ais likeRb
Tb
\TaThe response given in the tradition to this problem is to impose further con-
straints on the relation of likeness. Relevant or extrapolation-warranting like-
ness consists in the sharing of a property at least as narrow in extension as the
property to be extrapolated. The idea is clearly expressed and ably defended by
Din.na ̄ga, and for that reason he is rightly thought of as one of India s finest logi-
cians. I also think that the pre-Din.na ̄ga Naiya ̄yikas would not have developed
the theory of inference in the way they did unless they had some such idea in
their minds. But I do not believe we should be led by this to try to reinterpret
what they said as an unequivocal expression of the idea. Matilal puts the matter
well when he says that “the conception of a universal connection is being hinted
at on the analogy of a universal property.”^40 Thehintwas there, but it was for
Din.na ̄ga to lend that hint articulation (and perhaps it is only at moments when
a theory is being revised that a precise definition is needed). The important point
is this need not be read as the introduction of a new premise into the inference
pattern, but rathet as a condition on when an inference is admissible. The con-
straint is of the form: it is valid to infer Ta fromTbifais likeRbwhen b, the
example, is relevantly like a(i.e. when the property it shares with ais narrower
in extension than the property being extrapolated). An inference rule is not
another premise in the inference, but rather that in virtue of which the infer-
ence is valid or invalid. And the treatment of the early Nya ̄ya theory as a theory
of inference from sampling shows how the rule that there be a “universal con-
nection” (vya ̄pti) of this kind between the properties is not an enthymematic
premise, but a genuine inference rule of an informal logic.
Five sorts of bogus reason (hetva ̄bha ̄sa) are mentioned in Nya ̄yasu ̄tra 1.2.4:
thewandering, the contradictory, the unproven, the counter-balanced,andthe
untimely.^41 Of special importance here is the one called the “wandering” (for the
unproven” and the “counter-balance,” see section 1.9). This was interpreted by
later Naiya ̄yikas as meaning that the reason property deviates from the target,
and so as a case in which the criterion enunciated in the above paragraph is vio-
lated. If a faulty reason is one which is present somewhere the target property
is not, then by contraposition, a proper reason is one which is not present some-
where the target property is not, and so is at least as narrow in extension. The
Nya ̄yasu ̄tra definition is, however, less than explicit, and Va ̄tsya ̄yana’s explica-
tion of this most important su ̄tra is all over the place:
The wandering is that which does not remain at only one end. (NS 1.2.5)
An example is: sound is eternal because it is intangible. A pot is tangible and is
seen to be non-eternal. Sound is not tangible in the same way. What then? It is
intangible. One might say, therefore, that because it is intangible, sound is eternal.
However, in this example tangibility and non-eternality are not grasped as stand-
ing in a prover–proven relationship. For example, an atom is tangible and it is436 jonardon ganeri