An Introduction to the Middle-Length Agama

The Middle-Length Agama (MA) is Taisho No. 26, which spans 60 fascicles and 388 pages of the Taisho Daizokyo, Vol. 1. To put this in perspective, a page of Taisho prose typically translates to 1,500 words of English, so a complete translation of MA will probably be 600,000-700,000 words in length. In addition to this complete translation, there are over 70 individual translations of sutras found in MA that still survive as Taisho Nos. 27-98.

History of Chinese Translations

The history of the Chinese translation of MA is well documented in BDK English Tripiṭaka Series’s volume 1, which was published in 2013. It appears that MA, as it survives today, was actually the second complete translation to Chinese. The first was a collaboration between an Indian bhiksu named Dharmanandin and Zhu Fonian in the 384-5 CE. A second translation was produced by a team headed by Gautama Samghadeva only a decade later in 397-8 CE. Chinese records suggest both translations existed for several centuries before the first was lost.

As late as the 4th c. CE, Indian Buddhist texts were still preserved in memory rather than in writing, and we see direct evidence of this in the Chinese descriptions of translation teams. Typically, a bhiksu who had a text memorized recited it in the original language, a bilingual bhiksu then provided a Chinese translation verbally, and scribes wrote down the translation. Thus, the lingual and writing skills were typically divided between two or more people, which makes sense given the difficulty involved. Literacy in Chinese, for example, was a skill separate from the spoken language given that writing was  ideographic rather than phonetic.

Comparing the Agama to the Nikaya

The Middle-Length Agama is organized in a similar way as the Majjhima Nikaya (MN) in the Pali canon, but it contains more texts arranged in a different order. MA is a collection of 222 sutras in five divisions, while MN contains 152 suttas in three divisions. Each division is subdivided into thematic chapters that typically consist of 10 texts each, though MA has a couple larger chapters ranging from 15 to 25 texts.

It’s clear that MA was recited in a program that lasted five days because its divisions are called Recitations titled “First Day Recitation” and so on. These recitations appear to divide the collection into roughly equal segments (60-100 pages), and they sometimes split chapters for this purpose.

When we compare the texts in MA to MN, it quickly becomes clear that the two collections substantially overlap, but about 65% of MA’s texts either have no equivalent in the Pali Canon or they are found in a different Pali collection. Unlike MN, which doesn’t contain brief suttas, MA has many sutras that correspond to texts in the Anguttara Nikaya. Scholars have made efforts to index the parallels between the Agamas and the Nikayas, but it becomes a complex task when we also consider Sanskrit and other Chinese sources. Indexing the major parallels is complete, but the indices aren’t 100% comprehensive yet. At this point, it appears that 10% (23 sutras) of MA is unique, lacking any surviving parallels in other Agamas or the Pali canon.

Actual side-by-side comparisons of MA and its Pali parallels are difficult to generalize beyond the fact that many are clearly variations of the same text. In some cases, one text seems stripped down and simple compared to the other; or they only share certain sections and not others. Numerical lists usually differ only in order, but the actual definitions can be quite different.

Thoughts about the Origin of Variant Sutras

Most of the texts in MA are single topic presentations that center around a parable or numerical list of ideas. Originally, they were lectures given as part of the Buddha’s attempt to educate his students from day to day. These lectures were memorized by bhiksus in attendance and recited regularly to preserve them, becoming a canonical oral tradition after his death. It seems reasonable to assume a couple things about these original lectures during the Buddha’s time.

The Buddha himself probably repeated some lectures more than once to different audiences, and he may well have varied the presentations from one occasion to the next. He traveled from place to place in a repeated circuit for several decades. Thus, some of the variations we see were likely different versions of the same lecture from the beginning. Two variant sutras that claim to have been taught at the same location may have been taught there during different visits, possibly years apart.

As the Buddha gained followers, they spread out across ancient India. Some traveled with him, but more and more probably did not. Lectures that had been memorized at a given teaching were recited in his absence and “copied” into the memories of other bhiksus. Different versions of the same lecture likely were created when recitations weren’t exact or portions of a lecture were forgotten and recreated. This might account for slight differences and texts with the same structure but very different content.

Evidence of Recall Errors and Correction in Early Sutras

One of the common differences I see between versions for a sutra are numerical lists that differ in order. The beginning and end of a list will usually be the same in each version, but the items in the middle of the list will be jumbled. This to me is evidence of recall errors. The beginning and end of lists or narratives are easier to recall than the order of items or events in the middle of them.

Indeed, some texts are structured in a way to make these kinds of recall errors less likely. As written documents, they read in a repetitive and stilted way, but the purpose was to provide error detection. An example is a list that incrementally adds more detail to each item. It resembles a common way to memorize a long series of numbers:

  • 56
  • 56-87
  • 56-87-90
  • 56-87-90-34
  • 56-87-90-34-76

When we recite this series, it’s easy to realize where something was forgotten, and we can reconstruct a forgotten portion by going back to an earlier item. The Water Parable Sutra is a good example of this method of building error correction into a text. Its items are recited like this (to paraphrase):

  • A person lays in water
  • A person comes up and looks around
  • A person comes up, looks around, and stands
  • A person comes up, looks around, stands, and wades across
  • A person comes up, looks around, stands, wades across, and reaches the other shore
  • etc.


The upshot is that when we take all of these early texts in Pali and Chinese as a whole, we get a sum total of surviving lectures by the Buddha, however much they had evolved over the centuries before they were codified as written canons. I don’t think we can know for certain which are more original than the others when we group the parallels together. In some cases, a larger version might have been an amplification; in another case a smaller version might have been restored from an incomplete memory. In still other cases, we get two or more opinions about what something means.

As I continue my translation efforts, I intend to collect the versions of each sutra into a single document as individual studies. I plan begin working out a good format for this next month.

The Place of the Chinese Āgamas in Buddhist Literature

The Chinese Buddhist canon is unique in that it contains a record of the scriptures and philosophical texts that were circulating in Central Asia and India during the peak of Buddhist literary creativity between 0 CE and the arrival of Islam. Most of the major works were translated several times over a period of about 500 years or so, meaning that we can note their historical development by comparing these “snapshots” preserved in Chinese. It’s also thanks to Chinese historical records that we can date many of these texts.

The Chinese Āgamas are an important piece of the historical puzzle of Indian Buddhist history. Most of the original Sanskrit or Middle Indic canons that the Chinese Āgamas preserved were lost when Buddhism was disbanded in India. The only complete early Buddhist canon that still exists is preserved in Pali by Theravada Buddhists in countries outside of India. To my knowledge, we otherwise only have fragments of original language Āgama texts discovered by archeologists and quotations in secondary works like Abhidharma and Mahāyāna texts.

Like the Theravada Buddhists, the Indian schools of early Buddhism maintained a canon of five divisions:

  • Dīrgha (Long)
  • Madhyama (Middle)
  • Saṃyukta (Thematic)
  • Ekôttara (Incremental)
  • Kṣudraka (Minor)

These are Sanskrit equivalents of the five Pali Nikāyas: Dīgha, Majjhima, Saṃyutta, Aṅguttara, and Khuddaka. Each Āgama or Nikāya is a collection of individual texts that relate a teaching by the Buddha or a chief disciple.

These collections had their genesis from the initial sermon that the Buddha gave to his first five disciples. Throughout his teaching career, the practice of memorizing teachings and events and preserving them as an oral tradition built up a living canon that continued on after his death. It was several centuries later that these oral traditions were set down in writing. I say oral traditions because the Chinese Āgamas suggest that, by the time they arrived in China (circa 4-5th c. CE), each major branch of early Buddhism had an alternate version of the original oral tradition.

Scholars have begun to compare them to the Pali and to fragments of Indian works that exist. What they have found is that while the basic teachings and historical events are shared between each canon, they are seldom identical and vary a great deal in presentation and wording. Indeed, there are even differences in understanding the meaning of basic words because of Sanskrit rather than Pali derivations.

What has been striking to me is how certain parallel texts are very close in wording when we compare the Chinese to the Pali, while others are very different. There is also only a partial overlap: That is, that the Nikāyas have texts that are absent from the Chinese Āgamas and vice versa. It’s clear that the early Buddhist canon was not a word-for-word preservation like the Christian Bible, but rather a fuzzy body of literature handed down through multiple lines of transmission.

Efforts to translate and publish the Chinese Āgamas to English began during this decade, most notably with publications by the BDK translation project and also Dharma Drum Buddhist College. I had explored the Āgamas in the early 2000s during a period of study and translated a few of them in 2004. I am taking this effort back up because of the need for a careful rendering these difficult texts to English. It will give scholars and practitioners alike a way to get a better understanding of early Buddhism as a whole.