The Selection Bottleneck in
Early Buddhist Textual Transmission

Pre-Transmission Loss in Early Buddhist Orality

 

The Overlooked Problem of Pre-Transmission Filtering

An absolutely crucial point that seems to get glossed over in traditional accounts and even in much critical scholarship is this: the selection process to decide what would later be memorised in oral transmission had many difficulties. What follows draws entirely from the Pāli Canon itself—the same source used to support traditional views—but follows its logical implications to conclusions that challenge those established interpretations. Once considered, those difficulties immediately overturn the standard presentation of oral transmission. Chief among them being the overlooked implication:

The process of selection requires candidates from which to select.

There was no paper in use at the time so all candidates had to be recited. Candidates included not just those that would be selected, but also those that would be rejected. That means, in order to be considered, they had to be recited out loud which in turn meant they each had been memorised. There had to be a serious commitment already invested before rejection. The uncomfortable fact is that each individual candidate had already been memorised by saṅgha who believed it was definitely authentic. No one would memorise something they thought was false.

This observation highlights a fundamental gap in how we think about the formation of the Buddhist canon. Discussions of oral transmission typically focus on preservation mechanisms and degradation over time, but pay insufficient attention to the initial gatekeeping moment when suttas were being selected.

This also makes it clearer why the standard scholarly focus on “transmission fidelity” misses the point—the real epistemological problem occurred before the transmission pipeline even began.

The Logical Paradox of Selection

This creates a fascinating logical problem. If we accept that some filtering occurred (and it must have done, even without accepting the First Council narrative), then:

1. Multiple monks had memorised different versions or different teachings
2. Each monk who memorised something presumably believed it was genuine Buddha-word
3. Some of these memorised versions were rejected
4. The people whose versions were rejected must have then been convinced (or coerced) to believe that their strongly held memories were wrong

The council legend makes this problem even worse. The traditional narrative claims that only arahants were present at the First Council—500 of them according to the story. This device was meant to guarantee the authority and reliability of the selection process. But if only arahants were present, that means every candidate for consideration was memorised and recited by an arahant, which in turn means that some arahants had memorised teachings that were ultimately rejected.

So the story’s attempt to create authority shoots itself in the foot: perfect arahants, with their supposedly flawless memory and direct realisation of dhamma, had memorised candidate suttas that were rejected as wrong. Either these arahants had faulty memory (undermining the basis for trusting any of them), or they correctly remembered teachings that the assembly chose to exclude for other reasons (undermining the claim that the council simply “preserved what the Buddha taught”).

The Psychological and Social Dimensions

This is psychologically and socially remarkable. Arahants are being asked or required to abandon something they’ve previously committed to memory as sacred teaching, that had previously formed part of their practice and understanding.

The traditional council narrative obscures this conflict by presenting a harmonious gathering of arahants who simply “recited” what they remembered. But any actual selection process would have involved dispute, negotiation, and the exercise of power—whether through persuasion, authority, or political dominance within the early saṅgha. Alternatively, if there was no actual selection, there was no need for the Council—it was purely symbolic and the Council was an invention.

The Impossibility of Objective Criteria

But if there really were candidates, the criteria problem is even more interesting. What could the basis for selection have been when there was no external record to compare against? It couldn’t have been “accuracy to the original” in any objective sense. It must have been something like:

• Majority agreement (it’s still just the opinion of the majority and why would that be more reliable?)
• Authority of the reciter (prestige, seniority, and not consistent with Vinaya rules)
• Internal alignment with other accepted material (but this is circular—selection is based on a previous selection—who made that selection authoritative?)
• Doctrinal palatability to the dominant faction (which raises questions about whose Buddhism survived)

Each of these criteria is epistemically problematic when there’s no baseline text to verify against. The circularity is particularly troubling: teachings could be rejected for not fitting with accepted teachings, but the “accepted teachings” were themselves part of the same contested selection process.

Before Parinibbāna:
Dispersed Communities and Local Traditions

There was never one entry point of chosen texts to be memorised, as saṅgha communities were well dispersed before the Buddha’s death and would have chosen what suited their specific purposes.

This is not theoretical—the Canon itself provides direct evidence. The Mahāparinibbāna Sutta (DN 16) shows the Buddha and Ānanda travelling to visit various established saṅgha communities in different locations. The text depicts functioning, autonomous communities that existed independently during the Buddha’s lifetime. There is no indication that this situation is anything but normal to him. Elsewhere, the Buddha is shown visiting groups in Rājagaha, Nālandā, Pāṭaligāma, Vesālī, and numerous other locations, often giving teachings specific to those communities. Saṅghakarma in the Vinaya indicates that self-governing processes existed for this purpose before parinibbāna.

Multiple independent memorisation traditions developed in different regions during the Buddha’s lifetime. Communities in Kosala, Magadha, Vajji, and elsewhere were memorising and transmitting teachings locally, shaped by:

• Which monks happened to be present in that area
• What teachings those particular monks had heard directly
• What teachings served the needs and questions of local lay supporters
• The doctrinal emphases and practice preferences of local senior monks

No mechanism for coordination existed to ensure these geographically separated communities were memorising the same material. By the time of the Buddha’s death, multiple established oral traditions were already functioning, each with its own authority structure and each convinced of its authenticity.

There is no indication in the Pāli Canon that the Buddha wanted formal scriptures prepared. Everything in the Canon points the other way. He favoured understanding over words.

After Parinibbāna: The Problem of Lost Decades and Selection

Things changed after he died. At the point of selection for oral transmission, we cannot know the criteria on which material that was already memorised was rejected or accepted.

The point about the initial loss of rejected candidates is particularly strong: Even if oral transmission preserved what entered the pipeline, the real haemorrhaging happened before anything entered that delivery pipeline. There is no sign in the Canon that any of the Buddha’s teachings from the first 20-25 years were preserved.

It is traditionally held that Ānanda perfectly recalled the last 20 years when he was the Buddha’s personal attendant. But nothing in the Pāli Canon even remotely supports the idea that Ānanda had perfect recall. In fact, several suttas imply gaps or need for repetition. Ānanda asks for clarification or repeats questions to the Buddha. Certain discourses say “he did not understand until explanation”.

Where would the recall of 45 years of teaching come from? There is no plausible human mechanism by which forty-five years of dispersed, situational teaching could be recollected as a unified body. None is flagged in the Canon. The first two decades of the Buddha’s teaching career would have been known only through monks who may have died, dispersed, or never joined any formal recitation effort.

Two Stages of Loss

This suggests we should think of early Buddhist textual loss in two distinct phases:

Phase 1: Pre-memorisation loss
The teachings from the Buddha’s first two decades that no living monk had heard, the teachings heard by monks who died before any systematisation, the teachings given in contexts where no one present later joined the recitation effort, and variations that were simply forgotten before anyone thought to preserve them systematically.

Phase 2: Selection loss
Material that had been memorised and transmitted but was actively rejected during whatever process(es) established the early recitation traditions. This is arguably more significant because it represents teachings that survived initial memory but were deliberately excluded.

Evidence from Gāndhārī and Other Fragments

The Gāndhārī texts are indeed evidence of this “post-pipeline” diversity—teachings that someone thought worth writing down but that didn’t survive in later traditions. They show that the post-oral textual universe was once much larger.

Similarly, the texts preserved in Chinese translation that have no Pāli equivalent, the fragments found at various archaeological sites, and the divergences between early Buddhist schools all point to a much more diverse textual landscape in the centuries after the Buddha’s death. Traditions did not survive. What we have now represents a winnowed selection, not nearly a complete archive.

There is the idea of a communal choice that involves loss of unfavoured material as traditions evolved (whose content could never be known), but Gāndhārī shows texts that were still in use post-transmission but were later lost.

The Gāndhārī evidence is particularly telling here. These texts show sustained transmission—they were important enough to be written down centuries after the Buddha’s death, which means they had continuous oral and/or written transmission. Yet they didn’t survive into any modern tradition. This demonstrates that texts could:

• Enter oral/written transmission successfully
• Be maintained for centuries
• Still ultimately disappear

This wasn’t “failed” transmission in the sense of degraded memory. These were texts that were successfully transmitted but whose transmission lineages eventually died out, likely when the communities maintaining them disappeared or merged with other groups.

Decentralised Selection, Decentralised Loss

If we abandon the fiction of a central council, we get a much messier but more plausible picture:

Different communities developed their own collections based on local needs and available teaching. Some communities died out entirely, taking their unique collections with them. Some communities merged, with dominant groups imposing their textual preferences. Some texts were maintained in multiple communities and thus had redundancy. Some texts were unique to single communities and vulnerable to loss.

Geographic factors mattered enormously. A teaching transmitted only in the northwest could vanish entirely if Buddhism declined in that region, while a teaching known across multiple regions had better survival odds. Political changes, patronage shifts, and economic disruptions could wipe out entire local traditions.

This challenges the entire notion of a central selection moment. The council narrative implies a gathering where “the” canonical texts were determined, but the textual and geographic reality the Canon itself presents suggests an environment that is something quite different.

This means the Canon we have represents not “what was chosen” but “what happened to survive” in the communities that maintained continuous institutional existence until the tradition was eventually written down and standardised.

Questions This Raises

This analysis opens several important questions:

• What were the actual mechanisms of selection in multiple regional communities, and how did these differ?
• To what extent did doctrinal concerns (what the Buddha “must have” taught) influence what was accepted as authentic memory in different locations?
• How much of what we consider early Buddhism reflects the views of particular monastic communities that happened to survive, rather than a historical individual’s teaching?
• Can we identify traces of excluded material in the canonical texts themselves—references to teachings or practices that seem disconnected from their current context?
• What role did geographic concentration play in textual survival—did texts known across multiple regions have better survival odds?

The fact that anything like consensus emerged at all is remarkable, precisely because those presenting rejected material must have believed in their own versions, and because there was no central authority to impose uniformity. Understanding how that partial consensus developed across dispersed communities—or how certain regional traditions came to dominate—tells us something crucial about how religious traditions form and what “authenticity” meant in that context.

The Scale of Survival: A Quantitative Reality Check

One further point sharpens the entire problem of selection and loss and deserves to be made explicit.

If we treat the Buddha as a historical teacher operating over approximately forty-five years, even extremely conservative assumptions about how often he taught lead to a striking conclusion: only around four hours of teaching per year have survived in the Pāli Canon.

The four Nikāyas, when read aloud at an ordinary speaking pace (including their extensive repetitions), amount to only a few hundred hours of speech. Set against several decades of active teaching, instruction, dialogue, correction, and debate, this represents a vanishingly small fraction of the Buddha’s spoken output. Even generous adjustments to the estimate do not alter the basic picture: well under one percent of what was spoken can plausibly be said to survive.

This fact alone makes it impossible to think of the Canon as anything like a comprehensive record. The overwhelming majority of teachings were never preserved, regardless of how faithfully later material may have been transmitted. Loss is not an unfortunate by-product of oral transmission; it is the dominant feature of the historical situation. This means that selection pressure, not transmission accuracy, is the primary explanatory variable in the formation of the Canon.

Once this scale is acknowledged, the problem of selection becomes unavoidable. The question is no longer how accurately teachings were transmitted, but why these particular teachings survived at all, while tens of thousands of hours of equally sincere, equally authoritative instruction did not.

The Pāli Canon itself shows that the Buddha anticipated loss, degeneration, and the emergence of new teachings. In DN 16 he outlines how later teachings are to be evaluated against the Dhamma he taught, implicitly acknowledging that variation and distortion would occur. What appears to have survived this process is not a comprehensive body of instruction, but a small set of recurring commitments: impermanence, dependent arising, non-self, suffering, and compassion.

What millennia of winnowing seem to have preserved is not doctrinal completeness, but a persistent resistance to reification—something that can be realised through investigation. Results generate confidence in further inquiry and practice, whereas absolute belief rests on confidence in the revealer, himself.