Introduction

In this chapter I study some aspects of the early Buddhist oral tradition as exemplified by the different versions of the Brahmajāla. I begin with the significance of the introductory phrase "thus have I heard". Then I examine the opening narration and the exposition on morality in the first part of the discourse. In the final part of the chapter, I attempt a general assessment of the nature of the early Buddhist oral tradition.

---

The Introductory Phrase of the Brahmajāla

The standard opening of a discourse introduces what follows with the phrase: "thus have I heard". The relationship of this in itself innocuous marker of orality to what follows has led to considerable discussion among scholars. [n.1] Cf., e.g., Burnouf 1925: 286, Staël-Holstein 1933: iv and xiii notes 7 and 8, Brough 1950, Samtani 1964, von Hinüber 1968: 84–87, Samtani 1971: 68f, Wayman and Wayman 1974, Kajiyama 1977, Schopen 1978: 162–164, Silk 1989, Harrison 1990: 5 note 3, Galloway 1991, Tatz 1993/1997, Vetter 1993: 65 note 48, Bongard-Levin 1996: 90 note 1, Galloway 1997, Tola and Dragonetti 1999, Klaus 2007, Sander 2007: 174–176, and Nattier 2014. Several of these contributions examine relevant information in śāstra and commentarial literature, which in the context of my present survey I am not able to cover. My ignorance of Japanese has also prevented me from benefitting from research published in that language on topics taken up in this or subsequent chapters. On "thus have I heard" as a marker of orality cf. also below p. 473f. The point at stake is whether the subsequent phrase "at one time" should be considered as qualifying the preceding phrase "thus have I heard", or rather what follows in the text.[n.2] The first to have broached the subject appears to be Burnouf 1925: 286, who reasoned that "si on supprime la formule complète, y compris les mots êkasmin samayê, on a le commencement d'un Avadâna, classe de livres qui … ne diffère guère de celle des Sûtras que par l'absence de la formule, 'Il a été ainsi entendu par moi' … ne peut-on pas dire que les mots êkasmin samayê ne se trouvant pas en tête des Avadânas, appartiennent forcément à la formule êvam mayâ crutam?" The same point has then been made again by Staël-Holstein 1933: xiii note 7, pointing out that in several avadānas "evaṃ mayā çrutam is missing, and wherever evaṃ mayā çrutam is missing ekasmin samaye is also absent … this seems to indicate that the words evaṃ mayā çrutam ekasmin samaye constitute one phrase." Staël-Holstein 1933: iv also notes that "in all Tibetan and Mongolian preambles known to me a punctuation mark separates the equivalent of ekasmin samaye from the following words. The question as to whether ekasmin samaye belongs to çrutam or to viharati is discussed in a number of Buddhist commentaries attributed to Indians, and most of them seem to regard ekasmin samaye as belonging to the preceding words evaṃ mayā çrutam." A full paper dedicated to a detailed discussion of the same topic then concludes with Brough 1950: 426 stating that "from every point of view, then, it seems to me that the punctuation as preserved in the Tibetan is to be preferred." That is, according to him the specification "at one time" should qualify the phrase "thus have I heard".

This discussion is of relevance to the Brahmajāla, as a variant in some editions of a full discourse quotation of the Brahmajāla in the Abhidharmakośopāyikā-ṭīkā presents a slight but significant difference in formulation. The Narthang and Peking editions of this discourse's introductory phrase "thus have I heard" read ’di skad bdag gis thos pa’i dus gcig na.[n.3] N (mngo) tu 151a6 and Q 5595 tu 162b7: ’di skad bdag gis thos pa’i dus gcig na, whereas C (mngon pa) ju 142a4, D 4094 ju 141b5, and G 3598 tu 211a6 read: ’di skad bdag gis thos pa dus gcig na (due to the significance of this difference, in this case I consulted five editions, whereas elsewhere I only cover the Derge and Peking editions); this is Up 3050 in Honjō 1984, who also identifies partial citations of the Brahmajāla as follows: Up 2035, Up 2036, Up 2045, Up 3007, and Up 5005. The reading in C, D, and G is also found in the discourse version preserved in Tibetan, Weller 1934: 6,4 (§1): ’di skad bdag gis thos pa dus gcig na. On the first part of the Tibetan expression in general cf. Hahn 2006: 237f; other occurrences of the formulation thos pa’i have been noted by, e.g., Silk 1989: 160, Harrison 1990: 5 note 3, and Klaus 2007: 314 note 17; cf. also Galloway 1991: 92. The corresponding expressions in the other discourse versions of the Brahmajāla (all given on purpose without punctuation) are as follows: DN 1 at DN I 1,1: evam (Be and Ce: evaṃ) me sutaṃ ekaṃ samayaṃ, DĀ 21 at T I 88b13: 如是我聞一時, and T 21 at T I 264a23: 聞如是一時; on the standard Chinese renderings of this expression cf. also Qingzhi 2010: 494 and Nattier 2014; on some formal aspects of the Pāli formulation cf. Allon 1997b: 195 and 246f. Taking the addition of the particle -’i (kyi) to thos pa here to be indicative of the genitive,[n.4] According to Galloway 1991: 92, "the kyi is verbal and indicates a general connection between thos pas (śrutam) and the whole following sentence, whose verb is viharati, and not with dus gcig na alone." Klaus 2007: 314f note 17 in reply argues that this would only work if kyi were to follow directly after the bare verb thos, but not for the present case, where it follows thos pa: "diese Auffassung scheitert jedoch daran, daß die Partikel nicht auf den bloßen Verbalstamm thos, sondern auf das Verbalnomen thos pa folgt." Moreover, even if kyi were to follow thos directly, it would still have to be considered as a genitive particle, "selbst wenn die Variante nicht thos pas’i, sondern thos kyi lautete, wäre kyi … trotzdem als Genitivpartikel zu betrachten", with reference given to the discussion of the genitive particle in Hahn 1996: 133f. the discourse in these two editions would begin by qualifying the hearing referred to with the phrase "thus have I heard" to have taken place "at one time" (or "at a certain time"). This raises the question whether the specification "at one time" in the introductory phrase of the Brahmajāla, and by implication in other early discourses, should indeed be understood in this way.

When examined from the viewpoint of Pāli grammar, the assumption that "at one time", ekaṃ samayaṃ, qualifies the expression "thus have I heard", evaṃ me sutaṃ, is problematic, since one would not expect an accusative of time to stand after the verb.[n.5] This point has already been clarified by von Hinüber 1968: 86 (§72) who, after noting that, e.g., in Vin II 296,26 ekaṃ … samayaṃ occurs on its own, without being preceded by evaṃ me sutaṃ, points out that it would conflict with Pāli word order if the accusative of time were to come after the verb, "zudem würde die Stellung des acc. der Zeit nach dem Verb gegen die Worstellungsregeln des Pāli verstoßen … alles sprich also dafür, im Pāli in evaṃ me sutaṃ einen abgeschlossenen Satz zu sehen." On the accusative of time cf. also, e.g., Duroiselle 1906/1997: 156, Wijesekera 1936/1993: 56–58, and Warder 1963/ 1991: 18. Brough 1950: 423 sees an instance corroborating the possibility of positioning the accusative of time after the verb in MN 21 at MN I 124,7: ārādhayiṃsu vata me, bhikkhave, bhikkhū ekaṃ samayaṃ cittaṃ. In reply, Klaus 2007: 311 note 8 comments that due to the initial position of the verb this is not a conclusive precedent, "ist wegen der Initialstellung des Prädikats in diesem Zusammenhang nicht beweiskräftig". This would also hold for instances where the discourse itself is not attributed to the Buddha.[n.6] DN 34 at DN III 272,1 begins with the standard introduction, before referring to the location where the Buddha was dwelling, and then at DN III 292,6 concludes by indicating that the discourse was spoken by Sāriputta. Schopen 1978: 164 comments that "this makes it clear that what the speaker heard, and was reporting, was not that 'at one time the Blessed One dwelt at Campā, etc.', but what Sarīputta (sic) said when he and the Buddha were there. In this case – and by extension all other cases – ekaṃ samayaṃ could be attached to either what precedes or what follows it without affecting the meaning." Yet, even when the discourse was believed to have been spoken by Sāriputta in the presence of the Buddha, the introductory indication that at the time of the discourse's delivery the Buddha was living at Campā has the same function. Thus, to take ekaṃ samayaṃ as qualifying evaṃ me sutaṃ in the case of DN 34 would face the same problems as for DN 1 or any other discourse attributed to the Buddha or to a disciple. In fact the phrase "at one time", ekaṃ samayaṃ, occurs in some Pāli discourses without being preceded by "thus have I heard". In such contexts it must be referring to the time of the event reported.[n.7] Klaus 2007: 312 notes that "im Sutta- und Vinayapiṭaka der Theravādin sind zahlreiche Stellen enthalten, an denen ein Bericht über eine Begebenheit in der jüngeren oder ferneren Vergangenheit mit einem Satz eröffnet wird, der mit ekaṃ samayaṃ beginnt und mit viharati oder viharāmi endet"; for which he provides the examples MN 5 at MN I 31,27 and AN 3.90 at AN I 237,18. In the case of MN 5, the parallel MĀ 87 at T I 569b15 has the corresponding expression 一時, whereas two other parallels instead speak of the reported event being "in the past"; cf. T 49 at T I 841c27: 昔時 and EĀ 25.6 at T II 633c17: 昔. In the case of AN 3.90, the corresponding phrase in the parallel SĀ 830 at T II 213a20 differs and thus does not give any temporal specification. Another example that also has a Tibetan parallel would be MN 121 at MN III 104,6: ekamidaṃ, bhante, samayaṃ bhagavā sakkesu viharati, with counterparts in MĀ 190 at T I 737a2: 世尊一時遊行釋中, and in the Tibetan parallel in Skilling 1994: 148,1: dus gcig na bcom ldan ’das shā kya rnams kyi nang na shā kya rnams kyi grong rdal grong khyer zhes bgyi ba na bzhugs pa. The same suggests itself also from a standard pattern in the Pāli discourses where after the phrase "thus have I heard", evaṃ me sutaṃ, and "at one time", ekaṃ samayaṃ, the next sentence begins with "at that time", tena kho pana samayena. It seems safe to assume that in such cases both occurrences of samaya refer to the time of the event described in the discourse.[n.8] As pointed out by von Hinüber 1968: 144 (§134): "ekaṃ samayaṃ und tena samayena meinen denselben Zeitabschnitt, der verschieden gesehen wird. Der acc. bezeichnet den gesamten Zeitabschnitt, der instr. bestimmt die Zeit einer Handlung, die mit dem Verlauf dieser Zeit eintritt." The need to consider the significance of ekaṃ samayaṃ in relation to the subsequent tena kho pana samayena has also been highlighted by Tola and Dragonetti 1999: 55.

Moreover, in some Pāli discourses the expression "thus have I heard" stands for hearsay in contradistinction to what one has personally experienced.[n.9] This has already been pointed out by Klaus 2007: 318, who mentions as examples DN 5 at DN I 143,16 and MN 127 at MN III 152,7, where the fact that the speaker does not use the expression "thus have I heard" leads to the conclusion that he must have personally witnessed it. The parallels to DN 5 convey a similar sense, but without using the expression "thus have I heard"; cf. fragment 408r2, von Criegern 2002: 35, and DĀ 23 at T I 100b21. In the case of MN 127, the parallel MĀ 79 at T I 551c7 proceeds differently. This makes it improbable that the function of this phrase at the outset of a discourse is to designate that an eyewitness "at one time heard thus". Instead, the phrase "thus have I heard" simply has the function of highlighting that the discourse is something that has been heard thus, evaṃ, in this form, instead of being the product of one's own creative improvisation.

Several Vinayas report that the expression "thus have I heard" was already used by Ānanda at the time of the first saṅgīti that according to tradition took place soon after the Buddha's demise,[n.10] According to the Mahāsāṅghika Vinaya, T 1425 at T XXII 491c2, the Mūlasarvāstivāda Vinaya, T 1451 at T XXIV 406c1, and the Sarvāstivāda Vinaya, T 1435 at T XXIII 448b13, Ānanda used the expression 如是我聞 followed by 一時, which then each time led to an indication of the location at which the Buddha was believed to have been dwelling at that time. Brough 1950: 419 refers to Przyluski 1926: 18, 84, and 128 for accounts of the saṅgīti that do not have a reference to such a location (as well as to the exegesis on the introductory phrase in T 1509 at T XXV 66a27, translated by Lamotte 1944/1981: 80ff). Two of the texts mentioned by Przyluski provide a location: T 384 at T XII 1058b16 (here given in general terms as 說佛所居處) and T 2027 at T XLIX 6c13 (reference given to Vārāṇasī). Of the remaining two instances (which are two parallels to DN 16), T 6 at T I 191a17 indeed has only "thus have I heard" together with "at one time", without mentioning a location. T 5 at T I 175b26 then also does not mention "at one time" and has only "thus have I heard". In both cases, the audience interrupts at this point by acclamation or by breaking out in tears. Thus both are abbreviations caused by the reaction of the audience, without a deeper implication for the significance of the phrase itself. whereas according to other Vinayas he did not use this expression.[n.11] The Dharmaguptaka Vinaya, T 1428 at T XXII 968b19, the Mahīśāsaka Vinaya, T 1421 at T XXII 191a19, and the Theravāda Vinaya, Vin II 287,16, report that Ānanda's recital of the discourses at the first saṅgīti was prompted by a question regarding the location, in reply to which he then gave the required indications. Thus in these versions he does not use the standard phrase, which Klaus 2007: 321 note 29 sees as probably reflecting an earlier form of presentation. The use of the standard phrase by Ānanda at the first saṅgīti when reciting the Brahmajāla-sutta is then reported in the commentary on the discourse; cf. Sv I 26,1; on the significance of this phrase as authentication of the teaching given cf. also, e.g., Levering 1989: 61. In the case of the Brahmajāla, to take the phrase "thus have I heard" as expressing the hearing of the discourse by Ānanda at the time of its original delivery would in fact not work so well, since the events described in the introductory narration would have been directly experienced by him, instead of being something he heard.[n.12] This problem has already been noted by Klaus 2007: 317, who points out that "es muß ja doch einfach jeder, der die Worte evaṃ me sutaṃ (ekaṃ samayaṃ) hört, glauben, daß der gesamte folgende Bericht und nicht nur die in ihm enthaltene Lehrdarstellung das ist, was da 'so gehört' wurde"; cf. also Przyluski 1926: 346, who makes a similar observation regarding the concept of what is considered canonical: "dans un sūtra commençant par ces mots: 'Ainsi j'ai entendu. Une fois le Buddha demeurait à …', le cadre même du récit ne peut avoir été prononcé par le Buddha." Brough 1950: 425 takes the fact that parts of the discourse could not have been heard by those present at its original delivery to support a combining of "at one time" with "thus have I heard". As pointed out by Galloway 1991: 97 in reply, however, this issue is independent of how one relates "at one time" to what precedes and follows it. Again, in the final part of the discourse Ānanda poses a question.[n.13] Ānanda is on record for enquiring after the name of the discourse in DN 1 at DN I 46,19, Weller 1934: 64,31 (§220), D 4094 ju 153b4 or Q 5595 tu 177a6, DĀ 21 at T I 94a10, and T 21 at T I 270c15. The narrative description that introduces him and the words he is believed to have spoken would both not be something he could himself refer to as "thus have I heard". [n.14] Already Franke 1913a: 1 note 3 commented on the phrase evaṃ me sutaṃ in DN 1 that "dieser 'ich' kann nicht Buddha's Lieblingsjünger Ānanda sein, wie die einheimische Überlieferung annimmt", and then pointed to another passage in which Ānanda is referred to in the third person singular. Similarly, the concluding section's report that the monks delighted in what the Buddha had taught would have to be considered something Ānanda directly experienced and took part in, and thus also not something he merely heard.

For Ānanda to use the expression "thus have I heard" to introduce a discourse that has any kind of introductory narration or conclusion would only work if he had not been present at that time, but had heard the discourse from someone else. In the case of a discourse like the Brahmajāla, the "I" in the phrase "thus have I heard" therefore needs to be understood to stand for the various members of the lineage of reciters that according to tradition was started by Ānanda's original recital of the text.

Now in the case of these subsequent generations of reciters a problem arises if "at one time" is taken to qualify "thus have I heard". In the case of a long discourse like the Brahmajāla, it seems highly improbable that an average reciter heard the discourse only at one time, simply because it requires more than one hearing to be able to learn such a complex discourse in an oral setting. Except for an exceptionally gifted case like Ānanda, who in the tradition features as foremost among outstanding disciples for being of much learning (which in an oral setting of course requires excellent memory),[n.15] AN 1.14.2 at AN I 24,32 and its parallel EĀ 4.7 at T II 558a26; cf. also the Divyāvadāna, Cowell and Neil 1886: 396,18, and the Saṅghabhedavastu, Gnoli 1978: 54,18. Nyanaponika and Hecker 1997: 151 summarize the traditional belief regarding Ānanda as follows: "he could immediately remember everything, even if he had heard it only once. He could repeat discourses of the Buddha flawlessly up to sixty thousand words, without leaving out a single syllable"; cf. also Th 1024 and the Avadānaśataka, Speyer 1909/1970: 155,7. to master a discourse like the Brahmajāla can safely be expected to require several hearings.

Given the length of the discourse in terms of recitation time, such hearings would probably not be continuous, but rather be spread out over several days, with the prospective reciter being allowed time to rehearse what has already been learned before proceeding to learn new material. Even after having mastered the text, a reciter would still from time to time participate in group recitation with other reciters who know the same discourse collection,[n.16] Such group recitation is reflected in Vin II 75,31 which, as part of a description of the task of allotting accommodation to incoming monks, indicates that those who recite the discourses are allocated together so that they can do group recitation with each other, aññamaññaṃ suttantaṃ saṅgāyissanti. The need to have those who recite and teach the discourses stay together is also reflected in the corresponding passage in the Dharmaguptaka Vinaya, T 1428 at T XXII 587b21, the Mahāsāṅghika Vinaya, T 1425 at T XXII 394c12 (correcting the punctuation to read 經唄經唄共, etc.), the Mahīśāsaka Vinaya, T 1421 at T XXII 15a28, the Mūlasarvāstivāda Vinaya, T 1442 at T XXIII 695c9, and the Sarvāstivāda Vinaya, T 1435 at T XXIII 22a12. thereby again coming to hear the discourse earlier learned and ideally getting lapses of memory rectified in this way. Thus, at least inasmuch as the Brahmajāla and discourses of comparable length are concerned, it would not make much sense for an average reciter to qualify his hearing as taking place "at one time".

In sum, it seems to me that although with later texts the alternative interpretation that relates "at one time" to "thus have I heard" certainly needs to be taken into account,[n.17] According to Brough 1950: 421 and von Hinüber 1968: 87 note 1, the use of the accusative ekaṃ samayaṃ would be earlier than the locative ekasmiṃ samaye found regularly in Sanskrit texts. Vetter 1993: 65 note 48 then suggests that "the accusative was the older expression and was replaced by the locative in order to repair a long felt shortcoming, viz. that hearing the tenets a text contained was not sufficiently marked as also having happened at the time when the Lord dwelt there and there and met this or that person, etc." In line with the suggestion by Vetter, the Mahāsāṅghika, Mūlasarvāstivāda, and Sarvāstivāda Vinayas, mentioned above p. 56 note 10, appear to reflect the same need for authentication. This would explain a gradual shift in meaning, whereby the phrase "at one time" was increasingly seen as qualifying not only the event reported, but also the hearing of the discourse, resulting in what Harrison 1990: 5 note 3 has referred to as an "apo koinou construction". in the case of the early discourses,[n.18] For a more detailed discussion of the notion of "early discourses" and their historical value, especially of the Pāli discourses, cf. above p. 15ff. preserved in the Nikāyas and Āgamas, it can safely be assumed that the qualification "at one time" begins a new sentence introducing the events described in the discourse.[n.19] Klaus 2007: 320 concludes that the Pāli phrase evaṃ me sutaṃ originally did not function as an indicator that one was present personally at the time of the discourse. Instead, it served to mark that the text originated from oral transmission, rather than being the speaker's own composition, "all das zusammen macht es unmöglich, die Worte evaṃ me sutaṃ als eine Formel zu betrachten, mit der man kundtat, Ohrenzeuge der im folgenden Bericht enthaltenen Lehrdarlegung gewesen zu sein. Diese einleitenden Worte können ursprünglich nur als Hinweis darauf gemeint gewesen und ursprünglich auch nur so verstanden worden sein, daß es sich bei dem folgenden Bericht nicht um einen vom Sprecher selbst verfaßten, sondern um einen überlieferten Text handelt." The variant reading thos pa’i, found in some Tibetan editions of the Abhidharmakośopāyikā-ṭīkā quote of the Brahmajāla, is best considered the result of later influences.

---

The Introductory Narration of the Brahmajāla

After the standard phrase at the outset of a discourse, the Brahmajāla proceeds with an introductory narration. According to this narration, found with considerable similarity in the parallel versions,[n.20] The similarity between the introductory narrations in DN 1 and DĀ 21 had been noted already by Beal 1884: 34–36. the Buddha and his monks were on a journey and were being followed by a teacher and his pupil. This teacher kept speaking in dispraise of the Buddha, the Dharma, and the Community, whereas his pupil spoke their praise. The monks then had a discussion about this contrasting behaviour. Having become aware of that, the Buddha decided to join them. Once he had joined the monks, according to all versions he asked what they had been discussing.

A minor but noteworthy difference at this juncture can be found in a version of the Brahmajāla preserved in the Chinese Dīrgha-āgama.[n.21] Regarding the title 梵動經 of this discourse, Karashima 2006: 361 explains that "the translator(s) confused -jāla ('net') with -cāla ('moving'), both of which may become -yāla in the underlying language, as is common in Middle Indic, including Gāndhārī"; cf. also Weller 1971: 207, who introduces his translation of DĀ 21 as a "Verdeutschung des Brahmacālasūtra" and below p. 176 note 128. The Dīrgha-āgama version differs from its parallels in so far as it indicates that when the Buddha asked the monks what they had been discussing, he did so knowingly.[n.22] DĀ 21 at T I 88b29: 大眾前坐, 知而故問. None of the other versions accompanies the description of the Buddha's actual enquiry with an explicit specification that he did so knowingly. Instead, the other parallels merely report that he sat down and then asked the monks what they had been talking about.[n.23] DN 1 at DN I 2,23: nisajja kho bhagavā bhikkhū āmantesi, Weller 1934: 8,8 (§6): bzhugs nas kyang dge slong rnams la bka’stsal pa, D 4094 ju 142a4 or Q 5595 tu 163a7: dge slong gi dge ’dun gyi dbus su gdan bshams pa la bzhugs te, bcom ldan ’das kyis dge slong rnams la bos te, and T 21 at T I 264b6: 佛則 坐, 佛問諸比丘言.

A specification comparable to the one in the Dīrgha-āgama version can be found, however, in the Pāli commentary on the Brahmajāla-sutta of the Dīgha-nikāya. According to this commentary, the Buddha indeed asked knowingly. [n.24] Sv I 49,18: evaṃ nisinno pana jānanto yeva kathāsamuṭṭhāpanatthaṃ bhikkhū pucchi. The point of making such a qualification would be to make it unmistakably clear that the Buddha did not ask out of ignorance. Instead, his question should be understood as merely an expedient means in order to start a conversation with the monks. The felt need to make such an additional qualification would be related to the growing tendency to consider the Buddha as omniscient, leading to the concern that even circumstantial information be presented in such a way as to leave no doubt about the Buddha's all-knowing condition. Such a concern is not yet a prevalent concern of most of the early discourses, which appear to reflect a period when the Buddha was not yet seen as omniscient.[n.25] On the attribution of omniscience to the Buddha, based on a survey of relevant early discourses and publications on this issue, cf. Anālayo 2014c: 117–127.

Since among the parallel versions only the Chinese Dīrgha-āgama discourse has such an explicit qualification at this point, it seems likely that this remark is a later addition. This impression finds confirmation in the fact that the same expression occurs regularly in other discourses in the same Dīrgha-āgama, where again the parallels do not have such an indication.[n.26] The expression 知而故問 occurs also in DĀ 1 at T I 1b24, DĀ 2 at T I 19a29 and T I 25c1, and DĀ 30 at T I 114b14 (the last has no known parallel; cf. below p. 423ff). In the case of DĀ 1, the corresponding passages in the discourse parallels describe the Buddha's enquiry without an explicit indication that he asked knowingly; cf. Fukita 2003: 32,11: niṣadya bha(ga)vā(ṃ) bhikṣūn āmantrayati, DN 14 at DN II 1,12: nisajja kho bhagavā bhikkhū āmantesi, T 2 at T I 150a14: 世尊問曰 (preceded by describing that the monks paid respect to him), T 4 at T I 159b4: 佛即到, 諸比丘所問言, and EĀ 48.4 at T II 790a15: 在中央坐, 爾 時世尊告諸比丘. The same holds for the discourse parallels to the first passage in DĀ 2 at T I 19a29 (reading 我知而故問), where again the corresponding passages do not mention that the Buddha asked knowingly; cf. Waldschmidt 1951: 274 (§28.28): (tam aham evam ā)mantraye, DN 16 at DN II 131,32: taṃ purisaṃ etad avocaṃ, T 5 at T I 168b16: 我問, T 6 at T I 183c27: 吾問, and T 7 at T I 198a24: 我即問言. These sources follow the same pattern for the second instance in DĀ 2 at T I 25c1 (reading just 知而故問), Waldschmidt 1951: 296 (§32.15): bhagav(ā)n bhikṣūn āman(t)r(ayate), DN 16 at DN II 143,24: bhagavā bhikkhū āmantesi, T 5 at T I 169b13: 佛問比丘, T 6 at T I 184c25: 佛問比丘, and T 7 at T I 200b9: 世尊問餘比丘. In this case, however, an exception can be found in another parallel to this particular passage, EĀ 42.3 at T II 751a23, where the corresponding passage does mention that he asked the monks knowingly, 世尊知而告諸比丘曰 (whereas the above mentioned EĀ 48.4 does not have such a remark, another comparable instance can be found in the same discourse collection in EĀ 40.5 at T II 739b16: 知而問阿難曰, of which no discourse parallel appears to be known). On such formulations in Vinaya texts cf. Waldschmidt 1926: 47. Thus the addition of such a specification seems to be characteristic of this Dīrgha-āgama.

Now the circumstance that such a specification is found in the Pāli commentary on the Brahmajāla makes it possible that a similar remark could have been found in a Dharmaguptaka commentary on the Dīrgha-āgama as well. Both traditions can be expected to have incorporated in their respective commentaries material from a common ancient Indian commentarial tradition that would from a relatively early time have begun to develop alongside the discourses.[n.27] For a study of the Indian source material of the Theravāda commentaries cf. Endo 2005 and 2009. In this case, it would not be surprising if such a remark eventually came to influence the wording of the discourse. This would be in line with a general tendency discernible elsewhere in the early discourses, where during the course of oral transmission commentarial material appears to have become part of the discourse on which it commented.[n.28] For a detailed discussion with further references and a critical reply to the suggestion by Norman 1997: 158–160 that the commentaries were transmitted independently from the discourses cf. Anālayo 2010c.

---

The Exposition on Morality in the Brahmajāla

The influence of commentarial exegesis on the actual discourse can also be discerned in the next portion of the discourse, which takes up the topic of the Buddha's moral conduct.[n.29] For a comparative study of the subsequent examination of sixty-two standpoints for views in the parallel versions, including a discourse quotation in T 1548 at T XXVIII 656b19 to 661a7, cf. below p. 115ff.

In the discourse version of the Brahmajāla preserved in Tibetan translation, the relatively short exposition on morality lists the first two precepts, followed by an abbreviation which gives the impression that the remaining precepts should be supplemented. Then it briefly describes how some recluses and brahmins gain their livelihood in wrong ways. Even after supplementing the precepts that appear to have been abbreviated, the whole exposition is still rather short.[n.30] Weller 1934: 12,6 to 12,30 (§§18 to 21).

The same holds for the discourse quotation in the Abhidharmakośopāyikā-ṭīkā, which in fact only refers to the first precept and then abbreviates.[n.31] D 4094 ju 143a1 to 143a2 or Q 5595 tu 164a6 to 164a8. It seems fairly clear that the exposition of morality in the Mūlasarvāstivāda version of the Brahmajāla could not have been as long as what is now found in their Dharmaguptaka and Theravāda parallels.

The considerably longer coverage of the same topic of morality in the Dīgha-nikāya and Dīrgha-āgama versions falls into three distinct parts. The Brahmajāla-sutta of the Dīgha-nikāya explicitly introduces these three as a "short" (cūla) exposition on morality, followed by a "middle-length" (majjhima) exposition, and then a "great" (mahā) exposition.[n.32] The smaller section on morality in DN 1 ranges from DN I 4,1 to 5,26, the middle-length section then goes from there up to DN I 8,33, and the long section from there up to DN I 12,14.

The short exposition on morality starts in both versions with the first four of the five precepts. Following the fourth precept on abstention from falsehood come other forms of wrong speech, thereby completing the standard set of wrong forms of speech, which comprises divisive speech, harsh speech, and frivolous speech.[n.33] DN 1 at DN I 4,13 and DĀ 21 at T I 88c24; the same is also the case for T 21 at T I 264b28.

The Dīrgha-āgama discourse then continues with the fifth precept regarding abstention from alcohol, instead of which the Brahmajāla-sutta of the Dīgha-nikāya takes up the need to refrain from harming seeds (§10).[n.34] DĀ 21 at T I 89a5: 沙門瞿曇捨離飲酒; cf. also T 21 at T I 264c5: 不飲酒. The absence of such a reference in the corresponding section in DN 1 is in line with the observation by Reat 1996/1998: 49f that "abstinence from intoxicants is not included in elucidations of right action in the Pali sūtras, and is thereby not nearly as prominent an ethical issue as it came to be in later Buddhism"; cf. also the discussion in, e.g., Schmithausen 1991: 8f note 42, Nattier 2003a: 109 note 11, and Anālayo 2011b: 190f note 245. Both versions continue with the remaining eight precepts, which are presented together with several basic aspects of ethical conduct incumbent on a recluse in the ancient Indian setting (such as not possessing animals, abstaining from barter, etc.).

Such a basic listing of the precepts together with a few aspects of the proper livelihood of a recluse appears to have been a common starting point of the parallel versions, including the Mūlasarvāstivāda discourse preserved in Tibetan translation. This much in fact corresponds to the compass of the section on morality in the exposition of the gradual path in the Cūḷahatthipadopama-sutta of the Majjhima-nikāya and its Madhyama-āgama parallel, which corresponds to what the Brahmajāla-sutta reckons the short exposition on morality.[n.35] MN 27 at MN I 179,22 to 180,19 and MĀ 146 at T I 657a14 to 657b27.

The middling section on morality in the Brahmajāla-sutta of the Dīgha-nikāya then works in detail through the topics of injuring seeds and plants (§11), storing up food (§12), visiting shows and games (§§13 and 14), using luxurious beds, etc. (§15), and using various adornments (§16).[n.36] DN 1 at DN I 5,28 to 7,26, following the paragraph numbering given in Ee. The same or closely related topics occurred already towards the end of the smaller section on morality (§10).[n.37] DN 1 at DN I 5,4: bījagāmabhūtagāmasamārambhā paṭivirato … ekabhattiko … naccagītavāditavisūkadassanā paṭivirato … mālāgandhavilepanadhāraṇamaṇḍanavibhūsanaṭṭhānā paṭivirato … uccāsayanamahāsayanā paṭivirato.

The smaller section on morality mentions not harming seeds, which corresponds to the topic taken up again in §11. Next the smaller section takes up eating a single meal at the right time, which can be seen to stand in a close thematic relationship to the storing up of food in §12. Then the smaller section refers to not visiting shows, not using adornments, and refraining from high beds and seats, which correspond to §§13 to 16 on visiting shows and games, using high beds and seats, and using various adornments (with a difference in sequence in so far as adornments and high seats are in the opposite order).

A comparable pattern of relationship can also be observed in the Chinese Dīrgha-āgama version, where the middle-length section on morality describes lack of contentment with food and robes, undertaking commercial transactions and seeking profits, adorning oneself, and engaging in various forms of recreation. These various themes have already been briefly mentioned in the preceding section.[n.38] After the reference to alcohol (cf. above p. 65 note 34), DĀ 21 at T I 89a6 mentions not using adornments, not seeing shows, not sitting on high seats, not eating at the wrong time, and not taking gold and silver, 不著香華, 不觀 歌舞, 不坐高床, 非時不食, 不執金銀. The topics of eating at the wrong time and taking gold and silver can be seen to receive a more detailed treatment at T I 89a16 in terms of discontentment with food and robes, 衣服飲食無有厭足, undertaking commercial transactions, 自營生業, and seeking profits, 求諸利養, which also refers to high seats, 高廣大床. The problems of adornments and seeing shows recur in T I 89a23 in terms of trying to adorn oneself, 求自莊嚴, and engaging in various forms of recreation, 嬉戱. Similar correspondences can be observed in the individual translation T 21, where the corresponding section at T I 264c4 begins with the topic of not sitting on high seats, 不坐高 綺好牀, a topic then taken up in more detail at T I 264c17.

In this way the middle-length sections in the Dīgha-nikāya version and its counterpart in the Dīrgha-āgama discourse appear to have their origin in a commentary on the preceding smaller section.[n.39] This has already been suggested by Franke 1913a: 7 note 2, who commented that the degree of repetition found "spricht wohl dafür, daß die Listen von verwerflichen Beschäftigungen in unseren Paragraphen aus allerlei vorhandenen Schemata zusammengelesen sind. Manche Elemente treten ja auch mehr als einmal, in verschiedenen Paragraphen auf, vielleicht deshalb, weil sie in verschiedenen Schemata vorkamen." In relation to the same section, Weller 1935: 41 note 46 similarly noted that "wahrscheinlich liegen kommentarielle Erweiterungen vor, die einem älteren, einfacheren Textbestande zugewachsen sind." This has also been proposed by Meisig 1987a: 59f, based on a comparison of the smaller and middle-length sections on morality. This impression is further strengthened if one turns to the commentary on the Brahmajāla-sutta as well as the commentary on the Cūḷahatthipadopama-sutta of the Majjhima-nikāya (whose entire exposition of morality, as mentioned above, corresponds to the smaller section on morality in the Brahmajāla-sutta). In both cases the commentarial gloss on the need to refrain from harming seeds lists precisely those five types of seeds that are also mentioned at the outset of the middle section on morality in the Brahmajāla-sutta itself, when taking up the topic of injuring seeds and plants (§11).[n.40] Ps II 208,23 lists mūlabīja, khandhabīja, phalubīja, aggabīja, and bījabīja as the five types of seed not to be harmed (the corresponding passage in Sv I 77,6 differs in so far as it has aggabīja as its third and phalubīja as fourth); the same five seeds recur at the outset of the middle section on morality in DN 1 at DN I 5,31. Thus this middle section does indeed seem to stem from a commentarial gloss.

That this apparent commentary is an addition to an already existing shorter exposition of morality finds support in a shift to a different introductory phrase employed in the middle-length section. In both versions this phrase refers to others whose wrong conduct provides a contrast to the proper behaviour of the Buddha,[n.41] In the case of DĀ 21 at T I 89a15 the phrase reads: 如餘沙門, 婆羅門, 受他信 施. Similarly the middle-length section on morality in DN 1 at DN I 5,28 starts with yathā va pan' eke bhonto samaṇabrāhmaṇā saddhādeyyāni bhojanāni bhuñjitvā. In the discourse version preserved in Tibetan, Weller 1934: 12,14 (§20), this reads: ji ltar ’di na dge sbyong dang bram ze kha cig dad pas byin pa yongs su spyod cing, and in T 21 at T I 264c19: 譬有異道人, 受人信施食. In relation to the phrasing in DN 1, Franke 1913a: 7 note 4 comments: "die folgenden Paragraphe fangen eigentlich alle mit yathā 'wie' an. Dieses 'wie' erklärt sich wohl daraus, daß sie teilweise Erläuterungen, Beispielsanführungen zu § 10 enthalten." a reference not made in the preceding shorter section on morality.

Moreover, the Brahmajāla-sutta of the Dīgha-nikāya explicitly marks the end of the preceding section by indicating that the smaller section on morality is concluded.[n.42] DN 1 at DN I 5,27: cūlasīlaṃ niṭṭhitaṃ (Be and Se: cūḷasīlaṃ, Ce: cullasīlaṃ). Since the context of the discourse would not require a subdivision of the section on morality, it seems that this division is the result of an assembling of textual pieces that are of different origins.

In sum, the introductory narration and the exposition on morality appear to be examples of a general tendency during the process of oral transmission for commentarial exegesis at times to influence the wording of the discourse, or even become part of it.

---

The Early Buddhist Oral Tradition

The above surveyed features of the Brahmajāla exemplify patterns at work in the early Buddhist oral tradition in general, where at times major variations can be found between parallel versions. This can at first sight give the impression that improvisation is characteristic of the oral transmission of the early discourses themselves.

A considerable degree of improvisation is in fact a feature of the oral transmission of the ancient Greek and modern Yugoslavian epic.[n.43] Lord 1960/2000: 4f describes that in the case of the Yugoslavian epic literature studied by Parry and himself, "in a very real sense every performance is a separate song; for every performance is unique, and every performance bears the signature of its poet singer." Thus "improvisation is not a bad term for the process, but it too must be modified by the restrictions of the particular style", i.e., by the use of fairly fixed formulas that help the singer to compose his song rapidly during performance. However, as Lord 1960/2000: 36 points out, the singer "does not 'memorize' formulas" in the sense of "repeating something that one regards as fixed". Instead, he learns to employ these creatively for his performance purposes, somewhat similarly to learning the words of a language. Applying the research findings in relation to this type of oral tradition to the early Buddhist case, perhaps variations between parallel versions should simply be seen as the result of different creative performances, instead of being divergent versions of a text committed to memory?[n.44] Cousins 1983: 2 and 5f argues that "if we compare the Pali recensions of the nikayas with other surviving versions, the differences we find are exactly those we might expect to discover between different performances of oral works", "these divergences … are too frequent to arise from the natural variation of a manuscript tradition or even from a rigidly memorized oral tradition"; cf. also McGovern 2013: 364–401.

A problem with applying the findings made in relation to the oral tradition of the ancient Greek and modern Yugoslavian epic to the transmission of the early Buddhist discourses, however, is that the characteristics of the respective oral traditions differ rather substantially.[n.45] Allon 1997a: 42 points out that "many factors can influence the character of an oral literature and its method of composition and transmission: the nature of the information being relayed; the attitude towards this material and the extent to which accuracy is required; the character of the performers or composers, their status in society, the type of training they have undergone and the circumstances under which they perform; the nature of the audience and its expectations and therefore its demands on the performer or performers; the medium used (verse or prose) and whether the performance requires musical accompaniment. The Buddhist and Yugoslav-Homeric traditions differ in virtually all of these factors." Thus the problem of applying the findings by Parry and Lord to the early Buddhist case is not merely one of using research done on verse for texts that are predominantly in prose, pace Cousins 2013a: 99f. In fact Lord 1960/2000: 5 himself clearly recognizes that oral epic differs from the oral transmission of material where exact transmission matters, commenting that "if the reader interprets oral learning [of the epic type] as listening to something repeated in exactly the same form many times, if he equates it with oral memorization by rote, then he will fail to grasp the peculiar process involved in learning oral epic." So "if we understand thereby the transmission of a fixed text or the kind of transmission involved where A tells B what happened and B tells C and so on with all natural errors of lapse of memory and exaggeration and distortion, then we do not fully comprehend what oral transmission of oral epic is." Cf. also below p. 473ff. From a methodological viewpoint, it is questionable whether research done on the oral performance of epic material can just be applied to the oral transmission of material that is not epic and which has been passed on in a different cultural setting.[n.46] Graham 1987: 138 warns that the "oral use and even oral transmission of scripture should not be confused with folk oral tradition in which verbatim accuracy is not aspired to (i.e., in which 'formulaic composition' predominates: see, for example, Albert B. Lord, The Singer of Tales …)." As Gombrich 1990a: 21 points out, "the early Buddhists wished to preserve the word of their great teacher … the precise wording mattered." Norman 1997: 49 sums up that contrary to "oral literature which … is essentially of an epic nature where … no two performances are ever identical because the reciter is free to insert, at any point, material", in the case of "the great majority of Pāli canonical texts, however … complete accuracy of reproduction is required at each recitation. In these circumstances the findings of modern investigators of oral epic literature seem to have little relevance."

Looking for precedents for the early Buddhist oral tradition within the ancient Indian cultural setting, a rather different mode of oral transmission comes into view, namely the transmission of Vedic texts through memorization.[n.47] Gombrich 1990a: 23 notes that "the Buddhist canon has left us more clues that it is modelled on Vedic literature than has been generally recognized"; cf. also Lévi 1915: 441 and von Hinüber 1991: 123. In methodological terms, it would certainly be preferable if the early Buddhist oral tradition were to be approached from the perspective of its Vedic predecessor,[n.48] This methodological problem is particularly relevant to the case of McGovern 2013: 364–401 who, in spite of the overall theme of his study being the relationship between early Buddhists and brahmins, does not evaluate the early Buddhist oral tradition in the light of its Indic antecedents, but instead turns to the culturally and genre-wise unrelated Greek and Yugoslavian oral epic. given that the two share the general cultural setting of ancient India and both are concerned with texts considered sacred, whose correct wording matters, and recitation is undertaken by religious professionals whose primary aim is not entertainment.[n.49] Yugoslavian oral epic is instead performed by singers who come from various walks of life; cf. Lord 1960/2000: 18: "no particular occupation contributed more singers than any other." The actual performance then relies heavily on the singer's ability to entertain; cf. Lord 1960/2000: 17: "the length of the song depends upon the audience", as at times "the singer will realize shortly after beginning that his audience is not receptive, and hence he will shorten his song so that it may be finished within the limit of time for which he feels the audience may be counted on. Or, if he misjudges, he may simply never finish the song."

Comparing the Vedic and early Buddhist oral traditions, an important difference between these two is that brahmins were trained from their childhood onwards in memorizing, whereas training as an early Buddhist reciter could have begun only after ordination, which in the average case would have taken place at a later age.[n.50] This has already been pointed out by von Hinüber 1989: 67f. As I have shown elsewhere, the differing memory abilities that result from this basic dissimilarity do indeed explain why even a considerable degree of variation between parallel versions of a discourse could come into existence in an oral tradition that aims at correct and accurate transmission.[n.51] Anālayo 2011b: 867–876. As already noted by Allon 1997b: 366, "there is, however, scope for considerable change to occur within a tradition of the transmission of fixed, memorized texts." Similarly, in relation to variations between the above mentioned DN 34 and its parallels, Wynne 2004: 106 points out that such "differences could just as easily have been produced by the natural variations of a relatively fixed oral transmission."

Nevertheless, perhaps we should at least consider improvisation characteristic of the earliest phase of the coming into being of the early discourses?[n.52] Cousins 1983: 9f proposes "in the early period … the possibility of a strong improvisatory element. This can be confirmed by comparisons between the surviving versions derived from different sects", which was then followed by "the gradual fixation of the material at a later period". To be sure, the degree of formalization now found in some of the early discourses would indeed have come into being gradually, but some degree of formalization must have been used right from the outset when a particular discourse came to be orally transmitted. In fact, even the original delivery of a discourse in an oral setting can be expected to employ some degree of formalization, thereby ensuring that the audience better keeps in mind what has been said.[n.53] Ong 1982/1996: 34f even argues that "in a primary oral culture … you have to do your thinking in mnemonic patterns, shaped for ready oral recurrence. Your thought must come into being in heavily rhythmic, balanced patterns, in repetitions or antitheses, in alliterations and assonances, in epithetic and other formulaic expressions", because "in an oral culture, to think through something in non-formulaic, non-patterned, non-mnemonic terms, even if it were possible, would be a waste of time, for such thought, once worked through, could never be recovered with any effectiveness."

Besides, we simply have no evidence that would support a shift from an early period of fairly free improvisation to a subsequent period of strictly formalized transmission, except for variations found between parallel versions of a discourse.[n.54] As pointed out by Wynne 2004: 106, "it is hard to imagine that oral improvisation in the transmission of literature was the norm in the early period of sectarian Buddhism." Yet in the case of the Brahmajāla, for example, a rather substantial difference in the exposition of morality appears to have come into being only at a time when the Mūlasarvāstivāda version of this discourse was transmitted independently from the Dharmaguptaka and Theravāda versions. Such a difference could not result from a very early period of improvisation.

In fact at times substantial differences can even be found between discourses of the same Theravāda tradition. Thus pericope descriptions of the same event in versions of the same Pāli discourse found in different Nikāyas can show quite substantial variations.[n.55] Cf. Anālayo 2011b: 18f and below p. 477ff. Another and particularly telling example can be found in the two versions of the Kasibhāradvāja-sutta, found in the Saṃyutta-nikāya and the Sutta-nipāta, which differ in the effect the otherwise same instruction had on its brahmin protagonist: in the Saṃyutta-nikāya he takes refuge as a lay follower, but in the Sutta-nipāta he goes forth and becomes an arahant.[n.56] SN 7.11 at SN I 173,22 and Sn 1.4 at Sn 15,23 (reference is to the page). Such variations appear to have come into being within the Theravāda transmission lineage and thus would not seem to be just the result of an early improvisation period.

Instead of giving any evidence that during an early period improvisation was considered acceptable, the texts themselves tend to emphasize the importance of accurately memorizing the Dharma. Thus, for example, the Pāsādika-sutta reports an instruction by the Buddha that the teachings given by him should be recited together [comparing] meaning with meaning and phrasing with phrasing.[n.57] DN 29 at DN III 127,16: atthena atthaṃ vyañjanena vyañjanaṃ saṅgāyitabbaṃ. The Chinese Dīrgha-āgama parallel to this discourse takes up the same issue in more detail, describing how the monks should behave if there is a disagreement regarding the meaning and the phrasing, regarding only the meaning, only the phrasing, or regarding neither.[n.58] DĀ 17 at T I 74a19: 說句不正, 義理不正, followed by the alternatives 說句 不正, 義正, then 說句正, 義不正, and finally 說句正, 義正. In both versions the teachings referred to are those that constitute the thirty-seven qualities or practices that are conducive to awakening (bodhipakkhiyā dhammā).[n.59] Sanskrit fragment parallel 281v3, DiSimone 2016: 84, instead mentions the aṅgas.

A concern with correct phrasing is also evident in a Vinaya rule which prohibits teaching a layman recitation of the Dharma. Such recitation is qualified in the code of rules (pātimokkha) of the Lokottaravāda-Mahāsāṅghika, Mūlasarvāstivāda, Sarvāstivāda, and Theravāda traditions as being done "word by word".[n.60] Cf. below p. 495 note 74. The fact that monks should not teach a lay person in this way of course implies that this was precisely the way they would teach recitation to each other: word by word.[n.61] Wynne 2004: 109 comments that this rule implies that "Sutta portions of the early Buddhist literature were learnt verbatim among the ordained."

The code of rules is in fact perhaps the best example one could choose for assessing the nature of the early Buddhist oral tradition, since its regular recital can safely be expected to have had a determining influence on the early Buddhist oral tradition. The recital of the code of rules obviously involves a text with fixed wording, as is the case for all group recitals, making it clear that an improvisatory model cannot be applied to this text.[n.62] As noted by Wynne 2004: 108, the two "pātimokkhas, for example, can hardly have been subject to an improvisational method of oral transmission, for their content (monastic rules) is hardly the sort of material suitable for improvisation."

Nevertheless, the different versions of the code of rules that have come down to us show the same type of variation as found between parallel versions of the early discourses: the rules often come in a different sequence, they show variations in terminology, and at times rules can be found in some traditions that are not found in others.[n.63] Cf., e.g., the study by Pachow 1955. The degree of variation between parallel versions of the code of rules clearly requires an explanation that goes beyond the parameters of the oral transmission of the Greek and modern Yugoslavian epic. Since the code of rules was transmitted by members of the same Buddhist monastic reciter circles that were responsible for the transmission of the early discourses, it is also clear that similar mechanisms must have been at work. Not only the code of rules, but also the discourses could be performed in group recitation, which makes improvisation a practical impossibility.[n.64] As already pointed out by Allon 1997b: 366, "communal or group recitation or performance requires fixed wording. It is not possible for more than one individual to perform at the same time in the manner described by Parry and Lord without producing utter chaos, for in that method each individual creates his compositions anew each time he performs." Thus for the reciters of the discourses to perform together in the way described in the Vinaya passage cited above p. 59 note 16 would not allow for personal improvisation. Coward 1986: 300 notes that "group listening to check for errors is still an accepted method of verification in rural India today."

In relation to the recitation and transmission of the code of rules, keeping in mind the Vedic background is again helpful. In the case of a monk from a brahmin background transmitting the code of rules it is quite possible that he was trained in memorization in his youth.[n.65] The presence of a rather substantial percentage of brahmins among the Buddhist monastic community suggests itself from the surveys in Sarao 1989: 93– 139, Chakravarti 1996: 198–220, and Nakamura 2000a: 360–362. The case of a nun would be different, however. Even if she should stem from a brahmin family, as a woman she would not have stood a comparable chance of receiving such training. As one might expect, the different versions of the code of rules for nuns exhibit more pronounced differences from one another than is the case for the different versions of the code of rules for monks. [n.66] Cf., e.g., the study by Kabilsingh 1984. An important factor influencing this would quite probably have been the lack of memorization training among nuns responsible for transmitting their own code of rules.

The emphasis on accurate transmission, word by word, in the early Buddhist oral tradition can even be seen in some of the transmission errors that emerge through comparative study, which appear to be the result of lapses of memory rather than the outcome of improvisation. This is particularly evident in those cases where the counterpart to a particular term shows close phonetic similarity, but has a considerably different meaning.[n.67] For examples cf. below p. 480 note 15. Such errors can only occur in a tradition that aims at accurate memorization of texts.

Whereas the early discourses themselves would not have been considered the proper place for personal creativity, such would have been possible in relation to the commentarial explanations that a reciter might give alongside the main text in an oral teaching situation. The giving of a commentary on the discourse would of course have been open to and influenced by personal notions and ideas as well as views and opinions held by the textual or religious group to which a reciter belonged. Once in the course of time such a commentary has become to some degree fixed and is passed on alongside the discourse, it is easy to see how such material could eventually become part of the discourse itself. This would explain how even substantial differences can come into being, such as the exposition on morality in the different versions of the Brahmajāla, namely as the result of the integration of commentarial material into the discourse.

Another aspect that clearly was open to some degree of creativity was the allocation of discourses to and within a particular collection. Thus relatively similar discourse parallels are regularly found in substantially different positions in the same or even in different discourse collections. Just as the giving of a commentary, the grouping of discourses for memorization purposes was clearly open to personal preferences.

In sum, a proper appreciation of the oral legacy of the early discourses needs to be based on contextualizing their transmission within the ancient Indian setting, taking into account the precedent set by Vedic reciters and the limitations faced by the early Buddhist reciters in their attempt to preserve their sacred texts as accurately as possible. The resultant understanding of early Buddhist orality can accommodate even substantial variations between parallel versions, without needing to resort to foreign models based on improvisation. Such understanding in a way draws out the significance of the phrase found at the beginning of a discourse, which with the terms "thus have I heard" signals that what follows is not the product of personal improvisation, but much rather results from an attempt to transmit a text as it has been heard.