Indian Logic by Raffaele Torella

Unlike Western thought, where logic’s field of action appears much wider, defining the nature of the logic peculiar to Indian thought essentially amounts to defining the nature of inference. Indeed, Indian philosophers have systematically made use of ‘linked reasoning’, particularly in establishing the principles of the necessary connection between a given property in a certain entity and another property of the same entity, momentarily or permanently inaccessible to direct perception, so as to go back to one from the other. The Sanskrit term commonly translated as ‘inference’, anumāna, tells us only that we are dealing with knowledge (-māna) that is not primary, but comes behind (anu-) something else — as a rule, direct perception — which has occurred previously and from which it proceeds for further development, i.e. reaching where perception alone does not, or assessing the validity of the perception itself, since Indian thought as a whole, surprisingly for those who blame it for excessive abstractness and vagueness, is unanimous in assigning epistemological primacy to direct perception. As Abhinavagupta strongly affirms in the Abhinavabhāratī (vol. I, p. 281), even though, regarding a certain object, various indirect means of knowledge might be available — such as authoritative testimony or inference — knowledge does not come to its definitive resting point (na viśrāmyati) but after direct perception. The source of Abhinavagupta is a passage from Nyāyabhāṣya (p. 9), dealing with the possibility of applying different means of knowledge to the same object. In these cases, knowledge acquired through authoritative testimony then seeks confirmation in inference, and the latter in direct perception. Only with direct perception the wish for knowledge is at last fulfilled, and ceases (jijñāsā nivartate), since all knowledge culminates in direct perception (sā ceyaṃ pramitiḥ pratyakṣaparā).

Western logic, beginning from Aristotle, started by distinguishing inductive from deductive inference, the former proceeding from the particular to the universal and the latter from the universal to the particular. Whereas for the former the path is always problematic and risky, but capable of producing a real increase in knowledge, for the latter it is much more linear, albeit susceptible to tautology. For deduction, a risk still more insidious than tautology is to lose itself in the correctness of the method used to draw conclusions, which may be formally correct, but factually wrong. As a first approximation, we should note that no kind of Indian inference allows any such duplicity of evaluation: inference is ‘correct’ only if confirmed as so being by reality.

In the earliest stages of Indian logic, represented in particular by the great medical schools, early Nyāya and Vaiśeṣika up to the fifth century C.E., inductive inference clearly predominates (Katsura 2000, Tillemans 2004), the burden of proof being placed on the accumulation of examples, thus leading to results that always risk being invalidated by a contrary example (which is the objection of the Lokāyatas). This is what is known in Western thought as probable inference or inference by analogy, at one end of which we have the limit case of the induction of a general truth reached by examining the totality of particular cases.

Like Western thought, in India too the form most employed — and also with greater cognitive relevance — is induction of the ampliative kind, which leads to an increase (albeit not absolutely certain) of knowledge by recourse to a limited number of examples, selected for their particular significance. A quality leap is marked by the introduction in Nyāya of the criterion of agreement and difference (anvaya-vyatireka), which represents the passing from a mere accumulation of examples to the coordinated employment of positive and negative examples. An interesting Western parallel is the so-called ‘Mills’ method’ (cf. Matilal 1999: 17-18), elaborated in the second half of the nineteenth century by the English philosopher for inductive inference: the method of agreement; the method of difference; the joint method of agreement and difference; the method of remainders and the method of concomitant variation.

A further step, also within the confines of reasoning of an inductive kind, is taken by Dignāga’s doctrine known as the ‘triplicity of logical reason’ (liṅgatrairūpya): property ‘A’, to be demonstrated, must be present in property ‘B’, occurring in the same substrate; it must be present in those cases in which ‘B’ is present and absent in those cases in which ‘B’ is absent. This is linked with the introduction of the concept of ‘pervasion’, called upon to replace and make more precise the previous generic notions of ‘relation’, ‘co-presence’, ‘concomitance’, etc. The extension of class ‘B’ must be greater than (or at least equal to) class ‘A’. The ideal trajectory of inductive reasoning may be said to end with the work of Dharmakīrti (600-660 C.E.). Dharmakīrti, deeming that Dignāga’s reform was incomplete, reformulated the ‘triplicity of logical reason’ in more restrictive terms. In particular, also unsatisfied by his master’s attempt, Īśvarasena, with the intent of making logical reason more consistent (the introduction of the clause of not noticing an invalidation by contrary cases; adarśanamātreṇa, ‘by virtue of mere non-seeing’), restricts admissible logical reasons to just three types, their validity being first established as intrinsically and factually existent (principle of causality, principle of identity and absence of perception). The path of reasoning is thus inverted from inductive to substantially deductive, as shown by the loss in importance — for the first time — of examples. Although the positive example may still appear, albeit reduced from an element of proof to a rhetorical artifice or additional information for the simple-minded, the negative example no longer has any raison d’être. The relevance — or lack of it — of the example is also referred to by the notions of ‘external pervasion’ (bahirvyāpti) or ‘internal pervasion’ (antarvyāpti), used by medieval scholasticism as one of the classification criteria for inference theories. External pervasion has to be proven ‘from the outside’, i.e. through examples, while internal pervasion is proven ‘a priori’ within the subject of inference itself (pakṣa). The Buddhist Ratnākaraśānti, in upholding the latter position with his Antarvyāptisamarthana (‘Demonstration of Internal Pervasion’; cf. Kajiyama 1999), accepted also by the Jains, does not realise that all its presuppositions were already contained in Dharmakīrti’s work (Bhattacharya 1986). In Dharmakīrti, the leap from induction to deduction remains however imperfect: deductive reasoning is, for example, guaranteed by the causality relationship, but what can guarantee the guarantor beyond any doubt except recourse, yet again, to a (problematic) induction?

This article forms the “Excursus 2: Logic” of the book The Philosophical Traditions of India: An Appraisal by Raffaele Torella, Indica Books, Varanasi, 2011, pp. 180-183.