Introduction

With this chapter I return to the topic of the oral transmission of the early discourses, already broached in the first chapters of this book. This results in a shift of focus from the Dīrgha-āgama to material that is of relevance to appreciating the character of the early discourses as orally transmitted texts in general. In what follows I take up the significance of the set of nine or twelve aṅgas of texts as reflected in early Buddhist discourse literature and in relation to the division of texts into āgamas or nikāyas.

---

The Term Āgama in the Singular

The term āgama used in the singular occurs in several Pāli discourses, often qualifying a learned monk to be āgatāgama, literally one to whom the coming down has come down, that is, one who has "obtained the āgama" in the sense of being versed in it.[n.1] Similar references can also be found in Vin I 119,22, Vin I 127,30, Vin I 337,12, Vin I 338,20, Vin II 8,28, Vin II 55,19, Vin II 98,5, Vin II 299,16, and Vin IV 158,18 (notably the last case involves a learned layman). The significance of the term āgama itself is succinctly captured by Silk 2015b: 7 as "what has come down to us, what has been transmitted, tradition", which "in a Buddhist context [is] also a term for a collection of the Buddha's sermons." Such references often form part of a description of learned elders whom one would approach to receive clarifications regarding the teachings given by the Buddha.[n.2] AN 3.20 at AN I 117,28, AN 6.51 at AN III 361,23, and AN 10.11 at AN V 15,30. Alternatively, the passage in question may take up the need for such learned elders to give teachings.[n.3] AN 4.160 at AN II 147,29 and AN 5.156 at AN III 179,2.

Most of these references occur in discourses in the Aṅguttara-nikāya, of which no parallels have been preserved. In the case of one such reference in the Mahāgopālaka-sutta of the Majjhima-nikāya, however, three parallels extant in Chinese translation also mention such elders. Yet they only qualify them as "learned", without bringing in the other epithets used in the Mahāgopālaka-sutta and thus also without a counterpart to āgatāgama.[n.4] MN 33 at MN I 221,21 (=AN 11.18 at AN V 349,16) describes the elders as "learned, have obtained the āgama, bearers of the Dharma, bearers of the discipline, and bearers of the summaries", bahussutā āgatāgama dhammadharā vinayadharā mātikādharā. SĀ 1249 at T II 343a25 and a discourse quotation in T 1509 at T XXV 74a28 use the expression 多聞; an individually translated discourse, T 123 at T II 546c4, instead employs the expression 學問 to qualify the elders as "learned".

In the Mahāparinibbāna-sutta elders who are āgatāgama feature as possible sources of a teaching that requires further verification in order to determine whether it is indeed in accordance with the teachings already known and accepted.[n.5] DN 16 at DN II 125,6 (= AN 4.180 at AN II 169,18), where this forms part of the formulation of the four great standards, mahāpadesa. A Sanskrit fragment parallel has preserved only a reference to such elders in terms of their being upholders of the discourses and the Vinaya, to which the Tibetan version adds also the upholders of the summaries.[n.6] Fragment 360 folio 181 V6, Waldschmidt 1950: 24: sūtradharā vina[yadharā] and folio 182 R2 [vi]nayadh[arā]; Waldschmidt 1951: 243 (§24.16): mdo sde ’dzin pa ’dul ba ’dzin pa ma lta bu ’dzin pa (the transliteration style has been adjusted). A Dīrgha-āgama parallel extant in Chinese translation describes these elders as being learned and alternatively as being upholders of the discourses, the Vinaya (as a whole), or just its code of rules.[n.7] DĀ 2 at T I 17c15: 多聞 (T 7 at T I 195c23 and 196a3 has comparable references to being learned) and DĀ 2 at T I 17c27 and 18a11: 持法, 持律, 持律儀者. In a counterpart in the Ekottarika-āgama preserved in Chinese translation, the elders qualified as learned are capable in reciting the discourses and upholding the Dharma.[n.8] EĀ 28.5 at T II 652b17, 652c3, and 652c10: 誦經, 持法 … 博學多聞. In sum, none of the parallels to the Mahāparinibbāna-sutta has a counterpart to āgatāgama.

This does not mean, however, that the expression āgatāgama is only attested in Pāli discourses. Two discourses in the Madhyama-āgama have the same phrase as part of a description of a learned monk. In these two cases, the Pāli counterparts do not have the corresponding expression.[n.9] MĀ 1 at T I 421b19, 421b21, and 421b23: 阿含及所得 and MĀ 95 at T I 577b8, 577b11, and 577b14: 阿含及其所得; these references to being āgatāgama have no counterpart in the listing of otherwise comparable qualities in the parallels AN 7.64 at AN IV 114,3 and AN 10.53 at AN V 96,6.

In a discourse in the Aṅguttara-nikāya a monk, who has not yet memorized the discourses, is encouraged to "learn the āgama", āgamaṃ pariyāpuṇassū ti.[n.10] AN 10.44 at AN V 80,23; a recommendation also given in Vin II 249,16. Another Madhyama-āgama discourse somewhat similarly speaks of being capable at āgama recitation, which here is part of the description of a learned monk who might pride himself on his learning.[n.11] MĀ 85 at T I 561b27, 561b28, 561c1, and 561c2: 諳阿含.The corresponding passage in the parallel MN 113 at MN III 39,18+31 has two separate cases, where a monk could be "learned", bahussuto, or else an "upholder of the Vinaya", vinayadharo, none of which involves a reference to āgama.

In sum, the expression āgama in the singular serves as one of several terms to express that a learned monk was familiar with the orally transmitted discourses by the Buddha and his disciples.[n.12] Barua 1923: 359 comments that "in the Pāli discourses, ascribed to the Buddha himself, the expression Āgama is often met with, no doubt in the sense of a floating body of Buddhist literary traditions."

Although this type of usage is attested in discourses of more than one tradition, actual instances of its occurrence are as a rule not supported by their respective parallels.

A usage of the term āgama comparable to these instances in Pāli texts and Madhyama-āgama discourses can be seen in Asoka's Rock Edict XII, which combines a reference to being learned, bahusrutā, with kallāṇāgamā (Girnār version).[n.13] Bloch 1950: 123,29, Girnār: bahusrutā ca assu kallāṇāgamā ca, Kālsī: bahuṣṣuta cā kayyānāgā ca, Ṣāhbāzgarhī: bahuśruta ca kalaṇagama ca, Mānsehrā: bahuśruta ca kayaṇagama ca. Similarly, the Jain Vavahāra refers to a learned monk as babbhāgamaṃ.[n.14] Vavahāra 1.35, Schubring 1918: 15,4+6; cf. also Caillat 1965: 50.

---

The Four Āgamas

Accounts of the first saṅgīti in the Dharmaguptaka, Haimavata (?), Mahāsāṅghika, Mahīśāsaka, and Mūlasarvāstivāda Vinayas employ the plural form āgamas in their description of the compilation of the teachings that according to tradition took place under the leadership of Mahākassapa soon after the Buddha's demise.[n.15] The Dharmaguptaka Vinaya, T 1428 at T XXII 968b19, the Haimavata (?) *Vinayamātṛkā, T 1463 at T XXIV 820a23, the Mahāsāṅghika Vinaya, T 1425 at T XXII 491c16, and the Mahīśāsaka Vinaya, T 1421 at T XXII 191a24, agree on using the expression 阿含; the Mūlasarvāstivāda Vinaya, T 1451 at T XXIV 407b27, uses the expression 阿笈摩 with its counterpart in lung in D 6 da 314a7 or Q 1035 ne 297a4; on lung cf. the discussion in Eimer 1983: 23. Although the account of the first saṅgīti in the Sarvāstivāda Vinaya does not use either āgama or nikāya to refer to the textual collections, elsewhere the expression 阿 含 occurs repeatedly in this text in evident reference to textual collections that are to be memorized and recited; cf., e.g., T 1435 at T XXIII 453c17 for an occurrence, alongside a reference to the Vinaya, in relation to the undertaking of recitation. On the school affiliation of T 1463 cf. Anālayo 2011d: 270f note 11. The different Vinayas reporting this event agree that the Buddha's attendant Ānanda recited the discourses on this occasion.

They further report that the resultant textual material was divided into groups, presumably to facilitate oral transmission, by collecting long discourses and discourses of middle-length into corresponding āgamas; and short discourses were further separated into those assembled according to topic and those assembled according to a numerical principle (a discourse can discuss one or more items and can accordingly be allocated to the Ones, the Twos, the Threes etc.).

This division corresponds to the four Āgamas, which comprise a Dīrgha-āgama containing mostly long discourses, a Madhyama-āgama that assembles predominantly middle-length discourses, a Saṃyukta-āgama with discourses that tend to share a common topic, and an Ekottarika-āgama that follows an incremental numerical principle from Ones to Elevens (earlier perhaps only up to Tens).[n.16] For a survey of the four āgamas cf. Anālayo 2015a.

The accounts of the first saṅgīti in the Mahāsāṅghika and Mahīśāsaka Vinayas adopt the above sequence of dīrgha, madhyama, saṃyukta, and ekottarika,[n.17] T 1425 at T XXII 491c16 (another sequence can be found in a different context in T 1425 at T XXII 492c18, which lists the Vinaya, the Abhidharma, and then the Saṃyukta-āgama, the Ekottarika-āgama, the Madhyama-āgama, and the Dīrgha-āgama) and T 1421 at T XXII 191a24. which corresponds to the order of the corresponding four collections adopted in the Theravāda canon.[n.18] The account of the first saṅgīti in the Theravāda Vinaya, Vin II 287,16, only mentions the Brahmajāla (DN 1) and the Sāmaññaphala (DN 2) as the first two discourses recited and thus does not explicitly indicate the order of the four collections, except for positioning the Dīgha-nikāya in first place. The Dharmaguptaka Vinaya is closely similar, in so far as it adopts the sequence dīrgha, madhyama, ekottarika, and saṃyukta, thereby having only the last two in the opposite order.[n.19] T 1428 at T XXII 968b19. These four traditions thus agree on beginning with the long discourses, followed by those of middle-length and then the shorter discourses.

The Haimavata (?) Vinaya instead adopts the sequence ekottarika, madhyama, dīrgha, and saṃyukta,[n.20] T 1463 at T XXIV 820a23. and the Mūlasarvāstivāda Vinaya has the exact opposite by listing the four collections as saṃyukta, dīrgha, madhyama, and ekottarika.[n.21] T 1451 at T XXIV 407b27 and D 6 da 314a7 or Q 1035 ne 297a5. Similarly to the case of the Theravāda Vinaya, the Sarvāstivāda Vinaya does not give a full list, but only mentions the discourse recited first, which in its account is the first discourse spoken by the Buddha, parallel to the Dhammacakkappavattana-sutta (SN 56.11); cf. T 1435 at T XXIII 448b13. The prominent place given to the Buddha's first discourse here might in part be related to what appears to be a reworking of this discourse in the Mūlasarvāstivāda and Sarvāstivāda traditions. As a result of this apparent reworking, the three turnings become the essence of the first discourse, with the perplexing result that Kauṇḍinya (Pāli Koṇḍañña) attains stream-entry after hearing about these three turnings to be applied to the four truths, but only afterwards receives information on what these four truths actually mean; cf. in more detail Anālayo 2012a. These two versions thereby do not follow the pattern observed in the other versions of proceeding from long to short discourses, nor do they adopt the opposite pattern of moving from short to long discourses.

The central point conveyed by the account of the first saṅgīti in the different Vinayas is that the distinction into four āgamas was considered a creation by the reciting elders soon after the Buddha's demise, a shared division whose sequential order underwent some independent evolution in the different traditions. The discourse passages mentioned earlier that use the term āgama in the singular as a referent to what appears to be the whole body of discourses might then reflect a stage preceding this basic structural division into four collections, a time when the body of orally transmitted texts had not yet been systematically structured in the way reported in the accounts of the first saṅgīti.

The Theravāda Vinaya account of this first saṅgīti differs in so far as it does not use the expression āgama, but instead refers to the same type of collections of discourses as nikāyas.[n.22] Tournier 2014: 25 note 95 notes that "among the southern Mahāsāṅghika subschools, which transmitted a canon in Prakrit, there is epigraphical evidence that at least the Aparamahāvinaseliyas also called the divisions of their Sūtra-piṭaka nikāya." Another difference is that the Theravāda Vinaya speaks of five nikāyas.[n.23] Vin II 287,27 speaks of the recitation of the pañca nikāye. In addition to the four nikāyas that correspond to the four āgamas, this expression includes the Khuddaka-nikāya, a miscellany of texts also known in other traditions, where this fifth grouping is at times rather considered to be a piṭaka.[n.24] For the classic study on the topic cf. Lamotte 1956.

Not only do the Vinaya accounts of the first saṅgīti differ in their usage of āgama or nikāya, but also accounts of the first saṅgīti in the Samantapāsādika and its Chinese counterpart differ. The Pāli version's reference to the "four nikāyas" (as distinct from the fifth) has as its equivalent the "four āgamas" in its Chinese counterpart.[n.25] Sp I 16,14: cattāro nikāye and T 1462 at TXXIV 675b22: 四阿鋡, translated in Bapat and Hirakawa 1970: 9; a difference already noted by Lamotte 1958: 167. Heirman 2004: 385 comments that "when the translator adapted the text to the Chinese environment, he hereby translated all five nikāyas as āgama." The usage of the term āgama nevertheless continues in Pāli commentarial literature. Thus, e.g., Vism 442,30 defines knowledge of āgama to be mastery of the teachings of the Buddha, be it only the Chapter on Similes (the third chapter in the Majjhima-nikāya), and Mp II 189,17 equates the nikāyas with the āgamas and concludes that āgatāgama refers to mastery of one of these.

It is noteworthy that the Pāli Vinaya and its commentary give preference to the term nikāya, which is not employed in the Pāli discourses as a referent to scriptural collections. Instead, the Pāli discourses rather employ the expression āgama.

This is significant in so far as it gives the impression that expressions like āgatāgama or āgamaṃ pariyāpuṇassu, after having been introduced, were kept in place without being adjusted to the eventual predilection in the Theravāda tradition for the term nikāya.

The Nine Aṅgas

A recurrent reference in the Pāli discourses provides a list of nine items that eventually came to be known under the heading of being aṅgas, although the term aṅga itself is not used for these in the discourses. In what follows I list the nine items, translating them in a way that reflects my current understanding of the probable implications of each, without any pretence at presenting a definite solution to their individual significance:

• discourses [involving lists] (sutta),[n.26] According to Przyluski 1926: 341, the use of the expression sutta in the context of the aṅgas has the specific sense of an exposition that begins with an enumeration of a particular item, "un sūtra était un sermon commençant par un exposé numérique" (e.g., "there are four things … what are the four", etc.). Nyanaponika 1977: 13f explains that sutta in its Buddhist usage refers to a presentation of the Dharma that is internally connected by a thread, as it were, "eine zusammenhängende Lehrdarstellung … durch die sich ein gemeinsamer Faden hindurchzieht." The need to 'string together' material for recitation also emerges from a reference to the nine aṅgas in Vin III 8,7, according to which the teachings of former Buddhas who did not give much instruction in terms of the nine aṅgas were quickly lost, comparable to flowers not held together by a string; for a survey of publications relevant to the alternative explanation that derives sūtra from su + ukta, "well spoken", cf. Anālayo 2011b: 150 note 22.
• prose combined with verse (geyya),[n.27] Jayawickrama 1959: 12 comments that "geyya (from √gai gāyati, to sing), seems to represent the ākhyāna-type containing stanzas punctuated with narrative prose." According to Mayeda 1964: 24, geyya "is not, however, a simple juxtaposition of prose and verse. The prose section which comes first is repeated once again in the verse section which follows. This repetition of similar contents is the key point of geyya"; cf. also Burnouf 1844/1876: 47.
• explanatory expositions (veyyākaraṇa),[n.28] For a detailed discussion of this term cf. Anālayo 2008h.
• stanzas (gāthā),
• inspired utterances (udāna),
• quotes (itivuttaka),
• birth stories (jātaka),
• marvels (abbhutadhamma),
• answers to questions [between disciples] (vedalla).[n.29] Karashima 2015b: 136 explains that "the most original form … could have been *vedulla, a Middle Indic form corresponding to vaitulya (> vetulla > *vedulla), which might mean 'not' (vi) 'of the same kind' (tulya, MW, s.v.), i.e. 'unusual, irregular'. Scriptures consisting of repeated questions and answers, not always between the Buddha and another person, but rather between two disciples, might have been labelled as *vedulla, because they were 'unusual, irregular'. This form presumably changed in the Pāli tradition to vedalla."

Comparison with similar listings in the discourse parallels of other traditions shows that these usually take the form of an expanded listing of twelve.[n.30] Lamotte 1956: 263 note 2 explains that this twelvefold presentation prevails in the Āgamas, in the Chinese Vinayas (except for the Mahāsāṅghika Vinaya), in the main treatises of the Sarvāstivāda, Sautrāntika, Vaibhāṣika, and Yogācāra traditions, and in most Mahāyāna sūtras. For studies of listings of twelve cf., e.g., Hirakawa 1963: 61–65, Lamotte 1980: 2281–2305, and Nattier 2004. This expanded listing in turn suggests an adjustment to the sequence of the nine aṅgas as reflected in Pāli sources. On adopting this adjustment, marvels (abbhutadhamma) come to stand in last position, preceded by answers to questions (vedalla).[n.31] Cf. the survey provided in table form in Mayeda 1964 and Skilling 2013b: 157.

For abbhutadhamma to be the last member of the list would conform to the principle of waxing syllables, which I study in more detail in the next chapter. Suffice it for the present context to say that, according to this principle, several terms in a list tend to be arranged in such a way that words with fewer syllables are followed by words with an equal or a higher number of syllables.[n.32] Cf. below p. 474ff.

This in turn implies that the assumed shift of vedalla to last position would have taken place only at a time when the demands of facilitating oral transmission no longer made themselves felt as acutely as earlier. The possible demotion of vedalla to the last place in Pāli listings of the nine aṅgas might be related to the negative associations that vedallakathā carries in a discourse in the Aṅguttara-nikāya as leading to the future corruption of the Dhamma and Vinaya,[n.33] AN 5.79 at AN III 107,4; discussed in Skilling 2013b: 87. and to the even more negative connotations of the term vetulla in Theravāda chronicles.[n.34] Skilling 2013b: 88 explains that in the Ceylonese chronicles and later texts "Vetulla in Vetullavāda is used only negatively for unacceptable ideas or theories, in connection with doctrinal controversies that arose from the third to the second centuries BCE onwards."

Regarding the principle of waxing syllables, items in longer lists can be arranged into subgroups that are often based on some thematic or formal connection, and these subgroups then internally follow the principle of waxing syllables. From the viewpoint of oral transmission,[n.35] For a more detailed discussion cf. the next chapter, below p. 473ff, and for other aspects of the early Buddhist oral tradition above p. 15ff. a conveniently structured pattern results in this way that provides a rhythm for catching one's breath during recitation and at the same time facilitates memorization.[n.36] Cf. the discussion below p. 476f. Applying this principle to the adjusted listing of the aṅgas, the resulting subsections together with their syllable count would be as follows:

1. sutta, geyya, veyyākaraṇa, 2+2+5,

2. gāthā, udāna, itivuttaka, 2+3+5,

3. jātaka, vedalla, abbhutadhamma, 3+3+5.

Following the model of other subgroups in such long listings, it could be expected that the resultant three subsections have some thematic or formal connection. In view of the uncertainty that surrounds the exact implications of each of these aṅgas,[n.37] On the significance of the nine aṅgas cf., e.g., Jayawickrama 1959, Kalupahana 1965, and von Hinüber 1994a; for a summary cf. Anālayo 2011b: 150f. however, thematic continuity is less easily determined. Nevertheless, perhaps the following themes could be taken as approximate summaries of the import of each of the three subsections:

1. basic modes of exposition: "discourses, prose and verse, expositions",

2. shorter textual pieces: "stanzas, inspired utterances, quotes",

3. others: "birth stories, marvels, answers to questions".

---

The Function of the Nine Aṅgas

My above presentation of the nine aṅgas is to some extent based on the assumption that the original import of the aṅgas is to designate textual types instead of collections of texts.[n.38] According to Dutt 1957: 89, "the list [of aṅgas]… rests on an analysis of different forms of composition found in the canon." Jayawickrama 1959: 11 states that "it is a mere description of the literary types". Kalupahana 1965: 616 similarly indicates that "this classification … does not refer to nine different groups of literature, but to nine types of composition." Nyanatiloka 1952/1988: 193 explains that the aṅga system "is a classification according to literary styles, and not according to given texts". Lamotte 1980: 2282 clarifies that the aṅgas are not literary genres, but types of composition for forming texts, "ces Aṅga ne sont pas des genres littéraires, mais simplement des types de composition concernant la forme des textes." Norman 1983: 16 points out that "despite the fact that books called Jātaka, Udāna and Itivuttaka actually exist in Pāli, it is probable that the list of nine aṅgas did not originally refer to specific works in the canon." It is in fact hard to see how this list of nine, or even the more evolved list of twelve, could have functioned as an organizational principle for allocating discourses into textual collections. The problem here is that dividing the texts according to individual aṅgas would not yield groupings of material suitable for division of labour among reciters, which requires portions of at least roughly comparable size that provide the respective reciters with a representative selection of teachings.

An aṅga like marvels, for example, is a fairly rare occurrence in the early discourses and would hardly have sufficed for creating a corresponding textual collection. Moreover, one might wonder what would happen if some reciters were to memorize a textual collection with only marvels. Even the division into four Āgamas or Nikāyas led to differences of opinion and understanding among their respective reciters.[n.39] Cf., e.g., Adikaram 1946/1994: 27–32, Goonesekera 1968: 689, Dutt 1978: 42, Mori 1990: 127, and Endo 2003a and 2003b. A distribution of the texts to be memorized over different reciters according to the aṅga system would have inevitably resulted in rather lopsided understandings and quickly become a source of conflict.

The Pāli commentaries explain aṅgas like udāna, itivuttaka, and jātaka as corresponding to the respective collections in the Pāli canon.[n.40] Cf., e.g., Sp I 28,18. Yet among the discourses in the four Nikāyas (and thus apart from the canonical Udāna, Itivuttaka, and Jātaka collections) udānas explicitly identified as such occur repeatedly,[n.41] For a survey of occurrences cf. Anālayo 2008e: 381f note 1. making it clear that this textual type is not confined to the Udāna collection. Similarly, quotations marked with the help of the quotative iti and introduced as something said, vutta, are not confined to the collection of discourses known as the Itivuttaka. Instead, quotations explicitly marked in this way are a recurrent feature in the discourses in the four Nikāyas.[n.42] One of two examples to illustrate this type of usage can be seen when a statement to be explained is introduced with the construction iti kho pan' etaṃ vuttaṃ, kiñc' etaṃ paṭicca vuttaṃ, the explanation of which then concludes with idam etaṃ paṭicca vuttaṃ, found, e.g., in MN 54 at MN I 361,1. Another example can be seen when quotes are introduced by stating vuttaṃ kho pan' etaṃ bhagavatā, and are concluded with iti, a usage even employed in a passage that shows the Buddha quoting himself, cf., e.g., MN 3 at MN I 13,11. On the use of iti in commentarial literature cf. the study by Kieffer-Pülz 2014. In the case of jātakas found among the early discourses, these actually fall into a distinct class by being all in prose, whereas the Jātaka collection is entirely in verse and the stories are only found in its commentary.[n.43] Cf. von Hinüber 1998: 187. Clearly the aṅga of jātakas could not originally have referred to the transmitted Jātaka collection.[n.44] Jayawickrama 1959: 13 comments that "there is no justification for equating the Aṅga called Jātaka with the extant Jātaka collection … firstly, the stories themselves have no Canonical status, which is reserved for … the stanzas only. Secondly, there is no reason why Jātakas of Canonical antiquity such as those incorporated in other suttantas … should be excluded." This makes it safe to conclude that the commentarial explanation is better not taken as reflecting the original significance of these aṅgas.

---

Shorter Lists of Aṅgas

In what follows I turn to theories that assume the list of nine or twelve aṅgas to have evolved from a supposedly earlier stage with a shorter listing of aṅgas. Such theories do not avoid the problem that the aṅgas are not easily amenable to a system of textual divisions. Adopting such theories implies that the supposedly earlier listings eventually evolved into the listings of nine and twelve. This in turn would entail that at a very early stage a presumably workable scheme subsequently lost its presumed function and usability by being expanded to nine, and in spite of that the list subsequently still continued to expand to twelve. This seems hardly convincing.

Such shorter listings of aṅgas are found in two Pāli discourses. One occurs among the Fives of the Aṅguttara-nikāya. In the passage in question, a brahmin proclaims that one will no longer be interested in the teachings of others once one has heard the Buddha's teaching in the form of sutta, geyya, veyyākaraṇa, and abbhutadhamma.[n.45] AN 5.194 at AN III 237,17+23. Bodhi 2012: 1744 note 1196 comments about the brahmin protagonist, "for some reason, he cites only four of the nine divisions of the Dhamma. Perhaps it was only these with which he was familiar" (this is one of two possible explanations proposed by him). I take this to imply that, since in this instance the speaker is neither the Buddha nor one of his well-known disciples, the brahmin protagonist of this discourse could perhaps on purpose have been depicted as not being fully acquainted with the whole set of nine aṅgas. In line with a general tendency in the discourses to present brahmins as particularly concerned with the marvellous qualities of the Buddha, such as his physical marks, in this instance he might presumably be shown to remember only abbhutadhamma out of the six aṅgas mentioned usually after sutta, geyya, and veyyākaraṇa. This reference may well reflect the presumably original sequence of the listing of the nine aṅgas, attested in other traditions, when abbhutadhamma was still occupying the last place. The fact that the full list is not given would then simply be a case of abbreviation, following a standard pattern where the first three and the last member of a list are given in full.[n.46] In reply to the hypothesis by von Hinüber 1994a (repeated in 2015c: 38f) that this passage points to an early stage in the evolution of the aṅgas, Choong 2010: 60 argues that it is "likely that the unique Pāli list of just four aṅgas … is, rather, an abbreviation of the entire set of nine aṅgas in their original sequence; that is 'sutta, geyya, veyyākaraṇa, … abbhuta-dhamma.'" The idea that these four were an early division of the textual material is also not easily reconciled with the problem I mentioned above, in that a reciter who specializes in marvels would have relatively little material to learn and would moreover stand good chances of acquiring an unbalanced understanding of the teachings.

Only sutta, geyya, and veyyākaraṇa are mentioned in the other instance to be discussed, which occurs in the Mahāsuññata-sutta.[n.47] MN 122 at MN III 115,18. For a critical reply to the suggestion by Sujato 2005: 62 that Sanskrit fragments of the Mahāparinirvāṇa-sūtra support the notion of a special emphasis being accorded to the first three aṅgas cf. Anālayo 2011b: 698 note 69.

The same three are found in a parallel preserved in the Madhyama-āgama extant in Chinese translation, whereas a Tibetan parallel has the full set of twelve.[n.48] MĀ 191 at T I 739c4 and Skilling 1994: 242,13.

According to the context of this reference in the Mahāsuññata-sutta and its parallels, one should not follow the Buddha for the sake of sutta, geyya, and veyyākaraṇa (or the twelve aṅgas). Instead one should follow him for the sake of beneficial types of talks. The Mahāsuññata-sutta and its parallels agree that such beneficial types of talk are on the topics of morality, concentration, wisdom, liberation, and knowledge and vision of liberation.[n.49] MN 122 at MN III 115,25, MĀ 191 at T I 739c8, and Skilling 1994: 244,13.

In view of this narrative context, the earlier reference to sutta, geyya, and veyyākaraṇa (or the twelve aṅgas) could not be reflecting a scheme that stood representative of the entirety of the textual collections that comprise the teachings given by the Buddha and his disciples.[n.50] Pace Choong 2000: 9f and Sujato 2005: 61f. If all the textual collections comprising the teachings given by the Buddha and his disciples are set apart as insufficient grounds for following him, there would be nothing left for the sake of following him. His teachings on morality, concentration, wisdom, liberation, and knowledge and vision of liberation are of course contained in precisely those textual collections that would all have been already mentioned earlier if the reference to sutta, geyya, and veyyākaraṇa did indeed stand for some scheme of division of the whole corpus of the textual collections.

It seems therefore reasonable to assume that this passage was not originally about a listing of textual divisions that represented the whole of the canonical texts, however short or expanded it may have been. Instead, the original point of what now is a reference to sutta, geyya, and veyyākaraṇa must have been more limited in scope.

Perhaps at an early stage the present passage was just about following the teacher merely for the sake of ever more explanations, veyyākaraṇassa hetu, instead of putting the teachings already received into practice.[n.51] This would correspond to the understanding reflected in the Pāli commentary on the present passage, Ps IV 164,9, translated Ñāṇamoli 1982: 30, which explains that, even though to acquire much learning has been compared by the Buddha to a soldier acquiring weaponry, the learning acquired does not function as weaponry if it is not put into practice. During oral transmission such a reference could have led the reciters to supplement the first two aṅgas, something happening at a stage early enough in the transmission of the discourses to affect both the Majjhima-nikāya and the Madhyama-āgama versions. In continuity of the same tendency, such a reference could then have led to further supplementation, as evidenced in the Tibetan version's twelve aṅgas.

Although this is of course hypothetical, it would yield a meaningful reading in the present context. The point would have been to contrast following the teacher for ever more explanations of the teachings, a quest not necessarily related to the questioner's personal progress on the path, to following him just for the sake of those teachings that directly help one develop morality, concentration, wisdom, liberation, and knowledge and vision of liberation.

Be that as it may, on taking into account the narrative context it does not seem convincing to consider the reference to three aṅgas in the Mahāsuññata-sutta and its Madhyama-āgama parallel as originally reflecting a scheme that organizes the texts transmitted orally by the Buddha's disciples. In other words, the reference to three aṅgas does not appear to be reflecting an early stage in the evolution of the aṅgas, instead of which it is more probable that this reference is simply the result of a textual corruption.

According to another theory proposed by Japanese scholars, a listing of five aṅgas rather formed the starting point for the nine or twelve aṅgas.[n.52] Mayeda 1964: 26 and 34 and Nakamura 1980/1999: 28. In the *Karmavibhaṅgopadeśa a listing of six aṅgas can be found.[n.53] Another reference to three aṅgas can be found in Nett 78,9, which here comprises sutta, veyyākaraṇa, and gāthā. Closer inspection shows this to be the result of a textual corruption, as the same *Karmavibhaṅgopadeśa continues right away by summing up its presentation as being about the nine aṅgas.[n.54] Lévi 1932a: 161,8: sūtraṃ geyaṃ vyākaraṇam itivṛttam gāthodānam, evaṃ (Kudo 2012: 106: evan) navāṅgaśāsanaṃ. Variations can also be found in the Jain tradition, where the Viyāhapaṇṇati has a reference to eleven aṅgas only,[n.55] Lalwani 1973b: 177,14. differing from the standard count of twelve that forms the foundation for the Śvetāmbara canon.[n.56] For a survey of the twelve aṅgas according to the Śvetāmbara Jain tradition cf. Dundas 1992: 64f.

In sum, it seems fair to conclude that references in Pāli discourses and elsewhere that in their present state involve fewer than nine aṅgas are probably best seen as the results of errors or abbreviation during oral transmission, instead of being evidence of a supposedly early stage in the evolution of the nine and twelve aṅgas.[n.57] In relation to the hypothesis by von Hinüber 1994a that the listing of four aṅgas reflects an early attempt at organizing the texts, Klaus 2010: 518 points out that such hypotheses are not supported by the texts, which do not present the aṅgas as an attempt at ordering the texts, but rather as attempts at classification or just enumeration, "mir kommt es auf die Feststellung an, daß Vermutungen in diese Richtung sich nicht an die Texte anknüpfen lassen. Die Texte präsentieren uns die verschiedenen Aṅga-Listen nicht als Versuche, einen wie auch immer gearteten Gesamtbestand an Texten zu ordnen, sondern als Versuche, die verschiedenen Arten von Dhamma-Texten zu klassifizieren oder auch nur aufzuzählen." Cousins 2013a: 105 concludes that "short versions are sometimes interpreted as earlier lists of 'Aṅgas', but that seems quite anachronistic to me." Skilling 2017: 293 note 55 comments on my discussion in the article that forms the basis for this chapter: "I am in accord with Anālayo's main point, which is that the Aṅgas are not actual collections of texts."

In what follows I survey Pāli discourse references to the nine aṅgas within their narrative and doctrinal contexts, and in comparison with their parallels, wherever extant. Such contextual considerations can help to corroborate or else counter the significance of the nine aṅgas suggested above, in so far as the narrative context can indicate whether listings of these nine were meant to refer to textual divisions or textual types.

A discourse among the Fours of the Aṅguttara-nikāya mentions the nine aṅgas as part of an exposition of the benefits of memorization.[n.58] AN 4.191 at AN II 185,7. In this context the nine aṅgas stand representative of what one would learn by heart. The same sense recurs in a discourse among the Fives, where not learning the teachings in the form of the nine aṅgas is a factor that leads to the decline of the Dharma.[n.59] AN 5.155 at AN III 177,6.

A discourse among the Sevens of the Aṅguttara-nikāya employs the nine aṅgas in a definition of knowledge of the Dharma,[n.60] AN 7.64 at AN IV 113,13; the same contrast recurs in Vibh 294,22 in terms of dhammapaṭisambhidā and atthapaṭisambhidā. distinct from knowledge of the meaning. The same distinction is drawn in parallels to this Aṅguttara-nikāya discourse found in the Madhyama-āgama, in the Ekottarika-āgama, and in a discourse translated individually into Chinese (although in terms of the twelve aṅgas instead).[n.61] MĀ 1 at T I 421a17, T 27 at T I 810a11, and EĀ 39.1 at T II 728c3. A Sanskrit fragment parallel has preserved part of the listing of aṅgas; cf. SHT III 878 R4, Waldschmidt et al. 1971: 127. According to this distinction, mere memorization does not suffice for true knowledge, a theme that continues with other references to the aṅgas.

Such a contrast between mere learning and true understanding becomes particularly evident in the Alagaddūpama-sutta, which features the monk Ariṭṭha obstinately holding on to a mistaken view. His obstinacy features in the different Vinayas as the occasion for a pācittiya regulation against such behaviour,[n.62] The Dharmaguptaka Vinaya, T 1428 at T XXII 682a9, the Mahāsāṅghika Vinaya, T 1425 at T XXII 367a3, the Mahīśāsaka Vinaya, T 1421 at T XXII 56c12, the Mūlasarvāstivāda Vinaya, T 1442 at T XXIII 840b21, with its Sanskrit and Tibetan counterparts in Yamagiwa 2001: 86,7 and 87,9, the Sarvāstivāda Vinaya, T 1435 at T XXIII 106a3, and the Theravāda Vinaya, Vin IV 133,32. giving the impression that his stubbornness was considered a serious problem. According to the Pāli commentary on the Alagaddūpama-sutta, Ariṭṭha had been well learned in the Dharma.[n.63] Ps II 103,1. This suggests his behaviour to be in part due to the arrogant belief that by mere learning he had actually understood the teachings.

It is against this narrative background that the reference to the nine aṅgas falls into place as part of the famous simile of the snake.[n.64] MN 22 at MN I 133,24. According to this simile, someone trying to catch a snake would incur harm if he were to grasp it by its tail. This predicament illustrates the situation of someone who learns the teachings in the form of the nine aṅgas without examining their meaning and without cultivating wisdom, just for the sake of being able to debate with others. Conversely, one who learns the teachings in the form of the nine aṅgas not for the sake of debating with others, but rather to understand them and cultivate wisdom, is comparable to someone who catches the snake by its neck, making sure that it cannot bite.

In this context the nine aṅgas stand representative of a way of learning the teachings which could have wholesome or unwholesome consequences. The same holds for the parallel versions to the Alagaddūpama-sutta preserved in the Madhyama-āgama and the Ekottarika-āgama (again with the difference that they employ the listing of twelve aṅgas instead).[n.65] MĀ 200 at T I 764a14 and EĀ 50.8 at T II 813a16.

Similar implications are relevant for a range of other occurrences of the nine aṅgas. A discourse among the Fours of the Aṅguttara-nikāya distinguishes those who have learned the nine aṅgas into those who do not understand the teachings and those who do.[n.66] AN 4.6 at AN II 7,2; a distinction that recurs in Pp 62,33. Another two discourses among the Fours restate the same contrast in terms of those learned in the nine aṅgas who either have or lack an insight into the four noble truths.[n.67] AN 4.102 at AN II 103,8 and AN 4.107 at AN II 108,3; the presentation in AN 4.107 recurs in Pp 43,29. The first of these two discourses has an Ekottarika-āgama parallel which, instead of mentioning the four noble truths, distinguishes those who learn the aṅgas into those who do and those who do not teach others.[n.68] EĀ 25.10 at T II 635a10, which illustrates mere learning of the twelve aṅgas with the same simile of a cloud that thunders but does not rain, found also in AN 4.102. Another parallel, EĀ2 10 at T II 877b10, also employs the same simile, but does not mention the aṅgas. Here the one who is like a cloud that thunders but does not rain learns the discourses but does not understand the Dharma himself.

Two consecutive discourses among the Fives express the same basic contrast by way of defining one who is a dhammavihāri. Mere learning of the teachings in the form of the nine aṅgas is not sufficient for being a dhammavihāri, as such a one might still neglect seclusion and the cultivation of tranquillity of the mind or else not understand with wisdom the meaning of what has been learned.[n.69] AN 5.73 at AN III 86,25 and AN 5.74 at AN III 88,7.

Another discourse in the Aṅguttara-nikāya highlights that, even though the Buddha had taught much in the form of the nine aṅgas, for being "learned" it suffices to have penetrated the meaning of a four-line stanza and practised accordingly.[n.70] AN 4.186 at AN II 178,12. Parallels to this discourse in the Madhyama-āgama, in an individually translated discourse, and as part of an Udāna collection preserved in Chinese translation make the same point.[n.71] MĀ 172 at T I 709b6, T 82 at T I 901c19, and T 212 at T IV 643b25 (each lists twelve aṅgas). These references highlight that one need not be familiar with a variety of teachings in the form of the nine (or twelve) aṅgas, instead of which thorough understanding of a single specimen from one of these aṅgas, a gāthā, can make one become truly learned.

This conveys the impression that the contrast to the single stanza mentioned is not the entirety of texts in nine (or twelve) textual collections, but rather the variety of teachings the Buddha had given as exemplified by these aṅgas. Out of these, a single stanza can fulfil the purpose of becoming truly learned.

One more occurrence in a discourse in the Aṅguttara-nikāya brings me back to the topic taken up at the outset of this chapter, namely the term āgama. This discourse lists several praiseworthy qualities, one of which is having learned the teachings in the form of the nine aṅgas. Another quality is to spend the rainy season with elders who are āgatāgama, who have "obtained the āgama", and to use this opportunity to get the teachings clarified by asking them questions.

In this discourse the aṅgas occur side-by-side with a reference to being āgatāgama, each expression applied to different persons. This gives the impression that the two terms were not seen as being in conflict with each other.

In line with the conclusion arrived at earlier, it seems safe to infere that the aṅgas do not function as a supposedly early system of apportioning the discourses which eventually fell into disuse, to be replaced by the system of the four āgamas or nikāyas. Instead, all of the passages surveyed above are well compatible with an understanding of the listing of aṅgas as referring to textual types.[n.72] Cousins 2013a: 106 sums up that "there is no indication anywhere that any of this has anything to do with an arrangement of the canonical literature in some kind of earlier recension."

Besides this corroboration, what emerges from the above passages is that the nine aṅgas reflect the variety of the teachings that can be learned and as such function as a convenient reference to a considerable degree of learning, which in the ancient oral context of course required memorization. Many of these references draw attention to the shortcomings of mere rote learning without penetrating the meaning of the teachings. A particularly stark example for such drawbacks is the case of Ariṭṭha in the Alagaddūpama-sutta and its parallels, where mere learning of the teaching in the form of the nine aṅgas compares to catching a snake in such an unskilful way that one will get bitten. In short, learning the different types of presentations of the teachings in the form of the nine aṅgas does not necessarily imply that one has understood the meaning of what the Buddha taught, which instead can take place even with a single stanza that is well understood.

---

Conclusion

The overall impression that suggests itself from the evidence surveyed in this chapter is that an at first somewhat undifferentiated body of discourses, the āgama (singular), developed into separate āgamas (or nikāyas), a development which as far as I am able to ascertain did not involve an intermediate period during which the aṅgas fulfilled the purpose of forming textual collections. Instead, the aṅgas appear to stand for textual types, for kinds of compositions, and their main function, as reflected in the early texts, is to highlight the importance of a penetrative understanding of the meaning of the teachings over mere rote learning of its different manifestations. "One is not wise, Because one speaks much. Being peaceful, free from anger and fear, One is called wise."[n.73] Dhp 258.