The argument is quite subtle, but the main point is this.
Asanga in his Yogacarabhumisastra spoke of how the texts were organized primarily based on the Samyukta pattern. (Asanga is technically a Mahayanist, but the relevant section deals only with EBTs, and Asanga in general is sometimes characterized as an “Agamist”; he only introduced Mahayana texts when talking about specifically Mahayana ideas). Asanga’s root Agama texts were from the Sarvastivada ( or a closely related subschool).
The Samyutta, according to him, is comprised of the first three Angas: sutta, gāthā, veyyākaraṇa.
Yin Shun took his ideas seriously and used them to analyse the texts, hypothesizing that this threefold system was the earliest way of organizing the texts. Yin Shun’s analysis has become extremely influential in Taiwanese Buddhism, and among those interested in this area it is virtually the orthodoxy.
The success of the analysis is not just because it has a basis in old texts, but because it enables a clear and informative way of understanding the formation of the early text collections.
In my A History of Mindfulness I developed Yin Shun’s ideas further, especially challenging the identification of the veyyākaraṇa Anga. Yin Shun followed Asanga in identifying this with the minor samyutta collections. However, I believe that it means the discourses in “Q & A” format within the major samyuttas (and of course elsewhere).
This analysis allows us to make sense of many odd characteristics of the early texts. For example, the Bala Samyutta in SN is almost residual; the texts are little more than repetition series. The expected substantial texts on the balas are instead found in AN. But—and this is just my memory!—SA has more substantial section on balas. So the hypothesis would be that at an earlier time (as recorded in SA), the substantial bala texts were found in the samyutta, then in the Pali tradition were later removed to AN (presumably because the doctrinal content is similar to the Indriya Samyutta, so it is more useful for the AN reciters to learn it).
My hypothesis was that SA generally speaking retained a structure that remained closer to an ancestral samyutta. A few caveats:
- this doesn’t tell us anything about the age of the content, only the structure in which the content was placed
- there doesn’t have to be just one ancestral collection; on the contrary there will have always been some variation (as there is even within the Pali collection today!)
- it doesn’t mean the structure of SA is always older.
I don’t regard this analysis as proven; I think it’s likely, and a statistical analysis of the structural variations might serve to either confirm or deny it.
I went into some detail as to the relations between chunks of texts in SN and SA, and showed that the structural similarities were not merely on a samyutta level, but often on a vagga or sutta level. This analysis of the Samyutta was largely informed by Rod Bucknell’s work, itself based on Yin Shun.
Under this analysis, the cores of AN, MN, and DN were originally seeded from texts found within the Samyutta. This makes sense in cases like, say, the first chapter of DN, where all the texts have the same theme (threefold training), a major topic that is not found in the samyutta.
However there are many texts and features of the collections that are not explained by this hypothesis. For example, how do the long narrative suttas fit it, especially those that include discussions on many different topics? Or what of the many numerical suttas in AN that deal with minor miscellaneous matters?
Clearly it is not a complete theory, but it is a way to get a handle on how the canon may have been formed.