Inferential Reasoning

An excerpt from Science and Philosophy in the Indian Buddhist Classics, Volume 2: The Mind

John Dunne

John Dunne holds the Distinguished Chair in Contemplative Humanities at the University of Wisconsin–Madison. He received his PhD in Sanskrit and Tibetan studies from Harvard University.

The following is his introductory essay to Inferential Reasoning, Part 5 of Science and Philosophy in the Indian Buddhist Classics, Volume 2: The Mind.


The Crucial Role of Reasoning

One of the keen insights that runs throughout the Buddhist tradition is that the ultimate authority for knowledge claims must be one’s own experience. While an enormous corpus of scripture is readily available to any Buddhist philosopher, no scriptural passage—even those attributed to the historical Buddha himself—can in itself be used to establish empirical claims, at least according to the Buddhist epistemologists. Even reasoning, which inevitably requires the use of concepts, must eventually give way to direct experience, in part because conceptual cognitions are always distorted, in ways that we have previously discussed. And as we saw in the previous part, for the purposes of personal transformation and behavior change, only direct experience—not inferential or conceptual cognition—has the kind of visceral impact necessary for effective change.

Yet this emphasis on direct experience stands alongside another key theme: ordinary experience prompts intuitions that do not conform to the way things really are. At the very least—as noted previously—our ordinary experience suggests that we have an enduring, autonomous self that is the perceiver of perceptions, the feeler of feelings, and the owner or controller of the mind-body system. Thus, even though in the end the Buddhist paradigm would have us arrive at a direct experience of, for example, our mind-body system as devoid of any such absolute self, we are starting from a place where a direct experience of selflessness is inaccessible. To get to that point, we must engage in intensive observation and analysis guided by reason. And as a result, the Buddhist traditions examined in this volume have placed a tremendous emphasis on the crucial importance of rigorous, irrefutable reasoning.


Three Domains

One way to understand the role of reason in these Buddhist traditions is to examine the notion of three domains of inquiry articulated by Dignāga, Dharmakīrti, and subsequent Buddhist epistemologists. The first domain includes things that are directly available to our perceptual experience, or more precisely, knowledge claims we can justify directly based upon our experience. To take again the typical Indian example of the pot or water jug, my visual experience of an item on the table in front of me is enough for me to justify the claim “There is a pot on the table.” There are other claims, however, that I may not be able to make directly from my perception alone. For example, when I see black smoke billowing out of a house, I cannot directly see that there is a fire present; the fire is epistemically hidden (Skt., viprakṛṣṭa) for me, in that it is not directly available to my senses. Nevertheless, I can rationally infer—based on what I can perceive—that there is indeed some incendiary source in that location. This is the second domain: knowledge claims about epistemically remote objects that I can justify by using an inference—or a chain of inferences—based in what is directly perceptible.

These first two domains—the directly perceptible and the remote—­constitute what we might call an “empirical” sphere of knowledge, inasmuch as the evidence of the senses is what we fall back on when we are making knowledge claims in these contexts. There is, however, a third domain that, transcending the empirical, concerns objects that are extremely hidden (atyantaparokṣa). These are things that, at least given one’s current capacities, cannot be examined in any empirical way. For the Buddhist epistemologists such as Dharmakīrti and his followers, the only way to justify claims about such trans-empirical things is through the testimony of a reliable person (āptavacana), which paradigmatically consists in the words of the Buddha himself as transmitted in the Buddhist scriptures. Importantly, however, Dharmakīrti maintains that scripture cannot be used to justify empirical claims. In other words, if some matter of concern can be adjudicated by ordinary perception and empirical inference, then one cannot use scripture to support one’s claims. And if there is a conflict between some claim made in scripture and claims that can be empirically justified, then the scriptural claim must be rejected. Moreover, according to at least some interpretations, Dharmakīrti holds that claims justified by reference to scripture are not truly justified at all. In other words, according to this interpretation, scripture cannot be used to definitively prove or disprove trans-empirical claims. They only provide a provisional justification that enables one to address some important issue relevant to contemplative practice that, as Dharmakīrti puts it, require some answer but cannot be resolved any other way.

For our purposes, the upshot of the schema of these three domains is that, if we are to detect and correct the false intuitions that emerge from our ordinary experiences, then we cannot simply rely on scriptural claims. Of course, the teachings of the Buddha or some other reliable person can help direct us to an area of inquiry, such as a false sense of an absolute self, but from the Buddhist perspective articulated in this volume, that inquiry itself must eventually become a purely empirical one. And again, since our ordinary perceptions prompt intuitions that are not the way things truly are, we become especially reliant on empirical inferences that, using the reliable aspects of our perceptions, can help us see more clearly what we are actually experiencing.


Inference and Its Structure

Given the importance of empirical inference (Skt., vastubala­pravṛttā­nu­māna), it is no surprise that an entire section of this volume is devoted to inferential reasoning. This topic appears in early Buddhist sources, but Dignāga and Dharmakīrti are the key sources for the material presented here. To lay the groundwork for our authors’ articulation of the facets of inferential reasoning, it may be helpful to summarize the core theories concerning the structure and content of inferential reasoning.

An inference, or strictly speaking an inference for oneself (svārthā­nu­māna), is a type of cognition that enables one to know something by virtue of knowing something else. For Buddhist epistemologists, the basic structure of an inferential cognition looks like this:

S is P because E

Here “S” stands for a Subject or, more literally, a property possessor (dharmin) to which some Predicate or property (dharma) is being attributed. The statement or conceptual cognition “S is P” is thus the thesis (pratijñā) or proposition that is being established by the inference. The justification for formulating that thesis—the warrant for attributing that predicate to that subject—is the presence of another property that acts as the Evidence (hetu or liṅga), namely, the property that indicates the presence of the property to be proven. An example of all this, presuming that I am looking at some house with smoke billowing out of it, would be:

This house (S) is a locus of fire (P) because of being a locus of smoke (E).

In order for an inference of this kind to be valid, three relations must be in place: S must have E as its predicate or property (the house must have smoke billowing out of it); the presence of the evidence E must invariably indicate the presence of the predicate or property P (wherever smoke occurs, fire must also occur); and the absence of predicate P is invariably concomitant with the absence of the evidence E (where there is no fire, there can be no smoke). These three relations constitute the three “modes” (naya) of a reliable inference.

"Importantly, however, Dharmakīrti maintains that scripture cannot be used to justify empirical claims."

The Three Modes

Introduced by Dignāga and refined by Dharmakīrti, the theory that a reliable inference must exhibit the three modes of relation among its terms is a distinctive feature of their account of inferential reasoning. The theory emerges in part from Dignāga’s interest in categorizing the cases where inferences succeed, and our authors acknowledge the importance of Dignāga’s efforts by concluding this part with an entire section on his schema of successful and unsuccessful cases. In the interest of simplicity, however, we understand this issue by examining just the three modes that underlie Dignāga’s cases.

As noted above, the first mode amounts to a requirement that the property or thing being used as evidence (E) can be construed as a predicate or quality of the subject (S) of the thesis “S is P.” In the example above, it must be true that the house is indeed a locus of smoke. One way this can go awry is if there actually is no smoke present at all; in that case, the evidence (namely, the smoke) is unestablished (hetvasiddha). Another way that this mode can fail is that, while the evidence may indeed be present, we are mistaken about the subject that bears that evidence as a property. In the case of our house-fire example, perhaps I have mistaken some other structure for a house, and my inference actually should be about a shed. This problem becomes more acute when we consider the cases where we need to construct inferences about things—such as the kind of self that Buddhist philosophers refute—that have never existed. Here, we run the danger of having no subject at all (āśrayāsiddha), and these kinds of cases are part of what motivates the extensive analysis of consequential reasoning provided by Buddhist theorists, as we will see below.

Moving on to the other two modes, they concern the relationship between the evidence (E) and the predicate to be proven (P)—the smoke and the fire, in our example. Both of these modes fall under the general rubric of the pervasion (vyāpti) between the evidence and the predicate. In short, every case where the evidence is present necessarily is “pervaded” by the presence of the predicate. The metaphor of one thing pervading or spreading though another suggests a Venn diagram:

Applying this diagram of overlapping sets to our example, all cases where fire (P) is present pervade all the cases where smoke (E) is present. Thus, even though there are cases where there is fire without smoke, any time there is smoke, there must be fire. And this means that smoke can act as evidence or as an indicator (gamaka) of fire even when we cannot perceive the fire directly.

This relationship of pervasion encompasses two of the three modes, and they are articulated as the positive pervasion (anvayavyāpti) and negative pervasion (vyatirekavyāpti). The simplest way to articulate them is simply to say, for the positive pervasion, “Wherever the evidence is present, the predicate is present.” And for the negative concomitance, one can say, “Where the predicate is absent, the evidence is absent.” Students of logic will recognize that these statements are logically equivalent, and while this is true, one must be cautious here because the relationship between the evidence and predicate is not simply a logical one. Instead, it concerns the very natures of these things themselves, and thus it is a nature relation (svabhāvapratibandha).

Without going into great detail, we can note that the notion of a nature relation emerges from a fundamental insight articulated especially by Dharmakīrti: namely, that the relationship between the evidence and the predicate is not a matter of mere co-occurrence (sahabhāva). Here, Dharmakīrti is concerned with cases where we mistakenly believe that something can stand as evidence for something else. For example, it might happen to be the case, on a particular mango tree, that every mango we picked from a particular branch was ripe. And we might just infer that the next fruit (S) on that branch is ripe (P) because it is on that branch (E). But this inference is not, for Dharmakīrti, a reliable one because it fails to eliminate the possibility that it is mere happenstance that the ripe fruit were on that branch. Instead, the nature of the evidence must relate to the predicate’s nature in such a way that it is impossible for the evidence to be present while the predicate is absent. In short, evidence and predicate stand in a relationship of what Dharmakīrti calls “the necessary relation of unaccompanied nonoccurrence” (avinābhāvaniyama). If the evidence is not “accompanied” by the predicate, then the evidence simply cannot occur.

For Dharmakīrti and all later Buddhist epistemologists, this necessary relation between the evidence (such as smoke) and the predicate (such as fire) occurs in just two ways: either the evidence is the effect of the predicate or the evidence is a property whose essential nature requires the presence of the predicate. These are known respectively as effect evidence (kāryahetu) and nature evidence (svabhāvahetu). Our authors discuss these at some length, and effect evidence is perhaps fairly straightforward. Of course, determining exactly when a causal relation is actually in place poses many problems, but we need not concern ourselves with that issue here. Instead, let us turn to the case of nature evidence. The meaning of this term is perhaps less obvious than effect evidence, yet this form of evidence plays an especially important role in Buddhist practice.

A typical example for nature evidence would be something like this: This thing (S) is a tree (P) because of being an oak (E). To unpack this in straightforward terms, we can say that the relationship between evidence and predicate amounts to this: something that has the property of being an oak invariably has the property of being a tree because the features of the object that are required for us to correctly refer to it as an oak already include all of the features that are needed for us to correctly call it a tree. Although much more can be said about nature evidence, the main point is that, as Dharmakīrti puts it, this form of evidence is especially helpful for those of us who are “confused” (muḍha) about the way we are using certain concepts. For example, on one level we may be quite correct to know that a pot is something that is causally produced, but we may not recognize that when we use the term causally produced, we already have grounds for saying that a pot is also necessarily impermanent. Of course, unless we have a problematic attachment to pots, it may not be so crucial to understand the full implications of their causal nature. But when we turn to, for example, the constituents of our mind-body system, we may readily acknowledge that they are causally produced, but we still somehow believe that they are not impermanent. Or to use a parallel example, we may not fully recognize that if we can speak of ourselves as “born,” which we readily acknowledge, then the very use of this term means that we must die. And given this example, it is perhaps obvious why nature evidence plays a critical role in practice. Recall that, while direct experience is the bedrock of our knowledge, our ordinary experience prompts numerous delusions. Nature evidence plays an especially prominent role in enabling us to see how what we readily acknowledge in our own experience has implications that we have failed to uncover. In another key example, while I may readily acknowledge that I engage in causal actions in the world, I may not recognize that, merely by seeing myself as causally engaged, I am thereby necessarily interdependent and not an independent, autonomous self.

Much more can be said about the various aspects of inference unpacked by our authors. Numerous issues of great importance come to mind: the type of entailment required by the notion of pervasion; the requirement to ground inference in concrete, observable cases; the problem of determining when a necessary relation of unaccompanied nonoccurrence is in place; and so on. To this list should also be added the presence of a third type of evidence, nonperception (anupalabdhihetu), which uses the principles of effect evidence and nature evidence to negate, rather than establish, some thesis. But these issues can be explored through our authors’ efforts themselves, and with that in mind, let us turn briefly to examining two other key aspects of inference: proof statements and consequences.

"One way this can go awry is if there actually is no smoke present at all; in that case, the evidence (namely, the smoke) is unestablished (hetvasiddha)."

Proof Statements

So far, we have been discussing inference as a type of cognition. In contrast, a proof statement is a way of inducing an inferential cognition in another person, and as such, it is also called inference for another (parārthānumāna). In the classical formulation found in Indian Buddhist texts, a proof statement has an intriguing feature: the thesis to be proven is not explicitly stated. Instead, one lays out the ingredients, so to speak, that should lead one’s listener themselves to have a kind of “aha” moment in which they come to realize the truth of the unstated—but clearly implied—thesis.
The proof statement begins with a statement of the pervasion, and generally, just the positive pervasion is stated. Consider the case where one wishes to induce an inference in another person such that their mortality is recognized. One would begin by stating the positive pervasion: “Whoever is human (E) is mortal (P).” Importantly, the pervasion itself must be accompanied by a concrete case that is acknowledged by the discussants as exemplifying the pervasion. And so one might add “as in the case of Socrates.” One of the purposes of offering an exemplifying case is that it compels a consensus about the pervasion, but it also prevents one from appealing to types of reasoning that cannot be grounded in experience. Having stated the pervasion and its exemplification, one then states the relationship between the subject of the evidence: “You (S) are human (E).” If one’s interlocutor has fully accepted the pervasion (“Whoever is human is mortal”), then simply by saying “You are human,” one should be able to directly cause them to experience an inferential cognition whose content is “I am mortal.” Traditionally, one then ends the proof statement by noting the type of evidence used; in this particular case, one would say, “The evidence used here is nature evidence.”

A proof statement, or inference for another, is often deployed in the context of philosophical debate, and historically, it seems that kings and other benefactors would actually sponsor formal debates between philosophers from different traditions. Since institutions large and small—temples and monasteries, for example—required considerable support, the motivation to win such debates was considerable. Indian philosophers in multiple traditions, including Buddhists, discussed at length the ways one can win debates by any means possible, even by confusion and misdirection. Yet the Buddhist approach to proof statements suggests that the main goal is not simply to win. Instead, the point of the proof statement is to provide another person with the information that they can use to come to a particular understanding on their own. In other words, to return to the example above, one does not directly say, “You are mortal!” Instead, one provides the conditions for the other person to conclude, “Ah, I see—I must be mortal.” The goal is thus not to simply win the debate by proving the ­interlocutor to be wrong. Rather, one seeks to induce an understanding in the other person in a fashion that might even be characterized as therapeutic.

One problem, however, is that this approach to proof statements does not enable one to engage with persons who hold beliefs about completely unreal entities. Recall that the evidence (E) must be observed as a quality or feature of the subject (S) or thing under discussion. But what if that subject is a unicorn? Or a square circle? Or what if the pervasion—the relation between the evidence and the predicate to be proven—that a person holds to be true actually has no exemplifying case, precisely because it is impossible? I might absurdly believe that if something is a square circle, then it is necessarily blue, but the theory of proof statements presented here does not enable one to dissuade someone from holding such beliefs. In such cases, one must turn to another mode of argumentation: the consequence.


As we have noted, the Indian Buddhist traditions that form the core of the volumes in this series are often structured in terms of an ascending scale of analysis, and the highest or subtlest level of analysis is a version of the antirealist Madhyamaka known as the Prāsaṅgika—literally, the Consequentialists. The problem of trying to bring others to a particular understanding of one’s own position is especially acute for the Consequentialists because, to put it in simple terms, Consequentialists hold that all ordinary persons are caught up in the delusion that things truly exist in the way that they appear. Thus, as an ordinary person, when I see a pot on the table in front of me, it seems to be objectively real—from its own side, so to speak—without in any way being dependent on the way I see it, or even on other causes and conditions. It simply exists on its own. For the Consequentialists, however, that kind of objectively real, autonomous pot does not exist at all, yet if they are to lead me to that understanding, how could they use a proof statement? For me, there is a truly real thing called a pot in front of me, but from their perspective, I might as well be saying that there is a square circle or a married bachelor in front of me. How then can we have a discussion about it? The fault of having no subject at all (āśrayāsiddha) would continually apply. This is where the style of reasoning that employs consequences (prasaṅga) plays a crucial role.

The case of the Consequentialists is an especially obvious one where consequential reasoning must be employed, but even issues at a lower level of analysis may best be conveyed to a conversation partner by using consequences. Our authors thus choose a simpler example—the failure to see that a causal entity, which comes from causes and produces effects, must necessarily be impermanent or in flux. Using this example, our authors offer a detailed discussion of reasoning through consequences, including a schematic analysis of eight different ways the pervasion (the relationship between the evidence and the predicate) can be configured. Without reproducing their rich analysis, we can simply note some key features of this style of reasoning.

Overall, consequential reasoning begins with the context of a false belief, for example, that a pot or water jug is permanent or unchanging. This harkens back to our ordinary intuition: when we see a carafe of water on the table, we do not generally believe that it is constantly changing in each moment. The Consequentialists would focus especially on the problem of the pot itself—why do we think that there is an objectively existent entity called a pot in the first place? But at a lower level of analysis, where the existence of something we call a pot is accepted without much contention, consequential reasoning addresses especially the relationships among the various properties that we attribute to it. For example, in a quite straightforward way, we might believe that a pot is permanent, and yet it is capable of changing state, such that in one moment it is empty, and a few moments later, it is filled with water. In this context, consequential reasoning brings us to an “aha” moment in which we recognize the way that this belief is incompatible with what we experience: obviously, a jug filled with water and one that is empty are not quite the same, even in terms of being a jug. The point of consequential reasoning, however, is not to force this reasoning upon one’s interlocutor directly. Instead, and again in an ­arguably therapeutic fashion, one simply enables one’s interlocutor to arrive, on their own, to the realization that their belief was false. Here again we see how the Buddhist approach to many issues, including styles of reasoning, is inextricably linked to the practical task of personal transformation.


John Dunne

"Overall, consequential reasoning begins with the context of a false belief, for example, that a pot or water jug is permanent or unchanging."

This, the second volume in the Science and Philosophy in the Indian Buddhist Classics series, focuses on the science of mind. Readers are first introduced to Buddhist conceptions of mind and consciousness and then led through traditional presentations of mental phenomena to reveal a Buddhist vision of the inner world with fascinating implications for the contemporary disciplines of cognitive science, psychology, emotion research, and philosophy of mind. Major topics include:

– The distinction between sensory and conceptual processes and the pan-Indian notion of mental consciousness

– Mental factors—specific mental states such as attention, mindfulness, and compassion—and how they relate to one another

– The unique tantric theory of subtle levels of consciousness, their connection to the subtle energies, or “winds,” that flow through channels in the human body, and what happens to each when the body and mind dissolve at the time of death

– The seven types of mental states and how they impact the process of perception

– Styles of reasoning, which Buddhists understand as a valid avenue for acquiring sound knowledge

In the final section, the volume offers what might be called Buddhist contemplative science, a presentation of the classical Buddhist understanding of the psychology behind meditation and other forms of mental training.

To present these specific ideas and their rationale, the volume weaves together passages from the works of great Buddhist thinkers like Asaṅga, Vasubandhu, Nāgārjuna, Dignāga, and Dharmakīrti. His Holiness the Dalai Lama’s introduction outlines scientific and philosophical thinking in the history of the Buddhist tradition. To provide additional context for Western readers, each of the six major topics is introduced with an essay by John D. Dunne, distinguished professor of Buddhist philosophy and contemplative practice at the University of Wisconsin. These essays connect the traditional material to contemporary debates and Western parallels, and provide helpful suggestions for further reading.


There are no products in your cart.