16 February 2024

History as Practiced by Philologists: A Response to Levman's Response to Drewes.

In 2017, David Drewes published an article that is now famous or infamous, depending on your viewpoint. Drewes argued for the thesis that we cannot connect the Buddha to any historical facts and concludes that historians should stop referring to "the historical Buddha". His article has no abstract, so let me cite a passage from his introduction that seems to sum up his argument:

On one hand, the Buddha is universally agreed to have lived; but, on the other, more than two centuries of scholarship have failed to establish anything about him. We are thus left with the rather strange proposition that Buddhism was founded by a historical figure who has not been linked to any historical facts, an idea that would seem decidedly unempirical, and only dubiously coherent. Stuck in this awkward situation, scholars have rarely been able to avoid the temptation to offer some suggestion as to what was likely, or ‘must’ have been, true about him. By the time they get done, we end up with a flesh and blood person – widely considered to be one of the greatest human beings ever to have lived – conjured up from little more than fancy. (2017: 1)

When Drewes says that "the Buddha is not linked to any historical facts", he means that there is no contemporary documentation of the Buddha. There are no eyewitness accounts of the Buddha, and there are no contemporary coins, inscriptions, or documents of any kind. There was no writing anywhere in India prior to the mid-third century BCE. This is indisputable. However, Drewes' article has engendered much disputation, of which Bryan Levman's (2019) response in the Canadian Journal of Buddhist Studies is a prominent example.

In this essay, I will review Levman's (2019) response to Drewes. I will let Levman introduce his own argument. The abstract of his article says:

This article is a response to David Drewes’ hypothesis (2017: 1-25) that the Buddha was a mythic figure who did not necessarily exist as a historical fact. The article suggests that there are four criteria by which the Buddha’s historicity can be established, none of which were discussed by Drewes: 1) the historical facts presented in the Buddhist canon which are corroborated by non-canonical sources, 2) the fact that there is no plausible alternative explanation for the provenance of the teachings 3) the humanness of the Buddha as presented in the canon belies the purported mythologization which Drewes asserts and 4) a core biography of the Founder can be discerned in the Buddhist canon, once later interpolations are removed.

Bryan Levman is an expert philologist who has specialised in the history of language in India. He has quite a chequered past, however, we are concerned here with his writing as a Buddhologist. 

One of the notable features of Levman's vehement disagreement with Stefan Karpik over what language the Buddha spoke was that, amidst deploying abstruse arguments and accusing each other of incompetence, neither of them expressed any doubt whatsoever about the Buddha as a historical character. They both took the historicity of the Buddha for granted.  

Both Drewes and Levman reference "historical facts" but in retrospect it's clear that they are using this phrase very differently from each other. Indeed, I would say that they are operating in quite distinct epistemes. So my first task is to define a "historical fact". I will take a historian's view of this issue. 

Historical Facts

Historians argue about methods and aims a great deal but they all broadly agree that history is the study of people and societies through documents. As historian, John Vincent (2006: 9), puts it:

"Historical study requires verbal evidence, with marginal exceptions. And this verbal evidence, with all respect to the fascination of oral history, is nearly all written evidence."

Documents are defined as broadly as possible. Any form of written evidence can be considered, including coins and inscriptions. Vincent (2006: 10), again, says: "History is about literate societies, and strongly tilted, at very least, towards literate people in literate societies".  Richard J. Evans (1997: 75) cites Sir Geoffrey Elton's definition:

A historical fact was something that happened in the past, which had left traces in documents which could be used by historians to reconstruct it in the present. 

Evans (1997: 76) notes that this view was expressed in direct contrast to E. H. Carr's view that "a past event did not become a historical fact until it was accepted as such by historians." Carr's view turns out to be untenable since he confuses "fact" with "evidence". This gives us a useful distinction: a fact is something that happened, and evidence is an attempt to use that fact to argue for a particular view of history. Evans (1997: 80) again:

What is at issue, therefore, is how historians use documents not to establish discreet facts, but as evidence for establishing the larger patterns that connect them. 

A "historical fact", then, is a documented fact. To be a historical fact about a particular time requires that the document be authored by someone who lived at that time. In effect, historical facts are eyewitness accounts preserved in documents. Determining the veracity or trustworthiness of such accounts is bread and butter for historians. 

Alexander Wynne (2019: 100) suggests that "Good evidence for the Buddha would perhaps be his mention in a non-Buddhist document from the fifth century BC." This is an example of someone confusing "facts" and "evidence". To provide us with facts about the Buddha, presuming he lived in the fifth century CE, a document must be from the fifth century BCE. Wynne admits that no such documents exist. If he were a historian he would admit that the absence of documents of any kind means that we cannot write a history of India in the fifth century. We have to step aside and let archaeologists and anthropologist do their work. Wynne continues to argue sans any relevant facts for another fifty pages. 

NB: Historians don't typically refer to facts or evidence as "good" or "bad". A fact may be true or false, but not "good" or "bad". Similarly a fact may constitute "salient" or "relevant" evidence for a particular argument or not. 

Importantly for this discussion, an inference is not a fact. At best an inference is an interpretation based on a fact or facts. Moreover, logical inferences are validated or invalidated against sets of axioms. It's all too obvious that for Levman, Karpik, and Wynne, the existence of the historical Buddha is axiomatic. Each of their projects is tendentiously seeking to prove what they take on faith. And each erroneously takes their own inferences, validated against their article of faith, to be "historical facts". 

Long before Drewes joined the fray, historian Jonathan S, Walters (1999: 248) wrote:

I think it fair to say that among contemporary historians of the Theravāda, there has been a marked shift from attempting to say much of anything at all about "early Buddhism". Whereas earlier scholars tended to ignore post-Aśokan Buddhist history as corrupt, more recent scholars have tended to regard early Buddhism as unknowable.

The Buddha lived in a pre-literate society and thus in a prehistoric society. A history of a pre-literate society or person is a contradiction in terms. In the context of history as a field or discipline, what Drewes says is entirely uncontroversial and in keeping with the theory and methods of modern historiography. (Note: I take historiography to mean "the act of, and methods used in, writing of history")

It is surprising that anyone who knows anything at all about historiography would take issue with this. It turns out that those who disagree with Drewes don't seem to know about historiography. In my conclusion, I will offer a possible explanation of why philologists and linguists, in particular, might disagree with historians' definition of "historical facts". However, we have first to address Levman's attempts to prove Drewes wrong.

Levman's Arguments Against Drewes.

1. Corroboration.

Levman's first objection is "the historical facts presented in the Buddhist canon which are corroborated by non-canonical sources". Leaving aside, for the moment, the problem of what, if any "historical facts" are presented in the Buddhist canon (and when they refer to), let's look at Levman's examples of corroboration:

"The Asokan rock edicts for example, contain numerous references to the Buddha, the earliest going back to shortly after his coronation in 268 BCE." (28).

However, even if the Asokan corpus does refer to the Buddha, it was composed after 268 BCE. Most scholars guess that the Buddha died around the year 400 BCE (see Norman 2008: 50-52) but this is far from certain and in conflict with all the existing Buddhist traditions which place his death at 486 CE or earlier. The Asokan documents reflect a view from a least 170 years after the putative lifetime of the Buddha (possibly considerably more). This is not an eyewitness account or even a second-hand account. Something that no one seems to have remarked on is that, by the time the edicts were composed, Asoka was a Buddhist convert who appeared to have a certain amount of convert zeal

The Asokan edicts are not evidence of the historicity of the Buddha. At best, they reflect beliefs about the Buddha from a later period, as expressed by a latter-day Buddhist convert, who dedicated his early life to brutal wars of conquest and had a lot of bad karma to make up for.

This is a clear example of Levman making a hypothetical inference based on the Asokan corpus and treating his inference as a "historical fact" based on his pre-existing belief in the historicity of the Buddha. By the way, no one argues against the idea of a community of Buddhists existing in the third century BCE. This is a historical fact. Levman's (2019: 29) next argument is:

The presumed historical existence of the Buddha is reflected in many of the early suttas where the Buddha is situated in actual historical places alongside real historical figures.

Note the phrase "presumed historical existence of the Buddha". This is Levman's presumption, not mine. As an example, he continues:

"We know, for example, from other sources, that the kings (Ajātasattu, Bimbisāra, Pasenadi) the Buddha meets with were real historical figures."

It is simply not true that Ajātasattu, Bimbisāra, and Pasenadi are historical figures. As with the Buddha, there are no contemporary documents connected with any of these names. As kings, they left no trace of historical evidence, because there was no writing in India at that time. Given this, Levman's attempts to back up his assertion are surprisingly half-hearted. For example, Levman casually mentions references to Bimbisāra and Ajātasattu in "Jain texts" (without any citation). However, this is to completely ignore the history of Jain literature. Johannes Bronkhorst (2020) comments:

Our most important sources of information regarding early Jainism are found in the canon preserved by the Śvetāmbara Jains. Unfortunately, this canon was given its definitive form at a late date, some 980 years after Mahāvīra according to a Jain tradition, that is, 454 or 514 CE.

The Jains themselves tell us that all of their early literature was lost and then later reconstructed. Jain literature is all considerably younger than, and owes a difficult-to-quantify debt to, Buddhist literature. Bronkhorst (2020) again:

We have already seen that the Sūtrakṛtāṅga Sūtra is one of the oldest texts contained in the Śvetāmbara canon. However, the contents of even this relatively old text date from long after Mahāvīra. This is clear from the following: the Sūtrakṛtāṅga Sūtra shows acquaintance with the innovations that had taken place in northwestern Buddhism in the 2nd century BCE. 

Jain literature can tell us nothing at all about the putative lifetime of the Buddha because, although it mentions events in the past, it was written down much later even than Buddhist literature. Moreover, the references Jain literature makes to the Buddha are vague. As Bollée (1974: 27) says:

It is only in the post-canonical period, and especially when the Jains begin to write in Sanskrit, that in our sources the railings at undefined opponents with more or less ambiguous statements about their views make way for more concrete philosophical arguing with different schools, among whom the Buddhists gradually come to the front to such an extent that śākyādayaḥ as a comprehensive expression for various heretics becomes dominating.

Similarly, Levman cites "Sanskrit genealogies... [in] Purāṇas, and so forth." Levman does not give an example from or even the name of a Purāṇa text, so it's difficult to know what he is referring to here. As far as the Purāṇa literature goes, it is impossible to accurately date the composition of any given Purāṇa text. The most plausible dating scenarios suggest they were composed well into the Common Era. 

So Levman's examples "corroborated facts" are not factual and are not corroborated. And the whole article follows this pattern. 

Levman goes on to discuss stories from various suttas as though they were evidence of historicity, but we've already seen that historians have long considered this to be folly. The suttas are not documents from the fifth century BCE. At best they reflect beliefs from the late first century BCE, but more likely even later. That idea the suttas reflect an earlier time is not a fact, it is an inference. Inferences about the past are not historical facts. 

There is another caveat here. The oldest extant Pāli document of any kind is a partial manuscript from the fifth century CE (Stargardt 1995). The next oldest is a fragment from the ninth century. There are no Pāli manuscripts from the first century, though there are Gāndhārī texts from that period. 

The idea that the Pāli texts were written down in the first century is based on uncorroborated claims made in the Dīpavaṃsa and Mahāvāṃsa, which are relatively late texts composed in Sri Lanka. The Mahāvaṃsa (33.100) states that the canon and its commentaries were committed to writing in the reign of King Vaṭṭagāmiṇi (29-17 BCE) at the Alu-vihāra in Sri Lanka.

The authors of the Mahāvaṃsa lived thousands of miles and hundreds of years distant from the events they purport to describe. Moreover, as Jonathan Walters (2000) has pointed out:

Scholars who have treated the Vaṃsas as history have ignored the indications that they were written within (and should be understood within) a temporal and causal framework different from that which we know in the modern West.

In other words, Levman is guilty of the fallacy of presentism since he apparently assumes his own, modern, linear sense of time and causality applies to this ancient religious text. Similarly, Kristin Scheible (2016) has cast doubt on the naive use of the Mahāvaṃsa as a historical source. The clear trend in scholarship on the Vaṃsa literature is towards dehistoricizing it. The majority of modern historians don't consider the Mahāvāṃsa to be a straightforward record of history anymore. To some extent, Levman anticipates this objection and his response is telling:

The alternative, that somehow a pseudo-historical figure was fabricated out of whole cloth or evolved on its own does not make rational sense. (Emphasis added)

This is an example of the informal fallacy of argument from incredulity. Wynne (2019) and others are similarly incredulous. We don't even learn why Levman thinks that it "does not make rational sense". Presumably, this is because the historicity of the Buddha is a given in his view. It's not irrational to believe that human storytellers might have invented a heroic figure to be the protagonist of their stories. Since this is exactly what storytellers do, it would be more surprising if Buddhists did not do it (as I will argue below, we see them doing exactly this at every stage of Buddhist literature). That such stories might have evolved as they were repeated orally for centuries, is exactly what I expect.  

Levman finally finds some purchase on historical facts seven pages into his article when he introduces the issue of how accurately the Pāli stories present geographical information, and accurately reflect the flora and fauna of the Ganges Valley. This strategy is also employed by Wynne (2019). However, the fact that Sāvatthī, for example, was a real city is not evidence that the Buddha was a real person. Rather, it is evidence that the Pāli authors knew Sāvatthī from first-hand experience or got reliable second-hand descriptions. 

2. Aetiology

Levman's second argument is to ask: "If the Buddha is indeed a mythic figure, how did his teachings arise?" He argues that if we say his explanation is not the explanation then we are bound to offer an alternative explanation. This is not the case.

The drift of Drewes's argument is to say that in the absence of historical facts (i.e. contemporary documentation) there is nothing that we can interpret to create a historical narrative. The absence of historical facts means that historians have to accept that they are ignorant and stop talking. Moreover, the old Roman legal principle applies:

Onus probandi incumbit ei qui dicit, non ei qui negat.
The burden of proof lies with the one who asserts, not the one who denies

Levman is making assertions, so the burden of proof lies with him. We've already seen that the standard of the "evidence" Levman cites is insufficient to make his case. Indeed, although he repeatedly mentions "historical facts", Levman has presented none. Rather he presents his speculations about what the facts might have been, validated against his axiomatic belief in the historical Buddha, and treats this mess as "facts". There is no documentation from the time he wishes to historicise. Historians don't engage in the reconstruction of facts. They use facts as evidence to construct a story about the past. 

In making the observation that there are no contemporary documents from which to construct a history of that period, Drewes has done his job as a historian. Explaining prehistory is not the job of historians; it is the job of archaeologists and anthropologists. For example, there are interesting archaeological accounts of the second urbanisation based on the distribution of pot-making technology, which gives us the "two cultures hypothesis" (see e.g. Samuel  2008: 48 ff.). Neither Levman nor Karpik mentions this hypothesis. 

Levman (36) continues

Over the twenty-five centuries since the Buddha lived and taught, billions of people have responded to his teachings of relief from suffering through the realization of selflessness; the four-fold saṅgha of upāsakas and upāsikās, bhikkhus and bhikkhunis has lasted in an uninterrupted continuum from then to the present day. Are we to say that these teachings were simply invented or evolved? Is that even possible?

It's apparent here, again, that for Levman the historicity of the Buddha is not in question. It is something that he takes for granted. He's not making an argument from facts, he is stating a belief about what the facts might have been. And there are at least two other fallacies involved here. 

"Billions" is probably an exaggeration. The fact that a million people believe a myth is not a reason to consider it historical. This appeal to the authority of the masses is called the bandwagon fallacy. Moreover, millions of people (more often than not, the very same millions of people) have also believed that the Buddha performed miracles. Levman does not consider this other testimony from the same source to be a "historical fact". If the bandwagon fallacy applies, then Levman should be arguing that the historical Buddha did miracles as a matter of historical fact. 

In "Is it even possible?" we also have another argument from incredulity. Levman has twice now asked his readers, "Could the teachings have been invented and then evolved?" So let's look at how we might answer him. 

From Buddhist literature, we know that Buddhist teachings evolved constantly while there was life in Indian Buddhism (and also that it continued to evolve outside of India). Even within the Pāli texts, we see clear evidence of the evolution of Buddhist doctrines, from archaic formulations later refined or abandoned, to the emergence of abhidharma-style lists. This evolution is frequently used as evidence for the antiquity and authenticity of the Pāli suttas. In fact, every documented Buddhist sect in history eventually abandoned Buddhavacana in favour of new doctrines.

We also know that ancient Buddhists invented new buddhas. We know, for example, that Buddhists invented the "buddhas of the past" to compete with the Jains and their lists of tīrthaṅkaras. And this happened early enough to become canonical. We also know that, before the Common Era, Buddhists were busy inventing new buddhas ex nihilo, e.g. Amitābha, Akṣobhya, and Bhaiṣyajagūru. In addition, they invented a whole new class of ahistorical awakened beings, i.e. bodhisatvas such as Mañjuśrī and Avalokitasvara (later Avalokiteśvara).

Since the invention of both doctrines and buddhas are observed at every point in documented Buddhist history, it makes no sense to argue that such processes were unknown before the advent of writing. Here we see how Levman's a priori beliefs skew his arguments towards tendentious conclusions. If later mentions of the Buddha are "evidence", the later inventions of buddhas are also "evidence" in the same way. Levman considers the former to be factual and does not consider the latter at all. 

Interestingly, Roy Norman (2008: 47), notes that the words buddha and jina are common to both Buddhism and Jainism, meaning that "there were buddhas and there were jinas before the beginning of both Buddhism and Jainism". If buddhas predate Buddhism then it is entirely possible, for example, that the protagonist of the Pāli suttas is a composite of numerous buddhas. This might explain variations in terminology. 

In answer to Levman's question—Is that even possible?—then, I would answer, that it is not only possible that Buddhists invented doctrines and that those doctrines evolved; it was the norm. The invention of buddhas was also normative. The Buddha and his doctrines could easily have been "fabricated out of whole cloth" and this would have been entirely in keeping with trends we see everywhere in Buddhism and in other world mythologies. So Levman's incredulity is not probative; it's just an expression of his ignorance. 

Levman finishes this section by recapitulating his assertion that the bandwagon fallacy applies. This tells us that his invocation of this fallacy was not a mistake. He appears to genuinely think that the bandwagon fallacy is a valid historical method.

3. The Humanness of the Buddha

Levman's third argument is that amongst all the many supernatural features of the protagonist of the Buddhist suttas, are some human details. These details he draws from Pāli texts that were not written down until some 400-500 years after the events that they purport to describe. Detailing all of the fallacies that this argument involves would be tedious, however, there is one informal fallacy here that it is worth focussing on since it also cropped up earlier. 

On any given page of the Pāli canon, we are likely to encounter both human details and superhuman details attributed to the Buddha by the author(s). By "superhuman" here, I mean qualities that involve magic or the supernatural, such as miracles, psychic powers, visiting god realms, and so on; anything that breaks the laws of physics as we know them.

From this body of literature, Levman cites examples of one type of detail and not the other, and the only examples he cites are those that support his view. But he does not tell us why or how he made this distinction and doesn't admit that there are a huge number of passages that don't support his view. In using selected examples that are not representative of the whole literature, he appears to believe that what fits his presuppositions is positive evidence for his conclusions and what does not fit is not evidence at all. This is called the cherry-picking fallacy.

In fact, the Pāli authors almost always included both kinds of details and there is little or no sign that they made the kind of distinction that Levman takes to be a given. The authors apparently didn't think in terms of "historical facts" and "extraneous magical thinking that can safely be ignored". As far as the authors of the Pāli Canon were concerned, it's all undifferentiated buddhavacana, including the miracles and magic. Levman seeks to impose his modernist distinctions on an ancient literature that definitely did not make that distinction. So this is also an example of the presentism fallacy.

All stories contain human details, even when they are about non-humans because this is how stories work. Drewes (2017: 19) notes that many other mythic figures are fleshed out by storytellers:

There may similarly have been an actual person behind the mythical Agamemnon, Homer, or King Arthur; Vyāsa, Vālmīki, Kṛṣṇa, or Rāma, but this does not make it possible to identify them as historical.

After many pages of fallacious argument in this style, we find Levman asserting:

If the Buddha were indeed a mythic character, surely this kind of human material, where the Founder is portrayed as old and weak, would be the first to go (44).

In Christian hermeneutics, unflattering details about Jesus—such as being betrayed to his death by his own followers—are given extra significance because of the principle of embarrassment (c.f. Meier's Historicity Criteria). Stories about real people would be expected not to include unflattering details unless they were true, so such details can be taken to be more likely to be factual. This is not a criterion that can be applied in isolation and we would want to see documentary corroboration from another source but, still, the inclusion of negative qualities makes a protagonist seem more real, not less. 

Compare some examples from Greek mythology. I think of the myth of Prometheus—almost certainly not based on a real person—who creates humans and steals fire from Zeus for us. Zeus doesn't take fire away from us, but he does punish Prometheus for eternity and creates Pandora's box (which introduces evil into the world via women's curiosity about the world). Zeus himself was guilty of numerous rapes and other forms of brutality. I think also of Hephaestus who was born lame, rejected by his mother, fell hopelessly in love with Aphrodite, and experienced overwhelming jealousy towards Ares. Or think of  capricious Yahweh who, enraged by human conduct, wiped out humanity and started again from Noah and his family, but who apparently still applied the doctrine of original sin to justify oppressing humanity with difficulty and pain. 

Does myth-making always exclude the negative? By no means. The gods have all of humanity's foibles, often in extreme forms. Suppose the inventive storyteller wanted us to believe in the historicity of the Buddha. The little negative details are exactly the kind of qualities they would include, be it historical fact or pious fiction. Human details make fictional characters relatable and memorable. 

So there is no reason to assume that human details attributed to the Buddha reflect historical facts. 

4. Biographical

Finally, Levman argues "But discoverable in the canon is evidence of an early, core biography preserving the authentic history of a real person in an unembellished state. Is this also invented?" (26).

Note again the incredulity. We have already established that the only documented history can be "authentic" and the documents that Levman cites are from at least 400 years after the period he wishes to historicise (and probably much longer). Levman's method here is no more than the interpretation of scripture, a procedure already long discredited amongst historians when Drewes wrote his article. Much of what Levman writes in this section takes the form of "hand waving", e.g.

This may or may not represent something close to the actual words of the historical Buddha; the simplicity and candor of the statement do seem to reflect a “certain genuineness” on the part of the speaker (47).

The idea of "a certain genuineness" is vague and subjective and Levman's use of scare quotes here suggests that he was aware of this. It's all too apparent that Levman finds passages to be "genuine" when they confirm his belief and when he does not find that confirmation he does not discuss them at all (cherry-picking fallacy). 

More importantly, how would anyone know if any words from any source reflected the "actual words" of the "historical" Buddha? Given the lack of contemporary documentation, what is the yardstick here? No one disputes that the Buddha was a non-literate person living in a non-literate society. There are no possible corroborating sources from the fifth century BCE. 

Identifying common elements in versions of a story does not make them truer, if anything it just makes them seem older. How old, we have no idea. The idea that older = truer is a fallacy known as appealing to tradition. The "simplicity" of an idea has never been a criterion for its historicity.

The problem here is that the further back in time we go, the more partial and fragmentary are our witnesses to history. Fewer and fewer sources may well give the illusion of increasing simplicity, when in fact it's just a paucity of sources. There is no a priori reason that the past should be any less complex than the present (at least in historical terms). As Graeber and Wengrow (2021) have amply demonstrated in their first two chapters, those people the Europeans described as "simple" and even "savage" were usually anything but. Arguably the indigenous Americans encountered by Europeans were far more socially and politically sophisticated than their European counterparts. Notably, it was Europeans who adopted American ideas—like individual liberty—rather than the other way around.

How does Levman know that any statement in Pāli is "candid"? He claims to be concerned with rational conclusions, but what rational criteria can he possibly apply to arrive at this "insight"? This is all just confirmation bias. This section finishes with a flourish of hand-waving

Of his true roots we know very little, beyond the few snippets which are buried in the canon, or can be reasonably surmised based on the evidence. All of the material I have been able to find is summarized in my 2013 article. But though his background has been mythologized, this does not make him a mythological character, just someone whose true roots have been obscured and excised for purposes of social and political acceptance.

This is what Drewes was referring to at the outset when he said, "scholars have rarely been able to avoid the temptation to offer some suggestion as to what was likely, or ‘must’ have been, true about him."

In point of fact, of the Buddha's "true roots", we know nothing. The snippets that conform to Levman's views are dwarfed by an avalanche of passages that do not. Levman systematically ignores the vast bulk of the Pāli canon because it doesn't support his argument. There are literally no documentary facts upon which any reasonable surmise might be based. And Levman has not introduced any new facts and inferences are not facts. 

Levman sums up by repeating the numerous fallacies already listed. It is difficult to avoid the conclusion that Levman doesn't understand the theories and methods of modern historiography. He completely misses the significance and importance of Drewes's argument. 

Summing up

At this point, I would characterise Levman's article as an example of what historiographer Carl R. Trueman (2010: 45) calls the aesthetic fallacy: “if it looks scholarly, then, agree or disagree with it, it is scholarly and must be taken seriously and allowed a place at the scholarly table”. Levman's article looks scholarly, but his methods are not scholarly. At least not from the point of view of a historian. 

History is the study of documents. There are no Indian documents before Asoka because writing was not used in India until he created his famous edicts. Attempting to write a history of a preliterate society is a contradiction in terms, at least as far as historians are concerned. This is the historian's episteme. This is how historians try to ensure the validity of their use of historical facts as evidence for reconstructing knowledge of the past. The epistemology of history is still a live topic and the impact of postmodernist critiques of the use of texts is still being felt. Still, the centrality of contemporary documentation has never been problematised. 

Levman appears to fundamentally miss Drewes' point and makes a series of irrelevant arguments. For example, Levman appears to be convinced that certain presumptions and subjective judgements about stories recorded in Pāli amount to historical facts about the Buddha. Or that his inferences about the past amount to historical facts. In his arguments, Levman relies heavily on unexamined assumptions, skimps on citations, makes factually incorrect statements, and employs numerous informal fallacies including, presentismargument from incredulitycherry-picking, the bandwagon fallacy, and confirmation bias.

Fallacies and biases aside, it's clear that Levman, Karpik, and Wynne are all doing something similar when they argue for the historicity of the Buddha. And I think I can shed some light on this. 

Two Epistemes

Most Buddhist Studies scholars are educated in the theories and methods of philology and/or historical linguistics; not in the theories and methods of history and historiography. Philologers routinely reconstruct lost ur-texts from surviving witnesses and historical linguists routinely reconstruct long lost proto-languages. My thesis is that, given these prominent activities it might seem natural for philologists and historical linguists to use similar methods to attempt to reconstruct historical facts via inferences. 

Nineteenth-century linguists, especially in Germany, were able to analyse the way that phonology changes over time and observe that only certain changes and certain types of changes occur. This allows philologists to define descriptive "laws" which limit how any Indo-European language is permitted to change. So we get Bartholomae's Law, Grimm's Law and so on. Since phonological change follows regular patterns that apply across locations and times, we can apply descriptive laws prescriptively and retroactively to reconstruct a universal mother tongue for all of the Indo-European family of languages. Given modern languages and a set of rules, the sounds of ancient languages can be retrospectively reconstructed with considerable confidence. The reconstruction of Proto-Indo-European is an awesome achievement. 

The practice of textual criticism has its roots in the interpretation of legal and religious documents. Formalised methods of recreating the ur-text of the author developed over centuries. Whether they know it or not, modern scholars rely on the method of Karl Lachmann (1793-1851), especially as expressed by his student Paul Maas (1927). As manuscripts are repeatedly copied, errors and amendments build up. By carefully comparing witnesses using Lachmann's method, the textual critic may restore the "original text" even though none of the surviving documents reflects that text. 

In both cases, scholars can infer reliable knowledge of the past based on extant documents. This should sound familiar because it also describes the method of Levman, Karpik, and Wynne (other biases and fallacious arguments aside). They are all making inferences about the past and treating these as historical facts. However, this is not a sound methodology for historiography. 

In contrast to the situation in which we have complete descriptions of dozens of modern languages and extensive descriptions of ancient languages, the Pāli texts don't constitute anything like a complete description of a culture or society. They are normative religious texts that are, for the most part, mythological in character, and only look "historical" after some very restrictive cherry-picking. There is no historical analogue of the lawful changes in phonology (or grammar). 

Historians have long acknowledged that history is not governed by laws analogous to those that govern phonological change. It is a truism that those who fail to learn the lessons of history are doomed to repeat the mistakes of the past. However, it is also a truism that knowledge of the past does not enable one to predict the future. Indeed, in history knowledge of the present does not allow us to predict the past either. If it did then we could simply observe Buddhists in the present, formulate some laws that govern change, and infer facts about the past. This method does not work for history. 

It's notable that in the absence of any general laws, Levman appears to substitute "common sense". I point, for example, to his repeated argument from incredulity and his use of subjective terms like "candour" or "genuineness". Even though Drewes pre-warns him that this is not a credible method, Levman goes ahead and does it anyway.

In effect, historians and philologists have different views on epistemology based on different methods applied to different bodies of knowledge. In this view, philologists appear to believe that the kinds of methods that allow them to reconstruct a proto-language or an ur-text can be applied mutatis mutandis to historical facts. Levman repeatedly treats his inferences as facts. 

While the philological approach to history fails, and fails badly, in terms of historiography, at least this explanation of Levman's method as rooted in philology and historical linguistics rather than history and historiography makes a certain kind of sense. I'm not sure this is correct, but this is the most charitable interpretation of Levman's method that I have been able to come up with. 

This view may also help to explain the (undeniable) controversy that Drewes' article caused amongst Buddhist Studies scholars and religieux. Perhaps Drewes's invocation of historical methods, while obvious to any professional historian, was a bit too casual for an audience of philologists and linguists with no background in historiography. Philologists confidently resurrect lost texts and linguists resurrect dead languages all the time, so resurrecting the Buddha may well seem straightforward to them, more especially if his historicity is axiomatic for them. Historians in their turn expect facts to emerge from documents of that time. They are puzzled that the evidence presented is all 500 years too late and of very mixed provenance and doubtful veracity. One side is shouting "What about the facts?" and the other is shouting back "What facts?". As far as I know, no one has previously observed that the two sides define the word "fact" in different ways. 


In the arena of academic historiography, Drewes is right to say "my argument is really a minor one" (19). In the context of modern historiographical methods, there is no such thing as "the historical Buddha" because there are no documents from that time. Drewes is absolutely right that historians should stop using this phrase. 

I think it's fair to say that the dispute over the historicity of the Buddha has been framed in ideological terms, i.e. as a conflict between traditionalists and modernists. This is unfortunate because ideological disputes are extremely resistant to resolution. Ideologues don't change their minds. The dispute is better framed as a dispute over methodology and epistemology. 

This is to say, the dispute hinges on the ability of different methods to give us reliable information about the past. Historians, who specialise in explaining the past, universally agree that history begins with contemporary documents, with the broadest possible meaning of document as any form of writing. A historical fact is a documented fact. 

The problem is that Levman is not a historian. Levman does apply a historian's methodology and does not cite any authorities on the theories or methods of historians. Rather, where Levman is not relying on some fallacy or other, he relies almost entirely on treating inferences as historical facts (analogous to PIE or some ur-text). The raw materials for his inferences are documents from a much later period, after writing began to be used. The validation of such inferences seems to rest on his axiomatic belief in the historicity of the Buddha (the same can be said of Karpik and Wynne) and appeals to incredulity, common sense, and so. 

As compelling as the rhetoric of a "middle way" might be at this point in trying to resolve a dispute, it's clear that historians have already established a "middle way". This is to restrict themselves to contemporary documents. This means that historiography is necessarily limited in scope and reach. 

In fact, the method of treating inferences as facts, as adopted by philologists like Levman, is not a reliable way to get information about the past. It works in the case of proto-languages and ur-texts, but it does not work in historiography. That Levman's attempt to apply this method is plagued by fallacious reasoning and bias should not distract from the problem that his method is fundamentally unsound. 

This is also an answer to the philosopher/philologer colleague who accused me and Drewes of practising "positivism" because we refuse to accept the philological method of treating inferences as facts. We are not "positivists" demanding scientific facts, we are historians using generally accepted methods in historiography to assess the salience and veracity of facts in documents. 

That said, I do not think the idea of a founder of Buddhism is impossible or unreasonable (as I understand Drewes he thinks the same). It actually seems quite plausible that the mythology of Buddhism might be based on a real religious leader. The problem here is that history is not about what we surmise or guess to be true. Inferences are not facts. 

History deals with documented facts and prioritises facts that can be corroborated. As such history is extremely limited in scope. As John Vincent (quoted above) says, "history leans towards literate individuals in literate societies". The Buddha is not a historical figure by any definition of "historical" used by historians precisely because he is not a literate figure and was not from a literate society. In attempting to historicise the mythical Buddha using other methods and without reference to the long history of historiography, Levman ignores the accumulated wisdom of historians. 

Notwithstanding the possibility of his being based on a real person, the Buddha as presented in Buddhist documents is clearly a mythological figure, who has human traits, but also does miracles and has supernatural powers. The term mythological is not intended to have any pejorative connotation. Myths are how preliterate societies encoded their views about the world and their values before the advent of writing. 



Allen, Charles. (2012). Ashoka: The Search for India's Lost Emperor. London: Abacus.

Bollée, W. B. (1974). "Buddhists and Buddhism in the Earlier Literature of the Śvetâmbara Jains." In Buddhist Studies in Honour of I.B. Horner, edited by L. Cousins, A. Kunst, K. R. Norman, 27-39. Dordrecht: Springer.

Bronkhorst, Johannes. (2020) "The Formative Period of Jainism (c. 500 BCE – 200 CE)" In Brill's Encyclopedia of Jainism Online. doi:10.1163/2590-2768_BEJO_COM_047082.

Drewes, David. (2017). "The Ideal of the Historical Buddha." Journal of the International Association of Buddhist Studies 40: 1-15.

Evans, Richard J. (1997). In Defence of History. London: Granta Books.

Levman, Bryan. (2019). "The Historical Buddha: Response to Drewes" Canadian Journal of Buddhist Studies 14: 25-56.

Maas, Paul. 1927. Textkritik. Leipzig: Teubner. 

Norman, K. R. (2008). A Philological Approach to Buddhism: The Bukkyō Denō Kyōkai Lectures 1994. Oxford: Pali Text Society. 

Olivelle, Patrick. (2023). Ashoka: Portrait of a Philosopher King. Yale University Press

Samuel, Geoffrey. (2008). The Origins of Yoga and Tantra: Indian Religions to the Thirteenth Century. Cambridge University Press.

Scheible, Kristin (2016). Reading the Mahāvamsa: The Literary Aims of a Theravada Buddhist History. New York:  Columbia University Press. 

Schopen, Gregory. (1996). "Two Problems in the History of Indian Buddhism: The Layman/Monk Distinction and the Doctrines of the Transference of Merit." In Bones, Stones, and Buddhist Monks: Collected Papers on the Archaeology, Epigraphy, and Texts of Monastic Buddhism in India. University of Hawaii Press. 

Stargardt, Janice. (1995). “The Oldest Known Pali Texts, 5–6th century: Results of the Cambridge Symposium on the Pyu Golden Pali Text from Śrī Kṣetra, 18–19 April 1995.” The Journal of the Pali Text Society 21: 199-213.

Trueman, Carl R. (2010). Histories and Fallacies: Problems Faced in the Writing of History. Wheaton, Ill: Crossway.

Vincent, John. (2006). An Intelligent Person's Guide to History. London: Duckworth Overlook. (first published 1995).

Walters, Jonathan. S. (1999) "Suttas as History: Four Approaches to the Sermon on the Noble Quest (Ariyapariyesana Sutta)." History of Religions 38.3: 247-8.

———. (2000). "Buddhist History: The Sri Lankan Pāli Vaṃsas and their Commentaries". In Querying the Medieval: Texts and the History of Practice in South Asia, 99-164. Oxford University Press. https://doi.org/10.1093/oso/9780195124309.003.0003

Wynne, Alexander. (2019). "Did the Buddha Exist?" Journal of the Oxford Centre for Buddhist Studies 16: 98–48.

09 February 2024

Guanyin Does Not Speak in the Heart Sutra

In this short essay, I will challenge a universal presupposition about the Heart Sutra, i.e. that the lines that appear to be spoken to Śāriputra in the core section are spoken by Guanyin. I will show that, by all the conventions of Buddhist literature, this is not true. Guanyin does not speak. This observation further undermines the already weakened historically dominant narrative about the Heart Sutra.

For some years, I have made a practice of reading every scholarly publication on this text (in English), as well as selected popular works. To the best of my knowledge, no modern scholars have previously noticed the absences I mark below. I think Chinese Buddhist commentators in the late seventh–early eighth centuries were aware of this. And there were subsequently efforts made to obscure this fact.

A Buddhist sutra is, above all else, a record of speech. In Buddhist texts, speech is almost always indirect speech and the forms of indicating who is speaking to whom are essential to understanding the text. Forms of present speech in Buddhist texts are highly formalized and standardized; to the point of being universal across genres and over time. And they are not complex. We can easily describe the main forms and note the Chinese reflexes of these forms. I will focus on how they appear in the early Prajñāpāramitā literature, if only because this is the appropriate context for thinking about the Heart Sutra.

To begin with, we expect to see the speaker “addressing” (āmantrayate) the audience. At the beginning of the Aṣṭasāhasrikā Prajñāpāramitā (Aṣṭa), for example:

Then the Blessed One addressed the senior Elder Subhūti...”
(tatra khalu bhagavān āyuṣmantaṃ subhūtiṃ sthaviram āmantrayate sma… Vaidya 1960 2).

Here āmantrayate sma is the "pleonastic past". Here adding sma to a present tense verb makes it a past tense, but is also used for the "present in the past" tense so commonly used in storytelling. In Kumārajīva’s translation—the Xiǎopǐn bānrě jīng «小品般若經» (T 227)—this becomes…

Then he Buddha addressed Subhūti
Ěr shí Fó gào Xūpútí 爾時佛告須菩提 (T 227: 8.537a29)
Here the verb is gào 告 "to address". Note that Kumārajīva omits the honorific ayuṣman "Elder" and Subhūti's monastic title sthavira "Senior [monk]". In Chinese we also sometimes see:
“Subhūti addressed the Buddha, saying…” 
Xūpútí bái Fó yán 爾時須菩提白佛言 (T 227, 8.537b06)
The character bái 白 is polysemic but here means “to make plain, to state clearly”, while yán 言 means “speak, talk”.

The extracts found in the Heart Sutra are from a version of the Pañcaviṃśātisāhasrikā Prajñāpāramitā (Pañc). An example of the same form from the Gilgit manuscript of Pañc:
The Bhagavan addressed Elder Śāradvatīputra...”
(bhagavān āyuṣmantaṃ śāradvatīputram āmantrayata... Zacchetti 2005: 375).

The use of āmantrayate (in various conjugations) usually marks the beginning of a passage of discourse, but within a given conversation, the speaker of individual passages is also marked. It is usual to spell this out laboriously, including the name and title of each participant. In the following passage from Chapter One of Aṣṭa, we find all of the most common forms of ongoing verbal address:

Then Elder Śāriputra said this to Elder Subhūti, “Elder Subhūti, does this mind that is a mind without mind actually exist?”
When that was said, Elder Subhūti said this to Elder Śāriputra, “Elder Śāriputra, concerning that which is without mind, is the existence of mindlessness known or apprehended?”
Śāriputra said, “Indeed not, Elder Subhūti”.
Atha khalv āyuṣmān śāriputra āyuṣmantaṃ subhūtim etad avocat - kiṃ punar āyuṣman subhūte asti tac cittaṃ yaccittamacittam?
Evam ukte āyuṣmān subhūtir āyuṣmantaṃ śāriputram etad avocat kiṃ punar āyuṣman śāriputra yā acittatā, tatra acittatāyām astitā vā nāstitā vā vidyate vā upalabhyate vā?
Śāriputra āha - na hy etad āyuṣman subhūte /
(Vaidya 1960: 3).

The forms I wish to highlight are “said this” (etad avocat), “when this was said” (evam ukte), and “said” (āha). Both avocat and ukte derive from √vac “speak”, while the āha is from the defective verb √ah “say”. The same forms are used in the same way throughout the Pāli Suttapiṭaka also. An electronic search of the Chaṭṭa Saṅgāyana Tipiṭaka (4.1) suggests that in the four main Nikāyas: āmantesi occurs about 590 times, etadavoca occurs over 2200 times; evaṃ vutte occurs some 530 times; while āha occurs about 100 times. A few other such forms are used, but these are by far the most common. 

When a person is addressed in Sanskrit or Pāli, their name occurs in the vocative case, e.g. “O Subhūti” (subhūte). In the CBETA version of the Taishō edition of the Chinese Tripiṭaka, the “vocative case” is represented by "!" following the speaker’s name. 

Note the repetitive use of the title “Elder” (āyuṣman) most of the time. Given that both characters use the title, they appear to be social equals. I speculate that āyuṣman is omitted precisely when Śāriputra acknowledges Subhūti’s superior insight.

These forms are largely preserved in Chinese translations. Turning to Huifeng’s (2017: 205) translation of this same passage in the Xiǎopǐn:

Thereupon, Śāriputra said to Subhūti: “Does this mind which is mindless exist?”
Subhti said to Śāriputra: “That mind which is mindless, is it apprehendable as either existing or not existing?”
Śāriputra said: “Indeed not!”

Here, two different words are translated as “said”: 語 (here translating avocat) and yán 言 (translating āha).

None of these conventions for indicating that someone is speaking or for who is addressing whom occurs in the Heart Sutra. The only indication we get that some parts of the text are indirect speech is the use of the vocative śāriputre “O Śāriputra” (Shèlìzi 舍利子!). This is how we know that Śāriputra is being addressed. The Heart Sutra does not say who is speaking and everyone assumes that it is Guanyin. 

The passage in which Śāriputra is addressed has been traced to the Large Sutra (see: T 223, 8.223.a13-a20 and Zacchetti 2005: 393). When we read the passage in this context, the lines are spoken to Śāriputra not by Guanyin, who has no speaking part in any Prajñāpramitā text, but by the Buddha. Interestingly, Woncheuk’s (613-696 CE) commentary (T 1711) appears to take the Buddha to be speaking as well.

Question: [Since] this [teaching of] prajñāpāramitā is the dharma for the bodhisattva, why does the World Honored One preach not to the bodhisattva but to Śāriputra? (Hyun Choo 2006: 149)


Therefore, in the [Heart Sūtra], the Buddha preached to Śāriputra and intended to lead the Hīnayāna to the Mahāyāna as well. (Hyun Choo 2006: 149)

It seems to me that Woncheuk could only have deduced this by looking at the source of the passage, i.e. Pañc.


In this brief essay, I have tried to show that the universal view that Guanyin is speaking in the Heart Sutra is based on presupposition and unexamined assumptions. I did this by outlining Buddhist conventions for expressing who is speaking to whom. I argued that such conventional expressions are universal in Buddhist texts (in Pāli and Sanskrit) and that such conventions are absent from the Heart Sutra. All indications of who is speaking have been omitted. Taking the Heart Sutra at face value, no one is speaking.

We can explain this by pointing out that the Heart Sutra is not, in fact, a sutra. It is not a record of speech, rather it's a compilation of ideas and extracts from existing speeches. This much was obvious to the earliest commentators, though subsequently forgotten. The lines in their original context were spoken by the Buddha, as accurately reflected in the commentary by Woncheuk. 

The (now disproved) "fact" that Guanyin was speaking has always been a problem for scholars since Guanyin plays no active role in any non-Tantric Prajñāpāramitā text (and the Tantric Prajñāpāramitā texts are more Tantra than Prajñāpāramitā). Various unsatisfactory explanations have been advanced (I've made several previous attempts to explain), but they have always been ad hoc or post hoc rationalisations, rather than real explanations (they all amount to hand-waving).

The most striking attempts to make sense of this situation are the two recensions of the extended Heart Sutra text (1. T 252; 2. T 253 and all rest), probably composed in the early eighth century, possibly in the oasis town Dūnhuáng 敦煌, on the edge of the Gobi Desert (from where it was transmitted to Tibet). 

In the extended texts, the two redactors have attempted to better integrate Guanyin into the narrative, by expanding the first paragraph. However, both recensions retained the rest of the standard Heart Sutra unaltered and in this part of the text, the verbs of speech are still absent. This means that despite more clearly articulating the reason for the presence of Guanyin, neither of the extended texts addresses the problem. Guanyin's presence seems more natural, but he/she still does not speak the lines directed to Śāriputra in the "core section".

The fact is that, in the Heart Sutra, Guanyin does not speak. Guanyin is invoked and then plays no further role in the text. In their original context, the lines are spoken by the Buddha. At this point, we can say that more or less all of the historical dominant narrative about the Heart Sutra is a post hoc invention. 



Huifeng. (2017). “An Annotated English Translation of Kumārajīva’s Xiǎopǐn Prajñāpāramitā Sūtra.” Asian Literature and Translation 4(1): 187-236.

Hyun Choo, B. (2006) “An English Translation of the Banya paramilda simgyeong chan: Wonch’uk’s Commentary on the Heart Sūtra (Prajñāpāramitā-hṛdaya-sūtra).” International Journal of Buddhist Thought and Culture 6: 121-205.

Vaidya, P.L.(1960) Aṣṭasāhasrikā Prajñāpāramitā Sūtra. Darbhanga: The Mithila Institute. (Via the Gretil Archive, 2014. Including Karashima, S. (2013) On the "Missing" Portion in the Aṣṭasāhasrikā Prajñāpāramitā. ARIRIAB, 16: 189-192. Accessed 6 Feb 2024)

Zacchetti, Stefano. (2005). In Praise of the Light: A Critical Synoptic Edition with an Annotated Translation of Chapters 1-3 of Dharmarakṣa’s Guang zan jing, Being the Earliest Chinese Translation of the Larger Prajñāpāramitā. (Bibliotheca Philologica et Philosophica Buddhica, 8). Tokyo: The International Research Institute for Advanced Buddhology, Soka University.

Related Posts with Thumbnails