2.2.1.1 (OC1E1) Perception (pratyakṣa)

The author of NB says that one perceives the object O when one’s self is connected with one’s mind (manas), and one’s mind with a sense faculty (indriya), and the sense faculty with O (NS 1.1.4 in ND 197). Due to its atomic dimension, the mind can be connected only with one sense at a time. Simultaneous perception is an illusion; it actually consists of momentary perceptions that take place very fast, like the fire-wheel created by a firebrand whirled in the air. The sense faculty becomes connected with O through a sensory connection (sannikarṣa). Essentially, this connection is the epistemic instrument for perceptual cognition.

NS 1.1.4 defines perceptual cognition as follows: perceptual cognition is (a) generated by the sense-object connection (indriyārtha-sannikarṣotpanna); (b) non-verbal (avyapadeśya); (c) true to its object (avyabhicāri); and (d) doubt-free, that is, certain (vyavasāyātmaka; for a detailed discussion, see Matilal 1986: 228–229). When a person’s eyes get connected with a cat, and they cognize it as a cat, the person attains a valid perceptual cognition. The cognition is true to its object. A doubt presents multiple contrary objects (e.g. ‘is this a cat or a civet?’). A valid perceptual cognition must be doubt-free, since it cannot have contrary objects (A and B are contrary when they both cannot be true at the same time). It is non-verbal because it occurs at the preverbal state. One has to see a cat prior to calling it a ‘cat’. Perception is a prerequisite for verbally defining an object.

2.2.1.2 (OC1E2) Inference (anumāna)

Every inference is based on perception (tat-pūrvaka NS 1.1.5 in ND 291). Take this example. First the epistemic subject observes the correlation between the reason (liṅga) and the target (liṅgī); for instance, ‘every locus of smoke is a locus of fire’. After having observed this, when they see smoke (i.e. the reason) on a hill, they remember its special relation with fire. Finally, they infer fire. Here, the hill (i.e. the locus of the target) is the site (pakṣa). The target (i.e. fire) is inferred, not perceived (apratyakṣo ’rtho ’numīyate; NB on NS 1.1.5 in ND 291). In this case, inference qua epistemic instrument is basically the cognition ‘this hill has smoke, which implies fire’, while the inferential cognition is, ‘this hill has fire’ or, alternatively, ‘this hill is a fire-possessor (vahni-mān)’ (thus ‘x is y’ is equivalent to ‘x has y-hood’).

NS 1.1.5 mentions inferences of three types:

1. cause-based (pūrva-vat), which is inferring the effect from the cause – for example, inferring future rain from dark heavy clouds (NB on NS 1.1.5 in ND 291);
2. effect-based (śeṣa-vat; more literal translations of pūrva-vat and śeṣa-vat could be ‘from the antecedent’ and ‘from the consequent’ respectively), which is inferring the cause from the effect – for example, inferring that it rained in the recent past from the rising water level in the river (NB on NS 1.1.5 in ND 291);
3. general-observation based (sāmānyato-dṛṣṭa), in which the occurrence of the target in the site cannot be perceptually known. A self (ātman) is not perceptible. Therefore, nobody can perceptually check whether it exists. Still, one can infer it from special attributes such as desires. Based on the observation that an attribute must be housed in a substance (e.g. a colour in an object), it follows that desire, aversion, etc. too must have a substance that houses them. That substance is the self (NB on NS 1.1.5 in ND 292).

2.2.1.3 (OC1E3) Analogy (upamāna)

Suppose John is well aware of leopards, although he has not seen a jaguar. Before going to the American mountains, he asks Xico, ‘What does a jaguar look like?’ Xico answers, ‘Like a leopard, although a jaguar has a barrel-like abdomen whereas a leopard has a slender build’. In a few days John sees an animal that looks like a leopard in the American mountains, and cognizes, ‘This must be a jaguar’. Through Xico’s comparison, John gets to know the referent of the word ‘jaguar’. In this way, according to the Naiyāyika, analogy brings about the knowledge of the relation between a noun and its referent (saṃjñā-saṃjñi-sambandha) through a statement of similarity (sādharmya; NS 1.1.6 and NB’s gloss on it in ND 335). The epistemic instrument for analogical knowledge is the knowledge of similarity (sādṛśya-jñāna).

2.2.1.4 (OC1E4) Testimony (śabda)

Suppose Ram is looking for his pet cat, and John has seen it in a grey vat. John – in this case – is someone who has direct knowledge of the object in question (sākṣāt-kṛta-dharmā; NB on NS 1.1.7 in ND 365). Also, John, who has the desire to convey this message to Ram, acts accordingly (cikhyāpayiṣayā prayuktaḥ) when he sees Ram (NB on NS 1.1.7 in ND 365). John is a reliable speaker (āpta) here. The words of a reliable speaker are testimony qua epistemic instrument (NS 1.1.7 in ND 365). Here, Ram is the epistemic subject, who attains a valid verbal cognition about something he does not have immediate access to.

3.2.2.1 The ontic triad

A thing is a complex having three elements: a property-holder (dharmin), properties (dharma), and relations (sambandha) through which a property Pi resides in the property-holder. Thus, a grey cat is a property-holder that has properties such as the grey colour and cat-hood. Both reside/are related to the cat through inherence.

3.2.2.2 The epistemic triad

A cognition is formless, but its object has got a form. The grey cat as the object of a cognition too is tripartite. In the cognition, ‘this is a cat’ (ayaṃ mārjāraḥ), this (i.e. the referent of the word ‘this’) is the qualificand (viśeṣya), and cat-hood (mājāratva) is the qualifier (prakāra), since, in this cognition, this appears to be qualified by the property cat-hood. Inherence, through which the qualifier is related with the qualificand, is the cognitive relation (saṃsarga). This triad may be represented as In(cat-hood)(this) where In = inherence.

3.2.2.3 Limitation, residing, and determination

When an entity x ‘becomes’ y by virtue of having the property Pi, Pi becomes the limiter/limiting property (avacchedaka) of y-hood. This is required since several essential properties P1, ..., Pn are resident (niṣṭha) in x. A stick (daṇḍa) causes a pot (ghaṭa). It does not do that all the time. When the stick causes a pot, the former becomes a cause (kāraṇa) for the latter. Thus, the stick, at that time, possesses an adventitious property (āgantuka-dharma) namely cause-hood (kāraṇatva), and the pot possesses another one namely effect-hood (kāryatva). The stick has many other properties such as substance-hood (dravyatva), corporeal-hood (mūrtatva), entity-hood (padārthatva), etc. But it becomes a cause for a pot only by virtue of being a stick. Therefore, its cause-hood would be limited by stick-hood. In other words, stick-hood would be the limiter of the cause-hood residing (niṣṭha) in the stick.

In the language of Nyāya: daṇḍa-niṣṭha-kāraṇatvāvacchedakaṃ daṇḍatvam. Substance-hood or entity-hood cannot limit this cause-hood, since every substance or entity is not entitled to cause a pot. On the other hand, the effect-hood residing in the pot would be limited by pot-hood. In Sanskrit: ghaṭa-niṣṭha-kāryatvāvacchedakaṃ ghaṭatvam. We can generally say that when x becomes y by virtue of having the property Pi (i.e. x qua Pi-possessor becomes y), Pi limits y-hood in x.

In the stick-pot case, the stick has the cause-hood limited by stick-hood, and the pot has the effect-hood limited by pot-hood. Intuitively we understand that there must be a thread between the cause-hood and the effect-hood. That thread, for the Naiyāyika, is determination (nirūpaṇa), which is symmetric. Thus, we can say that the effect-hood limited by pot-hood is determined by the cause-hood limited by stick-hood (ghaṭatvāvacchinna-kāryatvaṃ daṇḍatvāvacchinna-kāraṇatva-nirūpitam).

3.2.2.4 Describing an absence

Every absence has an absentee (pratiyogin) and a locus (adhikaraṇa). Consider a bag that has no cat. Here, the bag (syūta) is the locus for the absence of cat (mārjāra-abhāva), and cat-hood limits the absentee-hood (pratiyogitva) of the absence. Let ¬x denote ‘the absence of x’. The absentee has absentee-hood, and it is the absentee of an absence by virtue of having a specific property, i.e. cat-hood. Hence, cat-hood becomes the limiter of the absentee-hood residing in the cat (mārjāra-niṣṭha-pratiyogitvāvacchedakaṃ mārjāratvam). The cat is a substance too. Therefore, it is true that there is a substance that is absent in the bag. But saying that the limiter of the absentee-hood of ¬cat is substance-hood is saying that there is no substance in the bag, which is not true. There is an entity qua cat, and that is absent in the bag by virtue of having cat-hood. Now, there is ¬cat in the bag, and there is absentee-hood limited by cat-hood in the cat. Once again, the relation of determination comes here to glue them up. Thus, we say that the absentee-hood limited by cat-hood in the cat determines an absence in the bag. In technical Sanskrit: mārjāra-niṣṭha-mārjāravāvacchinna-pratiyogitva-nirūpitaḥ syūta-niṣṭhābhāvaḥ. Sometimes, ¬cat is described as the absence, whose absentee-hood is limited by cat-hood (mārjāratvāvacchinna-pratiyogitākābhāvaḥ; these technical terms are discussed by Guha 1968; Ingalls 1988; Goekoop 1967; Wada 2007; and many others).

3.3.1.1 Epistemic validity (prāmāṇya)

According to Gaṅgeśa, a cognition C is valid if and only if (i) C is non-recollective (anubhava, not smṛti), and (ii) the property that figures as the qualifier (prakāra) of x in C really belongs to x (tadvati tat-prakārakānubhavaḥ). When John sees his grey cat, and cognizes, ‘this is a cat’ (C1), his cognition is valid. C1’s qualificandum (viśeṣya) is this (i.e. the referent of ‘this’), qualifier cat-hood, and cognitive relation (saṃsarga) inherence (samavāya). Thus, its epistemic structure is: In(cat-hood)(this). This is valid since cat-hood really belongs to this. When, in a dull evening, John sees a civet in a bush and remembers his grey cat, he may attain a doubt-free – but invalid – cognition, ‘this is a cat’ (C2). Although C1 and C2 share the same epistemic structure, C2 is invalid since cat-hood does not reside in a civet through inherence. (For the original text and detailed discussions, see Phillips 2020a: 227–230; and Mohanty 1989: 34–44.)

Let Kc be the set of causal conditions for producing the cognition C. Gaṅgeśa thinks that the validity of C originates from another causal factor Kv. This means that C is valid only when Kc + Kv obtains. Kv is called epistemic excellence (guṇa). Each epistemic instrument has its own excellence. For a perceptual cognition P, the excellence is a causal connection between each part of P (i.e. the qualificandum, qualifier, and relation) and a sense faculty (avayava-indriya-sannikarṣa). For an inference I, it is a proper reason (yathārtha-liṅga). For a testimonial cognition T, it is the knowledge of the phrase (vākyārtha-jñāna). For an analogy A, it is the similarity (sādṛśya). Similarly, the epistemic invalidity (aprāmāṇya) is due to another factor Ki, which is called the epistemic defect (doṣa). For P and I, these defects may be bile etc. (pittādi) and an error about the reason (liṅga-bhrama) respectively (see Phillips 2020a: 172–173; and Mohanty 1989: 51–58). This is the Nyāya theory of extrinsic origin of validity (parataḥ-utpatti).

Let Rc be the set of factors that captures the cognition C. According to Gaṅgeśa, Rc does not capture the validity of C. If it did, then nobody would ever doubt their cognition. Even after cognizing that ‘this is water’ (W), one may doubt whether W is really valid. In that case, one goes there and tests the thing that seems to be water. If it quenches one’s thirst, then one’s testing activity (pravṛtti) is successful (saphala) and one infers W’s validity from this successful activity. This is a negative inference: had this cognition not been true in regard to the object, it would not cause a successful activity (for Gaṅgeśa’s text and its translation, see Mohanty 1989: 150–153). This activity captures the validity of W. This is the Nyāya theory of extrinsic apprehension of validity (parataḥ-jñapti).

3.3.1.2 Theory of causation (kārya-kāraṇa-bhāva)

Indian epistemology is an extension of the Indian theory of causation. An epistemic instrument is the causal instrument for valid cognition (pramā-karaṇa). What is a causal instrument for the Neo school? It is the most important (sādhaka-tama) cause. A cause of E is a factor that regularly precedes the production of E. Suppose among all the causal factors C1, …, Cn, Co is such that E takes place immediately after Co happens. An axe fells a tree, and the felling happens immediately after the dynamic contact between the axe and the tree (kuṭhāra-vṛkṣa-saṃyoga) takes place. A camera takes a picture, which is produced immediately after pressing the shutter-release button. These final causal factors are called causal operations (vyāpāra). For the Neo Nyāya, the factor that houses the causal operation is the causal instrument. In the abovementioned cases, they are the axe and the camera respectively.

In his TSD (ch. 2), Annambhaṭṭa presents the Neo Nyāya position:

According to the view that holds, ‘The cause that possesses the causal operation is the causal instrument’, the cognition of pervasion is the causal instrument for inference. It acts through consideration (parāmarśa). The causal operation is that which is caused by I and causes the (principal) effect of I. (Brahmacārī n.d.: 167, original emphasis)

The causal operation is not independent of the causal instrument I. The causal instrument causes both the effect and the causal operation. So the ‘I’ in the definition of causal operation (that which is caused by I and causes the principal effect of I) is actually the causal instrument. In the axe-tree case, both the dynamic contact and the felling are caused by the axe. But the dynamic contact is one of the causes of ‘the felling of the tree’. Thus the dynamic contact is caused by the axe and at the same time causes the effect of the axe, i.e. ‘the felling of the tree’. Therefore, the dynamic contact is the causal operation. In the case of visual perception, the eye is the causal instrument and the perception is the effect. The eye-object contact is caused by the eye and causes the perception as well. Thus, in this case, the eye-object contact is the causal operation.

3.3.2.1 (NE1) Perception (pratyakṣa)

Perceptual cognition – for Gaṅgeśa – is characterized by cognitive immediacy (sākṣāt-kāritva; Phillips 2020a: 307–309). When one perceives x, one naturally intuits, ‘I am immediately aware of x’ (sākṣāt-karomi). Gaṅgeśa adds that a perceptual cognition is that which has no other cognition as its causal trigger or instrument (Phillips 2020a: 311–313). For the inferential cognition, ‘S has T, since S has R’, the causal instrument is the cognition of pervasion: ‘every locus of R is a locus of T’. For the verbal cognition, ‘there is a cat in that vat’, the causal instrument is the cognition of the words like ‘there’, ‘cat’, and ‘vat’. Perceptual cognition does not depend on any other cognition. That is its immediacy.

There are two stages of perception. When somebody’s eyes establish a connection with a cat, first they capture three things: the object, the universal called cat-hood, and inherence. At this preverbal non-relational level, the perception is indeterminate (nirvikalpa) since its objects are discrete and unrelated. At the final level, the objects are combined, and the subject cognizes, ‘this is a cat’. This is determinate (savikalpa). This theory of two-tier perception is based on the assumption that the cognition of the qualified (viśiṣṭa, x qualified by y) is preceded by the cognitions of the qualificand and qualifier. Prior to having the knowledge of a man with a staff, one must know a man and a staff separately.

The Nyāya system presents a complex network of sensory connections (sannikarṣa). Some of those are ordinary (laukika) and some extra-ordinary (alaukika). First of all, the object of perception must be the size that can be perceived (medium size; madhyama-parimāṇa). Infinitesimal (aṇu-parimāṇa) and extremely large (mahat-parimāṇa) objects are not fit for perception. Secondly, one perceives the object O after the self-mind-O connection gets established (see section 2.2.1.1). NSM gives a detailed account of the sensory connections that connect the mind with O (NSM 257–271).

Let us consider the case of a black cow B. The black colour and cow-hood reside in B through inherence, and black-colour-hood is inherent in the black colour. The eye captures B through the connection called contact (saṃyoga). It captures the black colour and cow-hood through contacted-inherence (saṃyukta-samavāya) and black-colour-hood through the contacted-inherent-inherence connection (saṃyukta-samaveta-samavāya). Since the sound resides in the ether (ākāśa), and the ear in nothing but a part of the ether, the ear captures a sound through the inherence (samavāya) connection. It captures sound-hood (i.e. the generic property inherent in a sound) through inherent-inherence (samaveta-samavāya). An absence modifies or qualifies its locus; therefore one captures an absence through the qualifier-hood (viśeṣaṇatā) connection. It is worth noting the systematic nomenclature of the sensory connections here.

There are three extra-ordinary connections in the Nyāya system (NSM 272–275):

1. Through the mystical (yogaja) connection, a yogī can perceive distant objects, even past or future objects.
2. Through the connection of generality (sāmānya-lakṣaṇā), one knows the inferential pervasion (vyāpti; see section 3.3.2.2).
3. Through the connection of cognition (jñāna-lakṣaṇā), the subject cognizes an object of illusion.

The Nyāya theory of perceptual illusion is known as anyathā-khyāti-vāda (otherwise-cognition theory). Suppose somebody looks at a piece of nacre. Their eye gets connected with the object through the connection of contact, and with inherence through the connection of contacted qualifier-ness (saṃyukta-viśeṣaṇatā). However, it fails to capture nacre-ness due to some reason; the gap must be filled up in a perceptual cognition. So, they remember silver-hood (maybe because nacre looks like silver), and the recollective cognition itself – according to the Naiyāyika – serves as a sensory connection. It fills up the gap in ‘this is ___’ with ‘silver’. We may notice here that neither is the object of cognition qua ‘this’ illusory, nor is silver-hood illusory (since it exists in a real piece of silver), nor is inherence illusory. Thus, silver-ness is true ‘otherwise’ (anyathā), although it is not there in the object right in front of the subject. An error is basically cognitive heterogeneity according to the Nyāya view. Had ‘this’ been a piece of silver, all three objects of the cognition, ‘this is silver’ (namely ‘this’, inherence, and silver-hood) would be collected from the same source. Thus, the content of a valid cognition is homogenous (see NSM 280–282; Dravid 1996).

Sometimes one may look at a sandalwood from a distance, and perceptually cognize, ‘this sandalwood is fragrant’, even though the fragrance of the sandalwood cannot reach one at that distance. This too is a case of the connection of cognition. The sight of the wood invokes the memory of the sandalwood fragrance, and the recollection itself works as a connection. As a result, one perceives the recollected fragrance (NSM 281–282).

3.3.2.2 (NE2) Inference (anumāna)

The causal instrument for the inferential cognition ‘S has T, since S has R’ is the knowledge of the general rule, ‘every locus of R is a locus of T’, which is known as pervasion (vyāpti). It generates the inference through a reflective cognition (parāmarśa): ‘S has R, which is pervaded by T’. The site (S) (pakṣa) is the locus, where the target (T) (sādhya) is inferred from the reason (R) (hetu). Thus, one sees smoke on a hill and remembers the pervasion: ‘every locus of smoke is a locus of fire’, which is equivalent to ‘smoke is pervaded by fire’ or ‘fire pervades smoke’. Then, one cognizes, ‘this hill has smoke that is pervaded by fire’, and then finally infers, ‘this hill has fire’.

According to Gaṅgeśa, Y pervades X if Y is not limited by the absentee-hood determined by an absence, which coexists with X in a locus. Suppose every locus of smoke is a locus of fire. That being the case, ¬Z (i.e. the absence of Z), which coexists with smoke in some locus can have anything but fire as its absentee. Suppose smoke coexists with the absence of cats on a hill (i.e. there is no cat on that hill). Since cat in general is the absentee of ¬cat, ¬cat determines an absentee-hood limited by cat-hood. We must notice here that fire is not limited by cat-hood, since fire is not a cat. This applies to a bat or a rat or a mat, since smoke may coexist with the absence of any of these. Y, that is not limited by cat-hood or bat-hood or rat-hood or mat-hood, pervades smoke. Y happens to be fire. In sum, Y pervades X if Y is not absent in any locus of X (the absence, i.e. ¬Z, which figures in the definition must not be a partial absence, that is, one that coexists with its absentee in the same locus; for a detailed discussion, see Phillips 2020b: 623–624).

Pervasion is of two types. Y positively pervades X (or X has the positive pervasion or anvaya-vyāpti of Y) if every locus of X is a locus of Y. Y negatively pervades X (or X has the negative pervasion or vyatireka-vyāpti of Y) if every locus of the absence of Y is a locus of the absence of X. Inference is of three types based on the type of pervasion. The Nyāya believes in some omnipresent properties such as nameability and knowability. When one observes only the positive pervasion of these two, and infers, ‘this is nameable, since this is knowable’, one’s inference is purely positive. No case of negative pervasion exists in this case, since nothing has ¬nameability or ¬knowability. In contrast, one can only observe that anything that does not have abstract thinking is not rational before inferring, ‘a human has abstract thinking, since (s)he is rational’. Assume that only humans (i.e. the site) have rationality and abstract thinking. The confirmation of pervasion must take place outside the site although nothing other than the site has these two properties. Therefore – in this case – no positive observation (such as ‘every locus of rationality is a locus of abstract thinking’) is possible (see section 2.2.13). The third type is the mixed inference such as the smoke-fire case. One may observe that every locus of fire is a locus of smoke, and every locus of ¬fire is a locus of ¬smoke.

Gaṅgeśa thinks that one epistemically grasps pervasion through an extra-ordinary sensory connection called the connection of generality (sāmānya-lakṣaṇā pratyāsatti; for a detailed discussion, see Chakrabarti 2010: 149–168). After having known that every observed case of smoke has been a case of fire, and after having found no counterexample, one establishes an extraordinary connection with both smoke-ness and fire-ness. This is how, in a way, all smoke-individuals and all fire-individuals of the past, present, and future figure in one’s cognition of pervasion. Does one ‘perceive’ all these individuals? No, because through this connection one does not grasp the special features of each and every individual. One just knows that every smoke-individual is such that it must be accompanied by a fire-individual. Thus, one ascertains that every case of an entity having smoke-ness is a case of an entity having fire-ness.

Every inference having the form ‘S has T, since S has R’ makes three claims: (C1) the inference (anumiti): ‘S has T’; (C2) the pervasion (vyāpti): ‘every locus of R is a locus of T’ or ‘every locus of ¬T is a locus of ¬R’; (C3) the reason’s presence in the site (pakṣa-dharmatā): ‘S has R’. The defect of a reason or inferential fallacy (hetvābhāsa) is a fact that contradicts at least one of these claims. For Gaṅgeśa, the inferential defect of the inference I is the object of a piece of valid cognition that blocks the production of I (for the original Sanskrit definition, see Phillips 2020b: 868; Gaṅgeśa offers three definitions). For the Naiyāyika, the cognition about x and the cognition about the absence of x are mutual blockers (pratibandhaka). If one has the certain cognition, ‘this hill has no fire’, it blocks the production of the cognition, ‘this hill has fire’. As long as the former exists, the latter cannot come into being. Therefore, if somebody already cognizes the former, they cannot infer, ‘this hill has fire’.

Gaṅgeśa classifies fallacies into five types: (NF1) deviation (vyabhicāra), (NF2) contradiction (virodha), (NF3) un-establishment (asiddhi), (NF4) defeat (bādha), and (NF5) counterbalancing (satpratipakṣa; most of the names of the Neo Nyāya fallacies are borrowed from Phillips 2020b). Let us consider a few fallacies defined by Gaṅgeśa.

• (NF1) Deviation opposes the pervasion-claim, i.e. (C2). It is not very different from the deviation of the Old Nyāya. It is conceptualized as being in opposition to pervasion, not as the absence of a prescribed feature. The ‘common’ (sādhāraṇa) deviation is the reason’s coexistence with the absence of the target in a locus.
• (NF2) The defect called ‘contradiction’ is essentially the fact that any case of the target is a case of the absence of the reason. This too opposes the pervasion-claim. Let us consider the old example of (OF2, see section 2.2.13): ‘this is a man, since this is a horse’. Although (NF2) and (OF2) are the same concept, we may notice the difference in their formulations. Any case of being a man is a case of not being a horse, i.e. a case of the absence of horse-hood (the reason). Therefore, horse-hood is not a genuine reason for proving manhood; it is a contradictory reason.
• (NF3) Un-establishment opposes the presence-in-site claim, i.e. (C3). (NF3a) The un-established location (āśrayāsiddhi) involves an un-established or nonexistent site such as a golden mountain. As a consequence, the reason does not find a proper location; hence its presence-in-site remains un-established. Consider the example: ‘this golden mountain has fire, since it has smoke’. The reason (smoke) cannot be located in the site (the golden mountain), since there is no golden mountain. (NF3b) The un-established nature (svarūpāsiddhi) is the reason’s absence in the site. Suppose somebody sees a cloud of mist on a hill, and infers: ‘this hill has fire, since it has smoke’. Since no smoke-individual has been established in the hill through this inference, its reason suffers from the defect called svarūpāsiddhi.
• (NF4) The Neo Nyāya concept of the defeat is the untimed reason in the terms of the Old Nyāya. For the Neo Nyāya, the defeat of the inference ‘S has T’ is the absence of T in S. It clearly opposes the main inferential claim (C1).
• (NF5) Counterbalancing is largely the same as the counterbalanced reason of the Old Nyāya. Gaṅgeśa defines counterbalancing as the defect that involves a reason, whose function has been blocked by the presence of another equally strong reason that establishes the absence of the target (Phillips 2020b: 907), as in the Coccinistercus case (section 2.2.13, OF4). Gaṅgeśa’s general definition of a fallacy does not seem to cover counterbalancing; for the cognition of the rival reason of a counterbalanced reason need not be valid. In the Coccinistercus example, if the blackberry-inference is valid then the strawberry-inference must be false; for either the strawberry-pervasion is faulty (i.e. deviation) or the boy does not consume strawberries everyday (i.e. un-establishment). In this case, the fallacious strawberry-inference does not suffer from counterbalancing. If the blackberry-inference is invalid, then the strawberry-inference is not fallacious – because, by definition, a fallacy is the content of a valid cognition.

We shall briefly discuss an important inference the Naiyāyika makes: ‘[E]arth and the likes have an agential cause, since they are effects, like a pot’ (Phillips 2020b: 981). Here the assumption is: such things as Earth are not eternal; they have a beginning. The homologues for this inference are objects such as pots. The heterologues are the eternal things such as space and time. It is observed that a thing like a pot is made by an intelligent agent, who has knowledge of the relevant materials (such as clay) and has the power and intention to produce the pot. From that, one may infer that things such as the Earth, which are effects too, have an intelligent maker. (This is a simplified version of Gaṅgeśa’s argument; for a detailed discussion, see Vattanky 1984.)

3.3.2.3 (NE3) Analogy (upamāna)

For Gaṅgeśa, analogy qua epistemic instrument brings about the knowledge of the basis for semantic usage (pravṛtti-nimitta) of a word, especially a noun, in a certain way. First of all, let us understand the phrase ‘the basis for semantic usage’ (for a detailed discussion, see Ganeri 2011a: 129–158). When one knows the meaning of the word ‘cat’, one would use it upon seeing a cat. Any factor that triggers the usage of the word W is the basis for semantic usage (or semantic basis) of W. In other words, W must be applied to each of its semantic bases. When John asks Xico, ‘What is a jaguar?’, John basically asks for the semantic basis of the word ‘jaguar’. John analogically learns the meaning of ‘jaguar’ when Xico’s answer is based on similarity (sādṛśya) – something like, ‘A jaguar is that which is similar to a leopard’.

But what is similarity? According to Gaṅgeśa, x’s similarity to y is x’s having a significant number of properties P1, …, Pn that are possessed by y (tad-gata-bhūyo-dharma-vattva) too, and none of these is unique to x (asādhāraṇānya; Phillips 2020b: 1180). P1, …, Pn are ‘common properties’ (niyata-dharma). Finally, Gaṅgeśa thinks that analogical cognition is the cognition of the semantic basis of a word, when the factor that servers as the semantic basis coexists with the common properties (Phillips 2020b: 1211). Let us consider the sentence, ‘a jaguar is that which is similar to a leopard’. ‘A jaguar is similar to a leopard’ means ‘a jaguar and a leopard share a significant number of properties P1, …, Pn’. When John sees an animal x that has P1, …, Pn, he experientially grasps the semantic basis of the word ‘jaguar’. Notice that the semantic basis of the word ‘jaguar’ is coexistent with P1, …, Pn (niyata-dharma-samānādhikaraṇa) in x.

3.3.2.4 (NE4) Testimony (śabda)

For Gaṅgeśa, testimony (śabda) is defined as the words that are produced by proper knowledge (tattva-jñāna) of objects (artha), and the knowledge must be conducive to producing linguistic utterance (prayoga; for an excellent discussion, see Mukhopadhyay 1992: 26–73; see also Phillips 2020b: 1213). John sees a cat in a vat, and tells Ram, ‘There is a cat in that vat’. These words are produced by proper visual knowledge of a cat in a vat. Also the knowledge is capable of analysing its object-complex into a cat, contact, a vat, etc., and represent those in an intelligible linguistic expression. Thus, John’s words are the epistemic instrument called testimony. A linguistic phrase P (vākya) preserves the truth of its source S by representing S’s elements correctly.

John, who has seen Hari cook rice, tells Ram, ‘Hari cooks rice’ (hariḥ annam pacati = hari + su [nominative, singular] anna [rice] + am [accusative, singular] pac [cook] + ti [third person, present, singular]). What this sentence conveys is more than the mere concatenation of its morpheme-meanings (padārtha). Let the meaning of the morpheme m be [|m|]. The meaning of John’s sentence (vākyārtha) consists of the relations between [|Hari|], [|cook|], and [|rice|] (for the sake of simplicity, the inflectional endings are dropped). Let the meaning be [|Hari|] ↔ [|cook|] ↔ [|rice|]. Not only does the phrase ‘cooks rice’ present cooking and rice, it also presents rice that is being cooked. This semantic connection ‘↔’ or saṃsarga is the essence of the phrase-meaning. How are the semantic connections known? On the Nyāya view, Ram knows the semantic connections John’s sentence has because it has three features, verbal expectation (ākāṅkṣā), contiguity (āsatti), and semantic fit (yogyatā).

Gaṅgeśa defines verbal expectation as an incompleteness of reference (abhidhāna-aparyavasāna). In the phrase ‘cooks rice’, ‘cooks’ raises the expectation ‘cooks what?’ and this is fulfilled by ‘rice’. In other words, without ‘rice’, ‘cooks’ will not be able to make complete semantic sense. This applies equally to ‘rice’. Consider the ungrammatical phrase, ‘cooks rice you’*. Here, the standalone morpheme ‘you’ does not refer successfully, since it cannot connect itself with any other morpheme. Additionally, ‘cooks’ and ‘rice’ (in ‘cooks rice’) share a grammatical relation (anvaya); without one, the other is not able to refer to its own meaning as a semantic element that is connected with another (na svārtha-anvaya-anubhāvakatvam; Phillips 2020b: 1259). An unrelated meaning blocks the production of phrase-meaning, and makes the phrase ungrammatical.

The string ‘Hari cook s (third person, singular, present) rice’* is not grammatically well formed since it lacks contiguity. For Gaṅgeśa, contiguity glues all the discrete morphemes together, and causes morphophonemic conversions (e.g. ‘cook + s’ becomes ‘cooks’; Phillips 2020c: 1281–1282). Some Nyāya texts including TSD and NSM state that contiguity is the property of being uttered/presented in succession (avilamba; see Jhalakīkar 1928: 134–135).

According to Gaṅgeśa, the semantic fit of the phrase ‘X Y’ is the absence of the knowledge of counterevidence (bādhaka-pramā-viraha) that could prevent [|X|] from being semantically connected with [|Y|] (Phillips 2020c: 1271). Here is an example: ‘Today I saw a round square’. It is blocked by the knowledge that ‘there is no round square’, which prevents roundness to be predicated of a square. Thus, this string of words lacks semantic fit and fails to generate any semantic connection.

The Vaiśeṣika school, which aims to reduce testimony to inference, may now say that: when John tells Ram ‘X Y’, Ram infers [|X|] is related with [|Y|] through the relation R, since [|X|] is denoted by the word ‘X’ that is related with ‘Y’ through some syntactic relation Rs. On this view, Rs is part of the reason of the inference. If this inference is correct, Ram gets to know from John that [|X|]R[|Y|]. In Nyāya-mañjarī (NM, Ch III), Jayanta wrote:

A phrase, whose relational content is not known before [hearing it], is capable of generating its meaning. This is evident from the fact that one who knows the meanings of the words of a newly composed verse understands its sentential meaning. On the contrary, the relation that constitutes the base for an inference, has to be known before making the inference. Such being the case, how is testimony the same as inference? (NM 140)

Upon hearing ‘X Y’, one may at most infer [|X|] and [|Y|]. Where does the relational content R come from? Gaṅgeśa adds: ‘Before hearing “X Y”, one did not know the [syntactic] relation between X and Y’ (Phillips 2020c: 1215–1217). His point is: before inferring fire from smoke, one needs to experience both in several loci. But Rs is known only upon hearing the phrase, ‘X Y’. That being the case, Rs cannot serve as a part of the reason for inferring R (see Chakrabarti 1992 for a detailed discussion on the irreducibility of testimony).

The meaning of the sentence, ‘This white horse, that is running here, is not Bucephalus’ consists of a substance, in which the white colour and the activity of running reside through the relation of inherence, and it has the absence of the property of being identical to Bucephalus. Here the word-meaning relations (vṛtti) are straightforward. But philosophers have been debating over the reference of the common noun ‘horse’. According to Gaṅgeśa, it is a horse-individual qualified by horse-hood (Phillips 2020c: 1564–1565). The meaning of the white-horse sentence consists of all primary meanings (śakya) presented through the primary semantic power (śakti) of the words (see Ganeri 1999 for semantic powers of words). In contrast, a phrase such as ‘the cowherd village on the Ganges’ involves a secondary meaning (lakṣya) since ‘the Ganges’ here means ‘the bank of the Ganges’.