Professional Documents
Culture Documents
Gangesaon God
Gangesaon God
Stephen Phillips
University of Texas at Austin
Nyāya's epistemic theory of inference. Inference is one of four ways we gain knowledge.
Logic is not abstracted from its embeddedness in the “knowledge source,” and it is as a
knowledge source that inference (anumāna) is analyzed, not only by Gaṅgeśa but by
Naiyāyikas all the way back to Gautama, the sūtrakāra, and Vātsyāyana, who in the fifth
century wrote the oldest extant commentary on Gautama's foundational text, the Nyāya-
sūtra (c. 200). Perception, inference, analogy, and testimony are the knowledge sources
recognized by the Nyāya school. All are conceived as processes connecting knowers with
objects and facts known, objects and facts in a very real world (to include psychological
and linguistic facts). Perception is the fundamental link to objects, and is presupposed by
the operations of the other sources. Analogy is restricted in scope to the acquisition of
word meanings. Testimony is conceived in an externalist and anti-inferentialist fashion
(acceptance is embedded in uptake normally: we don’t have to have to have evidence that
the testifier is an expert and then infer what she says is true, nor do we have to be self-
conscious about our acquiring new information). Unfortunately, there is not space here in
this venue for further elaboration of sources other than inference. After a few more points
of overview and some central epistemological principles, we shall focus exclusively on
inference as a pramāṇa before turning to the specific theistic inference that Gaṅgeśa
defends.
True beliefs are formed by episodes of occurrent knowledge defined as veridical
cognition (embedding a true proposition, savikalpaka-jñāna) that is produced by a
veritable knowledge source (pramāṇa). False beliefs are formed by pseudo-knowledge-
sources (pramāṇa-ābhāsa), imitators of the real McCoys. All bits of non-occurrent
knowledge are formed by corresponding bits of occurrent knowledge. Remembering is
not, therefore, an independent knowledge source, its correctness being dependent on its
formational source, perception, etc. Knowledge sources are factive. No non-veridical
cognition is—by definition of “knowledge source” (pramāṇa)—knowledge-source-
produced. However, people are fallible and sometimes believe that p when it is false and
that a belief is knowledge-source-produced when it is not. “Apparent (but non-genuine)
knowledge sources,” pramāṇa-ābhāsa, generate non-veridical cognitions or sometimes
veridical cognitions in the wrong way, e.g., from mistaking dust for smoke a subject
infers fire where there is actually fire (in a Gettier-style example).
A subject S knows that p if and only if:
(1) p is true
(2) S believes that p
(3) S's belief that p has been produced by a genuine knowledge source.
S may well know that p without knowing that she knows that p: Kp ≢ KKp.
Certification (KKp) is mainly a matter of source identification. There is a difference
between mere knowledge, pramā, and certified knowledge, nirṇaya. The division
corresponds nicely with the “animal knowledge” contrasted with “reflective knowledge”
in analytic epistemology: without knowing that we know, we, like animals, pick up new
information non-self-consciously, but sometimes we self-consciously defend or justify
what we take ourselves to know. Philosophy is all about certified belief, and so we self-
consciously put forth inferences, which, however, do not differ in structure from those
made automatically in everyday life, we learn from Vātsyāyana and Nyāya tradition,
from other schools, too. Indeed, certification has itself pragmatic consequences. It raises
the barrier to reasonable doubt.
Inference as a knowledge source requires that the premises themselves be
pramāṇa-generated. A valid deduction with a false or unwarranted premise does not
count as a proper inference.
To sum up and anticipate our attempt to assess Gaṅgeśa's effort in its own terms,
here are four central epistemological principles according to his Nyāya:
2
(a) All knowledge originates in a knowledge source.
(b) Justification amounts to source identification or confirmation by another source.
(c) Putative perceptual judgments, inferential conclusions, etc., are defeasible but are
to be considered true unless defeated by counterconsiderations (bādhakaṃ vinā).
(d) Inferential defeaters include fallacies (hetv-ābhāsa), undercutters (upādhi), and
high-level theoretic flaws such as departure from the overall coherence of the
system (apasiddhānta) and lack of economy (gauvara) — the latter employed in
tarka, “suppositional reasoning.”
Example: Sb,Fb & ¬Sc,¬Fc (positive and negative inductive support for premise 2)
The theistic inference from effects. Let's look now at the Nyāya inference that Gaṅgeśa
elaborates and defends thinking in the classical terms, which are all represented in the
first sentence of a long section devoted to proving īśvara (the īśvara-vāda of the Tattva-
cintā-maṇi)1:
1Tattvacintāmaṇi, ed. Ramanuja Tatacharya: anumāna-khaṇḍa, vol. 2 (1999), p. 191. The entire īśvara-
vāda section with commentary by Rucidatta and subcommentary by Dharmarājādhvarīndra runs pp.
191-294.
3
evam anumāne nirūpite tasmāj jagan-nimārtṛ-puruṣa-dhaureya-siddhiḥ kṣity-
ādau kāryatvena ghaṭa-vat sakartṛkatvânumānāt
Now for earth and the like, there is a principle called pakṣa-dharmatā-balāt, “on
the strength of X being a property of the inferential subject.” In one sense, the phrase is
used to capture the existential premise in a standard inference, namely, that the inferential
subject does indeed possess the prover property. But the property that later Naiyāyikas
often have in mind, and so too Gaṅgeśa in several tracts within the īśvara section, is the
probandum property, indeed, the complex property of being-qualified-by-the-prover-as-
pervaded-by-the-probandum. And so “pakṣa-dharmatā-balāt” is used to introduce further
reflection on what the inferential subject has to be like in order that, as we now know
having made the inference, the probandum is also a property of the inferential subject.
And so with the theistic inference it is reasoned that only an omniscient īśvara
could be the agential cause. Gaṅgeśa devotes many lines to the issue of what the
inference requires us to hold about the Lord’s nature.2 This reflection on his part has rich
Naiyāyika precursing. Vācaspati, for example, says that the universe including its
2 For example (Tirupati 250): pakṣa-dharmatā-balān nityaṃ jñānaṃ sidhyad buddhir anityā eva iti vyāpti-
pratyakṣeṇa na pratibadhyate | asmad-ādi-buddhi-viṣayakatvena bhinna-viṣayakatvāt eka-viṣaya-virodhi-
jñānasya eva pratibandhakatvāt nityatvânityatvayor eka-jātīye dravye 'virodhāt | buddhi-mātre
'nityatvâvagamāt kathaṃ tad-viśeṣe nityatva-buddhir iti cet na | buddhi-mātra-padasya īśvarânīśvara-
buddhi-paratve virodhād vyabhicārāc ca | asmad-ādi-buddhi-mātra-paratve ca bhinna-viṣayatvena
apratibandhakatvāt | “By force of what it is for the probandum to be a property of the inferential subject,
permanent cognition is proved. This is not contradicted by the pervasion known by perception, “Cognition
is nothing but impermanent,” because as restricted to the cognition belonging to the likes of us it is about
something different. Opposition, contradiction, ruling out, occurs only when we are aware of opposition
concerning one and the same thing. Being-permanent and being-impermanent are not in opposition when
applied (not to one and the same thing but) to substances within the same category.”
4
dimension of simultaneity is what God as agential cause would explain.3 In another line
of thought important for Gaṅgeśa, Vācaspati also argues that the concept of a single God
is simpler than that of multiple divinities each responsible for something different—down
potentially to the individual atoms, as though there were to be one divinity per atom. This
is an uneconomical hypothesis because of problems of co-ordination. Simpler is to
suppose a single God.4
The “effects” argument, as we might call it, was a target of Buddhists (Patil 2009)
who pointed to counterexamples such as growing grass. Growing grass exhibits the
prover property, being-an-effect, but not the property to be proved, having-an-agential-
cause. The Nyāya reply is to point out that growing grass and all such examples are in
dispute, that is to say, fall within the domain of the inferential subject (pakṣa), here “earth
and the like,” to wit, anything that is an effect but whose agential cause is not apparent.
Nothing that falls within the domain of the subject can be used either as an example
supporting the rule of vyāpti or as a counterexample, since that would beg the question
either way. The whole point of inference is to make something known that was not
known previously, and here, as with the existence of atoms, the conclusion could not be
known perceptually. The theistic inference has the purpose of showing that earth, and
things like earth that are effects, such as growing grass, have an agent within the causal
complex (sāmagrī) that brings them about. This is not something we know without
making the inference. Gaṅgeśa devotes several long tracts within the section to questions
about the scope of the pakṣa. A complex of issues surrounding the pakṣa along with the
nature of the īśvara inferentially proved make up the largest portion of the text
traditionally designated pūrva-pakṣa, the “opponent's” or the “prima facie” view.
(c) Putative perceptual judgments, inferential conclusions, etc., are defeasible but are
to be considered true unless defeated by counterconsiderations (bādhakaṃ vinā).
Knowledge sources are factive, but human subjects are fallible, as experience shows.
Perceptual illusions occur, sometimes apparently through no fault of the perceiver's, and
3Nyāya-vārttika-tātparya-ṭīkā, ed. Thakur (1996), pp. 564 (commentary under NyS 4.1.21):
aparimeyâniyata-dig-deśêndriyakâtīndriyaka-trasa-sthāvarâdi-lakṣaṇa-kāryôtpāda-yaugapadyam: “(That
which is to be accounted for) is the co-ordinate simultaneity of production of effects throughout
immeasurable and unlimited space at every place and location, effects perceptible and imperceptible in
animals and plants and the organic world as a whole and so on.”
4 ibid., p. 566.
5
similarly with inference we know that people reason fallaciously much of the time. And
there are situations where we are misled by what appears to be a reliable testifier.
Here we are concerned with potential defeaters of inferential knowledge, and
there are two that Gaṅgeśa considers at length, counterinference (sat-pratipakṣa) and the
upādhi “undercutter.” It is no “straw man” that Gaṅgeśa considers with respect to the
theistic inference from effects; in each case the defeater on the face of it seems successful
in blocking the inference. The proposed upādhi is:
The proposed counterinference (Tirupati 261ff) is (following the usual pattern, with ‘a’
standing for “earth and the like” and ‘I’ = “not-produced-by-an-agent-with-a-body” and
‘S’ as in the targeted inference = “having-an-agential cause”):
Ia
(x) (Ix ➡ ¬Sx)
_______
¬Sa (Earth and the like do not have an agential cause.)
In a long section much earlier in the inference chapter, Gaṅgeśa examined the
traditional definition of an “inferential undercutter,” upādhi (U), and argued for a slightly
revised view (Phillips and Ramanuja Tatacharya 2002). Gaṅgeśa finds some exotic
upādhis that do not meet the standard definition. For the purpose of appreciating the
problem he sees with his theistic inference, however, the standard definition will suffice.
It is comprised of two conditions:
Therefore:
(∃x) (Hx & ¬Sx). (There is something that has the prover without having the
probandum.)
5 Slightly different formulations are considered depending on the precise formulation of the effects
inference at a particular spot in the text. The first, Tirupati 243, is “being-produced-by-bodily-employment”
which is proposed as an upādhi for the inference whose probandum is spelled out as “being-produced-by-
an-agent.” “Having-a-living-body” is introduced a little later, Tirupati 244.
6
Thus it is false, for example, that fire (H) entails smoke (S), since smoke entails wet fuel
(U) and there is something, e.g., a hot ball of iron, where there is no wet fuel but there is
fire—or, so it is thought, to use the stock example.6
So, everything that we know of that has an agential factor within the causal
complex sufficient to bring it about (S) has as that causal factor an agent with a living
body (U). But grass and many other things appear to be effects without having as a causal
factor an agent with a living body (H & ¬U).
In my judgment, Gaṅgeśa does not have a great answer to this objection. Part of
his response is that no truly good inference faces an upādhi, and that the īśvara is a
counterexample to the upādhi’s presumed general rule. He elicits the distinction between
a genuine upādhi and the pseudo-upādhi in cases of the cooked-up property, being-other-
than-the-inferential-subject, pakṣêtaratva. Whether this is really an upādhi or not does
hinge on the cogency of an original inference. This makes it seem as though
counterinference is the real problem, and Gaṅgeśa does take that defeater up after
considering the apparently devastating upādhi (Tirupati 261ff). The two are closely
related in any case, and a counterinference can be constructed out of an upādhi. The
difference is subtle, a matter of whether we are focused on contrary evidence (the upādhi)
or on a different prover purportedly supporting the negation of our original probandum.
We shall address the full response after reviewing counterinference and introducing
Gaṅgeśa's ultimate point of defense which involves tarka, “suppositional reasoning.”
Here is the putatively defeating counterinference that Gaṅgeśa considers, with a
general rule and its contrapositive:
Whatever is produced agentially (S) is produced by (an agent with) a living body
(L).
Whatever is not produced by a body (¬L) is not produced agentially (¬S).
¬L = I (not-produced-by-an-agent-with-a-body)
Thus the atheistic counterinference accords with the definition above. In order words:
Theistic inference: Earth and the like are produced by an agent (without a body),
since they are produced, like a pot and so on.
6Typically, demonstrating an upādhi does even more than block a putative inference; it suggests why the
effort may have looked good although it is flawed. The upādhi is typically an “additional condition”: Ha
would not entail Sa unless there were in addition Ua: (x) ((Hx & Ux) ➡ Sx). But for our inference this is
not the case because the īśvara’s having a body is ruled out by tarka (to be explained later).
7
Atheistic counterinference: Earth and the like are not produced by an agent,
since they are not produced by a living body, like an atom, ether, and so
on.
“Suppositional reasoning” to the rescue. Tarka is very much like Socratic elenchus as
exhibited in early Platonic dialogues, a matter of finding an intellectual flaw in an
interlocutor's views. With Socrates, elenchus is typically employed without defense of a
counterthesis, but in the Indian case it is normally presumed that a counterthesis is on the
table which then profits from the refutation of its rival. Hypothetical or “suppositional
reasoning” (the rendering is due to Jonardon Ganeri) is addressed by Gaṅgeśa in an early
section of the inference chapter but no list of forms is forthcoming. There are however
lists in other Nyāya texts. Let us take a quick look at tarka in the context of its actual use
in the Jewel and other philosophic treatises as well as in Gaṅgeśa's theoretic treatment.
Given that some of our beliefs sometimes face counterconsiderations and are
reasonably challenged by a counterthesis or simply counterevidence, sometimes tarka—
which is not research or finding new evidence by way of a pramāṇa but rather armchair
reasoning in a good sense—is useful in its capacity of reëstablishing a presumption of
truth in favor of one or another thesis brought into doubt. For example, by supposing the
truth of an opponent's thesis and showing how it leads to unacceptable consequences, one
repossesses a presumption of truth, provided one’s own thesis is supported by at least the
appearance of a pramāṇa and, further, that it has not itself been ruled out. Thus tarka
comprises arguments that are not in themselves knowledge-generators but capable
nonetheless of showing what we should and should not believe. In particular, reasoning
can show “untoward ramifications” (prasaṅga) of not accepting a conclusion that is
evidenced, Gaṅgeśa argues.
Some forms of tarka, or of what is shown by suppositional reasoning, according
to Udayana (c. 1000) and others whom Gaṅgeśa follows without himself providing a list:7
7Our list is taken from Viśvanātha's notes on NyS 1.1.40, the sūtra that defines tarka (Nyāyadarśanam, pp.
325–327).
8
• self-dependence (begging the question), ātmâśraya
• mutual dependence (mutual presupposition), anyonyâśraya
• circularity (reasoning in a circle), cakraka
• infinite regress, anavasthā
• unwanted consequence (contradiction), aniṣṭa-prasaṅga
• being presupposed by the other, prathamôpasthitatva
• (hasty) generalization, utsarga
• differentiation failure, vinigamanā-viraha
• theoretic heaviness, gaurava
It’s the last that, according to Gaṅgeśa, eliminates the atheistic rival so that the cogency
of the theistic inference is restored (i.e., lack of comparative theoretic economy is a
defeater-defeater, defeating the purported counterinference so that the theistic inference is
restored to its rightful status as knowledge-generating).
And so the key question seems to be: Why does the theistic inference show
greater theoretic economy (lāghava which is the opposite of the alleged defeater-defeater,
gaurava)? Answer: simpler specification (avacchedakatva) of the inferential relationship
(Tirupati 281):
9
last of the chapter) mukti, “liberation,” in particular its possibility. Advocacy of theism
seems not the main point.
The disarming of the putative upādhi, having-a-living-body, seems similarly
weak, and so again the point seems to be less the cogency of the theistic inference from
effects and more the process of philosophical reasoning, drawing on many ideas
elucidated previously, in the “non-applied” portion of the inference chapter.
Recall that an upādhi is defined, for most purposes, as an item that pervades the
probandum while failing to pervade the prover. There is an easy counter to the proposed
upādhi pervasion that however draws deep on the Nyāya metaphysics of interactive
dualism. A self can originate physical effects but is not itself something physical. Effort
to move your hand does not require another physical instrument but originates in you as a
non-physical substance. And so Gaṅgeśa cites (Tirupati 256) movement of the hand as an
example of an effect that although itself physical or bodily does not have a bodily cause,
being the direct result of the will and effort of a self. Thus it is not true that having-an-
agential-cause is pervaded by (entails) having-an-agent-with-a-body, since selves are not
essentially embodied but are capable of originating certain bodily actions without a
bodily intermediary. The problem with this is that the metaphysical dualism is
controversial, and normally Gaṅgeśa, like any good Naiyāyika, tries to find examples that
are widely accepted, not so disputed as with the metaphysics of the self, which may be
the most controversial of all the topics debated on the classical scene.
There is another point made in favor of rejection of the upādhi challenger: a good
inference is by definition upādhi-free. Consider the cooked-up property, pakṣêtaratva,
being-other-than-the-inferential-subject, which seems guaranteed to pervade the
probandum: it is everywhere other than the inferential subject itself, and that subject is
bracketed in the context of inference, not to beg the question (see Phillips and Ramanuja
Tatacharya [2009: 160-62] for an extended discussion). And so to block the challenge of
the putative upādhi, having-a-living-body, the question may be, as with pakṣêtaratva,
which sometimes is and sometimes is not a pseudo-upādhi, whether or not we are certain
about its pervading the original probandum. When we are certain, for example, with the
bad inference from fieriness to smokiness, being-other-than-that-fire is a veritable
undercutter. But we are entirely confident that smokiness is pervaded by fieriness, and so
with respect to the good inference pakṣêtaratva is a pseudo-upādhi. Similarly, with the
theistic inference. We are entirely confident (but are we really?) that being-an-effect is
pervaded by having-an-agential-cause, and so the putative upādhi has to be pseudo. Earth
and the like are counterexamples to the rule implicit in the upādhi of having-an-agent-
with-a-living-body which does then not pervade having-an-agential-cause. The theistic
inference informs us that they are produced, in part, by effort on the part of an agent. The
patent circularity is a concern that does not go away in Gaṅgeśa's long presentation. Of it
he is well aware. Indeed, the attention he gives to it leads me again to suspect that, in the
end, endorsement of the theistic inference is not the main point, albeit it is enormously
refined and elaborately defended by the end of the section.
10
A final lesson. The topic of the ātman, “self,” is controversial, and this is important for
interpreting Gaṅgeśa, I believe. He does indeed present an inference proving a self earlier
in the inference chapter, in the kevala-vyatireki section (reconstructed by me [Phillips
2012: 65-68], but there he concedes that it goes through only in-house and not, for
example, against a Buddhist opponent who does not accept all of its terms. So here in the
īśvara-vāda, he may be taking himself to address a wide audience, not only those
predisposed to accept interactive dualism, or at least thinking about non-Naiyāyika
perspectives on the inference. My sense is that he would not like to be forced, as here, to
rely on examples not acceptable to many parties in philosophy.
Let me repeat that it is my reading that our author's main goal in the long and
complex īśvara-vāda in his Jewel—where long stretches of text are devoted to the
general nature of an inferential subject and precise specification of the terms of the
inference and how they are known—is first and foremost simply to illustrate excellent
philosophical reasoning (nyāya), as well as to show—I grant as a secondary motive—
how precise we need to be with the considerations for and against Nyāya’s īśvara
inference from effects. After all, Gaṅgeśa explicitly rejects the argument that the
excellence of the Veda requires that we suppose that its author is īśvara, agreeing with a
Mīmāṃsaka position.8 And out of the dozen or so theistic arguments put forth by
Udayana in his Nyāya-kusumâñjalī, Gaṅgeśa examines at length only that from effects.
All this leads me to think that Gaṅgeśa is a little less committed to theism than has
been presumed (e.g., Vattanky: 1984). Furthermore, Gaṅgeśa closes his inference chapter
with long expatiation on the possibility of apavarga, “final beatitude,” much as
conceived by Gautama and Vātsyāyana (as opposed to Udayana and others more
theistically inclined). Like that of those two very early Naiyāyikas, Gaṅgeśa's
commitment to “liberation”—which has it as understood in Sāṃkhya, that is to say,
dualistically, without an īśvara—seems clearly deeper than whatever there may be to any
theistic thesis. As many scholars have emphasized, a paltry number of Gautama's sūtras
—three—have anything to do with a “Lord,” and there is paltry discussion of an īśvara
on the part of Vātsyāyana. My feeling is that Gaṅgeśa is first and foremost a naiyāyika in
the old but also technical sense of someone both expert in and devoted to “critical
thinking,” and his specific religious views are subordinate to—at least in the Jewel—
championing what we may sum up as “philosophical reasoning,” nyāya. That is, the
īśvara inference is mainly an opportunity to elucidate canons of epistemology—
especially use of inference as supplemented by tarka, “suppositional reasoning”—taking
pramāṇa as methods of inquiry. Gaṅgeśa's is a rich and complex epistemological theory,
and my sense is that his treatment of the theistic inference is most of all to be appreciated
as an exercise in nyāya. This is not to say that he does not believe that its conclusion is
warranted, but rather to identify what he is really up to.
BIBLIOGRAPHY
8 Tattva-cintā-maṇi, perception chapter (1973: 385). Phillips and Ramanuja Tatacharya 2009: 192.
11
Bagchi, Sitansusekhar. 1953. Inductive Reasoning: A Study of tarka and Its Role in Indian
Logic. Calcutta: Munishchandra Sinha.
Gautama. NyS. Nyāya-sūtra. Ed. A.M. Tarkatirtha, Taranatha Nyayatarkatirtha, and H.K.
Tarkatirtha. 1936–1944 (reprint 1985). Nyāyadarśanam (Nyāya-sūtra with four
commentaries, the Nyāya-sūtra-bhāṣya of Vātsyāyana, the Nyāya-vārttika of
Uddyotakara, the Nyāya-vārttika-tātpārya-ṭīkā of Vācaspati Miśra, and the Vṛtti of
Viśvanātha). Calcutta Sanskrit Series 18.
Matilal, B. K. 1998. The Character of Logic in India. Ed. Jonardon Ganeri and Heeraman
Tiwari. Albany: State University of New York Press.
Patil, Parimal. 2009. Against a Hindu God: Buddhist Philosophy of Religion in India.
New York: Columbia University Press.
12
Phillips, Stephen H. “A Defeasibility Theory of Knowledge in Gaṅgeśa's Nyāya,”
forthcoming in the Oxford Handbook for Indian Philosophy, ed. Jonardon Ganeri. New
York: Oxford University Press.
Phillips, Stephen H. Forthcoming. “Fallacies and Defeaters in Early Navya Nyāya.” The
Bloomsbury Research Handbook to Indian Epistemology and Metaphysics, ed. J Tuske.
London: Bloomsbury.
13