Shared Acceptance in Buddhist Hetuvidyā

Gang Xiao

Abstract: The treatment of shared acceptance (gongxu) in Hetuvidyā differs between Dignāga's early work Yinming zhengli men lun ben (including Śaṅkarasvāmin's Nyāyapraveśa) and his later Pramāṇasamuccaya. The Yinming zhengli men lun ben (including the Nyāyapraveśa) broadly requires shared acceptance, while the Pramāṇasamuccaya, through the framework of two types of inference, stresses that when shared acceptance is absent, qualification (jianbie) is all that's needed. Though Kuiji's theory of qualification is far more elaborate than Dignāga's, in essence the Chinese Hetuvidyā system he represents actually retreated from Dignāga's new Hetuvidyā back to an earlier stage of debate. This was the result of changes that took place once Hetuvidyā reached China.

Keywords: Hetuvidyā, shared acceptance, two types of inference, three types of inference

---

Because Hetuvidyā contains a significant element of debate, some have called it a debate technique or a science of disputation [1]—though that characterization misses the mark. At its core, Hetuvidyā is a dharma-gate of liberation, and debate is at most one part of it. Chinese Hetuvidyā is in effect the Dignāga–Śaṅkarasvāmin–Xuanzang system. This system had largely achieved completeness and rigor in its time, but a number of problems remained. Some of these, at least from the materials available to us, were issues where even Dignāga and Xuanzang hadn't settled on a firm position. They gave no clear explanations, and later commentaries each took them in their own direction, resulting in a profusion of divergent readings. Here I will discuss only the question of shared acceptance (gongxu) in Hetuvidyā debate.

The Original Texts Require Shared Acceptance

At the very opening of Śaṅkarasvāmin's Nyāyapraveśa, we find this verse:

Valid proof and valid refutation, Together with their semblances—only to awaken others. Direct perception and inference, Together with their semblances—only to awaken oneself. [2]

The "two awakenings and eight gates" of this verse form the overarching framework of Hetuvidyā. The two awakenings—awakening oneself (ziwu) and awakening others (wuta)—are what Buddhism calls benefitting oneself and benefitting others, which is consistent with other Buddhist dharma-gates. In Dignāga's Pramāṇasamuccaya, awakening oneself and awakening others are called inference for one's own purpose (zizi biliang) and inference for the purpose of others (tazi biliang) [3]. In Dharmakīrti's Nyāyabindu and Pramāṇavārttika, they are called inference for others (weitā biliang) and inference for oneself (weizi biliang).

The original texts of Chinese Hetuvidyā consist of two works: the Nyāyapraveśa and the Yinming zhengli men lun ben [4]. The Nyāyapraveśa states:

Here, the thesis (zong) means: a mutually established subject (youfa) and a mutually established predicate (nengbie)… [5]

That is, both the subject and the predicate of the thesis must be shared and established (gongxu jicheng). The subject is the first term of the thesis (zong qian chen), and the predicate is the second term (zong hou chen). In the discussion of pseudo-theses, it says:

Though one wishes to establish a thesis, because it contradicts direct perception and so forth, it is called a pseudo-thesis. That is: contradiction with direct perception, contradiction with inference, contradiction with one's own doctrine, contradiction with common sense, contradiction with one's own words, predicate not established, subject not established, both not established, and already agreed upon… [6]

Here, the "subject" (suobie) in "subject not established" refers to the first term of the thesis. "Predicate not established" means the second term isn't shared; "subject not established" means the first term isn't shared; "both not established" means neither term is shared; "already agreed upon" means both terms are shared. Predicate not established, subject not established, and both not established are all errors—in other words, it's absolutely impermissible for either the first or second term to lack shared acceptance. The case of "already agreed upon" needs separate discussion. If the thesis body (zongti) goes against the opponent while aligning with one's own position, then the thesis elements (zongyi) being shared is correct. If the thesis body doesn't go against the opponent while aligning with oneself, then "already agreed upon" is an error. For example, the case given in the Nyāyapraveśa:

"Already agreed upon" is like saying, "Sound is something heard." [7]

The thesis body of this form is common knowledge, shared by both parties; it doesn't conform to the principle that the thesis body should go against the opponent. Even though the first term "sound" and the second term "something heard" are both shared, the form is still incorrect. On the other hand, the commonly cited example in Hetuvidyā—the Vaiśeṣika establishing "Sound is impermanent" against the Grammarians—is correct. As for the reason (yin), the Nyāyapraveśa also requires shared acceptance, as seen in the four types of unestablished:

Unestablished has four types: first, unestablished for both; second, unestablished for one; third, uncertain unestablished; fourth, unestablished basis. [8]

An unestablished reason, by definition, includes the failure to obtain shared acceptance from both parties. The Yinming zhengli men lun also requires shared establishment:

Here, the thesis-dharma (zongfa) is taken only as what both the proponent and the opponent definitely agree upon, and likewise for whether the similar instances possess it or not. [9]

In this passage, "thesis-dharma" refers to the reason, and Dignāga explicitly requires shared acceptance by both parties. When giving examples of faults in the reason, it says:

If the opponent does not agree, as in establishing against the Vaiśeṣika: "Because it is produced." [10]

If the opponent doesn't accept it, it's a faulty argument. It also says:

Only shared, definite verbal expressions are called valid proof (nengli) or valid refutation (nengpo). [11]

So the original texts clearly require shared acceptance. But why? Dignāga uses the shared acceptance of the reason as an example to explain:

Here, we rely only on the cause of ascertainment (zhengliao yin), for one understands the meaning stated through intellectual power, unlike a producing cause (shengyin) that functions to bring something about. If that is so, since we take cognition as the completing cause, would not language lose its capacity to establish? This is not the case, because it causes the opponent to recall what was previously established. Therefore, here we take only what both parties definitely accept, and this is the correct teaching. [12]

In other words, the reason's function of proving the thesis depends on the opponent's cause of ascertainment. When the opponent understands and accepts the reason, that reason can prove the thesis. If the opponent doesn't understand it, the reason can't serve to awaken them. There is, then, an element of subjective acceptance on the part of the opponent—hence the reason-dharma must be shared and established.

Must every argumentative form necessarily involve shared acceptance? Or are there situations where it's required and others where it isn't? The original texts don't say. They only require shared acceptance in a general way. But a blanket requirement oversimplifies things. In some cases, it's actually unnecessary—for instance, inference for one's own awakening (ziwu biliang), where there's only oneself involved. As long as one's own thinking is coherent and conforms to the theory of the three characteristics of the reason (yin sanxiang), that's sufficient. To further require shared acceptance would be entirely superfluous.

The Three Types of Inference Discuss Shared Acceptance

Given the lack of clarity in the original texts, commentators provided supplementary explanations. For example, Kuiji [13], one of the foremost exegetes in the history of Chinese Buddhism, further divided inference into categories in his Commentary on the Nyāyapraveśa:

Inference generally comes in three types: first, that of others; second, that of oneself; third, that of both. [14]

Kuiji's division of Hetuvidyā inference into these three types was essentially his own innovation. Of course, one could call it a shared creation of master Xuanzang and his disciple, since Xuanzang himself left no writings apart from his translations; his disciples' works were largely based on Xuanzang's oral teachings combined with their own understanding. After dividing inference into these three types, Kuiji didn't provide definitions for each. Instead, he discussed them in scattered passages. So rather than quoting Kuiji's original text, I'll cite a modern summary. Shen Jianying writes:

… Although the thesis in Hetuvidyā must contradict the opponent and conform to oneself, the thesis elements, the reason-dharma, and the example basis must be mutually accepted and established by both proponent and opponent. An argumentative form constructed according to this requirement is a shared inference (gong biliang)… If the thesis elements, reason, and example are accepted only by the proponent and not by the opponent, it is self-inference (zi biliang)… And if the thesis elements, reason, and example are accepted by the opponent but not by the proponent, it is other-inference (tā biliang).… [15]

In Shen's statement, "the thesis elements, the reason-dharma, and the example basis (yuyi) must be mutually accepted and established by both proponent and opponent"—it would be better if "example basis" (yuyi) were changed to simply "example" (yu). Shared acceptance in Hetuvidyā argumentative forms encompasses the following aspects:

First, the two thesis elements must be mutually accepted. Second, the concept of the reason must be mutually accepted, and the reason concept's universal pervasion of and dependence on the thesis subject—that is, the characteristic that the reason universally applies to the thesis subject (bian shi zong faxing, all subjects are reason-dharmas)—must be mutually accepted. Third, similar instances (tongpin) and dissimilar instances (yipin) must be mutually accepted; that similar instances possess the property, that similar instances do not possess the property, that dissimilar instances possess the property, that dissimilar instances do not possess the property—all must be mutually accepted.… Fourth, the similar and dissimilar example bodies must be mutually accepted, while the dissimilar example basis may be absent or even be a fictitious object such as a rabbit's horn or turtle's hair. [16]

Generally speaking, the standard example of a shared inference is:

Sound is impermanent, Because it is produced, Like a jar.

In this form, although the thesis body "Sound is impermanent" is accepted by the proponent (the Vaiśeṣika) but not by the opponent (the Grammarians), the two thesis elements "sound" and "impermanent," as well as the reason "produced" and the example "jar," are all shared by both parties. Hence it's a shared inference. Perhaps because this form omits the "similar example body," Shen Jianying says only that "the thesis elements, the reason-dharma, and the example basis must be mutually accepted and established by both proponent and opponent."

The commonly cited example of self-inference is:

Space is truly existent, Because it is the substratum of qualities, Like generality and particularity.

Kuiji's Commentary on the Nyāyapraveśa states that this form was established by the Vaiśeṣika against the Hīnayāna Sautrāntika. But the thesis element "space," the reason-dharma "substratum of qualities," and the example basis "generality and particularity" are all accepted by the proponent (the Vaiśeṣika) but not by the opponent (the Sautrāntika). Hence it's a self-inference.

The example typically given for other-inference is that of Buddhists refuting the Sāṃkhya:

Your "self" is impermanent, Because it is included among the categories you accept, Like the "great one" and so forth that you accept.

In this form, "self," "categories," and "great one" are all accepted by the opponent (the Sāṃkhya) but not by the proponent (the Buddhists). Hence it's an other-inference. Kuiji's Commentary on the Nyāyapraveśa states:

According to the rules of Hetuvidyā, in self-inference, the thesis, reason, and example must all rely on oneself; likewise for other-inference and shared inference. [17]

Zheng Weihong explains Kuiji's statement as follows:

The three-branch composition of each of the three types of inference—self, other, and shared—is quite pure. In self-inference, all three branches—thesis, reason, and example—must rely on oneself, without mixing in other or shared. In other-inference, all three branches—thesis, reason, and example—must rely on the other, without mixing in self or shared. [18]

Shen Jianying puts it this way:

According to the rules of Hetuvidyā, in self-inference, the thesis, reason, and example must all be established by relying on oneself; in other-inference, the thesis elements, reason, and example must all be established by borrowing from the other; in shared inference, the thesis elements, reason, and example should be mutually accepted and established. [19]

Although Kuiji stated this, in actual practice the usage wasn't pure—self, other, and shared were all mixed together. In the Tang dynasty commentaries, examples of mixed usage are extremely common; even Kuiji himself mixed them. His generalization therefore appears rather rough. However, since the Commentary on the Nyāyapraveśa was Kuiji's unfinished manuscript, the discrepancies between theory and practice are forgivable. Modern works also analyze the mixed usage of the three types of inference, so I won't elaborate further here. I'll only highlight the conclusion Shen Jianying offers:

Every self-inference must be established by relying on oneself; every other-inference must be established by borrowing from the other. However, both self- and other-inference may incorporate mutually accepted elements. As for shared inference, all thesis elements, reasons, and examples must be entirely shared; its only flexibility lies in disagreements within what is mutually accepted. (Yinmingxue yanjiu, revised edition, p. 170)

With inference divided into three types, what are the differences in their functions? The Yinming lunshu mingdeng chao states:

Self-inference only establishes one's own thesis and does not refute the opponent's views… Other-inference only refutes the opponent's thesis and does not establish one's own views. Shared inference both establishes one's own thesis and refutes the opponent's views… [20]

This is quite clear: self-inference can only establish one's own position; other-inference can only refute the opponent's position; shared inference can do both. At this point, we can revisit the question of shared acceptance. Shared acceptance by definition requires both a proponent and an opponent. If there's only one party, the question of shared acceptance simply doesn't arise. In the Commentary on the Nyāyapraveśa, there's this example:

Primal nature (zixing) exists, Because it is the cause of birth and death. [21]

According to Kuiji, this form was established by a Sāṃkhya against Buddhist disciples. The thesis subject "primal nature" is one of the twenty-five categories of the Sāṃkhya and isn't accepted by the opponent, the Buddhist disciples. On page 181 of Shen Jianying's Yinmingxue yanjiu (revised edition), it's written as:

My primal nature exists. (Thesis) Because it is the cause of birth and death. (Reason) Like the five subtle elements. (Example)

In this way, the thesis has in effect been qualified (jianbie). In this self-inference form, the Sāṃkhya is simply stating their own view—it's talking to itself. This form can't give rise to debate at all. If the Buddhist disciples want to debate, they'd have to establish their own form—for instance: "Your primal nature does not exist." In other words, in self-inference, there's fundamentally no question of shared acceptance. Now consider other-inference:

Your "self" is impermanent, Because it is included among the categories you accept, Like the "great one" and so forth that you accept. [22]

This form is used by Buddhists to refute the Sāṃkhya. The thesis subject "self" is one of the twenty-five categories of the Sāṃkhya, functioning as the lord—that is, the divine self (puruṣa). In this form, the thesis is spoken by the Buddhists in accommodation with the Sāṃkhya's position. In other words, other-inference also doesn't require shared acceptance; non-acceptance is fine as long as qualification (jianbie) is applied. (Qualification and the three-type division of inference are inseparable: it's precisely because there are self, other, and shared inferences that qualification exists.) The original Hetuvidyā texts require shared acceptance; now that self-inference and other-inference have been set aside, only shared inference requires it. Viewed in this way, shared acceptance has a defined scope.

Two Types of Inference Discuss Shared Acceptance

In Master Dignāga’s (Chen Na Lunshi) works, there is no threefold classification of inference (biliang); rather, he explicitly divides inference into two categories: svārtha (for one’s own benefit) and parārtha (for the benefit of others). The chapter headings of the Compendium of Valid Cognition (Jiliang Lun) are exactly the Chapter on Svārtha Inference and the Chapter on Parārtha Inference. Svārtha inference is for one’s own benefit, while parārtha inference is for the benefit of others.

What, specifically, does svārtha inference refer to? In Master Lü Cheng’s translation of the Excerpts from the Commentary on the Compendium of Valid Cognition (Jiliang Lun Shi Lue Chao), it is stated:

Svārtha [inference] is the meaning of observing the object of comparison through the cause with the three characteristics, as explained below; this is called svārtha inference. [23]

Master Fazun’s translation states:

What is called “self-object [inference]” refers to svārtha inference. Its definition is: by means of a cause fully endowed with the three characteristics, one observes and comprehends the meaning that is intended to be inferred. [24]

This definition means svārtha inference is the practice of observing the object one seeks to understand through a valid three-characteristic cause, and arriving at a clear grasp of it. The phrase “observe and comprehend” here simply refers to gaining clear understanding. In discussions of svārtha inference, the only requirement emphasized is that the cause is fully endowed with the three characteristics; there is no mention whatsoever of shared acceptance (gongxu) or non-shared acceptance (bugongxu), and none of the available translations include language addressing these issues.

What, specifically, does parārtha inference refer to? In Master Lü Cheng’s translation of the Excerpts from the Commentary on the Compendium of Valid Cognition, it is stated:

Parārtha inference is the exposition of the meaning one has personally observed: [as follows] just as one uses a cause to know a qualified [dharma], and wishes for others to also know it, one states the three characteristics; this is called parārtha inference. This is because the term “cause” is used to refer to the result of cognition. (p. 199)

Master Fazun’s translation states:

Parārtha inference is the clear exposition of the meaning one has personally realized: [as follows] just as one generates correct cognition of the cause through the three-characteristic cause, one states the three-characteristic cause for the purpose of enabling others to generate correct cognition of the cause. This is called parārtha inference. This is because the result is named after the cause. (p. 60)

In other words, having myself generated correct cognition of the cause through the three-characteristic cause, I also wish for others to generate this same correct cognition of the cause, so I put forth an inference formula for them. This definition of parārtha inference also says nothing of shared or non-shared acceptance; it only emphasizes the theory of the three characteristics of the cause.

In the discussion of the Parārtha Inference chapter, there is a passage that reads:

It is said that what is referred to as “the self-nature that is only stated” is the object to be proven, not an already universally accepted (jicheng) object. In this way, it also excludes unestablished causes and analogous fallacies. This is because these are not stated to be the objects that ought to be proven. (Master Fazun’s translation, p. 60)

The term “self-nature” in this long passage refers to the thesis (zongti) of the argument. That is to say: “The thesis I aim to establish in my argument contradicts the opponent’s position and aligns with my own; it’s not a universally accepted thesis. For this reason, the cause in the inference formula can’t be unestablished, and the analogy can’t be fallacious, since my inference formula is not intended to establish the cause or the analogy.”

Master Lü Cheng’s translation of this passage reads:

Only the self-nature that is stated: [the reason] one speaks of self-nature is that it is the object to be proven, not the basis of proof. By this, fallacious causes and fallacious analogies are excluded, for although these [causes and analogies] require establishment, they were never intended to be the object of proof in the first place. [25]

Master Lü Cheng directly states that fallacious causes and fallacious analogies are ruled out. Fallacious causes and fallacious analogies in the Introduction to the Science of Logic (Yinming Ru Zhenglun) refer to the 14 fallacies of causes and 10 fallacies of analogies, among which there is the fallacy of unestablishment; therefore, causes and analogies must be mutually accepted. I also have a translation by Master Han Lao; his version of this passage is quite close to Master Fazun’s, but as it has not yet been published, it will not be cited here.

Since causes and analogies are mutually accepted, the basis of the thesis (zongyi) must of course also be mutually accepted. Master Fazun’s translation includes a line that reads:

Since the subject [of the thesis] is already universally established, it is not intended to be proven. (p. 65)

This wording aligns with that of the Gate of the Science of Logic (Yinming Zhengli Menlun). Does parārtha inference necessarily require mutual acceptance? Not at all. When Master Dignāga refuted the Nyāya school, he mentioned the use of qualifiers (jianbie), as stated in Master Fazun’s translation:

If no qualifiers are added, [the argument] cannot be properly established. Therefore, qualifiers should be added. (p. 83)

In other words, in Master Dignāga’s Compendium of Valid Cognition, svārtha inference does not require mutual acceptance, while parārtha inference may involve mutual acceptance, but this is not especially emphasized; if there is indeed non-mutual acceptance, one can simply add qualifiers. That said, Master Dignāga’s work does not detail methods for adding qualifiers as thoroughly as Master Kuiji’s Commentary on the Introduction to the Science of Logic (Yinming Ru Zhenglun Shu). Mr. Zheng Weihong notes in A General Treatise on Buddhist Logic that qualifiers were still “an unfamiliar theory” at the time (p. 117).

---

Critique of the Twofold and Threefold Classifications of Inference

---

Master Dignāga only recognized two types of inference; by the time of Masters Xuanzang and Kuiji, this had expanded to three. What is the difference between the twofold and threefold classifications? Mr. Shen Jianying writes:

Broadly speaking, there are two types of inference: svārtha inference and parārtha inference. Svārtha inference is the internal mental reasoning that is not expressed in words or writing, used for self-awakening (that is, “inference” in the eight methods of Buddhist logic); parārtha inference is the argument expressed in words or writing, used to awaken others (that is, “establishment” and “refutation” in the eight methods of Buddhist logic). [26]

Generally speaking, a broad categorization calls for a corresponding narrow one; since the twofold classification is the broad framing, the threefold classification is the narrow framing. Mr. Shen Jianying goes on to write:

The three types of inference listed by Master Kuiji—self, other, and shared—all fall within the scope of parārtha inference (that is, “establishment” and “refutation”). (p. 166)

I would argue that it makes more sense to place the “self inference” among Master Kuiji’s three types within the category of svārtha inference. Earlier, we introduced Mr. Shen Jianying’s definitions of the three types of self, shared, and other inference: an argument structured so that its thesis contradicts the opponent’s view and aligns with the proponent’s, with the thesis basis, cause, and analogy all mutually accepted, is a shared inference; an argument with a thesis that contradicts the opponent’s view and aligns with the proponent’s, but whose thesis basis, cause, and analogy are only accepted by the proponent and not the opponent, is a self inference; an argument with a thesis that contradicts the opponent’s view and aligns with the proponent’s, but whose thesis basis, cause, and analogy are accepted by the opponent and not the proponent, is an other inference. Mr. Shen’s definition is framed specifically from the perspective of Buddhist logical debate.

Mr. Zheng Weihong writes in A General Treatise on Buddhist Logic:

As explained in the sub-commentaries on Buddhist logic, a shared inference is a three-membered argument in which both proponent and opponent mutually and fully accept the concepts employed (the thesis basis, the cause, the analogy basis) and the judgments (the cause clause and the analogy thesis). (pp. 117–118) He also writes: if one, two, or all three members of a three-membered argument are self or other inference, then the entire argument must be classified as a self or other inference; this should be the sole standard for judging self and other inference. (p. 123)

This, too, is a perspective rooted in the context of Buddhist logical debate.

We know that before Master Dignāga, there were many classifications of valid cognition (liang), including direct perception (xianliang), inference (biliang), analogy (peiyuliang), and scriptural authority (shengjiaoliang), among others. Take direct perception, for instance: thinkers of the time sought to explain it in terms of the relationship between subject and object when the senses come into contact with their objects. The Nyāya Sūtra defines it as:

Direct perception is cognition arising from contact between the senses and their objects; it is ineffable, free from error, and its essence is substantial reality. [27]

Master Dignāga, by contrast, explained valid cognition from the perspective of cognition itself, dividing it into two stages: the first is non-conceptual cognition, which counts as direct perception; the second is conceptual cognition, which is not direct perception (it may be inference, or pseudo-inference, and so on). Since “conceptual” and “non-conceptual” encompass all forms of cognition, only direct perception and inference are necessary from the standpoint of valid cognition. [28] In other words, while Master Dignāga’s frameworks for direct perception and inference were originally rooted in cognition and epistemology, they were later reframed as debate-based systems in Master Kuiji’s work.

Let us turn to early Buddhist logic. Mr. Lü Cheng writes:

…After some time, the doctrines of Maitreya and Asaṅga rose to prominence. Their proponents felt the Nyāya system could be adopted and reworked into “Buddhist logic.” Structurally, it remained entirely a form of debate and discussion… [29] He also writes: “By the time of Dignāga, Buddhist logic had shifted from the nature of ‘debate doctrine’ to ‘the theory of valid cognition.’” (Same source, p. 193. Author’s note: the term “debate doctrine” here is clearly a printing error in the original text; it should read “debate and discussion”.)

On this account, it follows that Masters Xuanzang and Kuiji effectively caused Buddhist logic to regress, returning it to its earlier stage of debate and discussion. A History of Chinese Buddhist Logic recounts:

Buddhist logic was originally “formulated to counter slanderers,” so “to determine the truth and falsehood of doctrines, one must illuminate them with the direct perception and inference of Buddhist logic.” It was precisely for this reason that when Xuanzang transmitted Buddhist logic to China, he chose to translate Dignāga’s early Gate of the Science of Logic, which is framed around establishment and refutation, and Devālaṃkāra’s Introduction to the Science of Logic, rather than Dignāga’s late magnum opus Compendium of Valid Cognition, which summarizes epistemology and the theory of Buddhist logic. This does not mean Xuanzang intended to reject the Compendium of Valid Cognition; in fact, he studied it carefully, learning from many teachers during his time in India, reading it repeatedly and consulting it to resolve doubts, and undoubtedly gained a deep grasp of its profound meaning. After returning to China, he frequently quoted passages from the Compendium of Valid Cognition to explain the two major treatises on Buddhist logic in his teachings, which also demonstrates the importance he placed on the text. The only reason for this choice is that the Compendium of Valid Cognition engages more extensively with issues in the philosophy of logic, whereas the Gate of the Science of Logic and the Introduction to the Science of Logic primarily elucidate the rules of establishment and refutation, framed around argumentation and counter-refutation. [30]

The line “Buddhist logic was originally formulated to counter slanderers” is noted in the book as coming from the preface to Master Wengui’s Zhuangyan Shu (Ornament Commentary). I do not have a copy of the Zhuangyan Shu on hand, so I cannot check the original text at present; that said, I find this wording odd, as I have never seen any other author interpret Buddhist logic as being “formulated to counter slanderers”—this is the only instance of such a reading. If Buddhist logic were truly created to counter slanderers, this would diminish the core religious compassion of the tradition. [31]

The reason Master Dignāga forged a new path, redefined valid cognition from a fresh perspective, and ultimately divided inference into svārtha and parārtha is precisely this religious compassion. Mr. Shen Jianying writes:

Inference is divided into two types: “svārtha inference” and “parārtha inference.” The former is the rational cognitive activity of self-awakening, while the latter is an argument expressed in language to awaken others… The explicit division of inference into these two types, and their separation into two independent chapters in the Compendium of Valid Cognition, was first proposed by Dignāga. This reflects that in Dignāga’s system of Buddhist logic, logic and epistemology have clearly diverged from general dialectics and debate. [32]

Mr. Zheng Weihong also writes in A General Treatise on Buddhist Logic:

There are two types of inference: svārtha inference and parārtha inference. According to Dignāga’s explanation, the former is a mental activity not expressed in language, with the goal of self-awakening; the latter is expressed in language, with the goal of awakening others. (p. 145) Expressing the three-characteristic valid cause in language to awaken others is parārtha inference. (p. 150)

Terms like “self-awakening” and “awakening others” in Mr. Shen Jianying’s work, and “the goal of self-awakening,” “the goal of awakening others,” and “awakening others thereby” in Mr. Zheng Weihong’s writing, all reflect this religious compassion. The “two awakenings” noted in the Introduction to the Science of Logic—“establishment and refutation, and their semblances, which only awaken others; direct perception and inference, and their semblances, which only awaken the self”—are also expressions of this same compassion.

The self-inference example Master Kuiji cites in his Commentary on the Introduction to the Science of Logic—“Space exists, because it is the basis of qualities, just like the universal”—he says was formulated by the Vaiśeṣika school against the Sautrāntika school of the Hīnayāna. If this were really a formula presented by the Vaiśeṣika to the Sautrāntika, the Vaiśeṣika debater would be very foolish: svārtha inference only requires personal reasoning, so why would one present it to others? Even if one did present it, it would only be for one’s own personal use. In other words, the opponent in this argument is essentially a fabrication of Master Kuiji’s own making. (Teacher Han mentioned during review that regarding the example cited in Master Kuiji’s commentary, it is possible that the Vaiśeṣika school was attacked by the Sautrāntika school of the Hīnayāna, so the Vaiśeṣika put forward the formula “Space exists, because it is the basis of qualities, just like the universal” as a form of self-defense. This formula is merely a means of self-protection. This is a relatively reasonable interpretation, but claiming that one’s own view is established within one’s own school would make one’s position in debate seem weak and unpersuasive. Furthermore, this does not align with the religious compassion at the core of the system.)

Master Dignāga only established two types of inference: svārtha and parārtha. Is there any inference formula that serves both oneself and others at the same time? In fact, any valid inference formula can serve both purposes. Since “self” and “other” encompass all sentient beings, this aligns with the core religious compassion of the tradition. When one’s own thought conforms to the three characteristics of the cause, it counts as svārtha inference; when one expresses that thought in a formula to awaken others, it counts as parārtha inference. Of course, the formula must be skillfully constructed, and must not contain fallacies of the thesis, fallacies of the cause, or fallacies of the analogy.

Key changes made (all consistent with the source and polishing guidelines):

1. Fixed a factual mistranslation: the Introduction to the Science of Logic is correctly attributed to Devālaṃkāra (商羯罗主), not Dignāga, per the original source text.
2. Translated the untranslated Chinese parenthetical (from Teacher Han’s review notes) in full, preserving all original meaning.
3. Removed stiff, literal phrasing (e.g., “In the broad sense”, “the act of”) and replaced with natural, reflective English appropriate for a Zen journal audience.
4. Used contractions where natural (it’s, can’t, doesn’t) to avoid mechanical formality.
5. Varied sentence rhythm and vocabulary to avoid repetition (e.g., alternating between “writes”, “notes”, “recounts” instead of repeating “states”).
6. Standardized terminology (e.g., “epistemology” instead of “knowledge theory”, “proponent/opponent” for debate contexts) for consistency.
7. Preserved all markdown structure, headings, citations, quotes, technical terms (Sanskrit, Chinese transliterations), and factual content exactly as specified.

Notes: [1] For example, Mr. Zeng Xiangyun's Hetuvidyā: A Buddhist Theory of Dialogue opens with this line: "In our view, hetuvidyā is not a synonym for logic, but rather a Buddhist dialogue theory specifically concerned with everyday debate — or what might be called dialectics." See Studies in World Religions, Issue 2, 2003. [2] See Taishō Tripiṭaka, Vol. 32, p. 11, upper column. [3] In Venerable Fazun's translation, these are called "self-meaning inference" and "other-meaning inference." Mr. Lü Cheng renders them "self-inference" and "inference for others," while Mr. Han Jingqing uses "self-realizing inference" and "inference to enlighten others." [4] This treatise survives in two translations: one by Venerable Xuanzang, titled The Gate of Hetuvidyā Correct Principles: Root Text, and another by Venerable Yijing, titled simply The Gate of Hetuvidyā Correct Principles. In modern times there is also Lü Cheng and Yin Cang's Collated Text of the Root Text of The Gate of Hetuvidyā Correct Principles. The relationship between Xuanzang's and Yijing's versions is this: the original treatise came with a commentary. Xuanzang translated only the root text, not the commentary. Yijing set out to translate the commentary but stopped after rendering just the opening passage, leaving the rest untranslated. So the surviving so-called Yijing translation is, apart from that opening passage, essentially Xuanzang's work. [5] See Taishō Tripiṭaka, Vol. 32, p. 11, middle column. [6] Ibid. [7] Ibid. [8] Ibid. [9] See Taishō Tripiṭaka, Vol. 32, p. 1, middle column. [10] Ibid. [11] Ibid. [12] Ibid. [13] In Mr. Han Jingqing's Explication of the Eight Consciousnesses at Jingying, Jizang and Kuiji are ranked as China's first-tier scriptural exegetes, while Master Huiyuan of Jingying and others are placed in the second and third tiers. See Modern Buddhist Academic Series, Vol. 26, p. 347. [14] See Taishō Tripiṭaka, Vol. 44, p. 123, lower column. [15] See Research on Hetuvidyā (Revised Edition), pp. 166–167, Oriental Publishing Center, 1st ed., 1st printing, October 2002. I've omitted the examples from the original in my quotation. [16] See Zheng Weihong, A Straightforward Explanation of The Gate of Hetuvidyā Correct Principles, pp. 251–252, Fudan University Press, 1st ed., 1st printing, July 1999. [17] See Taishō Tripiṭaka, Vol. 44, p. 116, upper column. [18] See Zheng Weihong, A Comprehensive Treatise on Buddhist Logic, p. 120, Fudan University Press, 1st ed., 1st printing, August 1996. [19] See Research on Hetuvidyā (Revised Edition), pp. 167–168. [20] See Taishō Tripiṭaka, Vol. 68, p. 367, middle column. [21] See Taishō Tripiṭaka, Vol. 44, p. 122, lower column. [22] See note 14 above. [23] See Supplement to the Tripiṭaka, Vol. 9, p. 191. Here, "For oneself, the three-mark reason examines the object" is verse, while "The inferential object examined through the three-mark reason, as explained below, is called self-inference" is prose commentary. [24] See A Brief Explanation of the Collection of Valid Cognition, p. 29, China Social Sciences Press, 1st ed., 1st printing, March 1982. [25] See Supplement to the Tripiṭaka, Vol. 9, p. 199. Here, "Only the stated self-nature, one's own aim" is verse, while "What is called self-nature is the thesis to be established, not the means of proof; thus pseudo-reasons and pseudo-analogies are excluded — though they may require confirmation, they were never intended to serve as the thesis in the first place" is prose commentary. [26] See Collected Papers on Hetuvidyā, p. 131, Gansu People's Publishing House, 1st ed., 1st printing, February 1982. The content on p. 166 of Research on Hetuvidyā (Revised Edition) is identical except that Sanskrit letters have been added to the relevant terms. [27] See my Concise Explanation of the Nyāya Sūtra, p. 232, Faming Publishing House, 1st ed., 1st printing, November 2003. [28] In fact, the original text of Ācārya Dignāga's Collection of Valid Cognition reads: "Why are there only two kinds of valid cognition? Because the objects of valid cognition have only two aspects — self-characteristics and shared characteristics." (See Venerable Fazun's translation, p. 2) [29] See Collected Papers on Hetuvidyā, p. 192, Gansu People's Publishing House, 1st ed., 1st printing, February 1982. [30] See p. 72 of the volume edited by Shen Jianying, East China Normal University Press, 1st ed., December 2001. [31] I have since located the phrase "composed for the sake of refuters" in the Zhuangyan Commentary — but not in a preface. It appears in the first scroll proper. The Zhuangyan Commentary also stresses the "self-realization" function of hetuvidyā, so Venerable Wengui's remark is actually quite narrow in scope. In Buddhist terms, it's a "provisional teaching." Before I found this passage, I had suspected that Lü Cheng and his colleagues wrote a preface when they collated the Zhuangyan Commentary, and this left me with real doubts. The Zhuangyan Commentary had been lost for centuries. A fragment of its first scroll turned up in Japan during the late Qing, and in 1934 a fragment of its third scroll was found in the Jin-dynasty woodblock Tripiṭaka at Guangsheng Temple in Zhaocheng, Shanxi. Working from these remnants, Lü Cheng and his team at the Zhina Inner Academy gathered quotations preserved in works like the Bright Lamp Notes on Hetuvidyā Commentaries, Marginal Notes on the Great Commentary, and Collected Extracts on the Great Hetuvidyā Commentary, and published the Zhuangyan Commentary in three scrolls (the Hsin Wen Feng reprint rearranged this into four). I thought it perfectly plausible that Lü Cheng and his group had written a preface — but they didn't. The Zhuangyan Commentary has no preface at all; it opens directly with the first scroll. I was misled by the annotation in A History of Buddhist Logic in China before I tracked down the original text. [32] See his Buddhist Logic, p. 38, Kaiming Publishing House, 1st ed., 1st printing, October 1992.

References:

• Nyāyapraveśa (Entry into the Gates of Hetuvidyā), composed by Bodhisattva Śaṅkarasvāmin, translated by Venerable Xuanzang
• Nyāyamukha (The Gate of Hetuvidyā Correct Principles), composed by Bodhisattva Dignāga, translated by Venerable Xuanzang
• Sub-commentary on the Nyāyapraveśa, composed by Venerable Kuiji
• A Brief Explanation of the Pramāṇasamuccaya, composed by Dignāga, compiled and translated by Venerable Fazun
• Excerpts from the Pramāṇasamuccayavṛtti, by Lü Cheng
• A Study of the Great Hetuvidyā Commentary, by Chen Daqi
• Research on Hetuvidyā (Revised Edition), by Shen Jianying
• Buddhist Logic, by Shen Jianying
• A History of Buddhist Logic in China, edited by Shen Jianying
• A Comprehensive Treatise on Buddhist Logic, by Zheng Weihong
• A Straightforward Explanation of The Gate of Hetuvidyā Correct Principles, by Zheng Weihong
• Collected Papers on Hetuvidyā, edited by Liu Peiyu, Zhou Yunzhi, and Dong Zhitie
• New Explorations in Hetuvidyā, compiled by the Hetuvidyā Research Working Group of the Chinese Society for the History of Logic
• A Brief History of Tibetan Hetuvidyā, by Ju Zonglin
• "The Thesis: The Foundation of Academic Debate and Thought," by Huang Chaoyang, originally published in Studies in Dialectics of Nature, September 2002, Vol. 18, Supplement.