Dignaga On Trairupya Reconsidered

Author(s): Shoryu Katsura
Publisher: Shoryu Katsura
Catalog link: https://jainqq.org/explore/269439/1

Page #1

Dignāga on trairūpya Reconsidered:
A Reply to Prof. Oetke
Shoryu KATSURA
0. Introduction
It is most likely that Dignāga was the first Indian logician to employ two new terms, viz. 'inference for oneself' (svārthānumāna) and 'inference for others' (parārthānumāna). The former stands for inference (anumāna) as a means of valid knowledge (pramāņa) in the Indian epistemological tradition, while the latter corresponds to proof (sādhana) in the Indian debate (vāda) tradition. As a matter of fact, in his earlier work on logic, the Nyāyamukha (=NMukh), Dignāga still uses the terms 'anumāna' (HI) and 'sādhana' ( *), respectively. By renaming the proof and identifying it as a type of inference in the Pramāṇasamuccaya (=PS), he succeeded in integrating the two different traditions of Indian logic into his new system of epistemological logic.
Dignāga defines svārthānumāna as “knowledge of an object / a state of affairs (artha) through a valid inferential mark (linga) which possesses three characteristics (trirūpa)” (PS II. 1ab: anumānam dvidhā; svārtham trirūpāl lingato 'rthadrk ). He defines parārthanumāna as “revealing / communicating an object / a state of affairs which has been known for oneself” (PS III. 1ab: parārtham anumānam tu svadrsțārthaprakāśanam /). It is clear that svārthānumāna represents a mental process of inferring, which is a private experience of a cognizer, while parārthānu māna is a public act of communicating what he has experienced.”
Now what are the three characterisitics of a valid inferential mark? There have been many attempts to explain it systematically. In 1994 Claus

Page #2

242
Oetke published Studies on the Doctrine of Trairupya (Wiener Studien zur Tibetologie und Buddhismuskunde Heft 33, Wien) in which he criticized some of the previous studies, including my own Dignaga on Trairûpya". Journal of Indian and Buddhist Studies, Vol. XXXII, No. 1, 1983), and proposed his new systematic interpretations. I am yet to digest all the implications of Oetke's rather difficult interpretations and far from giving the final opinion on them. However, I feel obliged to answer his criticisms and present my own understanding of the trairupya again.
It is to be noted at the outset that my article mentioned above was written in order to show some possible improvements upon previous translations of PS-Vrtti II. Sed, namely, the Japanese version by Hidenori Kitagawa (Indo Koten Ronrigaku no Kenkyu, Tokyo, 1965, pp. 96-100) and the English rendering by Richard Hayes ("Dinnaga's Views on Reasoning (svarthänumāna)," Journal of Indian Philosophy 8, 1980, pp. 252-253). The main point of my criticism against Kitagawa's translation is that we should read the restrictive particle eva into both the second and the third characteristics of the trairupya. I also wanted to correct the imprecise renderings of the Tibetan phrase nges par gzung ba; namely, ''(what is to be determined) by Kitagawa and 'requirement' by Hayes. The most likely candidate for its Sanskrit original is avadharaṇa which means 'restriction' and which is the function assigned to the particle eva in Indian Grammatical tradition. When Hayes published Dignaga on the Interpretation of Signs (Studies of Classical India, Vol. 9, Dordecht/ Boston / London 1987), he seems to have adopted some of my suggestions.
Apart from my 1983 paper, I have discussed the trairupya on at least two other occassions, namely, "On Trairüpya Formulae," (Buddhism and Its Relation to Other Religions, Kyoto 1985), and "Ronri gakuha" (The school of Buddhist Logic), (Iwanami Kōza Tōyōshisō Vol. 8, Tokyo 1988). In the latter paper I fully discussed the development of the trairůpya formulae in some basic texts of Indian or Buddhist logic, but the paper seems to have evaded the attention of Prof. Oetke.

Page #3

Dignāga on trairūpya Reconsidered:
A Reply to Prof. Oetke
243
The controversy between Oetke and me lies in the question whether Dignāga intended to insert the particle eva in the formula of the second characteristic (or 'condition' in Oetke's terminology) or not, the solution of which depends on how we should read PS II. 5cd and its Svavrtti. Therefore, I would like to give the two Tibetan versions of the above texts as well as a new literal translation of my own.
1. PS II. 5cd & Svavrtti 1.1. Kanakavarman's version (Kitagawa pp. 455b-456b) (Peking 111a6-111b1) tshul gsum rtags las zhes gang bstan pa de bshad par bya ste /
rjes d pag bya dang de mtshungs la
yod dang med la med pa'o // (PS II 5cd) rjes su dpag pa ni cho khyad par can gyi chos can yin te / (a7) de la dus phyis chos kyi spyi'i tshul gyis mngon sum nam rjes su dpag pas mthong ba'o // de'i rigs la yang mtha' dag gam phyogs cig yod pa nyid do /
de gang las zhe na / de dang mtshungs pa kho na la yod (a8) pa zhes nges par gzung ba'i phyir yin gyi yod pa kho na zhes ni ma yin no /
de ltar na med pa lam (Read la) med do zhes brjod pas mi bya'o zhe na / 'di ni med pa nyid la med pa yin gyi / gzhan pa la'ang ma yin 'gal ba la (111 b1) ma yin no zhes nges pa'i don du 'gyur ro) 'dir tshul gsum pa'i rtags gang las rtags can rtogs pa yin no
1.2. Vasudhararakṣita's version (Kitagawa pp. 455a-456a) (Derge 29a4-7=Peking 30a1-5) tshul gsum pa'i rtags zhes gang brjod pa de brjod par bya ste /
rjes {P: su} dpag bya dang de mthun la //
yod dang med nyid la med pa'o // (PS II. 5cd) rjes su dpag par bya ba ni chos kyi {P: kyis) khyad par du byas pa'i chos can no //

Page #4

244
{30a2) de la mngon sum mam rjes su dpag pas mthong gi {P: gis} rjes la de'i rigs dang mthun (D29a5) pa la yang spyi'i tshul gyis phyogs thams cad dam phyogs gcig la yod par grub pa'o |
ci'i phyir zhe na de dang mthun pa kho na la yod {P30a3} ces nges par gzung ba'i phyir ro / de dang mthun pa la yod pa kho na'o zhes ni brjod par mi bya ba'i phyir ro || med pa nyid las med pa zhes pas ni slar yang (D29a6) de nyid nges par bya ba'i don du ste / med pa nyid la med pa nyid yin gyi (P30a4} gzhan la yang ma yin la 'gal ba la yang ma yin no zhes pa’i don no {P: to} // bsgrub par bya ba'i rigs dang mthun pa la yod pa dang med pa la med pa nyid do zhes bya ba de dag gis ni gang tshul gsum pa'i rtags las rtags can (D29a7= P30a5) rtogs so zhes bstan pa yin no
As Richard Hayes has done in his 1980 paper, I shall translate Kanakavarman's version except for the underlined portion where I followed Vasudhararakṣita's because the latter version definitely gives the better reading. The Sanksrit fragments will be supplied in parentheses wherever they are available.
1.3. Translation: “The phrase 'trirūpāl lingatah' [in PS II. 1] should be explained.
PS II. 5cd: [The three characteristics of a valid inferential mark (linga) consist in its] presence in the object to be inferred (anumeya) as well as in what is similar to it (tattulya), and absence in the absence (of the property to be inferred (anumeya) and / or of what is similar to the object to
be inferred (tattulya)). (anumeye 'tha tattulye sadbhāvo năstităsati)" The object to be inferred is a property-possessor / locus (dharmin) qualified by a propety (which one wishes to know). (dharmaviśisto dharmy anumeyaḥ)"
After having known, either by perception or by inference, [the presence of an inferential mark] there (i.e., in the object to be inferred=the object of

Page #5

Dignāga on trairūpya Reconsidered: A Reply to Prof. Oetke
245
inference), we also establish in a general manner [its) presence either in all or some members of the class similar to it (i.e., the object of inference).
[Question:) Why is it so (i.e., a valid inferential mark is present in all or in some members of the class, not in all members of the class)?
[Answer:) Because we restrict (nges par gzung baravadhāraņa) [the second characteristic] in the way that it is present only in what is similar to it (i.e., the object of inference) (de dang mtshungs pa kho na la yod pa=tattulya eva sadbhāvah), not in the way that it is necessarily present in what is similar to the object of inference (yod pa kho na=Page #6

246
Cf. PS-Vṛtti [Kanakavarman] 150b8: phyogs kyi chos nyid (=*pakṣadharmatva) bstan pa'i don du gtan tshigs brjod pa /
NMukh (Tasiho p. 2c):
5 f
Cf. PS-Vrtti [Kanakavarman] 148b4: de Ita na gtan tshigs mthun pa'i phyogs nyid la yod pa (*sapaksa eva sattvam) dang bsgrub bya med pa la med pa nyid (*sādhyābhāve 'sattvam eva) bstan par rus pa yin ../
It is to be noted that there existed several different versions of the trairupya formulae before Dignaga, as I showed in my 1985 paper. The most. standard version seems to be (1) pakṣadharmatva, (2) sapakṣe sattvam, and (3) vipakṣe 'sattvam. The above passages, however, clearly indicate that Dignāga intended to insert the restrictive particle eva in the formulae of the second and the third characteristics.
Oetke's Objections
2.1. Objection 1: the meaning of the phrase nges par gzung ba Oetke (p. 58) says:
First of all, it is not even certain that the expression nes par gzun ba has the meaning of "must restrict", as assumed by Katsura. It is equally well possible that it represents a Sanskrit expression that means "(must be) recognized with certainty" or "(must be) ascertained" so that the whole sentence could possess the following literal sense: "It is [so] because it must be ascertained / one must ascertain that it (the logical reason) exists only in what is similar to it, but it is not the case that it exists only (i.e. exists without exception) [in this realm]."
It is very easy to show that the expression nges par gzung ba is Tibetan equivalent of the Sanskrit avadharana (or niyama) which has the meaning of 'restriction'. In commenting upon the Nyayabindu II. 6: sapaksa eva sattvam Dharmottara states the reason why the second characteristic of a valid

Page #7

Dignāga on trairūpya Reconsidered: A Reply to Prof. Oetke
247
inferential mark (!) is restricted (avadhāraņa) by the particle eva in that particular manner.
sattvagrahaņāt pūrvāvadhāraṇavacanena sapakşāvyāpisattākasyāpi prayatnānantariyakasya hetutvam kathitam / paścādavadhāraṇe tv ayam arthaḥ syāt / sapakşe sattvam eva yasya sa hetur iti prayatnānantariyakatvam na hetuḥ syāt / (Bibliotheca Buddhica VII, p. 19)
It is put into the following Tibetan sentences:
yod pa smos pa las sngar nges par gzung ba smos pa ni yod pa yin yang mthun pa'i phyogs la ma khyab pa rtsol ba las byung ba dag kyang gtan tshigs nyid du bstan pa yin no / phyis nges pas (sic!) gzung ba byas na ni don 'dir 'gyur te / gang zhig mthun pa'i phyogs la yod pa nyid de gtan tshigs yin no zhes bya bar 'gyur bas / rtsol ba las byung ba nyid gtan tshigs ma yin par 'gyur ro (Bibliotheca Buddhica VIII, p. 43)
Thus, it is quite certain that the expression nges par gzung ba means 'restriction.' Consequently we cannot deny the fact that Dignāga intended to read the restrictive particle eva in the second characteristic of a valid inferential mark. And he was well aware of its implication, namely, the redundancy of the third characteristic. Incidentally, Oetke's rendering of the same expression '(must be) recognized with certainty' or '(must be) ascertained' seems to correspond to Sanskrit niścita which was added by Dharmakirti to each characteristic of a valid inferential mark (See e.g. Nyāyabindu II. 7) but which is missing from Dignāga's formulae.
Oetke (p. 58) further says: "But it is not even necesary to insist that the expression nes par gzun ba has another sense than the one asumed by Kat
ura.” He argues that the object of the restriction could be “the logical reason" or "the canon of Trairūpya," instead of “the second characteristic" as I assume. However, as the above-quoted passage of the Nyāyabindu and many other similar texts support, the context of the argument in PS-Vștti II. 5cd forces us to regard “the second characteristic” as the object of the restric

Page #8

248
tion.
2. 2. Objection 2: the purpose of insertion of the particle eva
Oetke (p. 59) says:
A natural assumption would be that the second and the third condition together guarantee that the logical reason occurs in at least one similar instance apart from the pakșa and only there, i.e. never in the vipakșadomain.
Furthermore he considers that the formulae of the trairūpya presented in PS 5cd: anumeye 'tha tattulye sadbhāvo nāstitā 'sati, namely, those without the particle eva, properly represent the above assumption. According to him it is awkward that Dignāga inserts the particle eva in the second condition, which leads to the redundancy of the third condition--the difficulty to be solved by Dignāga himself. Then Oetke goes on to investigate the possible purposes of Dignāga's insertion of the particle eva. In this connection I would like to give my own understanding of the trairūpya foșmulae with and without eva.
2.2.1. The trairūpya formulae without eva
PS II. 5cd can be re-formulated into the following three formulae:
(1) anumeye [lingasya] sadbhāvaḥ / (2) tattulye [lingasya] sadbhāvaḥ / (3) asati (lingasya] nāstitā /
The first formula is traditionally called pakşadharmatva” and the second and the third formula are later called anvaya and vyatireka respectively.) Although Dignāga does not apply the names anvaya and vyatireka in this context, he uses them in order to explain how a word (śabda) expresses its meaning in PS-Vștti Chapter V.

Page #9

Dignāga on trairüpya Reconsidered: A Reply to Prof. Oetke
249
PS-Vịtti ad PS V. 34: śabdasyānvayavyatirekāv arthābhidhāne dvāram, tau ca tulyātulyayor vrttyavrtti / “Association / continued presence (anvaya) and dissociation / continued absence (vyatireka) are the two ways that a word expresses its object / meaning. They consist respectively in applying to what is similar (tulye vșttih) and in not applying to what is dissimilar (atulye 'urttiḥ)."9)
For instance, the word 'cow' is applied to those animals which resemble each other in sharing some common features, viz. a dewlap and others, but it is not applied to those which do not share such features. In the above passage anvaya and vyatireka are formulated in the following manner:
(4) anvaya=tulye (sabdasya] vịttiḥ / (5) vyatireka=atulye [śabdasya] avrttiḥ /
If we compare them with the above formulae (2) (3), we cannot but see a close resemblance in their formulations. Thus it is safe to assume that the second and the third characteristics of a valid inferential mark can be named anvaya and vyatireka respectively. As a matter of fact it is not too difficult to see the reason behind such a correspondence, if we take into consideration Dignāga's fundamental doctrine that a verbal cognition (śābda, discussed in PS V) is only a kind of inference (anumăna discussed in PS II) and that a word (sabda) functions exactly in the same way as an inferential mark (linga), namely, by means of 'exclusion of others' (anyāpoha / anyavyāvștti / anyavyavccheda), in the realm of universals.
Now what are the roles of anvaya and vyatireka? Here I would like to refer to Geroge Cardona's thoery expressed in his article “On Reasoning from Anvaya and Vyatireka in Early Advaita.”10)
Indian thinkers have used a mode of reasoning that involves the related presence (anvaya 'continued presence ') and absence (vyatireka) of entities as follows: (1) a. When X occurs, Y occurs.

Page #10

250
b. When X is absent, Y is absent. (2) a. When X occurs, Y is absent.
b. When X is absent, Y occurs. If (1a, b) hold in all instances for X and Y, so that these are shown consistently to occur together, one is entitled to say that a particular relation obtains between the two. Either (1a) or (1b) alone will not justify this, and a claim made on the basis of either can be falsified by showing that (2a) or (2b) holds. One relation that can be established by (1) is that X is a cause of Y. A special instance of the cause-effect relation involves the use of given speech units and the understanding by a hearer of given meanings. If (1a, b) hold, the speech unit in question is considered the cause of one's comprehending a meaning, which is attributed to that speech element.
Cardona named anvaya and vyatireka the 'Indian Principle of Inductive Reasoning' and considered them to be the means for discovering a certain relationship between two items, such as a causal relation between a fire and smoke and a relation between a word and its meaning.
In the case of the trairūpya theory, anvaya and vyatireka establish the relation between sādhya and sadhana / lingin and linga / gamya and gamaka; in short, anvaya and uyatireka together are a kind of the inductive method by which we can determine what is a valid inferential mark. In other words, the second and the third characteristics without the particle eva express the necessary conditions for a valid inferential mark.
2. 2. 2. The trairūpya formulae with eva and the meaning of the eva
restriction In PS-Vștti II. 5cd, as I have shown above, Dignāga inserts the restrictive particle eva at least in the second characteristic of a valid inferential mark in the following manner:
(6) tattulye eva [lingasya] sadbhāvaḥ /

Page #11

Dignāga on trairúpya Reconsidered: A Reply to Prof. Oetke
251
In this connection I would like to explain what is the meaning of the 'evarestriction'. Here is an often-quoted paribhāṣā (No. 131) attributed to a Grammarian Vyādi:
yata eva-käras tato'nyatrāvadhāraṇam (Paribhāṣāsamgraha ed. by Abhyankar, Poona, 1967, p. 43) "When the restrictive particle 'eva' is attached to the word X, the other word Y is restricted.”
In a sentence {X eva Y} or “vrkșa eva simśapā” (simšapā is a tree), the domain of Y is restricted by the domain of X and the domain of simšapās is restricted by the domain of trees. The terms 'avadhārana' or 'niyama' express such a restrictive function of the particle eva. X and Y are called 'niyāmaka' and 'niyamya' respectively.") If we graphically represent the relation between niyamya (Y) and niyāmaka (X), it can be presented in the following two circles:
X=niyāmaka
Y=niyamya
The same relationship can be called 'vyāpti' (pervasion); namely, the domain of Y is pervaded by that of X when Y is vyāpya (the pervaded) and X is vyāpaka (the pervader). As a matter of fact, the Nyāyakośa, compiled by Jhalakikar and revised by Abhyankar (Poona, 1978), gives vyāpti as the first meaning of niyama' and Nārāyaṇa (ca. 1560-1656) presents 'niyama' as one of the synonyms of 'vyāpti' in his Mānameyodaya (ed. and tr. by Kunhan Raja and Suryanarayana Sastri, Madras, 1933, p. 50), other synonyms being 'pratibandha', 'avyabhicāra', and 'avinābhava'.
Dignāga refers to the notion of restriction again in PS II. 21:

Page #12

252
linge lingi bhavaty eva linginy evetarat punaḥ /
niyamasya viparyase 'sambandho lingalinginoh
"The marked (lingin) necessarily (eva) occurs where the inferential mark (linga) occurs, and the latter (i.e. the mark) occurs only (eva) where the marked occurs. If the restriction (niyama) [by the particle eva] is reversed, there is no [proper] relation (sambandha) between the mark and the marked."
In the above verse the term lingin is followed by the particle eva in relation to the term linga; therefore, the domain of the marked is restricted (niyata) by that of the mark. Such a relation is regarded by Dignāga as the proper relation (sambandha) between the mark and the marked. Although he does not mention the term vyapti as an equivalent of sambandha in this context (he refers to it in PS-Vṛtti IV. 4), I understand that PS II. 21 is the first attempt to define the relation of 'pervasion' in the history of Indian It may be supported by the fact that Dharmakirti phrases his definition of pervasion, following the model presented in PS II. 21.
vyāptir vyāpakasya tatra bhava eva vyapyasya va tatraiva bhavaḥ | (Pramāṇavārttika-Svavṛtti, ed. by R. Gnoli, Serie Orientale Roma 23, Rome 1960, p. 2)
"Pervasion is the pervader (vyāpaka)'s necessary (eva) presence in the domain [where the pervaded (vyapya) occurs] or the pervaded's presence only (eva) in the domain [where the pervader occurs]."
As I demonstrated in detail in my article "Indo Ronrigaku niokeru Henjugainen no Seisei to Hatten" (The Origin and Development of the Concept of vyapti in Indian Logic, Hiroshima Daigaku Bungakubu Kiyo, Vol. 46, Special Issue No. 1, 1986), it was Dignaga who first introduced the concept of vyāpti as the foundation of a valid inference. He was well acquainted with the Indian Grammartical tradition and applied their concept of niyama as the forerunning model of the logical concept vyapti.

Page #13

Dignāga on trairūpya Reconsidered: A Reply to Prof. Oetke
253
Thus I would like to conclude that Dignāga's purpose of inserting the particle eva in the second characterisitic is to restrict the domain of an inferential mark by that of what is similar to the object of inference (tattulya), which, I believe, gave the way to the formation of the theory of pervasion in Dignāga's logic. In any case, the second characteristic with the evarestriction expresses not the necessary but the sufficient condition for a valid inferential mark; hence, Dignāga felt it unavoidable to explain the necessity of the formulation of the third characteristic. I hope that the foregoing argument makes it clear why I referred to PS II. 21 in my 1983 paper, which has been questioned by Oetke (p. 70).
2.3. Objection 3: translation of the word 'asati' in the third char
acteristic Oetke (p. 65) says:
Despite the fact that the interpretation of năstităsati in the sense of "absence in what is not similar to it" is linguistically a very natural, perhaps even the most natural one in the given context it should be given up because of the undesirable objective consequences it involves. Thus the expression asati should not be taken as representing a participle of the copula 'to be', but as a negated present participle of the verb 'to be'
='to exist'.
Here Oetke is quite right in pointing out my misunderstanding of the expression asati. I simply followed Kitagawa's Japanese translation (p. 96): [ATIC] #3 DICITFLON EVL 53] (absence in what is not (tattulya)), though Hayes (1980: p. 252) correctly translated: "absent in their absence". In my new translation I accepted Oetke's criticism and adopted the meaning of 'in the absence' for asati.
Now the question is what is the logical subject of 'absence' in the third characteristic. Hayes seems to take it to be both anumeya and tattulya mentioned in the first and the second characteristics, which is perhaps the most

Page #14

254
natural reading, if we expect vipaksa in the third characteristic and if we can identify anumeya and tattulya with paksa and sapakșa, for the universe of discourse in Dignāga's logic is divided into three parts, namely, pakșa, sapakșa and vipaksa.
Oetke, on the other hand, seems to opt for anumeya as the logical subject of the phrase asati and says (p. 65):
Perhaps ... the author of the PS invited the reader to take the word anumeya in two different senses for the interpretation of the verse, first in the sense of 'substratum of the property to be proven' at the beginning of the kārikā and later in the sense of 'property to be proven' for the supplementation of the logical subject of the existence-predicate represented by asati.
Now I would like to support the above interpretation of Oetke as one of the possible meanings of the third characteristic. But his second alternative, to take anumeya (in the first characteristic) in the sense of 'property to be proven', is contradicted by Dignāga's own comment, viz. dharmaviśisto dharmy anumeyaḥ. First I would like to show that Dignāga possibly assigned not two but three meanings to the word anumeya.
2. 3. 1. The meaning of the term anumeya
When we read Dignāga's texts on logic, we must be careful to distinguish different meanings attributed to the same technical term and to identify different terms which are meant to express the same or similar logical concepts. For example, in NMukh (Taisho p. 1b) Dignāga attributes three distinct meanings to the word paksa ), namely, (1) a proposition consisting of the topic / property-possessor dharmin ( ) and the property to be proven dharma (£) (e.g. "The mountain possesses a fire"), (2) the property-possessor as the topic of the proposition (e.g. the mountain), and (3) the property to be proven (e.g. a fire).

Page #15

Dignaga on trairupya Reconsidered: A Reply to Prof. Oetke
豈不穗以樂所成立合説為宗,云何此中乃言宗者唯取有法。 此無有失。以其總聲於別亦轉,如言燒衣。或有宗聲唯詮於法。 [Objection:] Haven't you in a general manner (in the first kärikā] called by the name 'paksa' the synthesized [proposition consisting of dharmin and dharma] which somebody wants to prove? Why in this [second kārika] do you apply [the same term] 'pakṣa' only to dharmin (i.e. the topic of the proposition)?
[Answer:] There is no such a defect because the word for an aggregate can be employed to express its constituent as well; e.g., the expression 'burned robe' [can be employed even when only a part of the robe is burned]. Furthermore, sometimes the word paksa can be employed only to express [the other constituent of the proposition,] dharma (i.e., the property to be proven).
255
14)
In the first kärkikä of NMukh Dignaga defines 'proposition' (paksa) and in the second kārikā he enumerates nine possible cases for a reason (hetu) which is defined as a property of pakṣa (pakṣadharma).1 In the second kärikā the word paksa is used in the sense of the topic of the proposition. Dignāga justifies this move by referring to the general principle in Sanskrit semantics that the word for an aggregate (samudaya-sabda) can be metaphorically employed to express its parts (avayava). Consequently, the word paksa can be employed in the three different meanings.
In PS III. 10 Dingaga says the exactly same thing about the word sadhya:
samudayārthasādhyatvad dharmamătre 'tha dharmini/
amukhye 'py ekadeśatvät sädhyatvam upacaryate
"Since [the word] sadhya has the meaning of an aggregate (samudāya) [i.e., a proposition consisting of dharmin and dharma], both a simple dharma as well as dharmin, though not the direct meaning [of the word sādhya], can be metaphorically called 'sādhya', because they are parts [of the proposition]."

Page #16

256
The word sādhya, which literally means that which is to be proven', too can be employed in three meanings, viz., (1) a proposition to be proven consisting of dharmin and dharma, (2) the property-possesor (sādhya-dharmin) as the topic of the proposition, and (3) the propety to be proven (sādhyadharma). Thus we can conclude that both pakṣa in NMukh and sādhya in PS III have the same range of meanings and, consequently, they refer to the same logical concept in Dignāga's system of logic.
I believe that the word anumeya, which literally means 'that which is to be inferred', also possesses the same set of three meanings, viz. (1) a state of affairs or a proposition to be inferred consisting of dharmin and dharma, (2) the property-possesor or locus to be inferred (anumeya-dharmin), and (3) the property to be inferred (anumeya-dharma), although Dignāga does not explicitly mention this.
The reason behind my conviction is that the word anumeya occupies the place of paksa in the first formula of the trairūpya. As I mentioned earlier, Dignāga refers to the first characteristic of a valid inferential mark by the traditional expression pakşadharmatva in NMukh. In PS II. 5cd, however, he refers to it by his new phrase anumeye sadbhāvaḥ, as we have seen above.16) Therefore, it seems quite possible to assume that the word anumeya is employed by Dignāga in the same way as the words paksa and sādhya in NMukh and PS III.
The word paksa seems to stem from the old Vada-tradition. In PS Dignāga seems to distinguish between anumeya and sādhya, by employing the former in svārthānumāna and the latter in parārthānumāna. A similar phenomenon can be found regarding the words referring to a logical reason; namely, hetu (reason) and sādhana (dharma) (a proving property), which seem to stem from the Vāda-tradition, are used in parārtānumāna, while linga (an inferential mark) and anumāna (a means of inference), which seem to stem from the Pramāņa-tradition, are used in svārthānumāna.
Thus Dignāga would support Oetke's first interpretation regarding the meanings of the word anumeya in the trairūpya; namely, it means the sub

Page #17

Dignaga on trairupya Reconsidered: A Reply to Prof. Oetke
stratum of the property to be proven in the first characteristic and just the property to be proven in the third characteristic. Now, in order to identify the logical subject of the third characteristic, we must have a clear idea about two more technical terms, namely, sapakṣa and vipakṣa, for I understand that they correspond to tattulya and asat in the second and the third characteristics of trairûpya.
257
2.3.2. sapaksa and vipakşa
Dignaga first defined sapaksa and vipaksa in NMukh (Taisho 1c-2a) as fol
lows:
此中,若品與立法隣近均等說名同品。以一切義皆名品故。
若所立無說名異品。非與同品相違或異。
"In this connection, if an object (paksa) () is close (samipa?) or similar to (samana) the property to be proven (sādhyadharma), it is called 'sapakṣa' (F), for any object (artha) () can be called 'pakṣa'. If [an object] lacks [the property] to be proven, it is called 'vipakṣa' ()The latter is neither incompatible (viruddha) (F) with nor different. (anya) () from sapakṣa...."
In PS III Dignaga gives somewhat different definitions of the two terms:
PS-Vṛtti [Kanakavarman] 130a: de la don thams cad phyogs su byas pas nye ba'i phyir bsgrub par bya ba'i chos kyi spyis mthun pa'i phyogs so || (Cf.. Nyayabindu II. 9-Nyayapravesa §2.2: sadhyadharmasāmānyena samano 'rthah sapakṣaḥ)
"In this connection, since any object (don=artha) can be called 'paksa', [an object is called] 'sapakṣa' (similar to pakṣa) in respect of the universal similarity (samanya) which is the property to be proven, for [it is] close to [pakṣa]."
PS [Kanakavarman] III. 19-20a: de las gahan dang de 'gal ba | gnyis po yang mi mthun phyogs ma yin | gtan tshigs med dang 'gal ba las || rnam

Page #18

258
par gcod par thal bar 'gyur | de’i phyir mthun phyogs med pa nyid // "That which is different from or incompatible with it (i.e. sapakşa); neither of them is (called] vipaksa. .. Therefore, (vi paksa is) nothing but the absence of sapakșa."
In NMukh Dingāga defines sapakșa as any object which is similar to sādhyadharma and in PS III as any object which is close / similar to (paksa] (the topic of a proposition) in respect of sādhyadharma-sāmānya—though he simply says "nye ba'i phyir", I assume from the context that we can supply the word pakșa.
I must confess that I cannot make good sense of the definition in NMukh, namely, “any object which is similar to the property to be proven.” Incidentally, the idea comes close to the Tibetan translation of the Nyāyabindu II. 9: mthun phyogs ni bsgrub par bya ba'i chos kyi spyi dang don mthun pa'o, which influenced the dGe lugs pa's understanding of sapakșa as demonstrated by Tom Tillemans.") If we adopt Tillemans's suggestion (p. 59), the NMukh definition of sapaksa may mean "any object which has the property to be proven", in which case, however, it would be difficult to distinguish between pakșa and sapakșa.
The PS III definition clearly distinguishes sapakșa from pakṣa and more explicitly indicates that sapaksa is “any object which has the property to be proven”. For example in the proposition: “The mountain has a fire,” sapakșa consists of any object which has the property to be proven, in other words, the domain of presence of the property to be proven, i.e., fiery places.
The PS III version becomes the standard definition of sapaksa after Dignāga.
In NMukh Dignāga defines vipaksa as "an object which lacks the property to be proven", in other words the domain of absence of the property to be proven. If we adopt the above-mentioned definition of sapakṣa in PS III, namely, the domain of presence of the property to be proven, then the two domains are complementary to each other and together will make up the

Page #19

Dignāga on trairūpya Reconsidered: A Reply to Prof. Oetke 259 complete universe of discourse (minus pakșa, i.e. the topic under discussion) or what Richard Hayes once called 'induction domain ’. 18)
In PS III Dignāga defines vipakșa simply as 'absence of sapakşa'. If we follow the above-mentioned definition of sapakṣa in PSIII, it should mean 'absence of any object which has the property to be proven', which in the final analysis can be understood as the domain of absence of the property to be proven as in NMukh.
In both NMukh and PS III Dignāga emphasizes that vipakșa is neither what is different from sapaksa nor what is incompatible with sapakṣa and that it is 'absence' (abhāva) of sapaksa. I understand that by restricting the meaning of vipakṣa in such a way, Dignāga tried to establish the complemenatry relationship between sapakșa and vipakșa as mentioned above."
Now let us go back to PS II. 5cd: anumeye 'tha tattulye sadbhāvo nāsitāsati. As the logical subject of asati we have two alternatives, viz. 'anumeya' and 'tattulya'. If we take the first alternative, the third characteritic means "absence (of the inferential mark] in the absence (asati) [of the property to be inferred (anumeya)]”, which corresponds to the NMukh definition of vipakșa (i.e.,.absence of the property to be proven). If we take the second alternative, then the phrase means "absence [of the inferential mark] in the absence (of what is similar to the object of inference (tattulya)], which corresponds to the PS III definition of vipaksa (absence of sapakşa).”
Just like Oetke I prefer the first alternative because it fits well with the alternative meanings of the word anumeya as I have explained above.20) Jinendrabuddhi, however, supports the second alternative.21) In any case, I believe that there is no fundamental difference in the interpretation of the trairūpya in either alterative.
The more I read Oetke's examinations of the trairūpya formulae, the more I am inclined to admire his thorough logical thinking. On the other hand, perhaps the problem with his approach is that he tends to rely on a few selected passages of the relevant Indian texts which he analyses minutely

Page #20

260
and from which he derives rather bold conclusions. My approach to the problem of trairūpya and other subjects is that I try to understand them in the broader context of the whole text under investigation. For example, Oetke (pp. 71-72) questions the relevancy of my reference to PS-Vștti IV. 4. To meet his objection and emphasize my wholistic approach, in a future paper, I shall show the connection between the theory of trairūpya and that of sādharmya- and vaidharmya-drstānta, in which I will deal with the role of 'Example' in Dignāga's logic.22)
3. trairūpya and Hetucakra
In order to strengthen my position that Dignāga intended to insert the restrictive particle eva in the the formula of the second characteristic, I would like to add the following argument. In PS-Vịtti II. 5cd Dignāga stated: “...we also establish in a general manner [its] presence either in all or some members of the class similar to it (i.e., the object of inference).” We may assume that he is here alluding to the second and the eighth hetu of the Hetucakra (Wheel of Reasons), where hetu is present in all or some members of sapakşa, respectively.
Hetu No. 2: (pakşa) anityaḥ śabdaḥ / [hetu] krtakatvāt / [drșțānta) yathā ghațaḥ, yathā ākāśam /
Hetu No. 8: (pakşa] anityaḥ śabdaḥ / [hetu] prayatnānantarīyakatvāt / [drșțānta) yathā ghațo vidyuc ca, yathā ākāśam /
Then comes the next question of how we justify the second and the eighth hetu as valid ones. Dignāga answers by means of the restriction of the

Page #21

Dignāga on trairūpya Reconsidered: A Reply to Prof. Oetke 261 second characteristic. If we restrict it in the way "tattulye (=sapakşe) lingasya sadbhāva eva", then the domain of the property to be proven or inferred (sādhya / anumeya-dharma) is restricted by the domain of linga. In other words, the former is pervaded by the latter. In that case, linga is deviant or inconclusive (vyabhicărin / anaikāntika) in respect to what is to be proven (sādhya) and it cannot be regarded as valid.
linga
sādhya
If we restrict the anvaya formula in the way "tattulya (=sapakşa) eva lingasya sadbhāvaḥ”, then the domain of linga is restricted by the domain of the property to be proven. In this case, linga is neither deviant nor inconclusive and it is a valid inferential mark.
sādhya
linga
In the case of Hetu No. 2 the domain of sādhyadharma (=anityatva) and the domain of hetu (=kặtakatva) are co-extensive but at least the domain of hetu does not deviate from the domain of sādhyadharma. Therefore, it can be regarded as a special case of hetu's being restricted by sādhyadharma.
In the case of Hetu No. 8 it is quite clear that the domain of hetu (=prayatnānantariyakatva) is restricted by the domain of sādhyadharma (=anityatva). Consequently, it is regarded as a valid hetu.

Page #22

262
4. Conlcusion
In PS II Dignaga presents two different formulae of the trairūpya, one without eva and the other with eva. PS II. 5cd can be divided into the following three formulae:
(1) anumeye [lingasya] sadbhavaḥ/
(2) tat (=anumeya)-tulye [lingasya] sadbhavaḥ/
(3) [anumeya / tattulya-] asati [lingasya] nästitä/
They correspond to the following traditional formulae:
(1) [hetoh] paksadharmatvam/
(2) sapakse [hetoḥ] sattvam/
(3) vipakṣe [hetoḥ] asattvam/
The second and the third formulae are respectively called anvaya and vyatireka, which constitute what Cardona called the 'Indian Principle of Inductive Reasoning'. If we understand tattulya / sapakṣa and asat / vipaksa in the above formulae in the sense of 'the domain of presence of the property to be proven' and 'the domain of absence of the property to be proven', then anvaya and vyatireka can be reformulated in the following manner:
(1) anvaya (continued presence): in the domain of presence of the property to be proven (sadhyadharma) linga is present; e.g., where there is a
fire there is smoke.
(2) vyatireka (continued absence): in the domain of absence of the property to be proven linga is absent; e.g., where there is no fire there is no smoke.
anvaya and vyatireka constitue the inductive method or process by which we can determine what is a valid inferential mark or reason, and sapakşa and vipaksa make up what Hayes called the 'induction domain' where we can conduct such an induction. In modern terminology, anvaya and vyatireka without eva express the necessary conditions for a valid reason.

Page #23

Dignaga on trairupya Reconsidered: A Reply to Prof. Oetke
In PS-Vṛtti II. Sed Dignaga inserts the restrictive particle eva at least in the anvaya formula: tat (=anumeya)-tulye eva [lingasya] sadbhavaḥ, which can be reformulated as: sädhyadharma-bhave eva lingasya sadbhavaḥ. As a result of the insertion of eva the anvaya formula no more represents the inductive process; it expresses the relation between the domain of linga and that of the property to be proven (sadhyadharma), namely, the former is restricted by the latter; in other words, the former is pervaded (vyāpta) by the latter. As far as I can see, Dignaga insisted on inserting the particle eva in order to express the relation of pervasion (vyapti) between linga and sadhyadharma which he considered to be the foundation of any valid inference or reasoning.
The fact that he does not employ the word vyāpti in this particular context may suggest that he is still in the process of struggling toward a new terminology for his theory of logical proof. Dignaga was well acquainted with the Indian Grammatical tradition. He borrowed from them not only the method of anvaya and vyatireka in order to determine a valid reason but also the notion of 'restriction' (avadhāraṇa / niyama) by the particle eva in order to express the proper relation between the valid reason and its object, in short, 'pervasion'.
263
Although Dignaga does not mention the vyatireka formula with eva in PS-Vṛtti II. Scd, NMukh and the corresponding passages of PS-Vrtti quoted at the beginning of this paper seem to indicate that he intended to insert the eva-restriction in the vyatireka formula, too. Thus we can reformulate it as: asati [lingasya] nästità eva or sadhyadharma-abhave lingasya nästită eva. In this formulation, the domain of absence of the property to be proven is restricted by that of absence of linga, which is logically equivalent to the anvaya formula with eva. Dignaga seems to be aware of such a logical equivalence and tries to solve problems related to it in PS-Vrtti IV, which I shall deal with in a future paper. Dharmakirti calls anvaya and vyatireka with eva 'niyamavat' (restricted) and apparently regards them as logically equivaHe tries to justify the necessity of both anvaya and vyatireka by

Page #24

264
saying that either formula with eva actually expresses both of them at once.2)
anvaya and vyatireka with eva singularly express the relation of 'pervasion' (vyāpti) of a valid inferential mark or reason by the property to be inferred or proven. In modern terminology, they express the sufficient condition for a valid reason.24)
Notes
1) For Sanskrit fragments, see Jambuvijaya, Dvadasāram Nayacakram of Ācārya
Sri Mallavādi Kşamāśramana, Part I, Bhavnagar 1966: Appendix p. 122. 2) For Sanskrit fragments, see Jambuvijaya, op. cit., Appendix p. 125. A differ
ent recension: parārthānumānam tu svadrsțārthaprakāśakam/. Dharmakirti cleverly characterized them as 'knowledge' (jñāna) and 'verbal
expression' (sabda) in the beginning of the Pramāņaviniscaya Chapter II. 4) Sanskrit fragment is found in the Nyāyavārttika of Uddyotakara ad Nyāyasú
tra 1. 1. 5, and many other places. 5) Sanskrit fragment is found in the Pramāņavārttikälamkāra of Prajñākaragupta
(Tibetan Sanskrit Works Series, ed. by R. Sānkstyāyana, Patna 1953), p. 580. 6) Cf. Karņakagomin, Pramāņavārttika-Țikā, p. 14: kim punaḥ kvacit tarkasās
tre drstam niyamārtham vacanam ity ata āha / sajātiya evet yādi / tatra yaḥ san sajātiye dvedhā cāsams tadatyaye sa hetur ity (PS III. 22) atrācāryiye hetulakşaņe/ sajātiya eva sattvam ity avadhāraṇena siddhe’pi vijātiyād vipaksād dhetor vyatireke yad etad asams tadatyaya iti sādhyābhāve 'sattvavacanam tan niyamārtham äcäryena vyakhyātam asaty eva năstita yathä syān nânyatra na viruddha iti / tathehäpi dharmivacanam bhāvaniyamartham āśamkyate / (... By the restriction in the second characteristic of trairūpya:] "sajātiya eva sattvam” the exclusion (vyatireka) of the reason (hetu) from the dissimilar class (vijātiya=vipakşa) (i.e. the third characteristic) is established, yet the absence of [the reason) in the absence of what is to be proved (sādhya) (i.e. the third characteristic) is stated (again] by [PS III. 22:] “asams tadatyaya" [in the definition of a valid reason). The Teacher (Dignāga) justifies it as having the purpose of restriction (niyama) [of the meaning of the term 'asat' (i.e. sādhyabhāva)] by saying "asaty eva năstita yathā syān nănyatra na viruddhe.") Cf. also Sākyabuddhi's Țikā (Derge 7b1-b3): yang ci rtogs pa'i bstan bcos la la las grub bzhin du brnan pa'i phyir yang smos pa lta mthong ngam zhe na/ mthun pa kho na la zhes bya ba la sogs pa smos te / slob dpon phyogs kyi glang po'i mtshan nyid ni phyogs kyi chos (d: 7b2] gang zhig mthun pa la yod la de

Page #25

Dignāga on trairūpya Reconsidered: A Reply to Prof. Oetke
265
med pa la med pa de ni gtan tshigs yin no zhes bya ba yin no / de la ni mthun pa kho na la yod pa zhes bya ba'i nges par gaung ba 'di kho nas gtan tshigs mi mthun pa las ldog pa grub mod kyi / de med pa la med pa zhes bsgrub par bya ba med pa la med pa [763] smos pa gang yin pa de ni med pa kho na la med pa yin gyi 'gal ba la yang ma yin la de las gzhan pa la yang ma yin no zhes brnan pa'i phyir yin te / de lta bas na brnan pa yin par dogs pa
bstsal ba'i phyir nye bar btags pa yin no // 7) For references see my paper "On Trairüpya Formulae".
See for example Jinendrabuddhi's Țikā on PS II. 5cd (Derge 93b-94a): de mtshungs la yod gyur (=tattulye sadbhāvah) ni rjes su 'gro ba (=anvaya) mtshon par rig par bya ste / ... rjes su 'gro ba ni tshul gnyis pa'o zhes pa'o I... rjes su 'gro ba zhes bya ba ni rtags la rtags can yod pa kho na ste / ... med pa la med pa nyid (=asati năstită) ni Idog pa (=vyatireka) nye bar mtshon par yang blta bar bya'o / Idog pa yang tshul gsum pa'o.../. Cf. also Tarkabhāṣā (of Mokşākaragupta, ed by Rangaswami Iyengar, Mysore 1944) pp. 24-26: anumeye sattvam eva niścitam ... tad ekam rūpam pakşadharmatāsamjñakam / ... sapaksa eva sattvam niscitam ity anvayasamiñakam dvitiyam rūpam / ... tatra (=asapakse) asattvam eva niścitam vyatirekasamjñakam trtiyam rūpam / See my paper "Dignāga and Dharmakirti on apoha", Studies in the Buddhist Epistemological Tradition, ed. Steinkellner, Wien 1991, p. 140; see also Hayes,
Dignaga on the Interpretation of Signs, pp. 297-298. 10) In: Studies in Indian Philosophy, A Memorial Volume in Honour of Pundit
Sukhalji Sanghavi, L.D. Institute of Indology, Ahmedabad 1981, p. 79. 11) For the function of the particle eva, see Y. Kajiyama, "Three Kinds of Affir
mation and Two kinds of Negation in Buddhist Philosophy," Wiener Zeitschrift für die Kunde Südasiens, Band 17, 1973 and Brendan Gillon and Richard Hayes, "The Role of the Particle eva in (Logical) Quantification in Sanskrit," Wiener Zeitschrift für die Kunde Südasiens, Band 26, 1982. For Sanskrit fragment, see Het ubindu-Ţikā (ed. by Sukhlalji Sanghavi and
Shri Jinavijayaji, Gaekwad's Oriental Series CXIII, Baroda 1949) p. 18. 13) N.B.: N.Mukh does not contain such a definition of niyama or vyāpti. 14) NMukh k. 2: E RA ## ## # ### ==PS III. 9:
sa pakşe sann asan dvedhā pakşadharmaḥ punas tridhă / pratyekam asapkşe ca sadasaddvidhatvataḥ / (Quoted in the Nyāyavārttikatātparyaţikā ad NS 1. 1. 35 and PVA p. 580). NMukh (Taisho p. 1b): UA B Quoted in the Nyāyavārttika ad Nyāyasūtra 1.1.35: hetuħ tadābhāso vā prāyaḥ pakşadharma eva bhavati.
9)
5
12)

Page #26
266 18) Cf. PS III. 8: sadhyadharmo yato hetus tadabhasas ca bhuyasa / tasmat tad vistarah purvam hetvadyarthat pradarsyate | Quoted in the Pramanavart tikalamkara p. 580. 15) Quoted in the Pramanavarttikalamkara p. 580; Hetubindutika p. 11, Dharmot tarapradipa p. 97 reads: samudayasya sadhyatvad dharmamatre ca... 16) As a matter of fact, Dharmakirti went back to the old terminology 'paksa dharma' when he first defined hetu in Pramanavarttika I. 1abi: paksadharmas tadamsena vyapto hetuh. In Nyayabindu II. 5, however, he defines the first characteristic as lirgasyanumeye sattvam eva (niscitam), which echoes the definition in PS II. 5cd. 17) "On Sapaksa", Journal of Indian Philosophy, vol. 18-1, 1990. See for example Hayes, op. cit., p. 113. 19) It is to be noted that Dharmakirti does not agree with Dignaga's definition of vipaksa. According to him, vipaksa | asapaksa is what is not sapaksa, which includes what is different from or incompatible with sapaksa as well as absence of sapaksa. See for example, Nyayabindu II. 11-12: na sapakso 'sa paksah || tato'nyas tadviruddhas tadabhavas ceti ll. 20) Prasastapada's formulae of the trairupya (anumeyena sambaddham, tadanvite prasiddham, tadabhave nasty eva) seems to support our interpretation of the logical subject of asati. For reference see my paper "On Trairupya Formu lae", p. 163. 21) Jinendrabuddhi's Tika (D94a): 'dir med pa'i sgras de dang mtshungs pa'i med pa brjod par bya ba 'ba' zhig ma yin gyi / 'on kyang de las gzhan pa dang de dang 'gal ba yang ste de'i rnam pas dben pa'i phyir ro zhes rig par bya'o || 22) I have discussed the matter in my paper presented at the XIIth IABS con ference in Lausanne in August 1999. 23) See Hetubindu (ed. by Ernst Steinkellner, Wien 1967, Teil I) p. 34. One final remark on Oetke's book (p. 52): He seems to have some difficulty in identifying the two Chinese expressions which appear in N Mukh (Taisho p. 2c), namely, and they correspond to paryudasa and prasaj yapratisedha, which can be proven from the parallel passage in PS-Vitti IV.3. This fact has been pointed out by Yasunori Ejima, Chugan Shiso no Tenkai-Bhavaviveka Kenkyu--, Tokyo 1980, p. 117. 24) I would like to thank Prof. Jonathan Silk for correcting my English and giving me several important suggestions. (WE WAKIU**)