Let us first see what is the condition for the truth of the conclusion, i.e., the inferential cognition. According to the Naiyayikas, an inferential cognition is true if its immediately preceding causal condition, viz. the paramarsa, that is, the cognition that the paksa (the subject of inference) is characterized by the hetu (the inferential mark) which has the relation of pervasion (vyapti) with the sadhya (the inferable property), is true. This so because the Naiyayikas hold that a true cognition can be produced only by means which is qualified by some guna or excellence.7 In the case of inferential cognition the means is the paramarsa and its truth is the stipulated excellence.8

Now, the paramarsa cannot be true unless the other two cognitions, namely, paksadharmatajnana and vyaptijnana, which precede the paramarsa and are among the causal conditions of anumiti (the inferential cognition), are also true. If we call these causal conditions premises, we can say that the truth of the premise is the sufficient condition for the truth of the conclusion, i.e., the inferential cognition.
The other set of conditions is required to be fulfilled in order that the inference is legitimate, that is to say, the hetu of the inference is legitimate (sat). A hetu is legitimate if and only if it possesses five characteristics,9 viz.: (1)   paksa-sattva (to be present in the paksa); (2)   sapaksa-sattva (to be present in a thing other than the paksa, where the presence of the sadhya is known for certain); (3)  vipaksasattva (should not be present in a thing where the sadhya is known to be absent); (4)  abadhitatva (the contradictory of the sadhya must not be known to be present in the paksa) and (5)   asatpratipaksitatva (a hetu tending to prove the contradictory of the sadhya must not be present in the paksa). In our example of the inference of fire in a hill on the basis of smoke, the hill is the paksa, the kitchen is a sapaksa, and the hetu smoke is present in both the hill and the kitchen. A lake would be a vipaksa, and smoke is not present in it. The hill is neither known to be characterized by the absence of fire, nor is some hetu which is pervaded by the absence of fire is found to be present in the hill. So the hetu of this inference has all the five characteristics of a legitimate hetu. But there are certain types of inference the hetu of which cannot and need not have all the five characteristics, four of them suffice in such cases. In order to grasp these details we are to have an idea of the Nyaya classification of inference.
The Naiyayikas first divide inferences into two kinds: svarthanumana (inference for one’s own self and pararthanumana10 (inference for others). Svarthanumana is a way of knowing, but pararthanumana is something more, it is a way of proving a position. In this paper, however, we are concerned with svarthanumana (inference for one’s own self) only.

Each of these two kinds of inference can again be of three types:11 (a)  anvaya-vyatireki,  (b)  kevalanvayi  and  (c)  kevalavyatireki. Though there is a difference of opinion regarding the principle of this division, still it can roughly be said that if both a positive instance of the hetu being accompanied by the sadhya and a negative instance of the absence of sadhya being accompanied by the absence of hetu can be found, then the inference with such a hetu belongs to the type (a). An example of this type of inference is the inference of fire from smoke, because, smoke is found to be accompanied by fire in the kitchen and the absence of fire is also found to be accompanied by the absence of smoke in the lake. If only a positive instance of the concomitance of the hetu with the sadhya can be found, then the inference with such a hetu belongs to the type (b). An example of this type of inference is: A horse is nameable because it is knowable, just as a man is. Since, as the Naiyayikas maintain, every real thing is knowable and nameable, no instance of the absence of nameability being accompanied by the absence of knowability can be found. And if only the negative instance of the concomitance of the absence  of the sadhya with the absence of the hetu can be found then the inference with such a hetu belongs to the type (c). The following inference belongs to this type. A substance differs from other things because it has substancehood. Here no locus of substancehood other than substance can be found, but all the substances are included in the paksa, hence only a negative instance of the concomitance of the absence of the difference from things other than substance with the absence of substancehood is found.

Now, of the five characteristics of a legitimate hetu as mentioned above, the hetu of an inference of type (a) should have all of them; the hetu of an inference type  (b) should have four of them excluding only vipaksasattva (not being present in a thing where the sadhya is known to be absent), because, in that type of inference the sadhya is such that nothing can be found to have its absence. The hetu of the inference of the type (c) should have the four characteristic excluding only sapaksasattva (being present in a thing other than paksa where the presence of the sadhya is known for certain), for in such an inference the paksa alone is the locus of the sadhya.

So we find that having either five or four of the characteristics mentioned above makes a hetu legitimate (sat). And we have already said that according to the Nyaya theory an inference with a legitimate hetu would yield a true conclusion.

Each of the five characteristics of a legitimate hetu has either direct or indirect bearing upon the truth o the conclusion of an inference. Let us see how.

 

Next..