A Comparison of the Madhyama Agama and the Majjhima Nikaya (on an Organizational Level)

Essays
, ,
1

Introduction

Over the past couple years, I’ve written essays showing the organizational parallels between the Samyukta Āgama (SĀ) and the Samyutta Nikāya (SN) as well as that of the Dīrgha Āgama (DĀ) and the Dīgha Nikāya (DN). These comparisons showed evidence that the extant versions of these collections share common cores that may be remnants of older presectarian collections in both cases. The case was strong for DĀ/DN, which was a particularly interesting study because of the existence of a third version that could be set alongside the other two.

Here, I want to continue these studies with the case of the Majjhima Nikāya and Madhyama Āgama, if only to have a study available. The truth is, there is not much to say about these two collections in terms of a common organizing principle that might point to an older ancestor collection: They have little in common. This is surprising given that MĀ is generally considered part of a Sarvāstivāda canon, which is not distantly related to the Theravāda canon in terms of sectarian history. The agreement of individual parallel texts in MN and MĀ testifies to this close relationship; yet, the two collections are very different organizationally. I can think of a few reasons this might be:

  1. There was no ancestor collection. This would mean both MN and MĀ formed after regional and sectarian canons were established from existing canonical material.

  2. One of these two versions was reorganized while the other was not. It may be that one of these two collections has been given a new organizing principle that erased any commonality with the other. In this scenario, MĀ seems to be the more likely candidate for being the one that was reorganized.

  3. Neither collection descends directly from an earlier version. This would mean both of the extant collections were reorganized or created independently, but that does not rule out an older version. Other canonical collections may have better preserved an early MN/MĀ, but they are lost today.

The fact that we only have these two editions of MN/MĀ limits our ability to come to a firm conclusion about their history. One thing we do know, however, is that they share a body of middle-length sūtras that are parallels. There are some interesting insights about the other early collections to be gleaned by looking at MĀ and MN as well.

The Parallels to MĀ Sūtras in Theravāda Sources

The Madhyama Āgama that was translated to Chinese contains 222 sūtras, of which 98 are parallels to suttas in the Majjhima Nikāya. While this sounds like a high ratio for MN (98 out of 152 suttas = 64% overlap), it’s less than half of MĀ (98 out of 222 sūtras = 44% overlap).

This is not to say MĀ has few parallels with the Theravāda Suttapiṭaka, however. It also has 88 sūtras with a parallel in AN (40% of MĀ). Because of this mass inclusion of AN parallels, MĀ has a large number of short sūtras: 33 sūtras are only a half page of Chinese or less, and 64 are less than a full page of Chinese.[1] These brief sūtras do not seem to belong in a middle-length collection, and MĀ would be closer to the same size as MN if they were removed (222 - 64 = 158 sūtras). These short sūtras are often augmented compared to their counterparts in AN, which perhaps was part of a process of moving them into MĀ from a Sarvāstivādin Ekottarika Āgama.

The case that sūtras were moved between collections is strengthened when we look at the parallels to DĀ and DN that are found in MN and MĀ. As Hartmann and Bucknell noted when they compared the reconstructed Sanskrit DĀ, DN, MN, and MĀ, it seems that two groups of eleven and ten texts have been transferred between these collections. The groups of texts involved are different for DN and MN than for the Sanskrit DĀ and MĀ, but the comparison of the two canonical collections makes the transfers apparent. MĀ includes a set of ten sūtras (4.5% overlap) that today have parallels in DN. They may have been found in MN at some earlier time before they were moved into DN. MN similarly contains a set of suttas that are paralleled in the Sarvāstivāda DĀ instead of MĀ.

The parallels that we’ve described thus far account for 86% of the sūtras in MĀ (190 of 222).[2] There are also 12 sūtras with parallels in other Theravāda collections like SN, the Suttanipāta, or Itivuttaka. The remaining 20 sūtras in MĀ have no known Buddhist parallel (thus far).[3]

The Parallel Structures of MĀ and MN

Given this amount of overlap between MĀ and MN, we would expect to find parallel structures when we look at these two collections, but there are only a couple cases that appear to preserve older chapters. The table below demonstrates this by listing each MĀ sūtra with its MN parallel by chapter and also noting the chapters those parallels belong to in MN.

MĀ Chapter MĀ Sūtra MN Chapter MN Sutta
1 MĀ 9 3 MN 24
1 MĀ 10 1 MN 2
2 MĀ 14 7 MN 61
2 MĀ 19 11 MN 101
3 MĀ 26 7 MN 69
3 MĀ 27 10 MN 97
3 MĀ 28 15 MN 143
3 MĀ 29 1 MN 9
3 MĀ 30 3 MN 28
3 MĀ 31 14 MN 141
4 MĀ 32 13 MN 123
4 MĀ 34 13 MN 124
6 MĀ 63 9 MN 81
6 MĀ 64 13 MN 130
6 MĀ 67 9 MN 83
7 MĀ 72 13 MN 128
7 MĀ 75 11 MN 106
7 MĀ 77 7 MN 68
7 MĀ 78 5 MN 49
7 MĀ 79 13 MN 127
7 MĀ 81 12 MN 119
7 MĀ 85 12 MN 113
8 MĀ 87 1 MN 5
8 MĀ 88 1 MN 3
8 MĀ 89 2 MN 15
8 MĀ 91 1 MN 8
8 MĀ 93 1 MN 7
9 MĀ 98 1 MN 10
9 MĀ 99 2 MN 13
9 MĀ 100 2 MN 14
9 MĀ 101 2 MN 20
9 MĀ 102 2 MN 19
9 MĀ 103 2 MN 11
9 MĀ 105 1 MN 6
9 MĀ 106 1 MN 1
10 MĀ 107 2 MN 17
10 MĀ 108 2 MN 17
10 MĀ 109 12 MN 114
10 MĀ 115 2 MN 18
11 MĀ 131 5 MN 50
11 MĀ 132 9 MN 82
11 MĀ 133 6 MN 56
12 MĀ 143 11 MN 107
12 MĀ 145 11 MN 108
12 MĀ 146 3 MN 27
12 MĀ 150 10 MN 96
12 MĀ 151 10 MN 93
12 MĀ 152 10 MN 99
12 MĀ 153 8 MN 75
12 MĀ 161 10 MN 91
13 MĀ 162 14 MN 140
13 MĀ 163 14 MN 137
13 MĀ 164 14 MN 138
13 MĀ 165 14 MN 133
13 MĀ 166 14 MN 134
13 MĀ 167 14 MN 132
13 MĀ 168 12 (MN 120)
13 MĀ 169 14 MN 139
13 MĀ 170 14 MN 135
13 MĀ 171 14 MN 136
14 MĀ 173 13 MN 126
14 MĀ 174 5 MN 45
14 MĀ 175 5 MN 46
14 MĀ 178 3 MN 25
14 MĀ 179 8 MN 78
14 MĀ 180 14 MN 142
14 MĀ 181 12 MN 115
15 MĀ 182 4 MN 39
15 MĀ 183 4 MN 40
15 MĀ 184 4 MN 32
15 MĀ 185 4 MN 31
15 MĀ 186 5 MN 47
15 MĀ 187 12 MN 112
15 MĀ 189 12 MN 117
15 MĀ 190 13 MN 121
15 MĀ 191 13 MN 122
16 MĀ 192 7 MN 66
16 MĀ 193 3 MN 21
16 MĀ 194 7 MN 65
16 MĀ 195 7 MN 70
16 MĀ 196 11 MN 104
16 MĀ 198 13 MN 125
16 MĀ 199 13 MN 129
16 MĀ 200 3 MN 22
16 MĀ 201 4 MN 38
17 MĀ 203 6 MN 54
17 MĀ 204 3 MN 26
17 MĀ 205 7 MN 64
17 MĀ 206 2 MN 16
17 MĀ 207 8 MN 77
17 MĀ 208 8 MN 79
17 MĀ 209 8 MN 80
17 MĀ 210 5 MN 44
17 MĀ 211 5 MN 43
18 MĀ 212 9 MN 90
18 MĀ 213 9 MN 89
18 MĀ 214 9 MN 88
18 MĀ 216 9 MN 87
18 MĀ 217 6 MN 52
18 MĀ 221 7 MN 63

Looking at this table without any further context, there are some apparent affinities between chapters in MĀ and MN, but there is nothing particularly strong that could serve as a significant core of common chapters. We see pairs and sometimes three or four sūtras that occur together in MĀ that are found in the same chapter in MN.

The strongest correlation is between Chapter 13 of MĀ and Chapter 14 of MN. They share nine parallels, which is nearly an entire 10 sūtra chapter. In addition to this, the two chapters have similar titles: MĀ titles it “Root Analysis” (根本分別) and MN simply “Analysis” (vibhaṅga). Of interest, too, is that this chapter in MN has two extra suttas appended to it to make a chapter of twelve, but these additional suttas have parallels elsewhere in MĀ.

A second correlation is less coherent but serves as evidence of an older pair of chapters: Chapters 8-10 of MĀ contain 7 parallels from MN’s Chapter 1 and 8 parallels from MN’s Chapter 2. It would appear that these three chapters in MĀ and two chapters in MN are both expansions of older chapters they once had in common. The titles and arrangements of the resulting chapters are quite different in each collection.

Another correlation is MĀ’s Chapter 16 and MN’s Chapter 7. They are both collections featuring major disciples of the Buddha, but today they only share three sūtras in common.

There are also chapters dedicated to brahmins in MN (chapter 10) and MĀ (chapter 12) that share four parallels between them. In MĀ, this chapter has been doubled to twenty sūtras in order to collect more brahmins together.

Both collections share chapters with the theme of paired sūtras, but only the first four sūtras in MĀ’s Chapter 15 are parallels with suttas in MN’s Chapter 4. Oddly enough, they parallel the first and last pairs found in the MN chapter.

The last four suttas in MN’s Chapter 9 on kings are found in chapter 18 of MĀ, which does not have the theme of kings. Chapter 6 in MĀ has the theme of kings, but it bears little resemblance to MN’s Chapter 9 and includes only two parallels to it.

Another strange case is three suttas from the end of Chapter 8 on wanderers in MN. It has parallels in chapter 17 of MĀ, which does not have the theme of wanderers. They point out the oddity in MN that the smaller and greater discourses to Sakuludāyi have another sutta inserted between them when such pairs typically follow each other, as do their parallels in MĀ (nos. 207 and 208).

Thus, we see some evidence of three chapters that were once common to both collections. The other correlations can be explained as natural coincidences based on common themes found in other EBT collections like kings, great disciples, or brahmins. Taken together, though, they could be the remnants of some older version of MN/MĀ common to both. If so, these two later versions do not preserve much of its original organization.

The Parallel Structures between MĀ and AN

This is not the end of the story for MĀ because, as mentioned earlier, it contains a large body of texts that parallel suttas found in AN. The table below shows the AN parallels and the chapters to which they belong in MĀ and AN.

MĀ Chapter MĀ Sūtra AN Chapter AN Sutta
1 MĀ 1 7.7 AN 7.68
1 MĀ 2 7.7 AN 7.69
1 MĀ 3 7.7 AN 7.67
1 MĀ 4 7.2 AN 7.15
1 MĀ 5 7.7 AN 7.72
1 MĀ 6 7.6 AN 7.55
1 MĀ 8 7.7 AN 7.66
1 MĀ 10 6.6 AN 6.58
2 MĀ 11 3.10 AN 3.100
2 MĀ 12 4.20 AN 4.195
2 MĀ 13 3.7 AN 3.61
2 MĀ 15 10.21 AN 10.217-9
2 MĀ 16 3.7 AN 3.65
2 MĀ 18 8.2 AN 8.12
3 MĀ 21 2.4 AN 2.32-41
3 MĀ 22 5.17 AN 5.166
3 MĀ 24 9.2 AN 9.11
3 MĀ 25 5.17 AN 5.162
3 MĀ 29 9.2 AN 9.13
4 MĀ 32 4.13 AN 4.127
4 MĀ 35 8.2 AN 8.19
4 MĀ 36 8.7 AN 8.70
4 MĀ 37 8.2 AN 8.20
4 MĀ 38 8.3 AN 8.21-2
4 MĀ 40 8.3 AN 8.24
4 MĀ 41 8.3 AN 8.23
5 MĀ 42 11.1 AN 11.1
5 MĀ 43 11.1 AN 11.2
5 MĀ 44 11.1, 8.9 AN 11.3, 8.81
5 MĀ 45 11.1, 7.7 AN 11.3, 7.65
5 MĀ 46 11.1, 7.7 AN 11.3, 7.65
5 MĀ 47 11.1 AN 11.3
5 MĀ 48 11.1 AN 11.4
5 MĀ 49 5.3 AN 5.22
5 MĀ 50 5.3 AN 5.22
5 MĀ 51 10.7 AN 10.62
5 MĀ 52 10.7 AN 10.61-2
5 MĀ 53 10.7 AN 10.61-2
5 MĀ 56 9.1 AN 9.3
5 MĀ 57 9.1 AN 9.1
7 MĀ 73 8.7 AN 8.64
7 MĀ 74 8.3 AN 8.30
7 MĀ 76 8.7 AN 8.63
7 MĀ 82 6.6 AN 6.60
7 MĀ 83 7.6 AN 7.61
7 MĀ 84 10.8 AN 10.72
8 MĀ 90 10.3 AN 10.24
8 MĀ 92 10.3 AN 10.23
8 MĀ 94 10.9 AN 10.87
8 MĀ 95 10.6 AN 10.53
8 MĀ 96 10.6 AN 10.55
9 MĀ 105 10.8 AN 10.71
10 MĀ 109 10.6 AN 10.54
10 MĀ 110 10.6 AN 10.51
10 MĀ 111 10.7 AN 10.63
10 MĀ 112 10.7 AN 10.62
10 MĀ 113 10.6, 8.9 AN 10.58, 8.83
10 MĀ 116 8.6 AN 8.51
11 MĀ 117 3.4 AN 3.39
11 MĀ 118 6.5 AN 6.43
11 MĀ 119 3.7 AN 3.67
11 MĀ 122 8.1 AN 8.10
11 MĀ 123 6.6 AN 6.55
11 MĀ 124 8.3 AN 8.29
11 MĀ 125 6.5 AN 6.45
11 MĀ 126 10.10 AN 10.91
11 MĀ 127 2.4 AN 2.35
11 MĀ 128 5.18 AN 5.179
11 MĀ 129 7.6 AN 7.64
11 MĀ 130 6.5 AN 6.54
11 MĀ 137 4.3 AN 4.23
11 MĀ 138 7.6 AN 7.62
12 MĀ 142 7.3 AN 7.22
12 MĀ 143 3.6 AN 3.60
12 MĀ 148 5.4 AN 5.31
12 MĀ 149 6.5 AN 6.52
12 MĀ 155 9.2 AN 9.20
12 MĀ 157 8.2 AN 8.11
12 MĀ 158 5.20 AN 5.192
12 MĀ 160 7.7 AN 7.74
14 MĀ 172 4.19 AN 4.186
15 MĀ 172 10.12 AN 10.116
17 MĀ 202 3.7 AN 3.70
18 MĀ 213 10.3 AN 10.30
18 MĀ 215 10.3 AN 10.29
18 MĀ 220 7.6 AN 7.54

As can be seen in this table, several chapters of MĀ are primarily parallels to AN, but most of the selections are fairly randomized. The most striking exception is the first chapter of MĀ, which is titled “The Sevens” and showcases a set of sūtras featuring lists or parables with seven items. Five of them are found in the seventh chapter of AN’s Book of Sevens (AN 7). What’s more, other members of AN 7’s sixth and seventh chapters are found elsewhere. A little more than half of these two chapters have parallels in MĀ (12 of 21), as shown below:

AN Chapter AN Sutta MĀ Chapter MĀ Sūtra
7.2 AN 7.15 1 MĀ 4
7.3 AN 7.22 12 MĀ 142
7.6 AN 7.54 18 MĀ 220
7.6 AN 7.55 1 MĀ 6
7.6 AN 7.61 7 MĀ 83
7.6 AN 7.62 11 MĀ 138
7.6 AN 7.64 11 MĀ 129
7.7 AN 7.65 5 MĀ 45-6
7.7 AN 7.66 1 MĀ 8
7.7 AN 7.67 1 MĀ 3
7.7 AN 7.68 1 MĀ 1
7.7 AN 7.69 1 MĀ 2
7.7 AN 7.72 1 MĀ 5
7.7 AN 7.74 12 MĀ 160

Clearly, there was a relationship between the Madhyama Āgama and its Ekottarika Āgama counterpart, but its difficult to say whether sūtras were moved from EĀ to MĀ, given that that version of EĀ is lost.

There is also the question of the 54 suttas in MN that have no parallel in MĀ: Was some portion of them moved from AN to MN? If we had the Sarvāstivāda EĀ, it might provide us with evidence of this in the form of parallels for those suttas. It’s a plausible scenario, but one with no direct evidence today. Despite that, it seems very unlikely that they are all unique to the Theravāda tradition. It would also be interesting to compare MN to a Dharmaguptaka MĀ, but that collection is also lost.

Conclusion

The picture we are left with is more tenuous than when we compared the organization of DN/DĀ or SN/SĀ: There is a common core of parallels without any significant parallelism in organization, and there is evidence of large scale alterations.

The lack of organizational parallels between MN and MĀ makes it difficult to see a presectarian version preserved in them beyond a set of common sūtras. If we were to theorize that the entire collection was created from other collections, we would need to explain the core set of MN/MĀ parallels. Some MN suttas also have AN parallels, but only a handful of them. This leaves us with no direct evidence of the origin of the remaining MN suttas that lack any parallels. They may have belonged to an original presectarian middle-length collection that has been almost completely reorganized over time.

MĀ certainly looks like a good candidate for the reorganization scenario, given how many AN suttas parallel its content compared to MN. Being the smaller collection, MN seems more likely to be the older version of the two, but that doesn’t mean it hasn’t been reorganized as well. We would need more parallel EBT collections to decide which of them preserves an ancestor collection better (if either of them still does).

  1. Typically, a page of Taisho prose yields 1,000-1,400 words of English. ↩︎

  2. This number is less than adding the MN, AN, and DN parallels together because some sūtras have parallels in both MN and AN. ↩︎

  3. One sūtra, MĀ 159, appears to be a Buddhist version of a section of the Br̥hadāraṇyaka Upaniṣad. ↩︎

12 Likes
2

Having read carefully your presentation on the structure of the organisational parallels, I suggest you may delete the word “older”.

This is because the organisational parallels do not have any direct connections about older/earlier or later issues of the particular collection “Nikaya/Agama”. It is about the issues of how an original or ancestor collection should have been arranged or structured.

Regarding older/earlier or later issues, one needs to study the formation of the collections, such as the connection between Nikayas/Agamas and Angas, or the formation of the individual texts.

Note: The reconstruction of the sequence of the Saṃyukta-āgama in T 99 by Yin Shun is not about older or later of the collection. It is about the issues of how the original collection in T 99 should have been arranged.

3

Yes, correct, it is about the issues of an ancestor collection.

1 Like
4

That’s what I meant by “older” – a middle-length collection of sutras that predates the two we have. The sutras themselves would have to be evaluated on a case by case basis, and I’ve only looked at a sample of them closely myself.

5

If so, instead of “older collection” or “older presectarian collection”, I suggest it is clearer to say older or earlier arrangement, structure, or organisation of a collection “agama/nikaya”, such as a middle-length collection of sutras - MA/MN.

This is because the expression, an “older collection” “older presectarian collection”, could just mean an “older agama/nikaya” “older presectarian agama/nikaya”.

1 Like
6

Yeah. Well, when I said:

They [the suttas in MN without parallels] may have belonged to an older middle-length collection that has been almost completely reorganized over time."

I didn’t necessarily have a presectarian collection in mind. The Majjhima Nikaya may have been reorganized at some point in its lifetime since it became its own canonical lineage. It has existed as part of the Theravada tradition for over a thousand years. Their tradition asserts that it’s been over two thousand years. A lot can happen in thousands of years of human history. Wars, plagues, religious persecutions, natural disasters, cultural change - these are all forces that made maintaining a scriptural canon unchanged difficult at times.

So, yeah, it’s complicated. Sorting out relatively recent history (say, from 1000 to 1900 AD) from classical times (say, from 0 CE to 800 CE) and that from truly ancient times (500 BCE to 0 CE) is tough to do when a tradition doesn’t have any records about it and just says, “This is what the Buddha said.”

1 Like
8

Okay … I’m not sure what you had to say in reply (apparently it was such that someone decided to flag it?), but I did reconsider the wording and changed it. For the suttas without parallels, I guess the real issue is whether they might go back to an original version of the collection or not, so I changed it to “original presectarian collection.”