be preserved in a single version, against the majority of parallels. Regarding the seven cases surveyed above, I for one fail to see a definite reason why some of them could not have been part of the original. Vetter in fact words his judgements in the subjunctive mode, indicating that these are just possibilities. Yet, if these are mere possibilities, it follows that their lateness is not sufficiently well established to use this as a basis for assessing the historical value of the Pāli discourses in general. in each of the above surveyed instances the dates of the Chinese translations do not reflect the suggested stage of development of the respective texts. That is, not a single one of these seven instances conforms to the pattern proposed by Vetter that the Pāli version is

more 'developed' than the translations ascribed to An Shigao (late 2 c. C.E.) and in the Madhyamāgama (late 4 c.), but less than the versions in the Dīrghāgama (early 5 c.) and by Shi Hu (late 10 c.).

The same holds true even if the latest translation attributed to Shīhù is left aside.[n.48] Vetter 1994b: 141 notes at the outset of his examination the somewhat doubtful nature of the version translated by Shīhù. In the case of the remaining three Chinese parallels the pattern is still not invariably one of the version attributed to Ān Shìgāo and the Madhyama-āgama version preserving what Vetter holds to be earlier than the Pāli version, and the Dīrgha-āgama version reflecting further developments.

This illustrates what is in fact a general feature of Chinese translations, of which the present case merely provides an illustration. In principle there is no reason why a fairly early Indic text might not have been translated at a relatively late date in China, because it reached China only at that time. In contrast, a text that had come into being more recently in India may have reached China more quickly and therefore been translated at an earlier date. In other words, the dates of Chinese translations do not necessarily correspond to the chronology of textual composition in India.

Turning to the translation of Buddhist texts into European languages provides a convenient illustration of this principle. The fact that the Saddharmapuṇḍarīka-sūtra was translated already in the mid nineteenth century by Burnouf (1852) does not mean that this text must be earlier than the Aṣṭasāhasrikā Prajñāpāramitā translated more than a century later by Conze (1973).[n.49] This example was given by Douglas Osto in a paper delivered at the fifteenth IABS conference in Atlanta in 2008.

As Schopen (2000/2005: 4) rightly points out,

it is not impossible that some – if not a great deal – of what has been said on the basis of Chinese translations about the history of an Indian text has more to do with the history of

Chinese translation … than with the history of the Indian text itself.

Moreover, the date at which a text containing a presumably earlier version of a particular passage was translated into Chinese gives us a terminus ante quem only for changes in that same text, not for changes in any other text. Changes in the Pāli parallel could have occurred at any time after the respective textual traditions had begun to transmit their versions separately, be this because of geographical separation or for any other reason. This would be the terminus post quem for changes in the Pāli version, not the date when any parallel was translated into Chinese.

In other words, the Mahānidāna-sutta passages identified by Vetter as later changes could in principle have been present already in the Pāli material that according to the traditional account was brought to Sri Lanka during the reign of Asoka. There is no reason to think that these presumed alterations in the Mahānidāna-sutta could only have happened after the parallel version(s) had been translated into Chinese.

---

The Finalization of the Pāli Commentaries

The assumption by Vetter that the development he sees in the Mahānidāna-sutta would have come to a conclusion around the time of the finalization of the Pāli commentaries echoes a position taken by Schopen, hence in what follows I examine the relevance of the finalization of the Pāli commentarial tradition in relation to the Pāli discourses.

In the paper that also takes up the tale of the stūpa of the former Buddha Kassapa/Kāśyapa,[n.50] Schopen 1985: 10f also examines how far extant manuscripts and epigraphy reflect knowledge of the canonical collections. His presentation would need to be revised in light of the recent Sanskrit and Gāndhārī fragment findings, also taking into account the suggestion by Norman 1989: 33 that "it is clear from inscriptional evidence that the nikāyas had certainly received their names by the second century BC"; cf. in more detail Salomon 2017. For a reply to the suggestion by Schopen 2010: 127 that the early generations of Buddhist reciters would not have been able to transmit the amount of literature later considered canonical cf. Anālayo 2011b: 870 note 65. Schopen (1985: 10) proposes that

it is not until the time of the commentaries of Buddhaghosa,

Dhammapāla and others – that is to say, the 5 /6 Century A.D. – that we can know anything definite about the actual contents of this [Pāli] Canon.

He bases himself for this assessment on a quote from Malalasekera (1928/1994: 44) to the effect that "how far the Tipiṭaka and its commentaries reduced to writing at Alu-vihāra resembled them as they have come down to us now, no one can say." This remark needs to be considered within its original context, since on the same page Malalasekera continues by indicating that

in view of the great care with which the orthodox monks attempted to preserve the purity of the Word there is not much probability that the canon underwent any material changes in the interval.

Assessing Schopen's proposal requires defining what "knowing anything definite" refers to. If it refers to the type of certainty that comes from having direct access to manuscripts, then in the case of the Pāli discourses the fifth to sixth centuries C.E. do not offer us anything that enables such definite knowledge, because manuscripts of the Pāli discourses from that period are not extant. According to von Hinüber (1978: 48),[n.51] For a critical examination of a suggestion in the same paper by von Hinüber 1978: 57, regarding a case of alleged late influence on the Pāli discourses, cf. Anālayo 2011b: 779f note 118. in the case of Pāli canonical material we have in general "no manuscript older than about 400 years", with a few exceptions.[n.52] In the case of the Saṃyutta-nikāya, de Silva 1938: 20 lists a manuscript (no. 70) of this collection with the indication "original copy, copied in 1412 A.C.", although the precise dating seems uncertain, cf. von Hinüber 2000: 89f. There is, however, additional evidence for a long-standing transmission of the Saṃyutta-nikāya and its commentaries in Sri Lanka; cf. von Hinüber 1988: 178. Thus definite knowledge in relation to the Pāli discourses would need to be placed considerably later than Schopen proposes, perhaps as late as the seventeenth century in order to be able to have direct access to manuscripts of all four Nikāyas.

If, however, "knowing anything definite" refers to drawing a reasonable inference, then we would be able to go further back into the past. Here the question would be how far we are willing to trust the written transmission of texts in Sri Lanka. If this cannot be trusted in principle, we are back to the date of actual manuscripts of the main Pāli Nikāyas, somewhere around the seventeenth century.

Once we assume major changes were possible before the fifth century, affecting the written canonical material transmitted in various manuscript lineages in Sri Lanka, there would be no reason why such changes should not be possible afterwards. If the written transmission up to the fifth century is not trustworthy, there is no reason why it should be trusted after the commentaries were finalized in the fifth or sixth century, based on already existing commentarial traditions. That is, whatever happened in the fifth or sixth century does not offer us real certainty in our knowledge about the Pāli canon that decisively differs from what can be known about the same canon in, say, the third or fourth century.

If the early written transmission in Sri Lanka cannot be trusted, then there would be no reason to trust it in other countries. In the case of the Chinese canon, for example, taking the same sceptical attitude we would have to discount all information we have on different early translators. Distrusting the written transmission during the early centuries of Buddhism in China, one might then allocate all translated works equally to the time of the eighth century catalogue Kāiyuán lù, 開元錄 (or more precisely 開元釋教錄), which had a determining influence on the formation of the Chinese canon and thus would be the closest equivalent to the role the finalization of the Pāli commentaries is presumed to have played for the Pāli discourses.[n.53] Cf. esp. the 入藏錄, T 2154 at T LV 680b. As part of a study that highlights the value of studying Chinese translations for identifying changes in the corresponding Indic texts, Harrison 2008/2010: 218 offers the pertinent observation that "scepticism, then, is all very well, but it is no better than blind faith when it chooses not to look at the evidence." Based on a study of pillar inscriptions from Deorkhotar, Salomon and Marino 2014: 37 similarly warn against excessive scepticism, pointing out that "while caution and a healthy scepticism are not only desirable ingredients but absolutely necessary ones for good scholarship, there is always the danger of letting scepticism take over one's thinking, leading to the mindset of 'In the end, we know nothing'."

Yet, familiarity with the different Chinese translations makes it clear that a development in translation terminology from early to later translators has taken place, making such scepticism an unconvincing position.

The same holds for the Pāli material, where familiarity with the sources shows an evident development in language and content from the discourses via the Vinaya and the Abhidharma to the commentaries.[n.54] As von Hinüber 1996/1997: 26 points out, "the cultural environment of the first four Nikāyas of the Suttapiṭaka is markedly older than that of the Vinaya-piṭaka." Gethin 2006: 84 notes that "in terms of doctrinal development the four primary Nikāyas are clearly older than certain texts of the Khuddaka-nikāya and the texts of the Abhidhamma Piṭaka, which in turn are clearly older than the Pāli commentaries." Thus to allocate all this material to the fifth century is similarly unconvincing.

Schopen (1985: 19), in the context of his discussion of the Kassapa tale, in fact comes to the conclusion that the version found in the Dhammapada commentary "appears to be the latest of all the versions", the rest of which are, as mentioned above, canonical versions found in different Vinayas. That is, his own assessment of the sources conforms to the principle according to which the commentaries contain later elements than the canonical scriptures on which they comment.

Assuming that the written transmission, be it in Sri Lanka or in China, can be trusted to have preserved the essentials, with the inevitable minor errors and changes that are natural byproducts of textual transmission, then in the case of the Pāli discourses we could go back all the way to the first century B.C.E., when according to the traditional account the texts were written down at the Alu-vihāra.

Now the Indian commentaries that had been rendered into Sinhala were according to the Mahāvaṃsa written down together with the canonical texts in the first century B.C.E,[n.55] Mahāvaṃsa 33.100f. Hartmann 2009: 97 comments that "the wording of the passage appears to presuppose an already well-structured canon with commentary." and these were apparently passed on without substantial changes.[n.56] According to Endo 2005: 50, the commentarial tradition stemming from India, which forms the basis for the Pāli commentarial tradition, came to a closing point around the time of the first century B.C.E. He explains that "there is no literary evidence whatsoever to prove that the class of basic … sources of the present Pāli Commentaries … was expanded … [by adding] new elements after the time of King Vaṭṭagāmaṇī-Abhaya. This positively suggests that the expansion of the … 'Original Commentaries' had come to a completion by the time of this king, when the Buddhist texts were committed to writing." Even the additional commentaries by Sri Lankan elders and tales related to Sri Lanka, recorded in the Sri Lankan commentarial tradition, seem to reflect only a period up to the third century.[n.57] Regarding the opinions of individual Theras quoted in the commentaries, von Hinüber 1996/1997: 101 points out that "as far as these Theras can be dated, almost all of them lived before AD 100." Mori 1989: 5 explains that "all the datable Sri Lankans appearing in the Commentaries are restricted, with a very few exceptions, to those who were active in the period between the reign of King Devānampiyatissa (reigning: B.C. 250–210) … and that of King Vasabha (A.D. 65–109)." Mori 1988/1989: 99 adds that "in the reign of Mahāsena (A.D. 276–303) the formation of the sources ceases entirely: after that time nothing was added to them; and they were well preserved until the time when the Pāli commentators wrote their Aṭṭhakathā texts based upon those sources in the fifth and sixth centuries or thereabouts. Consequently the real and substantial date of the Aṭṭhakathā texts should be regarded not as the date of the writing of the texts, but as the date of their source materials." For instances where the Pāli commentaries clearly rely on old material cf. also von Hinüber 2015a: 369 and 376. Thus if the commentaries (aṭṭhakathā) on the Pāli discourses should be considered as in some way encapsulating the discourse material and helping to fix its final form, this would in the main be the case already for the 1 century B.C.E., with the time of definite closure apparently being the 3 century, instead of the 5 century.

Moreover, Endo (2009: 175) notes that there

is a general practice in the Pāli commentaries that, whenever authorities are sought for the sake of supporting, or going against, a view, or for mere additional information, the 'pāḷī' (Canonical Text(s)), the 'Aṭṭhakathā ' … are cited in this order of importance … this classification appears to have been established according to the historical evolution of Buddhist texts.

Endo (2008: 165) also explains that on those few occasions when Buddhaghosa expresses disagreement with positions held in the Mahāvihāra tradition, "his criteria were based on giving the first preference to the Tipiṭaka", that is to say, Buddhaghosa's regard for what he considered canonical was such that he would venture to disagree with the Mahāvihāra exegetical tradition if he felt that this was not in line with the canonical position.

In sum, if the ancient Indian commentarial tradition that formed the foundation for the Pāli commentaries as we have them now can indeed be considered as coming to a closing point in the first century B.C.E., and keeping in mind the respect given to the canonical material over the commentaries, it follows that the Pāli discourses on which the commentaries comment must also have come to a close by that time. It is in fact hard to imagine how, after the texts had been written down, changes could have been implemented consistently, once copies of the written texts had begun to spread to various monasteries in Sri Lanka.

This is not to propose that what was written down at that time is exactly what we have now. In the course of time most certainly minor changes and errors occurred during the copying of the texts, resulting in the variants to be found between the different extant editions.[n.58] A rather substantial 'variant' would be the long exposition on the four noble truths found in the Be and Se editions of the Satipaṭṭhāna-sutta, MN 10, which is absent from the Ce and Ee editions; cf. MN I 62,21. Nevertheless, this variant simply follows the example of the Mahāsatipaṭṭhāna-sutta, DN 22 at DN II 304– 313, and thus does not involve a real innovation. Trenckner 1888/1993: 534 notes that the long exposition on the four noble truths in MN 10 was in the Burmese manuscript from the India Office Library that he consulted; in fact the long section is already found in the Burmese fifth-council edition. But it seems to me realistic to assume that in terms of doctrinal content the Pāli discourses would have been in existence in a form roughly resembling what we have now by the first century B.C.E.[n.59] Reat 1991: 140 explains that "in recent years the emphasis of Buddhist studies in the West has fallen upon what cannot be ascribed to earliest Buddhism – i.e. that virtually nothing can be attributed with any certainty to earliest Buddhism. For many Western scholars, this position has become an indisputable maxim which justifies neglect of … the Pāli canon" and provides "an excuse for some Western academics to ignore … the Pāli suttas". Gómez 1995: 212 remarks that "common among contemporary scholars is the role of … the guardian of 'secular authority' … I refer to the scholar's interest in undermining the authority of the tradition he or she studies. Seldom is this role part of the scholar's public role."

The Political and Geographical Setting

The assumption that the Pāli discourses would have come to a close by the time of the writing down of the canon in Sri Lanka finds support in the fact that the world the discourses reflect in political and geographical terms seems to be earlier than the first century B.C.E.

In relation to knowledge of the political situation, as already pointed out by Rhys Davids (1903/1997: 174), it is significant that the Pāli discourses

make no mention of Asoka. Had they undergone any serious re-editing after the reign of the great Buddhist Emperor (of whom the Buddhist writers, whether rightly or wrongly, were so proud), is it probable that he would have been so completely ignored?[n.60] In the same vein, Geiger 1916: 7 comments that "es ist auch sehr zu beachten, daß der Name des in der buddhistischen Gemeinde so hochgefeierten Asoka nirgends im Kanon erwähnt wird. Dieser war eben zu Asokas Zeit inhaltlich so ziemlich abgeschlossen." Pande 1957: 16 similarly points out that "the silence of the canon over Aśoka, which contrasts so strongly with later tradition … suggests its antiquity", concluding that "it may be asserted that the growth of the Nikāyas falls between the 5th and the 3rd centuries B.C." Another relevant observation regarding political constitution has been made by Ling 1983: 69, who in relation to "the disagreement between canon and chronicles on the concepts of kingship and nationalism" notes that "a disagreement of this sort … is one of sufficient importance to the Sinhalese that for them to have transmitted it suggests that the canon had been transmitted without interference before Sinhalese historiography began."

The motif of the wheel-turning king, found repeatedly in the Pāli discourses, need not reflect knowledge of Asoka. A century ago Jacobi (1910: 336) pointed out that "the idea of a universal monarch … is very ancient in India", already found in the pre- Buddhist period. Others have come to similar conclusions. [n.61] Cf., e.g., Sastri 1940, Zimmer 1951: 129, Horsch 1957: 73, Wijesekera 1957: 265, Drekmeier 1962: 203, Armelin 1975: 6, Nanayakkara 1977: 592, Strong 1983: 48, Chakravarti 1996: 6, Collins 1998: 470, Mahony 2005: 1350, and Anālayo 2011g. Hence references to the cakravartin/cakkavattin need not be considered post-Asokan additions.

In regard to geography, Bareau (1984: 37) points out that the description of the realm of King Reṇu in the Mahāgovinda-sutta of the Dīgha-nikāya reflects a geographical awareness more limited than what would have been known by the time of Candragupta.[n.62] DN19 at DN II 235,19. Bareau 1984: 36f explains that the "authors" of the text "defined the kingdom of their hero from a real model which they knew and which was then the largest territory united under the authority of a king in the Ganges basin. If they had known an even more widespread empire encompassing that same region, there is no doubt they would have taken it as an example and consequently pushed the frontiers of Reṇu's … further back … In comparison to the territories possessed by the kings … during the life of the Buddha, this kingdom attributed to King Reṇu was indeed immense, but compared to that which Candragupta, the first sovereign of the Maurya dynasty, placed under his command, it was quite modest."

Regarding the reference to Pāṭaliputta in the Mahāparinibbāna-sutta, von Hinüber (2008b: 202 and 204) points out that

no mention is made of Pāṭaliputta as the capital (rājadhānī, e.g., D II 7,29) of the Maurya empire, in spite of the fact that Mauryas are indeed referred to at the very end of the same text … [which] makes a lot of sense if the text is pre-Mauryan.[n.63] DN 16 at DN II 87,34.

The Dīgha-nikāya collection also contains a listing of ten Indian countries,[n.64] DN 18 at DN II 200,4 lists the inhabitants of the countries Kāsi, Kosala, Vajjī, Malla, Cetiya, Vaṃsa, Kuru, Pañcāla, Maccha, and Sūrasena. On the degree of urbanization reflected in the Pāli discourses cf. also Witzel 2009: 298. which several discourses in the Aṅguttara-nikāya then expand into a listing of sixteen countries, additionally mentioning the inhabitants of Assaka towards the south, Avanti in the west and the Gandhāra and Kamboja regions in the north-west.[n.65] AN 3.70 at AN I 213,1, AN 8.42 at AN IV 252,5, AN 8.43 at AN IV 256,17, and AN 8.45 at AN IV 260,27, which besides Assaka, Avanti, Gandhāra, and Kamboja also mention Aṅga and Magadha, in addition to the locations listed in DN 18 (see above note 64). Whatever the precise boundaries of these regions, the geographical knowledge reflected in the Pāli discourses is certainly more restricted than what is found in Asokan edicts, which show a wider knowledge, extending to South India and even to Sri Lanka.[n.66] Rock edicts II and XIII refer to Taṃbapaṃṇi (with several variant readings); cf. Bloch 1950: 93 and 130. In contrast, references to Tambapaṇṇi are not found in the Pāli discourses, but only in other works of the Pāli canon, like the canonical commentary Mahāniddesa, Nidd I 415,8, an uddāna in Vin II 72,27, and repeatedly in the late Parivāra (e.g., Vin V 3,4, 3,30, 5,9, 8,3, 49,7, 49,24, 55,1, 56,4, 81,36, and 82,13); cf. also the discussion in Cousins 2013b. Regarding the significance of the text listed in Asoka's Bhābhrā Edict, the fact that he selected a few textual pieces that could easily be memorized by his subjects does not imply that only those texts were in existence and the four Nikāyas or Āgamas had not yet come into being at that time; cf. in more detail Anālayo 2011b: 866f.

Kamboja also occurs in conjunction with another location that according to several scholars reflects knowledge of the Bactrian Greeks and thus points to a time after the conquest of Alexander the Great: Yona.[n.67] Cf. Winternitz 1920/1968: 40, Lamotte 1947: 217, Bechert 1961: 41 note 2, Bronkhorst 2007a: 209 and 2011: 35f. The reference in question occurs in the Assalāyana-sutta and in its Madhyama-āgama parallel. The two versions agree that, in a discussion regarding the brahminical claim to class superiority, the Buddha asks if his visitor has already heard of a two-class society consisting of masters and slaves existing in some border countries, such as Yona.[n.68] MN 93 at MN II 149,4: sutan (Be and Ce: sutaṃ) te yonakambojesu (Se: yonakakambojesu) aññesu ca paccantimesu janapadesu dveva vaṇṇā, ayyo c'eva dāso ca and MĀ 151 at T I 664a18: 頗聞餘尼及劒浮國有二種姓, 大家及奴. Another parallel, T 71 at T I 877a13, differs inasmuch as instead of Yona it speaks of the Kuṣāṇās.

Now a well-known example of a two-class society consisting of masters and slaves was Sparta, and closer inspection of the historical context suggests that Indians at the time of the Buddha could indeed have had knowledge of Sparta.[n.69] I already drew attention to this in Anālayo 2009m.

The Spartan leader Leonidas is famous for valiantly resisting, with a small group of followers at the pass of Thermopylae, the army of Xerxes, who in the year 480 B.C.E. had begun an invasion of mainland Greece. This invasion was in revenge for the support the Greeks had offered to a revolt of the Ionians living in Asia Minor against Persian dominion.

According to Herodotus, among the large forces that Xerxes had mustered was a contingent of Indian archers.[n.70] Histories 7.65, Godley 1922/1982: 378: Ἰνδοὶ δὲ εἵματα μὲν ἐνδεδυκότες ἀπὸ ξύλων πεποιημένα, τόξα δὲ καλάμινα εἶχον καὶ ὀιστοὺς καλαμίνους· ἐπὶ δὲ σίδηρος ἦν. ἐσταλμένοι μὲν δὴ ἦσαν οὕτω Ἰνδοί, which describes the Indians as clad in cotton and armed with reed bows and iron-tipped arrows. During their advance into Greece, the Persian army had found a bypass and circumvented the Greek defenders at the pass of Thermopylae, attacking them from the rear as well. To enable an orderly withdrawal of the remainder of the Greek army, Leonidas remained with a few Spartan companions to defend the pass to the last man. Their heroic resistance deeply impressed the Persian army of mercenaries.

In view of this historical background, it would not be surprising if by the time the Indian survivors of the unsuccessful campaign of Xerxes had returned home, the rumour had spread of the valiant soldiers of Sparta, who lived in a two-class society of masters and slaves. Since this campaign was in revenge for the support given by the mainland Greeks to the Ionian revolt against Persian dominion, it would also not be surprising if the Indians referred to the Spartans as people from Yona, and allocated them to some border countries of which they had no clear geographical awareness.

The fact that according to the Assalāyana-sutta and its Madhyama-āgama parallel the Buddha first ascertains whether his visitor had already heard about this two-class society gives the impression that this may have been a fairly new piece of information. Thus, instead of being a sign of lateness, the reference to Yona rather seems to be testifying to the earliness of the passage in question.[n.71] That the reference to Yona or Yauna need not be taken as a sign of lateness has in fact already been pointed out by several scholars. Bühler 1895/1963: 27 note 1 comments that "it seems to me not wonderful that an author … whose countrymen, the Gandhāras, had furnished a contingent for Xerxes's invasion of Greece, should mention the old Oriental name of the Greeks … to me it seems absolutely impossible to make the mere occurrence of the word Yavana in Sanskrit works a mark, which proves that they must have been written after the invasion of Alexander." Sircar 1953: 101f comments that "the Greeks of Ionia in Asia Minor … and the people of north-western India certainly came into contact with each other … since the time of Darius I (522–486 B.C.)." Given that "Indian archers … formed an important element in the army of Xerxes and shared the Persian defeat … the people of Uttarāpatha at least had direct knowledge of the Greeks as early as the sixth and fifth centuries B.C.", so that it is "no wonder that the very word Yauna, meaning the Greeks, used for the first time in the records of Darius I, was borrowed by the Indians without any modification." Halbfass 1995: 199 sums up: "references to yona, etc., in the Pali canon need not be later interpolations, added after Alexander's campaign. They may well go back to the oldest layers of the canon."

Conclusion

In sum, it seems to me reasonable to consider the Pāli discourses as fairly closed, in doctrinal terms, by the time of the first century B.C.E., obviously with small variations of the type characteristic of written transmission still taking place.[n.72] Cousins 2012: 68 similarly comes to the conclusion that the evidence he has surveyed "is most compatible with the position that the Pali Canon, largely or wholly, dates back to the first century BC." That is, I would conclude that from that time onwards they were in existence in a form that roughly resembles what we currently have at our disposal. In terms of geographical awareness and in terms of doctrinal development, they seem to go back even further into the past, allowing us to catch a glimpse of Buddhist thought predominantly from the pre-Asokan period.

The reason for such early closure could well be that by the time of Asoka interests in the Buddhist communities began to develop in directions that to some degree led away from a concern with the discourses. The development of narrative material of the jātaka and avadāna type, evident in the different Vinayas and subsequently in autonomous collections dedicated to this genre, alongside the growing interest in abhidharmic philosophy, also leading to a genre of texts on its own, had by then emerged as the central venue for new thoughts and ideas. Soon enough early Mahāyāna texts would come to fulfil a similar role. Once any of these texts had acquired a canonical status and/or become closed in turn, commentaries and manuals or śāstra texts would have stood in their place, providing the arena for new developments.

In other words, I would hold that, although the Pāli discourses do reflect stages of development up to a certain point, their role was soon displaced by other texts.[n.73] [72] Examples for this would be, e.g., the first beginnings of what would eventually become the fully fledged bodhisattva ideal, on which cf. Anālayo 2010b and 2017a, or else the beginnings of abhidharmic thought, cf. Analayo 2014c, just to mention two examples where a comparative study of the early discourses enables the construction of informed hypotheses regarding early stages in the development of Buddhist thought and doctrine. This, so it seems to me, is precisely why the Pāli discourses, when studied in comparison with their parallels preserved in Chinese, Sanskrit, Tibetan, etc., offer a window on the early stages of the development of Buddhist thought and doctrine, inasmuch as this has left its traces in literature.