5. Prasaṅga and Proof by Contradiction in Bhāviveka, Candrakīrti, and Dharmakīrti
I. Two Types of Reductio ad absurdum
JON BARWISE and John Etchemendy, in their introductory logic textbook, give a handy mnemonic description of what it is to prove not-S by contradiction or, what is the same, by reductio ad absurdum: “To prove not-S using this method, assume S and prove some sort of contradiction.”1 We can see two ways in which we can “prove some sort of contradiction”:
A. Typically, in logical reasoning in philosophy and mathematics, one shows that if S is added to a number of premises P,Q,R, then the conjunction of P,Q,R,S would lead to a contradiction, and then one says that if (or because) P,Q,R are all true, S must be false and hence ¬ S (“not-S”) must be true. This is the type of proof by contradiction that, for example, the Greeks used to prove that the square root of two is irrational. Note that if we wish to prove the truth of S in this way, we must accept that P,Q,R are true. We would in that sense be committed to the truth of P, Q, and R.
B. We can reason that the conjunction of P,Q,R and S leads to a contradiction, and is in that sense internally inconsistent, but nonetheless remain unwilling or unable to say which of the conjoined statements are true and which must be negated. We are only prepared to say that P&Q&R&S implies some contradiction and hence that ¬(P&Q&R&S) is the case. We can leave the matter in suspense and go no further toward saying which of S, P, Q, and R is actually false. This is typically what happens when we feel that a theory, a story, or a testimony is contradictory but abstain from saying where exactly it goes wrong and what is really so. Or, if we know our interlocutor’s mindset and are intent on leading her to determine which of the conjuncts is to be abandoned, we simply present her with the internal inconsistency of her statements, knowing that the truth of P, Q, and R will be so deeply held by this person that she will naturally and inevitably end up doing what we want her to do — that is, to negate S. We, however, need not be committed to the truth of P, Q, and R. Let us speak of this method as being that of “internal inconsistency without commitments.” It is a method that a cross-examining attorney can use to make hostile witnesses reveal what they believe to be true. It suffices to point out that their stories are inconsistent, and then let them do the rest, with at most a few nudges here and there to keep them on track.
There is another variant on method B. Suppose that for a certain single statement S, it could be determined (a priori) that if S were true, ¬S would also follow from it. It then follows that S is not true and hence that ¬S is true. Such is the type of reasoning that we use when we want to show that square circles or barren women’s children do not exist; it is also used to show the falsity of certain self-refuting statements like “Everything I say is false.” In short: if so-and-so were a barren woman’s child, then she also would not be a barren woman’s child, and hence she isn’t. Note that here we prove ¬S simply from the internal inconsistency of S; other statements need not be accepted as true.
It is my contention that Bhāviveka, and Svātantrikas in general, promoted method A and that, in its essentials, method A is what is involved in moving from a prasaṅga (absurd consequence) to a prasaṅgaviparyaya — the contraposition of the consequence, in which it must be true that the reason qualifies the subject (pakṣa) and implies the property being proved (sādhyadharma). Candrakīrti in his famous debate with Bhāviveka is, however, advocating B, and this is what he himself seems to favor as a Prāsaṅgika. Finally we will turn to Dharmakīrti’s Pramāṇavārttika 4.12 and the related passages at the beginning of his Pramāṇaviniścaya 3, where Dharmakīrti describes how prasaṅga and prasaṅgaviparyaya operate. There he elaborates method A; interestingly enough, he used B but denied that it could yield any real knowledge.
II. Bhāviveka
We begin with the sixth-century thinker Bhāviveka.2 He accuses an earlier coreligionist, Buddhapālita, of wrongly commenting on Mūlamadhyamakakārikā 1.1, the programmatic verse that, inter alia, denies that anything can be produced from, or by, itself. Buddhapālita’s misstep was supposedly that he formulated the refutation of production from self as a faulty prasaṅga (reductio ad absurdum), one that when contraposed will imply a statement that no Mādhyamika can accept. The argument in Bhāviveka’s own commentary on the Madhyamakakārikā, his Prajñāpradīpa, and hence in the first chapter of Candrakīrti’s Prasannapadā, is often talked about but still baffles many. To see that prasaṅga and its contraposition in the needed detail, we need to backtrack a bit and go step by step.
For starters, what Buddhapālita said is this:
na svata utpadyante bhāvāḥ / tadutpādavaiyarthyāt / atiprasaṅgadoṣāc ca /
Things/entities are not produced from themselves because their production would be pointless [i.e., futile] and because there would be the absurdity [of an infinite regress of productions].
If this is taken as a prasaṅga, then, at least according to Bhāviveka’s understanding, we get:
It would follow absurdly that things are produced pointlessly and without end, because they are produced from themselves.
Now, this prasaṅga, according to Bhāviveka, must yield the following viparyaya, or contraposed reasoning:
Things are not produced from themselves because their production has a point and has an end.
From here it is a short step to a statement that Mādhyamikas cannot accept, for given that things are produced but not from themselves, they must be produced from something else. And that would result in siddhāntavirodha (contradiction with one’s philosophical tenets) in that it would contradict the fundamental Madhyamaka philosophical stance that the negation must be nonimplicative (prasajyapratiṣedha) and that things are neither produced from themselves nor from something else.3
Bhāviveka’s logic is in essence as follows. We all know as true that (1) things’ production has a point and has an end (it does not go on and on over and over again). We also all know that (2) if something existent were produced from itself, it would be produced pointlessly and over and over again. If we add the opponent’s principle (3) that things are produced from themselves, we get the absurdity stated in the prasaṅga: things’ production would be pointless and endless.
Clearly for Bhāviveka the first two statements must be conserved and are true, so that it is the third that is negated. And that is what we find in the prasaṅgaviparyaya: the truth of the first and second are established by pramāṇas (sources of knowledge), and that is why the third must be negated.
To look at Bhāviveka’s argument in Indian terms, the first statement is cited as the hetu (reason) in the prasaṅgaviparyaya, and the second is the vyāpti, the entailment or, more literally, “pervasion.” Although the second statement (i.e., the vyāpti) is not explicitly given by Bhāviveka himself in either the prasaṅga or the viparyaya, it is certainly accepted by him as true. Indeed, not stating vyāpti explicitly is nothing surprising: it is quite routine in Indo-Tibetan argumentation.4 In any case, the vyāpti unpacks as the following universal generalization: “if something existent were produced from itself, it would be produced pointlessly over and over”; its logically equivalent contrapositive formulation is “if anything existent is not produced pointlessly over and over — if its production has a point and an end — then it is not produced from itself.” These statements are essential to the prasaṅga and prasaṅgaviparyaya respectively.
Now let the first statement (the hetu) be P, the second statement (the vyāpti) be Q, and let the statement “Things are produced from themselves” be S. (We don’t need an R here.) In Bhāviveka’s hands, the prasaṅga is a reductio ad absurdum along the lines of our method A, where the conjunction of P, Q, and S yields a contradiction and where the proponent of the reasoning proves ¬S precisely because he is committed to P and Q being true. I might remark that when I speak of an absurdity or contradiction being derived, there is a slight unclarity here, because the prasaṅga itself only explicitly derives one statement, the assertion that things’ production would be pointless and without an end. To get a genuine contradiction we obviously need to conjoin the absurd statement with what I termed statement 1, that production has a point and so on. Then we would get a clear contradiction that things’ production would both have and not have a point.
III. Candrakīrti
Turning to the second kind of use of reductio ad absurdum, namely, what I have been terming the method of “internal inconsistency without commitments,” this is, I believe, what we find advocated by Candrakīrti in his reply to Bhāviveka. Candrakīrti, as is well known, replies that showing an internal contradiction in an opponent’s position is enough to convince someone who is not “out of his/her mind” (unmattaka) to give it up; he also replies that the Mādhyamika proponent is not obliged to put forth a contrapositive form, nor accept its component statements as being true. Such a stance is part and parcel of what it means for a Mādhyamika not to have any theses and thus not himself make truth claims. Let us take two representative passages:
Prasannapadā 1.15.7–10: tasmān nirupapattika eva tvadvādaḥ svābhyupagamaviruddhaś ceti / kiṃ tanmātreṇa codite paro nābhyupaiti yato hetudṛṣṭāntopādānasāphalyaṃ syāt / atha svābhyupagamavirodhacodanayāpi paro na nivartate / tadāpi nirlajjatayā hetudṛṣṭāntābhyām api naiva nivarteta / na conmattakena sahāsmākaṃ vivāda iti /
So your own position is simply illogical and contradicted by what you yourself accept. When criticized by [us saying] just this much, then why would the opponent not accept [that things are not produced from themselves], so that [the additional step] of using a reason and an example would then accomplish something? But suppose the opponent does not even desist when criticized as having a contradiction with what he himself accepts. Then too, due to his lack of any shame [at contradicting himself], he would not desist at all, not even because of a reason and example. Now, our argument is not with someone who is out of his mind.
Prasannapadā 1.23.3ff.: prasaṅgaviparītena cārthena parasyaiva sambandho, nāsmākaṃ svapratijñāyā abhāvāt / tataś ca siddhāntavirodhāsambhavaḥ /
It is only the opponent who is linked to the contraposition of the absurd consequence and not we, for we have no thesis of our own. And therefore we do not have a contradiction with [our] philosophical tenets.
In effect, Candrakīrti seems to be arguing that Bhāviveka has imposed upon him a type-A use of reductio ad absurdum (with the ensuing commitment to the truth of certain statements and hence the implied siddhāntavirodha), whereas he himself was only using a type-B method of internal inconsistency without commitments. The opponent — but not Candrakīrti himself — accepts that things are produced from themselves; from this and other statements acceptable to the opponent, a contradiction follows. Candrakīrti himself goes no further than to show that the conjunction of the statements accepted by the opponent yields a contradiction, but he does not himself say (or more likely, refuses to say) which statements are actually true and which must be negated. He does, of course, know what the opponent’s probable reaction will be, and therefore he himself can stand back and assert no more than that P&Q&R&S will imply ¬(P&Q&R&S) or, equally, that P&Q&R&S implies internal consistency, namely, P&Q&R&S and ¬(P&Q&R&S). No more than that. In short, we have a method designed to show internal inconsistency that is coupled with a (fallible) prediction that rational individuals in a particular dialectical situation will behave in such and such ways.
IV. Implicative Reversal and Contraposition
An exegetical excursus is badly needed at this point. For, before going on, we have to take up, in broad outlines at least, David Seyfort Ruegg’s carefully developed opinion that the talk of prasaṅgaviparītārtha or prasaṅgaviparyaya in Prajñāpradīpa and Prasannapadā 1 is not to be taken in the straightforward technical way of the contraposition of the consequence — that is, in the way I outlined above. Seyfort Ruegg takes up the prasaṅga and two contrapositions (formulated by Jeffrey Hopkins on the basis of a major Gelukpa author) and argues that in fact, in Bhāviveka’s objection, it is not contraposition that is at stake at all:
[Bhāviveka’s objection] apparently involves the idea not of contraposition but of implicative reversal, namely that a negation of production from self would imply the affirmation of production from an other.5
Seyfort Ruegg is thus perfectly aware that later Gelukpa writers did say there were prasaṅga and viparyaya at stake like those I have sketched out but sees this as an interpretative error of imposing a later understanding of viparyaya/viparītārtha on an earlier one. I would respectfully disagree: the Gelukpas and Jeffrey Hopkins had things essentially right.
Now, part of Seyfort Ruegg’s argument is that Buddhapālita’s own words tadutpādavaiyarthyāt / atiprasaṅgadoṣāc ca use the ablative and hence state reasons rather than absurd consequences. That is, they are being used to say “because …,” rather than “it follows that …”.6 I don’t think that this is a particularly strong argument, because the actual form of Buddhapālita’s reasoning is far from obvious and is controversial among scholastic writers in spite of Buddhapālita using two ablatives. Indeed the passage can be taken (as we even see in some Tibetan monastic manuals) as meaning “Things are not produced from themselves, because if they were, then their production would be pointless.” In effect “because if they were, then their production would be pointless” would express the consequence, “It follows that things’ production would be pointless, because they are [supposedly] produced from themselves.”
Although Seyfort Ruegg does not himself stress this, some of the strongest evidence for his position, as far as I can see, is that it concords with Avalokitavrata’s interpretation of Bhāviveka’s Prajñāpradīpa. Here is briefly how that interpretation goes. In his Prajñāpradīpa commentary, Avalokitavrata argues that Buddhapālita’s reasoning is a statement vulnerable to criticism (sāvakāśavacana, sāvakāśavākya = glags yod pa’i tshig) because it is not a valid reason, and that for it to be a valid reason, the property being proved (sādhya, i.e., “not being produced from self”) must be changed to a negative (viparītārtha) like “being produced from other things.” Similarly Buddhapālita’s bad reason “because its production is pointless” must be amended to “because its production has a point.” So Avalokitavrata is not talking about prasaṅgas and their contrapositions but about amending Buddhapālita’s argument so that it would have a valid reason and not be criticizable as badly formulated.
In an earlier article (Tillemans 1992a) I had sketched out this interpretation by Avalokitavrata and translated representative passages from his Prajñāpradīpaṭīkā — the interpretation was, incidentally, also known and even incorporated in a very complex fashion by the Gelukpas and others into some of their explanations of the Bhāviveka-Candrakīrti debate. A part of the problem was that Bhāviveka and Avalokitavrata did not actually speak of the argument against production from self as being a prasaṅgavākya at all but rather characterized it as a glags yod pa’i tshig — that is, a sāvakāśavacana or sāvakāśavākya, “a statement that offers an occasion [for a critique],” “a vulnerable statement.” Candrakīrti on the other hand did represent Bhāviveka as using the term prasaṅgavākya. It is significant that Candrakīrti does clearly speak of prasaṅgavākya and prasaṅgaviparītena arthena and not just the viparītārtha and prakṛtārthaviparyaya spoken about by Bhāviveka and Avalokitavrata.7 A natural reading is that, irrespective of what Avalokitavrata might have thought, Candrakīrti took the debate as being about prasaṅgas and contrapositions in the usual technical sense, especially so because a nontechnical, looser sense in terms of “implicative reversal” is not readily found elsewhere for these terms.
However, we can go considerably further than what I wrote in 1992. Toshikazu Watanabe (2013), independently of my earlier unpublished investigations, has looked at the passages in Prasannapadā 36.11–37.2 dealing with Bhāviveka’s criticism of Buddhapālita’s arguments. He examined in particular the prasaṅga refuting any production from “other” (paratra) — causes that would be radically other than their effects. Here too, as Watanabe shows, Bhāviveka’s criticism of the prasaṅgaviparyaya as leading (again) to contradiction with Mādhyamika philosophical tenets (siddhāntavirodha) presupposes a normal contraposition of the prasaṅga, one that a later logician (Dharmakīrtian or Gelukpa) would find fully familiar. The logical structure of this argument and its contraposition is thus the same as that of the earlier refutation of production from self. What one says about the one will hold for the other.
On the other hand, an interpretation of the whole controversy — be it Bhāviveka’s or Candrakīrti’s understandings — by relying on Avalokitavrata’s commentary is impossible to comprehend in that normal fashion. The problem is that an “implicative reversal” and Avalokitavrata’s proposed amendments are not generalizable logical moves at all, whereas contraposition is; it is one that is actually quite ordinary in both Western and Indian logic. To attribute “implicative reversal” or amendments here means in effect that we do not know why this move is being made, for either it seems to be an ad hoc tactic, or worse (but less likely), it is one that simply relies on errors in logic.
The best course of action, as far I can see, is thus to discard Avalokitavrata’s interpretation as creating more trouble than it is worth.8 As I hope the central arguments of this paper show, the Bhāviveka-Candrakīrti controversy can have a simpler, and logically understandable, explanation when seen as turning on the technical senses of prasaṅga and contraposition, an interpretation in which Candrakīrti’s statement that he is not bound to accept the prasaṅgaviparyaya becomes a justifiable stance in keeping with some recognizable and defensible features about the use of reductio ad absurdum. On a methodological level, I think that such a gain in theoretical elegance would be considerably more important than ambiguous philological evidence and counter-evidence from competing commentators. What is at stake is, in effect, an application of the principle of charity, privileging elegance of interpretation and rational reconstruction. When faced with complicated arguments in dense Sanskrit or Tibetan, philological analysis is easily blind: logic counts.9
Finally, I think that it is particularly relevant that on the technical interpretation of prasaṅga and prasaṅgaviparyaya, Candrakīrti can conserve his “thesislessness” and actually establish that the opponent’s position (i.e., the conjunction) is false by using what seem to be recognizable theorems in a classical logic. In other words, he is showing that P&Q&R&S, which is a conjunction of various individual statements, is false.10 The only thing he is not doing is committing himself to which individual conjunct is true and which is false. In that sense he himself stays thesisless and leaves the hard (but predictable) choice of what to reject to his adversary.
V. The Use of Recognizable Logical Theorems
I mentioned that on this interpretation of the Bhāviveka-Candrakīrti debate, we would see that Candrakīrti uses “recognizable theorems.” But which ones? The basic point is this: If P&Q&R&S also imply ¬(P&Q&R&S), then we can validly infer ¬(P&Q&R&S). If we look at the history of logic, as depicted by William and Martha Kneale, it is clear that variants upon this type of argument pattern were formulated in Stoic logic (“If the first, then the first; if the first then not the first; therefore not the first”). The theorem that a statement that implies its negation must be false is what figures implicitly in Plato’s rejection of Protagoras’s relativism (“If it were true, it would also be false; therefore it is false”). In short, the theorem relied upon is the following: (P→¬P)→¬P. This is a tautology, but its applications can be vast, as we shall see.
Note that this same theorem is what is at stake when someone argues against a liar-type sentence, like “Everything I say is false,” saying that if it were so, then it would not be so, and that therefore it is not so. Arguably too, this strategy seems also to be what we find in St. Thomas Aquinas’ De Veritate, “From the destruction of truth it follows that truth exists, because if truth does not exist, it is true that truth does not exist and nothing can be true without truth. Therefore truth is eternal.”11
There are other theorems that underlie arguments that proceed along the lines of “such-and-such a statement implies its opposite and hence the opposite is so.” We have seen (P→¬P)→¬P. We can also obtain a similar theorem by inserting the contradiction P&¬P. Thus, for example, (P→ (P&¬P))→¬P is equally a tautology. So in that sense, when I speak of Candrakīrti relying upon “recognizable theorems,” the plural does have a point: there may indeed be more than one, and it is admittedly hard to pin him down. Note too that if instead of these theorems, we redistribute the negations and take ¬P as the starting point and show that it implies P, we can thereby infer P, and with this redistribution of negation signs we get the infamous consequentia mirabilis. This seems to be what is behind St. Thomas’s argument, which states that the nonexistence, or destruction, of truth (if true) would imply that it exists undestroyed, and hence we can infer that truth exists undestroyed (= eternally). It also seems to be the theorem behind Descartes’ cogito ergo sum, where the reasoning is that if there were nothing, not even a thinking subject, there would at least be something, the thinking subject, and thus there is a thinking subject. Candrakīrti thus seems to be in good company, reasoning in recognizable ways, even if he certainly would have used these theorems only for negative ends and not to establish substantive metaphysical claims a priori. His quietistic reluctance to countenance substantive truth claims of how things are would probably be even stronger if someone argued that he could justify such a claim a priori by an argument that conjures a truth from the “fact” that its falsehood would imply its truth.
What about Bhāviveka? Was he rejecting these theorems outright? The worst-case scenario is that in imposing the demand for a viparyaya, he may have missed something fundamental about logic and hence may even have embarked on a sterile debate. But what is much more charitable and likely is that Bhāviveka had other deep-seated reasons, not of a purely logical order, to reject an argument strategy turning on simple internal inconsistency.
The problem is this. Method B, unlike method A, would not end up obliging an opponent to acknowledge any specific individual statement in the conjunction as true or false: it would only oblige him to acknowledge that the conjunction is not true and thus give up something in previously cherished positions. That may be all that a quietist advocate of no-theses, no-prapañca (no conceptual proliferations), like Candrakīrti, needs in order to guide people to the irenic standpoint that he considers genuine Madhyamaka. But the problem is that there is no guarantee that it will work to guide people to truth, to what they should believe. It was then probably not enough for Bhāviveka, who emphasized that Mādhyamikas advocated knowledge and positions; he advocated pramāṇas leading to determinate customary truths.
VI. Dharmakīrti
Finally, we turn briefly to certain thinkers of the pramāṇavāda school, notably the Buddhist logician Dharmakīrti and his successors, some of whom were not only logicians but also Mādhyamikas of the broadly Svātantrika orientation of Bhāviveka.
Which methods did they use? With a little analysis we can readily see that it is indeed the type-A use of prasaṅga and prasaṅgaviparyaya that is described and promoted. Dharmakīrti’s Pramāṇavārttika 4.12bcd is the classic source:
parakalpitaiḥ / prasaṅgo dvayasambandhād ekābhāve ’nyahānaye //
An absurd consequence is [drawn] by means of the opponent’s conceptual constructs (parakalpitaiḥ); as the [consequence’s] two terms are necessarily connected (sambandhād), it serves to negate the second term in the absence of the first.
The prasaṅga at stake in the parallel discussion in Pramāṇaviniścaya 3 and in Manorathanandin’s commentary on the Pramāṇavārttika is as follows:
sāmānyasya paropagatānekavṛttitvād anekatvam āpādyate
It follows absurdly that the universal is several things because, as the opponent accepts, it is present in several different things.
The contraposition (viparyaya) is:
The universal is not present in several different things (nānekavṛttitva) because it is not [itself] several things.
This then looks clearly like the type-A use of reductio. After deriving the absurdity from the conjunction P,Q,S, the key step is to say that S is false because P and Q are true.12
So did Dharmakīrti have any place at all for what we are terming the type-B use of reductio, the method of merely showing internal inconsistency of a conjoined ensemble of statements? No doubt, Dharmakīrti was extremely sensitive to inconsistency in treatises (śāstra) and scriptures (āgama) and in the tenets (siddhānta) they propound, and not surprisingly he thought that any inconsistent ensemble of statements was discredited. Indeed, he goes on at length about how certain Brahmanical scriptures and treatises are inconsistent in their accounts of liberation, in that they state that passions (kleśa) such as desire and the like (rāgādi) are the root causes of immorality (adharmamūla) and yet they also say that purification of such immorality is to be accomplished by bathing (snāna) in the Ganges and other such practices that have nothing to do with the passions. In short, the Brahmanical treatise is inconsistent in that one of its main statements about passions being the root of immorality is contradicted by statements about the would-be remedy — this “remedy” would, in effect, imply that passions are not the root causes at all but that something else is.13
Such a refutation of the trustworthiness of a scripture is essentially a use of method B, showing internal inconsistency of a conjunction of propositions. Dharmakīrti, like Bhāviveka, however, also holds that merely showing inconsistency of P, Q, R, and S does not establish any genuine knowledge of specific truths. However, the reason given why that is so is somewhat different from that given by Bhāviveka. For Dharmakīrti, an examination (parīkṣā) of inconsistencies is characterized as na vāstavī, not bearing upon real facts, or real entities (vastu), at all but only on words and concepts; in Devendrabuddhi’s commentary, we are told that merely finding a contradiction in a set of scriptural statements does not give any significant information (literally “genuine knowledge” don dam rtogs pa) about real entities or their actual properties.14 And in Pramāṇavārttika 4.98–101 Dharmakīrti argues that we need an extra procedure to determine what is actually so. What is needed to determine specific truth and falsity is a pramāṇa “grounded in reality” — that is, more literally, one that “operates due to real entities” (vastubalapravṛtta).15
VII. Some Questions and Speculations
Dharmakīrti’s position is complex. It is clear that he adds important metaphysical requirements about “grounding in reality” that a Mādhyamika would not easily endorse. And method B is wanting not just because it does not tell us which individual statements are true, but because genuine knowing also requires understandings that are grounded upon real entities. Now given that metaphysical requirement about grounding in reality, an interesting problem arises as to whether we can genuinely know the truth of ¬(P&Q&R&S) when that statement is only derived because the conjunction P,Q,R,S is shown to be inconsistent. Is it still a full-fledged fact about the real world? And could one know that fact a priori without relying on any other pramāṇas? Interestingly enough, the thrust of his philosophy of logic would seem to be to say “no” to both. One indicator is Dharmakīrti’s treatment of the liar-statement “Everything said is false” and other self-refuting statements (like “My mother is barren”) in Pramāṇavārttika 4.93–101. He, in effect, says that reasoning simply from words and concepts cannot establish that everything said is false nor can it establish that it is not so that everything said is false. The same holds for “My mother is barren,” “There are no sources of knowledge,” and so on. If we proceed purely a priori, we can only know that a statement like “Everything said is false” implies itself as well as its opposite, and that the two propositions “clash” or “obstruct each other” (pratibandhaka), but there is no specific invalidation (bādhā) of either of them; we do not actually know which is true and which is false unless there is an extra pramāṇa grounded in reality, such as an independent confirmation that people sometimes do say true things. We seem to have, in effect, a thoroughgoing rejection of a priori reasoning as being able to establish genuine truths. Instead, facts about reality need to be established by reasoning linked to reality (that is, involving a so-called “natural connection” svabhāvapratibandha), reasoning that ultimately relies on perceptual input. This stance would indeed leave logical tautologies, whose truth is determined purely a priori, as having no real purchase on reality.16 We could even go further and say that presumably other logical laws — like excluded middle, double negation, et cetera — would have to be of a similar status; it would not be a genuine fact that P v ¬P or that P if and only if ¬¬P; knowing the truth of a tautology is not like knowing the truth of an individual statement that is grounded in reality.
It is also an important question as to what truth status Candrakīrti himself would assign to ¬(P&Q&R&S). If he would not, like Dharmakīrti, invoke a metaphysical requirement of grounding to say that its truth is not genuinely known, would he then simply accept it as true? Significantly, Seyfort Ruegg 1983 has shown that Candrakīrti does himself seem to accept some negative statements — the conclusions of refutations — and argues that this does not violate his deliberate avoidance of substantive truth claims, or theses (pratijñā, pakṣa), about what there is. Philosophically, it would seem that this would have to be the case, as otherwise we have the oddity that Candrakīrti would not even accept the bona fides of his own arguments. I will leave that conundrum for another occasion.
VIII. Conclusions
Where does this three-way comparison of uses of reductio lead us? In the end, a sixth-century debate turns on quite different, but recognizable, conceptions of what it is to do philosophy. One way to look at this Svātantrika-Prāsaṅgika debate — and there are other significant ways too17 — is that it turns on the question of whether philosophy is essentially normative in nature.
Bhāviveka, and his successors in this Madhyamaka subschool as well as Dharmakīrti, seem to have recognized that a type-B method showing internal inconsistency of the conjunction of P, Q, R, and S does not constitute a proof of the truth of any one of the negated individual statements, in the sense that there is no clear normative account why the opponent must admit such and such a proposition to be true. (Method B may tell you that such and such a philosophy will not work, but it does not actually prove which specific propositions you should actually think to be true or false.) For a Svātantrika or Dharmakīrtian then, the normative dimension is missing in a purely prasaṅga method, and that is why the method is inadequate. Prāsaṅgikas who choose method B and the principle that rational individuals will behave in such and such predictable ways would seem to content themselves with psychology and rhetoric. Instead of proving truths that all people should think by a universally valid methodology, they lead people to changes of heart but not truths that they themselves endorse. They rely on techniques that simply do, as a matter of fact, lead to persuasion in specific individuals but cannot be universalized as a proper way to knowledge.
Issues of normativity are significant to how one does philosophy. There is a well-known Kantian opposition between conviction (Überzeugung), valid for all rational beings or for an (idealized) universal audience, and persuasion (Überredung), which particular groups of people in particular contexts may feel but which is solely a subjective ground for their personal judgments. One can imagine Bhāviveka’s and Dharmakīrti’s views transposed into the language of the first Kantian Critique. The persuasions to which the Prāsaṅgika leads his opponents via mere reductio ad absurdum are not understanding but only “a mere semblance” (ein blosser Schein), with “private validity” (Privatgültigkeit) and no more; where Candrakīrti goes wrong is that he cannot provide conviction, as he just “talks people into beliefs” and thus fails to meet the rational normative demands that make philosophy the activity that it is. I have no doubt that the Kantians would see Bhāviveka’s and Dharmakīrti’s points quite clearly.18 I can also imagine that Candrakīrti would reply in some way to the Kantians and the Svātantrikas that their normative standards are too high and are unattainable. Insistence on conviction would be the last illusion in the Svātantrika’s Madhyamaka, a disguised realist metaphysic about truth being thoroughly independent of thought.
Notes to Chapter 5
1. Barwise and Etchemendy 1992, 64.
2. Such is how he now is known ever since the work of Yasunori Ejima, in part because of the Chinese phonetic rendering Po pi fei jia; we’ll adopt “Bhāviveka” too, instead of “Bhāvaviveka” or “Bhavya.”
3. Here is the actual text as found on 14.4–15.2 of La Vallée Poussin’s edition of Candrakīrti’s Prasannapadā: athaike dūṣaṇam āhuḥ / tad ayuktaṃ hetudṛṣṭāntānabhidhānāt / paroktadośāparihārāc ca / prasaṅgavākyatvāc ca prakṛtārthaviparyayeṇa viparītārthasādhyataddharmavyaktau parasmād utpannā bhāvā janmasāphalyāt / janmanirodhāc ceti kṛtāntavirodhaḥ [or siddhāntavirodhaḥ] syāt //. “Here certain people [i.e., Bhāviveka] set forth the following critique: This [reasoning of Buddhapālita] is incoherent because (1) it does not state a [valid] reason and example and (2) it does not eliminate [certain] faults which the [Sāṃkhya] adversary states. (3) Since [Buddhapālita’s reasoning] is a statement of an absurd consequence (prasaṅgavākya), should one, by contraposing the terms in question [i.e., in the prasaṅga], then set out what is to be proved (sādhya) and its [pakṣa]dharma as the contrapositives (viparītārtha), one would in that case contradict one’s own philosophical tenets (kṛtānta = siddhānta), for entities would be produced from other [things] because their production would have a point and because there would be an end to [this] production.”
4. It is routine in Indo-Tibetan Buddhist argumentation not to formally state everything, even though the strict formal canons in Vāda manuals go on at length about exactly what you should and should not state! Theory versus practice: Vāda rules are regularly streamlined in practice. Indeed, one typically gets the complete formal reasoning (prayoga) with hetu and vyāpti and examples (dṛṣṭānta) only in so-called “inferences-for-others” (parārthānumāna). The most frequent “working reasoning,” however, is “A is, or would be, B because it is C.” The implicit vyāpti is “All Cs are Bs,” or equivalently, “All non-Bs are non-Cs.” This is the type of reasoning used in our debate. See Tillemans 2008b.
5. Seyfort Ruegg 2000, 253.
6. Seyfort Ruegg 2000, 254: “But the first two reasons adduced by Buddhapālita (utpādavaiyarthyāt and atiprasaṅgadoṣāt) in fact remain reasons in the form in which they appear in Bhavya’s text, even if the reason ‘because of futility’ has indeed been converted there to ‘because of usefulness’ (i.e., janmasāphalya in the PPMV) and the reason ‘because of over-extension’ (viz. because of being endless) has been converted there to ‘because of having an end’ (i.e., janmanirodha in the PPMV), so that (according to Bhavya) Buddhapālita will be in contradiction with the Madhyamaka siddhānta.”
7. In Tillemans 1992a, 320, I came to the following conclusion, which I would stick by now as a minimal position: “Personally, I think we must take Prasannapadā’s reading of prasaṅgavākyatvāc ca as showing that at least Candrakīrti took the passages from Prajñāpradīpa as arguing that the prasaṅgaviparyaya leads to a contradiction with the Mādhyamika’s siddhānta. Candrakīrti’s later discussion (Prasannapadā 23ff.), where he speaks of not having to accept the viparītārtha of the prasaṅga and thus avoiding contradiction with siddhānta, shows beyond reasonable doubt that Candrakīrti himself took Bhāviveka’s argument as turning on a prasaṅga and a prasaṅgaviparyaya.”
8. Watanabe does the same, presenting a “normal” version of the prasaṅga and viparyaya and then noting (Watanabe 2013, 177n14) that this interpretation is different from that of Avalokitavrata.
9. For a forceful application of this maxim, see also chapter 7.
10. We can go a bit further: ¬(P&Q&R&S) is true. Arguably, this would not be a substantive truth claim about reality. See the discussion in section VII below on “questions and speculations.”
11. Translation in W. and M. Kneale 1962, 202 and 746.
12. The details are as follows: the term dvayasambandha (the necessary connection between the two terms) in verse 12 means that there is an appropriate vyāpti, or general principle that holds because of a real/natural relation (i.e., sambandha = svabhāvapratibandha) between the terms; the vyāpti is thus true and hence its contraposition is also true. In the reasoning at hand, “presence in several different things” implies “being itself several things”; the contraposition of this general statement figures in the viparyaya and consists in the principle that what is not itself several things is not present in several different things. “In the absence of the first” is a true statement that must be accepted: it is the hetu of the prasaṅgaviparyaya, i.e., “not being [itself] several things.” The opponent’s conceptual construct (viz., the universal’s being present in several different things) is in effect the statement that leads to the absurdity (viz. the universal’s being several things). It is thus “the second term” that is negated in the absence of the first. See Tillemans 2000, 21–24, for translation and explanation of the verse and Manorathanandin’s commentary.
13. Pramāṇavārttika 4.107 with the commentarial additions of Manorathanandin (see Tillemans 2000, 150–51): virodhodbhāvanaprāyā parīkṣāpy atra tadyathā / adharmamūlaṃ rāgādi snānaṃ cādharmaśodhanam //. “In the case of this [treatise], examination, moreover, consists chiefly in pointing out contradictions [between prior and subsequent statements and is not concerned with facts (na vāstavī)]: for example, [contradictions such as when it is said that] desire and so on are the root of immorality and that bathing is what purifies immorality.”
14. Devendrabuddhi, in his Pramāṇavārttikapañjikā P. 346b5–7, is explicit in saying that this examination of an inconsistent treatise points out contradictions (’gal ba brjod pa) but does not concern what is really so (don dam pa ma yin ≈ na pāramārthikā; na vāstavī). Manorathandin clearly follows him on this in saying that the examination detecting inconsistencies does not concern real facts (na vāstavī). See Tillemans 2000, 151n522.
15. On vastubalapravṛtta and related concepts like svabhāvapratibandha (real/natural connections), see Tillemans 2000, 12. Classic sources for the latter term are to be found in Dharmakīrti’s Svavṛtti ad Pramāṇavārttika 1.14. See also Steinkellner 1971 and 1984.
16. In Tillemans 2000, 140, I was puzzled by his views on liars and the like and said: “This inference [establishing that it is not so that everything said is false] would be valid because the statement ((P→not-P)→not-P) is a theorem … [His discussion] here strongly suggests that either he did not know this basic logical theorem…. or, perhaps more interestingly from a philosophical point of view, that he did not accept it.” It is more the latter: he rejected purely a priori reasoning as leading to knowledge about reality.
17. See Tillemans 2003.
18. The Kantian distinction between conviction (Überzeugung) and persuasion (Überredung) is found in the Critique of Pure Reason II (Transcendental Doctrine of Method), chapter II (The Canon of Pure Reason), section III (Of Opinion, Knowledge, and Belief). See Kant 1998, 684–90. The Kantian pair of concepts was first put into the context of argumentation theory in Perelman and Olbrechts-Tyteca (1958), who saw it as pertaining to arguments destined to a universal audience and particular audiences respectively. The conviction-persuasion distinction was applied to the Svātantrika-Prāsaṅgika debate in Tillemans 2003 and will be developed further in my forthcoming article on Nāgārjuna’s quietism in the Oxford Handbook of Indian Philosophy, ed. Jonardon Ganeri.