The Problem of Induction in Indian Philosophy

·期刊原文
The Problem of Induction in Indian Philosophy
By Roy W. Perrett
Philosophy East & West
V. 34 (1984)
pp. 161-174
Copyright 1984 by University of Hawaii Press
Hawaii, USA

--------------------------------------------------------------------------------

p. 161

I.
One of the many interesting parallels between Indian and Western philosophy is the way in which the problem of induction arises in both. The skeptical position that in the West is associated with the name of David Hume was in India associated with the Caarvaaka materialists. The problem arose in Indian philosophy in the context of the inter-school debates about the number and status of the pramaa.nas or valid means of knowledge. Except for the Caarvaakas, all the schools accepted at least perception (pratyak.sa) and inference (anumaana) as valid means of knowledge, although there was considerable dispute as to the ultimate status of these pramaa.nas. The Caarvaakas, however, denied the validity of inference and only accepted perception as a pramaa.na. The reasons offered for this stand are fundamentally concerned with the supposed impossibility of justifying the inductive relation that is the basis of Indian inference forms.

To understand how the problem came to occupy the attention of Indian philosophers we need to know something about what was considered as coming under the rubric of "inference." More particularly, we need to make a few comments on the standard inference model in Indian logic. [1] Consider the following argument which is a standard example in Indian logic texts. It might occur if two people were standing together looking at a mountainside from which they could see smoke rising. One of the persons involved here remarks that there is a fire on the mountain. When asked for his reasoning he replies that he holds that there is a fire on the mountainside because there is smoke. He then appeals to familiar conjunctions of fire and smoke: as in a kitchen. Furthermore, he reminds his friend that one never sees smoke where there is no fire: as, for example, in a lake.

Of course, this example is as overly simplistic as any of the standard examples in Western logic texts. Nevertheless, the argument can be formalized as follows:

Hypothesis: That mountain is fire-possessing.
Reason: Because that mountain is smoke-possessing.
Examples: (a) like kitchen.
(b) unlike lake.

This inference has three members and five terms. The three members are the hypothesis (pratij~naa), the reason (hetu), and the examples (d.r.s.taanta). [2] The five terms are the pak.sa (that mountain), the saadhya (fire-possessing), the hetu or li^nga (smoke-possessing), the sapak.sa (kitchen), and the vipak.sa (lake). It is important to understand that each of the five terms italicized is to be considered as a class of things. Thus, in our example above, the pak.sa (that mountain) is a unit class with only one member (namely, the mountain the two people are looking at); the saadhya (fire-possessing) is the class of all fiery things; the hetu (smoke-possessing)

p. 162

is the class of all smoky things; the positive example, kitchen, is the class of all kitchens; and the negative example, lake, is the class of all lakes.

The Indian logicians discussed various rules of inference, the violation of which would involve mistakes in reasoning. Perhaps the most important of these is the rule of the saadhya-pervaded hetu. This requires that the hetu must fall completely within the saadhya. The pervasion relation meant here is the relation of class-inclusion. Thus class A can pervade class B only if all members of B are members of A, though not necessarily vice versa. Similarly, class A is pervaded by class B only if all members of A are members of B, though not necessarily vice versa. Hence our example above is a valid inference only if the class of fire-possessing things really does pervade the class of smoke-possessing things. This relation of pervasion or universal concomitance (vyaapti) can be reformulated as the major premise of a Western syllogism: for example, "All smoke-possessing things are fire-possessing things."

It should be clear by now that "inference" in Indian philosophy meant generally (though not exclusively) inductive inference. The focus of concern in Indian logic was the ascertainment of the truth of the universal proposition of an inference form and hence the establishment of the validity of the given inference. This is the import of the rule of the saadhya-pervaded hetu, as can be illustrated by considering the following argument:

Hypothesis: That mountain is smoke-possessing.
Reason: Because that mountain is fire-possessing.
Examples: (a) like kitchen.
(b) unlike lake.

This argument is invalid because the hetu (the class of fiery things), is not completely included within the saadhya (the class of smoky things). It is possible to give examples of things that are members of the first class but not of the second. Here the standard instance adduced by Indian logicians is a red-hot iron ball which is fiery but not smoky. This type of contrary instance is called an upaadhi in Indian logic. Discovering an upaadhi amounts to a denial of the relation of universal concomitance (vyaapti) that must obtain between the hetu and saadhya of an Indian inference in order for it to be a valid inference.

But now the question arises as to what happens if we do not find any contrary instances. Can we then assume invariable concomitance? It is in this manner that the problem of justifying induction arose for Indian philosophers. When the whole validity of inductive inference rests on the relation of invariable concomitance, how can we be certain that this relation actually obtains? The problem here was conceived as twofold. First, how to justify a generalization about the universal concomitance of As and Bs when we have not seen all past instances of As and Bs, let alone established that all non-Bs are also non-As. Second, how to justify the projection of past concomitance of As and Bs into the future when we have not seen any future instances of A's and B's.

The problem thus posed bears obvious and striking resemblances to what in

p. 163

Western philosophy has been viewed as the problem of justifying induction, a problem originally posed in the West by David Hume's analysis of causation in the eighteenth century. In India the problem arose much earlier, though in a different context. However, the general question of how to justify induction was seriously considered by Indian philosophers and various responses were elicited. We shall briefly examine four of these Indian responses: namely, the skeptical Caarvaaka response, the Naiyaayika appeal to saamaanyalak.sa.na perception, the Advaitin use of abaadhitva, and the twofold Buddhist reply.

II.
The Caarvaaka response to the problem of justifying the inductive vyaapti relation consists in simply denying that inference really is a pramaa.na. For the Caarvaaka there is only one valid means of knowledge: namely, perception (pratyak.sa). According to the summary of their views in Maadhava's Sarvadar`sanasa^mgraha they argued somewhat as follows. [3] Inference is dependent on universal concomitance (vyaapti). For inference to be a valid means of knowledge this relation of universal concomitance must be able to be known by one of the other pramaa.nas. However, it cannot be known by any of the pramaa.nas, as an examination of each of them shows. Thus perception (internal and external) cannot establish such a universal relation, since we never perceive all past particulars and no future ones are ever perceived. Neither can inference establish the universal proposition, since it is obvious that to appeal to inference to justify inference itself is to enter into a vicious regress. Testimony (`sabda) is also rejected as the means of knowledge of the universal proposition, since it ultimately depends on perception or inference. Finally, comparison (upamaana) is ruled out too because it can only establish a quite different relation, namely, the relation of a name to the object named. These four categories are the four kinds of pramaa.nas recognized by the Nyaaya school and, although the Vedaantins recognize two more (anupalabdhi and arthaapatti), these cover the valid means of knowledge admitted by almost all the Indian dar`sanas. Hence the Caarvaakas conclude that since vyaapti cannot be known by means of any of the pramaa.nas, it must be the case that inference is not a valid means of knowledge.

However, the Caarvaakas do offer an alternative account of inference. They claim that it is either based on a former perception or it is a mistake. The fact that it is sometimes followed by successful results is just an accidental coincidence. In other words, inference is a psychological process, not a logical one, and our reliance on such reasoning is due to psychological conditioning. It is sometimes accidentally successful, but there is no logical connection because, the Caarvaaka argues, it has been established that we can never really know the vyaapti on which inference is based. (The resemblance to Hume is quite striking here, for he also concluded that induction cannot be logically justified at all because it is not really a process of reasoning but rather a habit of expecting what has previously occurred in certain given circumstances to reoccur in similar circumstances.) [4]

p. 164

Of course, one difficulty with the Caarvaaka explanation here is that the rate of accidental coincidence seems inordinately high! However, their opponents did not pursue this line of objection, which might easily have given rise to the types of discussion about probability theory that have so engrossed Western philosophers working on the problem of induction. Rather, the other Indian schools were more concerned to press home a charge of self-contradiction. They accused the Caarvaakas of a self-contradictory use of inference to deny the validity of inference. It is not clear how far this charge is justified. If we invoke a distinction not present in Indian logic, the distinction between deductive and inductive argument, then it is possible to represent the Caarvaaka argument as a valid deductive argument. Thus, the Caarvaaka argues that if inference is a pramaa.na then vyaapti must be knowable. But vyaapti cannot be known. Therefore, inference is not a pramaa.na. This is an instance of Modus Tollens: "If A then B; not-B; therefore, not-A." In other words, the Caarvaakas were using a valid deductive argument to establish the invalidity of inductive arguments. Because they were not using an inductive argument themselves it seems that the charge of self-contradiction has no basis.

How far the Caarvaakas thought their strictures against inference were supposed to extend is now extremely unclear. Hiriyanna suggests that they were only concerned to attack inferences that moved from the material to the nonmaterial. [5] This would naturally eliminate theological inferences as exemplified, for example, in the Naiyaayika arguments for the existence of God, and this kind of criticism would seem in keeping with their reported antireligious tone. They may also have been willing to allow inference a certain practical usefulness while still denying its status as a pramaa.na. Nevertheless, even given these modifications, the Caarvaaka skepticism about the possibility of inference represented a serious challenge to the theoretical assumptions of the other dar`sanas. As Karl Potter points out, [6] it was a challenge they could not afford to ignore because the skepticism of the Caarvaakas raises the doubt as to whether there exists any form of universal regularity in the world, or whether such apparent regularities are merely projections of our own psychological conditioning. This skepticism naturally calls into question the very possibility of achieving the religious freedom (mok.sa) that was the avowed goal of the other schools of Indian philosophy. If there is no regular connection between events and actions, then it is impossible for a person to enter into the course of events as a conscious causal agent whose decisions and activities have predictable consequences. Karman is a fiction since it can only be established by inference which is not a pramaa.na, if the Caarvaakas are right. Thus, since it is not at all clear whether the attainment of mok.sa is even possible, it is obviously a waste of time engaging in religious activities and practising asceticism. Rather, the Caarvaakas advocated a policy of hedonism and the pursuit of worldly pleasures. According to the Sarvasiddhaantasa^mgraha, they held that:

p. 165

The enjoyment of health lies in eating delicious food, keeping company of young women, using fine clothes, perfumes, garlands, sandal paste, etc. [7]

Faced with such a challenge it was inevitably incumbent upon the mok.sa-oriented philosophers to defend inference as a valid means of knowledge, and we shall now examine some of these attempts.

III.
The Nyaaya school's response to this challenge was built upon their naive realist ontology. [8] They claimed that we can actually perceive (nonsensuously) the vyaapti relation. Jayanta Bha.t.ta presents the Naiyaayika view in his Nyaayama~njarii thus:

A man perceives that smoke and fire co-exist in the same locus. He comprehends by means of the method of difference that smoke is not present in the locus where fire does not exist. Then he synthesizes the results obtained by the joint method of agreement and difference and frames a judgment by means of the internal organ that smoke is the invariable concomitant of fire...
The relation of concomitance obtaining between the middle term and the major term may be determined by means of the universals inhering in them (these two terms). The relation of concomitance holding between smoke and fire amounts to that of concomitance subsisting between the universals of smoke and fire. The positive aspect of the relation may be grasped by extraordinary perception acknowledged by the Naiyaayikas. But the negative aspect of the relation should also be grasped in order to grasp its invariable character. Therefore, we should also know that smoke does not exist where fire does not exist. [9]

In other words, their account of the method of inductive generalization is as follows. First, we observe a uniform agreement in presence (anvaya) between two things A and B; that is, whenever A is present B also is present. Second, we observe that there is a uniform agreement in absence (vyatireka) between A and B; that is, whenever B is absent A also is absent. Third, we do not observe any contrary instance in which A is present without B being present, or vice versa. Given these conditions we conclude that there is a relation of invariable concomitance between A and B.

However, we still have to establish that this relation is independent of any upaadhi. Thus, in addition to the method of sampling by observation of agreement and difference, the Naiyaayikas also utilize the method of indirect proof (tarka). [10] The idea here is that we can indirectly prove a universal proposition like "All smoke-possessing things are fire-possessing things" by disproving its contradictory proposition. In other words, if the universal proposition is false then its contradictory "Some smoke-possessing things are not fire-possessing things" must be true. But this would be to claim that there could be smoke without fire, a conclusion which is absurd because it denies the well-known causal relation between fire and smoke. Hence we can conclude that since "Some smoke-possessing things are not fire-possessing things" is obviously false, then it must be the case that its contradictory "All smoke-possessing things are fire-possessing things" is true.

p. 166

The Naiyaayika method for establishing vyaapti as outlined to this point is basically simple enumeration supported by tarka. But, of course, this is not a sufficient reply to Caarvaaka skepticism at all. In an induction by simple enumeration we move from some observed cases of As and Bs to a generalization about all As and Bs. It is precisely this move that the Caarvaaka challenges. The real question is how it is possible for us to know from the observation of some As as related to some Bs that all As are related to Bs. The Naiyaayika reply here makes use of their doctrine of saamaanyalak.sa.na perception, that is, the perception of a universal characterizing of all members of a class, one of whose members is presented.

As we have already seen, Jayanta refers to this kind of "extraordinary" perception in the Nyaayama~njarii. However, it is in the Navya-Nyaaya that we encounter the fuller account of saamaanyalak.sa.napratyak.sa where it is classified as one of three kinds of "extraordinary" (alaukika) perception. (The other two are yogic perception and j~naanalak.sa.napratyak.sa, that is, the perception of the features of something previously known as here and now presented.) Vi`svanaatha presents the Navya-Naiyaayika position in his Siddhaantamuktaavalii thus:

... where smoke or the like is connected with the [sense] organ, and the knowledge that it is smoke has arisen, with smoke as its substantive, in that knowledge smokehood is a feature. And through that smokehood as the connection, there arises the knowledge "cases of smoke" comprising all smoke. [11]

In other words, when we perceive particular smokes and fires we also perceive the universals smokeness and fireness inhering in them. Through this sense contact with smokeness and fireness, which are generic properties equally shared by all cases of smoke and of fire, we can in turn (nonsensuously) perceive all cases of smoke and of fire. Thus the concomitance of smoke and fire is established through an "extraordinary" perception of the whole class of smoke-possessing things as related to fire. The objection that this alleged kind of perception would entail omniscience is forestalled by the claim that, although we could perceive all objects of knowledge comprehended under a generic character, they would still not be known in detail and we could not perceive their mutual differences.

Of course, this answer did not satisfy the Caarvaakas at all. In the first place they simply replied that it is not true that we perceive universals and through them general classes. We only perceive particulars and only those particulars available to our ordinary perception. [12] The Naiyaayikas object: "So how can there be knowledge of all smoke as smoke and of all fire as fire, without the help of the connection based on a common feature?" [13] But the Caarvaaka answers that this begs the question, for we do not in fact perceive all smokes and all fires. Moreover, the particular smokes and fires that we do perceive exhibit no common feature and hence even less so would the innumerable members of the class of all smokes and the class of all fires.

In the second place we have the Caarvaaka argument recorded in Maadhava's Sarvadar`sanasa^mgraha:

p. 167

Nor may you maintain that this knowledge of the universal proposition has the general class as its object, because, if so, there might arise a doubt as to the existence of the invariable connection in this particular case [as, for instance, in this particular smoke as implying fire]. [14]

This question of doubt occasions an important and ingenious Navya-Naiyaayika counterargument to the effect that without the admission of saamaanyalak.sa.na perception the arising of the doubt whether smoke is the concomitant of fire cannot be accounted for. This follows from their definition of doubt:

Doubt is a knowledge of contradictory features, viz. presence and absence, with regard to the same substantive.... The cause of doubt is the knowledge of attributes that are common to two things. [15]

Hence the argument runs:

For since the relation of fire to the smoke that is being perceived is already known, and no other smoke is known (at the time), the doubt whether smoke is the concomitant of fire or not is inexplicable. [16]

However, the Caarvaaka finds this argument uncompelling, since to have force it requires prior acceptance of Naiyaayika analyses of doubt and allied concepts. Caarvaaka skepticism requires only the logical possibility of things being otherwise. Because it is only contingently the case that smoke is accompanied by fire, the concomitance of all smokes with fire is doubtful. The Nyaaya logic, unconcerned with this kind of philosophical doubt which the Naiyaayikas pragmatically dismiss as empty of content, cannot adequately answer the Caarvaaka skepticism using the terms framed by its very different conception of inquiry. [17]

All in all, the Caarvaaka position here is clearly summed up in Jayaraa`si's Tattvopaplavasi^mha:

There is another reason why the knowledge of an invariable relation cannot be established. Is it the cognition of a relation between two universals, or between two particulars, or between a universal and a particular?
If it be the cognition of a relation between two universals, then that is incorrect, for the universal itself is not demonstrated (anupapatti)... Nor is it possible to conceive of such a relation subsisting between a universal and a particular object because of the indemonstrability [or impossibility, asa^mbhavaat] of universals.
Nor is it [possible to think of] such a relation between two particulars for there are innumerable cases of particular fires and particular smokes, and also because ... no common element exists among the many particulars. [18]

IV.
The Advaitin answer to the Caarvaaka challenge (classically presented in Dharmaraaja's Vedaantaparibhaa.saa [19] is somewhat different from that of the Nyaaya school, although there are important similarities. The criterion of validity in Advaita Vedaanta is unfalsifiedness (abaadhitva). Thus concomitance can be affirmed on the basis of a single instance, and the Naiyaayika method of agreement and difference is unnecessary. This concept of nonfalsification ties in closely with the Advaitin theory of the two levels of truth. Hence ordinary

p. 168

knowledge remains knowledge until falsified, but this falsification can take place in two ways. First, within the realm of ordinary knowledge there can be falsification through a negative instance. Second, the whole world appearance can be seen as illusion in the experience of mok.sa.

In common with the Naiyaayikas the Advaitins hold that agreement in presence (anvaya) and nonobservance of any exception are necessary conditions for establishing a vyaapti relation. However, they reject the Naiyaayika insistence on agreement in absence (vyatireka). They also reject the method of hypothetical argument (tarka) on the grounds that it is no use trying to test the validity of a vyaapti with the aid of a tarka, because a tarka itself involves another vyaapti which also requires proving, and so on. The third important difference between Nyaaya philosophy and Advaita Vedanta with regard to inference is to be found in their respective positions on the question of the perceptual knowledge of a vyaapti. The problem under consideration is how to justify an inference that moves from the observation of a limited number of As accompanied by Bs to the conclusion that all As are accompanied by Bs. The Naiyaayika answer was that the perception of a particular involved the perception of a universal inhering in it, and hence the perception of all members of the class characterized by the perceived universal. Thus vyaapti is supposed to be established through perception. The Advaitins, however, reject this account and argue in a somewhat different way. They claim that a general proposition like "All smoke-possessing things are fire-possessing things" is justified because by perceiving particular instances of smoke and fire we are enabled to establish a relation between the two universals smokeness and fireness. It is only this relation that can supply the foundation of a general relation between all smoke-possessing things and all fire-possessing things, just insofar as they are respectively constituted by the universals smokeness and fireness.

D. M. Datta calls the Advaitin account here "a connotative view of the universal proposition" and contrasts it with the Naiyaayika "denotative or enumerative view." [20] The Advaitins hold that a single observation can supply knowledge of a universal concomitance, provided that no exception is known of. That is, under certain optimum conditions a single observation can provide knowledge of a connection between two universals (for example, smokeness and fireness), and this is sufficient to justify the inference. One objection that the Advaitins did try to deal with was the apparent way in which a universal proposition like "All cases of smoke are cases of fire" seems to state a relation obtaining between individual smokes and individual fires. Is it not, after all, also just a "denotative view" of the universal proposition? Here the Advaitins reply that a universal proposition like "All cases of smoke are cases of fire" is actually reached by a deductive inference from the vyaapti between smokeness and fireness. Upon the observation of only one case of such a concomitance between universals we can thereby argue that all other past and future smoke is accompanied by fire by virtue of its possessing the characteristic smokeness.

p. 169

Of course, this account is completely unsatisfactory to the Caarvaakas who simply deny the existence of universals, even in the conceptualistic sense held by the Advaitins. Furthermore, some aspects of the Advaitin epistemological position were equally unacceptable to the other dar`sanas. In particular, their position on inference strongly reflects their belief that truth is to he identified with nonfalsification (abaadhitva). In Advaita Vedanta all knowledge gained through the pramaa.nas is valid so long as it is unfalsified by experience, yet none of it is ultimately "true" in that all the contents of the pramaa.nas are in principle falsifiable. Only knowledge of Brahman is nonfalsifiable. Against this view the Advaitin philosopher Madhva argued forcefully that a pramaa.na is supposed to give us knowledge of the world as it is. Hence it is nonsense to talk of a valid means of knowledge being ultimately falsified, which is what the Advaitin two-level theory of truth entails. Thus, in his Anuvyaakhyaana, Madhva argues that in the case of perception "it would be contradictory to impose any temporal limit on the validity of perception and restrict it to the 'present' moment of perception. If perception is to be invalidated later, how could it have any validity now?" [21] Similarly, the Advaitin use of nonfalsification in reply to Caarvaaka skepticism about inference just misconceives the whole problem of establishing the vyaapti relation and hence the status of inference as a pramaa.na. By a universal relation (vyaapti) is understood a relation of concomitance independent of all conditions (upaadhis). To talk of the contents of the pramaa.na inference as being ultimately falsified is simply to admit that the vyaapti relation is not independent of all upaadhis and thus that inference is not a pramaa.na. But this is precisely what the Caarvaakas assert and what the Advaitin use of nonfalsification (abaadhitva) was supposed to deny.

V.
The Buddhist reply to Caarvaaka skepticism involves two major lines of defense. In his Pramaa.navaarttika Dharmakiirti presents the first line thus:

Experience, positive and negative, can never produce (a knowledge) of the strict necessity of inseparable connection. This always reposes either on the law of Causality or on the law of Identity. [22]

And in the Nyaayabindu he reaffirms:

Because, as regards (ultimate) reality, (the entity underlying the logical reason) is either just the same as the entity (underlying) the predicate, or it is causally derived from it. [23]

Thus, according to this first line of defense, the Buddhists admit only two kinds of legitimate universal propositions. First, the vyaapti associated with causation is valid. Hence it is legitimate to assert an invariable association of smoke with fire because smoke is caused by fire and the law of causality is a universal law. The pa~ncakaara.nii test is used to determine whether two objects A and B are causally related or not. In brief: if it is the case (other things being equal) that the

p. 170

appearance of a given phenomenon A is immediately succeeded by the appearance of another phenomenon B, and the disappearance of A is immediately succeeded by the disappearance of B, then A and B are related as cause and effect. Once we know that A and B are causally related then we can assume that they are universally related. Second, the Buddhists also admit the vyaapti associated with identity as a legitimate universal relation. Thus to know that something is a `si^m`sapaa (a variety of tree) is to know that it is a tree because to deny the invariable concomitance associated with the genus-species relation is absurd.

Against this line of defense the Buddhists faced opposition on all fronts. On the one hand, both the Naiyaayikas and the Advaitins argued that there were other kinds of valid universal concomitances than just those based on the principles of causality and identity. For example, there is a universal relation of succession between day and night or between the seasons. On the other hand, the Caarvaakas remain unsatisfied that the original problem has been solved. True, the Buddhists deny that inference has anything to do with ultimate and unrelated reality, for Dignaaga maintains that "all inference ... (all relation between a reason and its consequence) is based upon relations constructed by the understanding between a substrate and its qualities, it does not reflect ultimate reality or unreality." [24] However, even if the Buddhists analyze the cause-effect relation in terms of the belief that two things are causally connected, the original question still arises. That is, how can we be sure about this causal connection in all past instances and all future ones? If the Buddhists reply that they are only talking about the way in which our minds order data, that is, that the cause-effect relation is simply a human ordering of perceptual data, then the Caarvaakas answer that it still has not been shown how it is that we can be sure that our minds will continue to order the data in the future as they have in the past.

This leads us to the second and final line of defense used by the Buddhists, a defense which is summed up in their appeal to the maxim of vyaaghaataavadhiraa`sa.nkaa. [25] The point here is roughly the absurdity of practical alternatives. Doubting must have an end when it results in conceptual contradictions or pragmatic absurdities. Although this defense was used by the Buddhists, it is not entirely exclusive to them. Thus Jayanta in the Nyaayama~njarii dismisses the Caarvaaka challenge in this way:

They cannot confute the validity of inference per se since its validity has been universally accepted.

A woman, a child, a cow-herd, a cultivator and such other persons know another object (lying beyond the ken of their sense-organs) by means of its sure mark with absolute certainty.

If validity is denied to inference then all worldly transactions cannot be conducted with the mere help of perception. All the people of the world should remain motionless as if they are painted in a picture. [26]

And his fellow Naiyaayika philosopher Udayana argues in his Nyaayakusumaa~njali that "doubt is permissible only so long as there is no contradiction." [27] This is to

p. 171

propose a behavioristic criterion of doubt. If someone claims to doubt the existence of a vyaapti between smoke and fire, then why does he light a fire when he wants to produce smoke? His own activity indicates that his doubt is not real. Of course, the Caarvaaka answer here is that we are so constituted psychologically that we expect a uniform regularity between instances of fire and instances of smoke, although there is no logical justification for such an expectation. The point of the Buddhist's second line of defense seems to be that naturally we cannot logically justify induction because it is a complete misconception to look for such a justification. This argument bears a striking resemblance to certain Western proposals for "dissolving" the problem of induction, and for the remainder of this essay we shall try to develop a version of this type of approach.

VI.
If the traditional "problem of induction" is conceived of as a demand for the logical justification of induction, then it can be shown that this is an incoherent demand. It is a priori impossible to supply a justification for inductive inference if what is being demanded is that any proposed valid inductive inference must meet the conditions of adequacy appropriate to deductive inferences. This is to take the criteria of adequacy appropriate to one area of inquiry and improperly impose them on quite another area of inquiry. Of course we cannot logically deduce valid conclusions about the future from premises that say nothing about the future. But this is only to say that inductive inferences are not deductive inferences. Why blame induction for not being deduction? We might as well lament the fact that deduction is not induction.

The whole demand for a general justification of induction is radically mistaken. It makes sense to ask of a particular belief whether or not its adoption is justified, for this is to ask whether there is any evidence for it. Out normal use of terms like "justified," "well-founded," and so forth involves implicit appeal to inductive standards. But if we ask whether the application of inductive standards is itself justified or well-grounded, then to what standards are we appealing? Failing any possible answer here we must conclude that the question makes no sense. P. F. Strawson highlights this point with a revealing analogy:

Compare it with the question: Is the law legal? It makes perfectly good sense to inquire of a particular action, of an administrative regulation, or even, in the case of some states, of a particular enactment of the legislature, whether or not it is legal. The question is answered by an appeal to a legal system, by the application of a set of legal (or constitutional) rules or standards. But it makes no sense to inquire in general whether the law of the land, the legal system as a whole, is or is not legal. For to what legal standards are we appealing? [28]

Strawson goes on to say that the only sense we can give to the question "Are all conclusions, arrived at inductively, justified?" is to take it as meaning "Do people always have adequate evidence for the conclusions they draw?" Then the answer is simple: Sometimes they do and sometimes they do not.

p. 172

Undoubtedly some people will feel unsatisfied with this line of argument. Perhaps they will object that if we are going to appeal to certain criteria for good reasons that are implicit in the rules of use for ordinary language, then surely we should try to break through this linguistic barrier and attempt to determine why these rules are considered to supply the criteria for good reasons, whether or not these actually are good criteria, and thus whether or not we should still accede to these rules. This objection is seriously misconceived for reasons that we shall attempt to delineate in what follows.

We, all of us as adult humans, utilize what are broadly the same procedures for making and assessing nondemonstrative inferences. In this sense we are all part of the complex interacting network of experience-based learning systems that Max Black has called "the inductive institution." [29] The philosophical problem of justifying induction can only arise for someone who is already a member of this inductive institution. It is not like the case of chess, which someone can understand without being a player. Rather we are all necessarily players of the "induction game" before we can possibly begin to pursue that kind of self-conscious reflection characteristic of philosophical inquiry. Piecemeal reform of the inductive institution is certainly possible and indeed characterizes the growth of modern science. However, although no particular feature of the institution is above criticism and reconstruction, the entire institution cannot be called into question without destroying the meaning of the very words used to pose the philosophical problem of induction. To be in command of the inductive language and hence part of the inductive institution is also to be subject to the norms of belief and conduct imposed by the institution. Thus the question "Why should we accept any inductive rules?" makes no sense.

It is instructive at this point to recall the views of Hume and the Caarvaakas. Both correctly regard inductive practices as social and contingent facts, for it is certainly true that it is a contingent fact that at a particular point in time there came into existence beings sufficiently rational to be able to use language and hence inductive concepts. However, what they both failed to realize is that the inductive institution is partly formed by normative inductive rules to which the skeptical philosopher is already committed by virtue of his being a rational human being. As Max Black puts it:

Thus, the encompassing social fact of the existence of the inductive institution includes within itself the means for appraisal and criticism of inductive procedures; we cannot regard inductive inference as something merely "given", as a natural fact, like the Milky Way, that it would be absurd to criticize. To understand induction is necessarily to accept its authority. [30]

In sum: The question "Why should any induction be trusted?" -- that is, a question about the ultimate or final justification of induction as such -- is without sense. Trying to raise such a question we reach what Wittgenstein called "the limits of language." What we have been concerned to argue here is that such questions are not the fundamental inquiries that some people imagine, but rather

p. 173

they are devoid of content, for even to be able to formulate the question requires prior commitment to the inductive institution. This being so, it follows that "the problem of induction" as conceived by both the Caarvaakas and Hume is really a pseudo-question, that is, not a problem to be solved but rather a confusion to be "dissolved." In the Indian philosophical tradition it seems that the Buddhist appeal to the doctrine of vyaaghaataavadhiraa`sa.nkaa yields an approach consonant with the type of argument we have been developing in the last few paragraphs.

However, although the skeptic's major propositions are false, it is also important to understand that what he is saying has some point. One very popular and misleading way of drawing attention to the fact that a particular variety of X is importantly different from another variety of X is to say that the second variety of X is not really X at all. In drawing attention to the important differences in the kind of knowledge we get through perception, as opposed to the kind of knowledge we get through inductive inference, the Caarvaakas overstated their case to the point of claiming that inference was not really a valid source of knowledge at all. What was rather required was a demonstration of the way in which inferential knowledge is dependent on perceptual knowledge and with this an associated cogent analysis of the invalidity of certain types of inferences current in Indian philosophical debates of the time. For example, there is a penetrating critique of the inductive-style inferences offered by the Naiyaayikas as proofs of the existence of I`svara. (In fact Indian philosophy had to wait for the theist Raamaanuja to supply such a critique of the Naiyaayika teleological arguments.) Nevertheless, what the Caarvaaka skepticism about inductive inference did do was to force the other dar`sanas to reexamine their presuppositions and thus to pursue the task of philosophical inquiry with increased rigor and subtlety. This in itself is no mean achievement. [31]

NOTES
1. What follows here draws very heavily on the excellent discussion in Karl H. Potter, Presuppositions of India's Philosophies (Englewood Cliffs, New Jersey: Prentice-Hall, 1963), p. 59ff.

2. Although it is not important for the purposes of this paper, it is worth making the historical remark that the Naiyaayikas did not accept this three-membered inference form, preferring instead a five-membered form. The Buddhists, Miimaa.msakas, and Advaitins all agree that the Naiyaayika inference form includes redundant members.

3. Sarvepalli Radhakrishnan and Charles Moore, eds., A Sourcebook in Indian Philosophy (Princeton, New Jersey: Princeton University Press, 1957), pp. 231-233.

4. David Hume, An Inquiry Concerning Human Understanding, sees. IV-V.

5. Mysore Hiriyanna, Outlines of Indian Philosophy (London: George Allen & Unwin, 1932), p. 188. And this does at least seem to have been the position maintained by one Purandara in the seventh century: see Surendranath Dasgupta, A History of Indian Philosophy, vol. 3 (Delhi: Motilal Banarsidass, 1975), p. 536.

6. Potter, Presuppositions, p. 49f.

7. Radhakrishnan and Moore, Sourcebook, p. 235.

8. For a very useful discussion of the Naiyaayika position see Satischandra Chatterjee, The Nyaaya Theory of Knowledge, 2d ed. (Calcutta: University of Calcutta, 1950), pp. 209-218, 243-252.

p. 174

9. Janaki Vallabha Bhattacharyya, trans., Jayanta Bha.t.ta's Nyaaya-Ma~njarii, vol. 1 (Delhi: Motilal Banarsidass, 1978), pp. 252-253.

10. For a study of the role of tarka in Indian logic see Sitansusekhar Bagchi, Inductive Reasoning (Calcutta: `Srii Munishchandra Sinha. 1953).

11. Swaamii Maadhavaananda, trans., Bhaa.saa-pariccheda with Siddhanta-muktaavalii by Vi`svanaatha Nyaaya-pa~ncaanana, 2d ed. (Calcutta: Advaita Ashram, 1954), p. 100. In taking Vi`svanaatha's writings as representative of the views of the Navya-Nyaaya philosophers I am following Ingall's usage and regarding all Naiyaayikas from the time of Ga^nge`sa (13th century) to the present as Navya-Naiyaayikas: see Daniel H. H. Ingalls, Materials for the Study of Navya-Nyaaya Logic (Cambridge, Massachusetts: Harvard University Press, 1951), pp. 1, 5. However, some scholars feel that Vi`svanaatha's work is more precisely classified as a syncretic manual combining the views of the old Nyaaya school with the terminology of the new school: see, for instance, D. N. Shastri, Critique of Indian Realism (Agra: Agra University Press. 1964), pp. 122-123.

12. Of course, the Buddhists, too, denied the existence of universals. However, it is interesting to observe that Jayanta's argumentative strategy against the Buddhists would beg the question if directed against the Caarvaakas, for he maintains that "the sum and substance of the argument of the Naiyaayikas is that the hypothesis of a universal is necessary for the possibility of verbal and inferential knowledge" (Bhattacharyya, pp. 628-629).

13. Maadhavaananda, Bhaa.saa-pariccheda, p. 103.

14. Radhakrishnan and Moore, Sourcebook, p. 231.

15. Maadhavaananda, Bhaa.saa-pariccheda, p. 215. For an interesting paper on the Nyaaya theory of doubt and its relation to philosophical skepticism see J. N. Mohanty, Phenomenology and Ontology (The Hague: Martinus Nijhoff, 1970), pp. 198-219.

16. Maadhavaananda, Bhaa.saa-pariccheda, p. 103.

17. The Nyaaya conception of inquiry as the construction of a philosophical system is discussed in an illuminating way in Karl H. Potter, ed., Indian Metaphysics and Epistemology: The Tradition of Nyaaya-Vai`se.sika up to Ga^nge`sa (Princeton, New Jersey: Princeton University Press, 1977), pp. 38-46.

18. Radhakrishnan and Moore, Sourcebook, p. 237. There is some doubt as to whether Jayaraa`si really was a Caarvaaka: see Dale Riepe, The Naturalistic Tradition in Indian Thought (Seattle, Washington: University of Washington, 1961), p. 62. Nevertheless, the argument here seems consonant with what little is known of the Caarvaaka position.

19. S. S. Suryanarayana Sastiri, ed. and trans., Vedaantaparibhaa.saa by Dharmaraaja Adhvarin (Adyar: Adyar Library, 1942), chap. 2.

20. D. M. Datta, The Six Ways of Knowing, 2d ed. (Calcutta: University of Calcutta, 1960). p. 209.

21. B. N. K. Sharma. Madhva's Teachings in His Own Words, 2d ed. (Bombay: Bharatiya Vidya Bhavan, 1970), pp. 50-51.

22. Th. Stcherbatsky, Buddhist Logic, vol. 1 (New York: Dover Publications, 1962), p. 260.

23. Ibid., vol. 2, p. 73.

24. Ibid.. vol. 1, p. 248.

25. Cf. Hiriyanna, Outlines, p. 200.

26. Bhattacharyya, Jayanta Bha.t.ta, p. 250.

27. Nyaayakusumaa~njali iii, 7, as translated in Debiprasad Chattopadhyaya and Mrinalkanti Gangopadhyaya, Nyaaya Philosophy, vol. 2 (Calcutta: Indian Studies Past & Present, 1968), p. 75.

28. P. F. Strawson, Introduction to Logical Theory (London: Methuen, 1952), p. 257.

29. Max Black, "Induction," in Paul Edwards, ed., The Encyclopedia of Philosophy, vol. 4 (New York: Macmillan, 1967), p. 178.

30. Ibid., p. 179. For a different but not totally unrelated approach see Nicholas Rescher, Induction (Oxford: Basil Blackwell, 1980) which presents a "methodological-pragmatic" justification of induction. Rescher argues that induction is a cognitive method, and hence its justification can be assimilated to those pragmatic devices used to justify methods in general. Perhaps this might be an approach attractive to a modern Naiyaayika, for in many respects it seems congruent with the Nyaaya conception of inquiry.

31. My thanks to the anonymous outside reader for Philosophy East and West who provided a number of useful suggestions with regard to an earlier version of this paper.