Part2–Studies

H E L M U T K R A S S E R

LogicinaReligiousContext:DharmakīrtiinDefenceofāgama1

1.INTRODUCTIONContrary towhatmanyhavebelieved,Dharmakīrtiʼs (~550)2 viewofscripture (āgama)3 is unambiguous and clear.4 In sharp contrast toVasubandhu,who,inaccordancewiththeBodhisattvabhūmiandotherearlierYogācāratexts,acceptedthreemeansofvalidcognition(pramā‐ṇa),namely,perception,inference,and–asthethird–eithertheBud‐dhaor the SonsofBuddhaor their teaching,5 and also in contrast toDignāga,whoacceptedscriptureasameansofvalidcognitionbutsub‐sumed it under inference, Dharmakīrti neither counts scripture as aseparatemeansofvalidcognitionnorsubsumesitunderinference.ForDharmakīrti no scripture, including even the Buddhist teaching, is a 1Ashortversionofthispaperwaspresentedatthe14thWorldSanskritConference,

September1–5,2009,KyotoUniversity. 2Fortheproposeddateofca.themiddleofthesixthcenturyforDharmakīrti’stime

ofactivity,seeKrasser2012. 3The translation “scripture” forāgama is problematic (see Eltschinger 2007a: 17–

20)andonlyadoptedforwantofabetterone.āgamashouldalsocomprise,e.g.,theeternalwordsoftheVedaortheoralteachingsoftheBuddhaorMahāvīrathathavebeenhandeddownwithoutinterruption.

4For a general overview of Dharmakīrti’s philosophical thought, see Eltschinger2010.

5See AKBh 76,22–23:kiṃkāraṇam/pramāṇābhāvāt/nahy eṣāṃdravyato ʼstitvekiñcidapipramāṇamastipratyakṣamanumānamāptāgamovā,yathārūpādīnāṃdharmānāmiti/;AKBh460,2–3:saddharmanītautupunarbuddhāevapramāṇaṃbuddhaputrāśca/. SeealsoBoBhū25,19:pratyakṣamanumānamāptāgamaṃpramāṇaṃniśritya.FortheacceptanceofthesethreekindsofpramāṇasintheVY,see Verhagen 2008; for their acceptance in other Yogācāra texts, see Eltschinger,forthcoming,n.79.

84 HELMUT KRASSER meansofvalidcognition,andtherecanbenocertaintybasedonscrip‐ture – quite a remarkable position for a Buddhistmonk! The reasonwhythisunambiguousunderstandingofscripturehasnotalwaysbeenrecognizedassuchisDharmakīrtihimself.Hedoesnotproclaimitveryloudly–for,again,itentailsthateventhebuddhavacanacannotbere‐garded as a pramāṇa – probably not only because it contradicts thedefinitionofscripturegivenbyDignāga,butalsobecauseitwasatoddswith the attitude of the Buddhist community. Such a view certainlywouldnothavebeenfavouredbythespiritualauthorities.

2.PVSV108,1–109,22ONPS2.5abDharmakīrtidealswithDignāgaʼsde initionofscriptureinthecontextof his theory of concept formation and language, apoha. One of thecornerstones of his apoha theory, as formulated in PV 1.213, is thatwordshavenoconnectionwiththeexternalworld;theycanonlymakeknown the intention of the speaker, not external objects.6 As thiscontradicts Dignāgaʼs de inition of scripture, immediately afterexplaining this verse in his Svavṛtti Dharmakīrti raises the followingquestioninordertodealwiththisproblem:7

How then [should it be understood]when [Dignāga byhis definition ofāptavāda,]“Thestatementofacrediblepersonisinference,becauseitisequalinnotbelying,”saidthatscriptureisinference?8

6PV1.213:nāntarīyakatābhāvācchabdānāṃvastubhiḥsaha/nārthasiddhis tatas tehivaktrabhiprāyasūcakāḥ//. “Becausewordsarenot invariably concomitantwithrealentities(vastu),therefore[they]donotestablishrealentities(artha).Fortheymakeknown[only]thespeakerʼsintention.”

7Dharmakīrtiʼs treatmentofDignāgaʼsāptavādadefinitionhasbeen translatedsev‐eral times. ForPV 1.213–217with PVSV, see Yaita 1987 (included in Yaita 2005:442–449); forPV1.214–223withPVSV,seeDunne2004:361–373; forPV1.213–268withPVSV,seeEltschinger2007a:217–385.

8PVSV 108,1–2: yat tarhīdam – āptavādāvisaṃvādasāmānyād anumānatā (PS2.5ab)ityāgamasyānumānatvamuktam,tatkatham/.Amoreaccuratetranslationwillbeofferedbelow.AsIintendtodealwithDignāgaʼsunderstandingofthisde i‐nitionelsewhere,inthefollowingIwilldiscussonlytheproblemsrelatedtoDhar‐makīrtiʼstreatmentofit.For(different)interpretationsofPS2.5ab,seeEltschinger2007a:70(withn.9),218ff.,andLasic2010.

LOGIC IN A REL IG IOUS CONTEXT 85

Dharmakīrtidevotes4verseswithcommentarytothesolutionofthisproblem, in which he provides two different interpretations of thisdefinition.9 Of these interpretations I will refer only to those state‐mentsthatarenecessaryformyargumentation.

2.1.WhydidDignāgaprovideadefinitionofscripture?

Immediately following the posing of the above question Dharmakīrtitellsuswhyweareinneedofāgamaatall.

Apersoncannotlivewithoutrelyingonthevalidityofscripturebecausehehas heard that in the case of certain [activities] the results ofwhichcannot be perceived, engaging or not engaging [lead to] extremelypraiseworthy or disastrous [results], and because he does not see any‐thingthatcontradictsthepresenceofthose[results].Therefore[Dignāga]taughtthatthevalidity[ofscripture]isduetoacriticalexamination[ofit,havinginmind]“giventhatonehastoproceed,itisbettertoproceedinsuchaway[i.e.criticallyexaminingthescripture]10.”11...Ifthat[treatise,]uponbeingexamined,isnotliabletomakingfalseas‐sertions(na…visaṃvādabhāk),[then]theonewhoproceedsmightshine(śobheta)!12

Inotherwords:Wedoneedreligionbecauseitistheonlymeanstoes‐capesaṃsāra.Thus,wehavetobaseourreligiousactivitiesonāgama.

9ForahelpfulsynopsisandconcisesummaryofthisexcursusonPS2.5ab,seeYaita

1987:2–3andDunne2004:240–241.Forproposedchanges,seebelowp.100n.38. 10Cf. PVSVṬ 390,30–391,9: tat sati pravartitavye varam evam āgamaṃ parīkṣyapravṛttaityāgamasyaparīkṣayāprāmāṇyamāhācāryaḥ.

11PVSV108,2–6:nāyaṃpuruṣoʼnāśrityāgamaprāmāṇyamāsituṃsamarthaḥ,atyakṣa‐phalānāṃkeṣāñcitpravṛttinivṛttyormahānuśaṃsāpāyaśravaṇātatadbhāvevirodhā‐darśanāc ca. tat sati pravartitavye varam evaṃ pravṛtta itiparīkṣayā prāmāṇyamāha.

a°āpāya°PVSVms25a2(TSP4,22–23;PVSVṬms140b6;ṅansoṅPVSVTibD322a4/P478b3;seeYaita2005:443n.159):°āpāpa°PVSV,PVSVṬ390,25.

12PVSV108,15–16:tadyadinaparīkṣāyāṃvisaṃvādabhākpravartamānaḥśobheta.

86 HELMUT KRASSER 2.2.PV1.214–215:Whichāgamashouldbefollowed?

Inordertodecidewhichāgamaweshouldfollow–thisisexplainedinPV 1.214with commentary –we have to look for a scripture that iscoherent (sambaddha)and teachesaproperhumanaim(puruṣārtha)aswellasasuitablemeans forobtaining it (anuguṇopāya). Ifwe findsuchascripture,forexamplethatoftheBuddhists,thenwehavetoex‐amineitinitsentirety,asexplainedinPV1.215withPVSV.Whateveritteachesthatcanbecheckedbyperceptionandinferencehastobetrue.Anything taught insucha scripturemustnotbecontradictedbyper‐ceptionor inference.Moreover, therecannotbeasingle internalcon‐tradictioninit.Suchinternalcontradictionsarecheckedwiththehelpof the so‐calledāgamāpekṣānumāna (PVSV109,1),13 i.e., an inferencethat, based on passages of the scripture, tests its consistency.14 If anāgamapassesthisthreefoldtest15–wherebyitistestedinalltestable 13Later(PVSV174,22–23;seeabove,p.69)Dharmakīrtiusestheformulationāgamā‐

śrayeṇacānumānena.Incontrasttothiskindofinferenceaninferencethatoperateswith visiblematters is calledanāgamāpekṣānumāna (PVSV108,24). In the fourthchapterDharmakīrtitermsthislatterkindofanumānatogetherwithperceptionasvastubalapravṛttapramāṇa(PV4.108≈PVin331,11:yacchāstraṃvastubalapravṛt‐tenapramāṇenasvavacanenacābādhitaṃdṛśyādṛśyayorviṣayayoḥ,tadgrāhyamiti).These two kinds of inferences are already distinguished by Pakṣilasvāmin (NBh3,12onNSū1.1.1:pratyakṣāgamāśritaṃcānumanam).

14OntheapplicationofthismethodasdescribedinPV1.214–215byŚāntarakṣitaandKamalaśīla,seeMcClintock2010:318ff.

15Apossiblesourceof this threefoldanalysis,whichcametobeknownasdpyadpagsuminthelaterTibetantradition,mightbeVasubandhuʼsVY;seeVerhagen2008:244–247(SectionFour,“Threepramāṇas”)withtheappendix(pp.253–258).Asi‐milar idea isencountered in thedefinitionofāptāgama inASBh153,5ff.:yatropa‐deśetatpratyakṣamanumānaṃcasarvathānavirudhyetenavyabhicarataḥsaāptā‐gamaḥsampratyayitvāt.“DieUnterweisung,zuderdieseWahrnehmungund[diese]Schlußfolgerung in keinerWeise imWiderspruch steht [und von der] diese nichtabweichen, istāptāgama,weilerVertrauenerweckt.”TranslationOberhammeretal.1991:122b–123as.v.āptāgama.OnthebackgroundoftheTibetan“threefoldana‐lysis (dpyadpagsum)” seeTillemans1993:10ff., andKeira2006:182withn.15,who informs us that the notion of “purified through the threefold analysis (tshulgsumgyibrtagpasyoṅssudagpa)”isalreadyfoundinKamalaśīlaʼsMĀ(D148b4–5):[deʼiphyir]bkaʼgaṅźigtshulgsumgyibrtagpasyoṅssudagpathogmadaṅthamadaṅbardudgebarṅespadenimislubaʼiphyirdemkhasparnamskyisbrtenparbyabañiddo//.“[Therefore,]sincestatementsascertainedasfaultlessbymeansofthreekindsofinvestigation(tshulgsumgyibrtagpas)andasexcellentforthefirst

LOGIC IN A REL IG IOUS CONTEXT 87

cases(śakyaparicchedāśeṣaviṣayaviśuddhi,PVSV109,3–4)–andisthusestablished to be without contradictions, its being pure or cleansed(viśuddhi)16 constitutes its non‐belying nature or reliability (avisaṃ‐vāda, PVSV109,4). “Non‐belying” isof course thedefinitional charac‐teristic of a valid cognition.17 If a scripture is “purified” in this way,thenwecaninferthatthisscriptureortestimonyofacredibleperson(āptavāda) might be true also with regard to the non‐empirical ortranscendent things18 it teaches, because the scripture or credibleperson is the same.And if then, basedon that scripture, one acts to‐wardsnon‐empiricalobjectsonemightbe luckyandobtainwhatonewaslookingfor(seen.12).

2.3.DoesPV1.216commentuponPS2.5ab?

UptothispointtheexistinginterpretationsandtranslationsofPV(SV)(seen.7)moreor lessagree, and Ihave followed them thus far.Thesituationchanges,however,withthenextverse,i.e.,PV1.216.Scholarsdisagree about themeanings of thewordsagatyā and sāmānyāt and

[stage of practice, i.e., hearing (śruti)], the last [stage, i.e., meditation (bhāvanā)],and the middle [stage, i.e., consideration (cintā)], are non‐belying (avisaṃvāda),scholarsshouldrelyuponsuch[statements].”OntheresemblanceofDharmakīrtiʼsthreefoldtesttoĀryadevaʼsCŚ12.5(=12.280inTillemanscounting),seeTillemans1990:I.29–32.

16ThechoiceofthetermviśuddhimighthavebeeninspiredbyasimilaruseintheŚlo‐kavārttika,where śuddhatva of the causes guarantees thevalidityof the resultingcognition; see ŚVcodanā 44ab: tasmātkāraṇaśuddhatvaṃ jñānaprāmāṇyakāraṇam.“Therefore(tasmāt),thecauseofvalidityinacognition(jñānaprāmāṇyakāraṇam)istheexcellenceofcauses(kāraṇaśuddhatvam).”TranslationKataoka2011a:248.SeealsoKataokaʼsn.189onthetranslationofk.46,wherethetermśuddhyasambhavaisused.

17SeePV2.1ab1:pramāṇamavisaṃvādijñānam.“Validcognitionisnon‐belyingcogni‐tion.”

18In this context Dharmakīrti refers to these kinds of things as “not knowable byperceptionor inference” (pratyakṣānumānāgamye ʼpyarthePVSV109,8).Laterhealsousesthetermatyantaparokṣa(PVSV153,7,155,14,175,3–4,PV1.314c,1.316a,3.94,4.210).Forhisusageofparokṣainthesenseofatyantaparokṣa,seeTillemans1986:44n.14(=1999b:34n.13).ThetermatyantaparokṣaisalsousedbyUddyo‐takara(NV204,12onNSū2.1.31)andoccursalsoinYD70,15onSK4ab1,99,12onSK6cd,and101,7stillonSK6cd.

88 HELMUT KRASSER about the overall purport of the verse. Thus, it seems appropriate tohaveacloserlookatthedifferentpositions.

2.3.1.PreviousinterpretationsofPV1.216Let me begin with Tom Tillemans, to whom we owe much of ourunderstandingoftheBuddhistepistemologistsʼapproachtoscripture.In his article “Dharmakīrti, Āryadeva, and Dharmapāla on scripturalauthority” he provides uswith the following translation of the state‐mentofDignāgaʼswehavebeenconsidering:

PS II, k. 5a: Because authoritative words (āptavāda) are similar [to aninference] in not belying, they are [classified as] inference. (Tillemans1986:32=1999b:27ff.)

InhissummaryofDharmakīrtiʼsPV1.216,which if Iunderstandhimcorrectly19heconsiderstogetherwithPV1.215tobethecoreofDhar‐makīrtiʼs explanation of PS 2.5ab, Tillemans gives a slightly differentaccount of Dignāga’s statement. But first, let us have a look at Tille‐mans’translationofPV1.216.

āptavādāvisaṃvādasāmānyādanumānatā/buddheragatyābhihitāparokṣeʼpyasyagocare//PV1.216Asauthoritativewordsaresimilarinbeingavisaṃvāda, theunderstand‐ingoftheirimperceptible(parokṣa)objectisalsotermedaninference,for[otherwise] there would be noway [to know such objects]. (Tillemans1993:11)20

On the basis of this translation of PV 1.216 Dignāga’s position in PS2.5abissummarizedbyTillemansasfollows:

(d) v. 216: Dignāgaʼs point in saying that authoritative words were aninference was that when an authorityʼs words (=scripture) have beenfoundtobenon‐belyingonrationallydecidablematters,thenwearejus‐

19IderivethisunderstandingfromthefactthatTillemanspresentsDignāgaʼsPS2.5ab

togetherwithPV1.215–216twiceinordertooutline“theepistemologicalschoolʼsposition”onscripture(Tillemans1990: I.24ff.)andthebackgroundof theTibetan“threefoldanalysis(dpyadpagsum)”(Tillemans1993:10ff.).

20AlsotranslatedinTillemans1990:I.25.

LOGIC IN A REL IG IOUS CONTEXT 89

tified to understand radically inaccessiblematters based on that scrip‐ture....(Tillemans1999a:399–400=1999b:41–42)

I would subscribe to Tillemans’ translation of PV 1.216, though weshallhavetocomebacktotheexpressionagatyālater(below,pp.108–111). Nevertheless, Dharmakīrti’s understanding of PS 2.5ab in thewayTillemansissuggestingposesaproblem.IcannotseeanybasisinthewordingofPS2.5abitselfforsayingthatourunderstandingofthetranscendentobjectofscriptureistermedaninference.Itisonethingtosaythatāptavacanaisanumānaandanotherthatourcognitionofitsobject isanumāna. Inotherwords,didDharmakīrtiover‐interpretorevenmisinterpretDignāga,ormustweunderstandDignāgainadiffer‐entway?

Astheāgama‐sectionofDharmakīrtiʼsPVSVhasbeentranslatedbyHideomi Yaita in 1987, by John Dunne in 2004, and by VincentEltschingerin2007a,wemaygainsomehelpfromconsideringtheirin‐terpretations.Beforelookingattheiranalysesitshouldbementionedthat PS(V) 2.5ab is incorporated byDharmakīrti in his PV(SV) twice,and thathe interprets it somewhatdifferentlyeach time.First,here‐fers to itwhen initially raising the question of itsmeaning, and thenaboutapagelaterintheGnoliedition,inconcludinghisfirstinterpre‐tationofPS(V)5ab.Onitsfirstoccurrenceitismarkedasaquotation,21onthesecondDignāgaʼswordingisusedwithoutexplicitlycharacteris‐ing it as a quotation. The latter passage, namely PV 1.216, is com‐menteduponbyDharmakīrtiasfollows:

āptavādāvisaṃvādasāmānyādanumānatā/buddheragatyābhihitāparokṣeʼpyasyagocare//PV1.216tasyacāsyaivambhūtasyāptavādasyāvisaṃvādasāmānyādadṛṣṭavyabhi‐cārasyapratyakṣānumānāgamye ʼpyarthepratipattes tadāśrayatvāt tad‐anyapratipattivadavisaṃvādo ʼnumīyate/tataḥśabdaprabhavāpisatīnaśābdavadabhiprāyaṃnivedayatyevetyarthāvisaṃvādādanumānamapi/(PVSV109,8–11)

21PVSV108,1–2:yattarhīdamāptavādāvisaṃvādasāmānyādanumānatetyāgamasyā‐numānatvamuktam,tatkatham.

90 HELMUT KRASSER Dunneʼstranslationoftheversewithcommentaryreads,22

[Dignāga]saidthat,sincethestatementsofacrediblepersonaregener‐ally trustworthy, a cognition fromsuch statementsof those statementʼsobjectisawellformedinferenceofthatobject,eventhoughtheobjectisepistemically remote. The cognition is said to be an inference becausethereisnootherwaytoknowthatobject.[PV1.216]These kinds of statements of a credible person – those [described byDignāga]andthose[accordinglydelineatedabove]–aregenerally(sāmā‐nya) trustworthy.Hence,notobserving thosestatements tobemislead‐ing, one infers the trustworthiness of a cognition of those statementsʼobject,eventhoughitisnotknowablebyperceptionandempiricalinfer‐ence.One infersthatsuchacognition is trustworthybecause it isbasedonthosestatements, just liketheothercognitionsbasedonthosestate‐mentsthatcanbeverifiedbyperceptionandempiricalinference.Hence,eventhoughthatcognitioncomesfromlanguage,itdoesnotmakeknown just the speakerʼs intention like a cognition coming from [ordi‐nary] languagebecause in thiscase thecognition isalsoan inferenceofthe statementʼs objects, since it is trustworthywith regard to thoseob‐jects(artha).(Dunne2004:363ff.)

Hereitisclear,aswehavealreadyseeninthetranslationofTillemans,that in theverseourcognition of a transcendent objectderived fromthestatementsofacrediblepersonissaidtobeaninference.However,Dunne’stranslationoftheearlieroccurrenceofPS2.5ab,whereDhar‐makīrtiquotes it (PVSV108,1–2) to introducehisdiscussionofscrip‐ture,construesthetestimonyofacrediblepersonasasourceforanin‐ference.

[Dignāga]said,“Thetestimonyofacrediblepersonisthesourceforanin‐ferencebecauseitisgenerallytrustworthy,”...(Dunne2004:361)

Inasectionearlierinhisbookwhichgivesageneralaccountofscrip‐turalinference,titled“ScripturalInferenceandDharmakīrtiʼsRejectionofCredibility,”Dunnetranslatesthesamepassageinanotherway. 22In the translation offered by JohnDunne his explanatory remarks are not always

writteninbracketsinordertopresentasmoothtextreadablealsofornon‐special‐ists.However,theyeasilycanberecognizedassuch.

LOGIC IN A REL IG IOUS CONTEXT 91

Since the statements of a credible person are generally trustworthy, acognitionarising from them is an instrumental inference. (Dunne2004:239)

In the footnote to the translation Dunne justifies this interpretationwithreferencetoPV1.216.

Note,first,thatalthoughthephrase“acognitionarisingfromthem”mustbesuppliedbycontext,Dharmakīrticlearlyunderstandsthisstatementinthatfashion(seePVI.216,translatedalongwiththerestofthesectiononscriptureintheappendix).

Interestingly enough, thisdoubleunderstandingofoneand the samepassage – as referring to either scriptureor the cognition that arisesfrom it asan inference– is also tobe found inHideomiYaitaʼs treat‐mentofDharmakīrtiʼsviewofscripture.Inhissynopsishesummarisesthepassagewehavejustreferredtoasfollows:

Opponent(108,1f.):DharmakīrtiʼsstatementisnotconsistentwithDignā‐gaʼsstatementinPSII5ab,whichrecognizedāgama(āptavāda)asapra‐māṇa,i.e.anumāna.(Yaita1987:2)

HistranslationofPVSV108,1–2reads,(Opponent:)Then,how[doyouexplain] the fact that [Dignāga]said thesacredtradition(āgama) is[apramāṇa, that is]anumāna [inhisfollow‐ing statement:] “Because the words of a credible person (āptavāda)generallydonotdisagree,[thecognitionbasedonit]isaninference(anu‐māna)”?(Yaita1987:6)

The above interpretation of PS 2.5ab in thispūrvapakṣa is obviouslybasedonPV1.216.

Becausethewordofacredibleperson(āptavāda)generallydoesnotdis‐agree [with our experience], the cognition [basedonāgama] evenwithregardtoits(=āgamaʼs)objectbeyondtherange[ofanymeansofknowl‐edge]issaid[byDignāga]tobeaninference,forthereisno[other]possi‐bility(agatyā)[toexplainit].(Yaita1987:8)23

23VanBijlert(1989:123–124)isalsooftheopinionthatPV1.216isanexplanationof

PS2.5ab:“InthenextverseDharmakīrtiusesthisgeneralconceptionoftrustwor‐thinesstogivehisfirstexplanationofPSII.5ab…(PVI.216).”

92 HELMUT KRASSER Eltschinger (2007a:219,224)doesnotreadPVSV108,1–2 in lightofPV 1.216 but treats both passages independently. Thus, the questionremains:DidDignāgasay inPS2.5abthatscripture isakindof infer‐enceordidhetermourunderstandingoftheirtranscendent(parokṣa)objectsaninference,ordidhesayboth?

2.3.2.OnthecontextofPV1.216withPVSVAsIcannotseehowonecanharmonizethesetwointerpretations,theone clearly saying thatāgama is inference (PVSV108,1–2), theotherthat our cognition is inference (k. 216), it seemsworthwhile to haveanotherlookatthecontextofPS2.5ab.24Thishasbeennicelysumma‐rized by Tillemanswhen arguing against Hayesʼ interpretation of PS2.5abinhismonographDignāgaontheInterpretationofSigns.

2) In the immediately preceding passage to k. 5, Dignāga had made adistinctionbetweentwotypesofinference,dependinguponwhethertheobject isempiricalornon‐empirical, arguing that in the formercasewecanapplynamestowhatisempirical,butinthelattercaseweonlyhaveaconcept (rnamparrtogpa=vikalpa)anddonotcognize thesvalakṣaṇaobject.3)Anopponentthentriestofindanabsurdity,sayingthatinthatcaseau‐thoritativestatementsaboutimperceptibleobjectswouldjustexpresstheconceptuallyinventedobjectandnottherealparticularatall:hencetherewouldbenodifferencebetweenauthoritativeandunauthoritativestate‐ments.4)Dignāga then replies that authoritative statementsaboutheavenandthe like do not express just the conceptually invented object: they aresimilartonormalinferencebecausetheytooarenon‐belyingwithregardtotherealparticular.For,althoughtheheavensandsofortharebeyondoursenserange,authoritativepeoplehavedirectlyseenthemandhencewereabletoapplythewords“heaven,”etc.(Tillemans1990:I.22)

Tillemanshad to base his summary on the two barely intelligibleTi‐betan translations of the PS(V) by Vasudhararakṣita and Kanakavar‐man. With the help of Jinendrabuddhiʼs commentary in Sanskrit we 24OnthewiderstructuralpositionofPS2.5ab,seeLasic2010:514–522.

LOGIC IN A REL IG IOUS CONTEXT 93

nowcangetamorenuancedpicture.TheopponentreferredtobyTil‐lemans in (3) in fact asks how this latter cognition operatingwith anon‐empiricalobjectwhich isamereconceptof theobjectcanbe in‐ference(kathaṃtarhitasyānumānatvam).25Dignāgaintroduceshisan‐swertothisquestionbymentioningwordssuchasheaven, therefer‐entofwhichweneverhaveseen:“For(hi)wordslike‘heavenʼdonotexpress themere object (arthamātra).”26 Next follows PS 2.5abwiththeVṛtti.

āptavādāvisaṃvādasāmānyādanumānatvam/PS2.5abāptavacanaṃgṛhītvārthāvisaṃvādatulyatvāttasyānumānatvamuktam.27

25A possible Sanskrit can be reconstructed as follows: adṛṣṭārtha arthavikalpamā‐tram,naviśiṣṭārthapratītiḥ.kathaṃ tarhi tasyānumānatvam.nahisvargādiśabdairarthamātramucyate.(=PSVK111a2–3:mamthoṅbaʼidonladondurnamparrtogpatsamyingyidongyikhyadparrtogspanimayinno// jiltarderjessudpagpañidyintemthorislasogspaʼisgrarnamskyisdontsambrjodbanimayinno//.)TheSanskritisavailableinPSṬms62b7:adṛṣṭārthesvargādāvarthavikalpamātramiti(PSṬTibD92a2/P103b1–2:mamthoṅbaʼidonlanimthorislasogspalaste/dondurnamparrtogpa tsam źespa), quoted inPVSV37,27;PSṬms63a5: tataścedarthavikalpamātraṃbhavati,naviśiṣṭārthapratītiḥ(PSṬTibD92a7/P103b7–8:delasgaltedonrnamparrtogpatsamduʼgyurgyidongyikhyadparrtogsparmiʼgyur na); PSṬms 63a4: kathaṃ tarhi tasyānumānatvam iti (PSṬTib D92a6/P103b7:ʼonaciltarderjessudpagpayinźenaźespa/);PSṬms63a6:āha–nahītyādi (PSṬTibD92b1–2/P104a2:bśadpa/mayin źespa la sogspa ste/);PSṬms63a7: tasmānnāptaprayuktaiḥ svargādiśabdairarthamātramucyate.kiṃ tarhi.laukikaśabdāsādhāraṇebāhyeʼpyarthaiti(PSṬTibD92b2–3/P104a3–4:deʼiphyiryidchespas rab tu sbyarbaʼimtho ris la sogspaʼi sgra rnamskyi (read:kyis)dontsambrjodpamayingyi/ ʼonaciźena/ ʼjigrtenpaʼisgradaṅthunmoṅmayinpaʼiphyirolgyidonyaṅṅoźespaʼo//).

26nahisvargādiśabdairarthamātramucyate,seeabove,n.25. 27ForthereconstructionoftheSanskrittext(followingmainlytheTibetantranslation

ofVasudhararakṣita),seebelow,p.94.TheTibetantranslationreads:yidchestshigkyaṅmislubar//mtshuṅsphyirrjessudpagpañid//PS2.5ab

yidchespaʼi tshigñidbzuṅnaskyaṅmibslubarmtshuṅspaʼiphyirdeyaṅrjessudpagpañiddubrjoddo//PSVV29a2–3.

yidchestshignimisluba//spyilasrjessudpagpañid//PS2.5ab yidchespaʼitshigñidbzuṅnasdonde lamibslubaʼiphyirdaṅ/mi ʼdrabaʼiphyir

rjessudpagpañiddubśadpayinte//PSVK111a3–4.

94 HELMUT KRASSER ...(PS2.5ab).Havinggraspedthestatementofacredibleperson,it(tasya)hasbeensaidtobeinference,because[it]isequalinnotbelyingwithre‐gardtotheobject.

IfIunderstandDignāgaʼswordinginthecommentarycorrectly,thisisthe answer to the initial question kathaṃ tarhi tasyānumānatvam,referringtoourcognitionofsvargaetc.,theobjectofwhichneverhasbeen seen by us but only by the āptas. This interpretation goessmoothlywithJinendrabuddhiʼscommentswhenhesaysthattheques‐tion,nowreformulatedaskasmātpunastasyānumānatvamtoconformtothewordinghehaspreviouslyused,28hastobeansweredbyarthā‐visaṃvādatulyatvāt.Jinendrabuddhiʼscommentsread,

āptavacanaṃ gṛhītvetyādi. āptavacanaṃ gṛhītvāvisaṃvādādhigamahe‐tuṃtasyānumānatvamuktama.kasmātpunastasyānumānatvamarthā‐visaṃvādatulyatvād itivyākhyeyam.PSṬms63b5–6 (yidchespaʼitshigbzuṅnasźespalasogspasteyidchespaʼitshigbzuṅnasmibslubarrtogspaʼirgyur/derjessudpagpañiddugsuṅsso//ciʼiphyirderjessudpagpañidyinźena/donlamibslubarmtshuṅspañidkyiphyirźesbśadparbyaʼo//PSṬTibD93a1–2/P104b2–4)auktamem.(gsuṅssoPSṬTib):yuktaṃPSṬms“Havinggraspedthestatementofacredibleperson”andso forth. It(tasya)hasbeensaidtobeaninference,havinggraspedthestatementofa credible personwhich is a logical reason (hetu) for anunderstandingthat[it]isnotbelying(avisaṃvāda).[Thequestion]whynowthis(tasya)isaninferenceistobeexplainedby“because[it]isequalinnotbelyingwithregardtotheobject”(arthāvisaṃvādatulyatvāt).

tasyainJinendrabuddhiʼsexplanationthuswouldhavethesamemean‐ing as it has in Dignāgaʼs initial question and refer to the inferencewhich is a mere concept, which would correspond to Dharmakīrtiʼsbuddheḥ inPV1.216orpratipatteḥ in theSvavṛtti.AndDharmakīrtiʼsāptavādāvisaṃvādasāmānyāt of PV 1.216, finally, would be hisrephrasing of Dignāgaʼs arthāvisaṃvādatulyatvāt. Thus, I propose to

28PSṬms63a4:kathaṃtarhitasyānumānatvam iti (PSṬTibD92a6/P103b7: ʼonaciltarderjessudpagpayinźenaźespa/);seeabove,p.93n.25.

LOGIC IN A REL IG IOUS CONTEXT 95

understandPV1.216 togetherwith theSvavṛtti as commentinguponDignāgaʼsVṛttionPS2.5abandnotonPS2.5abitself.29

2.4.Onsāmānyāt

Thevariousinterpretationsofsāmānyāthavealreadybeenaddressedby Lasic (2010: 511–514), whose main concern is a possible inter‐pretationofPS2.5abinitsownright,notDharmakīrtiʼsinterpretation.Thus,hedoesnotofferasolutionforhowtounderstanditinDharma‐kīrti.

2.4.1OnthemeaningoftheablativeendingTheablativeendingofsāmānyahasbeentranslatedintwoways:(1)as“insofar,”etc.,and(2)as“because,”etc.Hayesʼ(1980:252,1988:238)interpretation of the ablative as “insofar” has been refuted convinc‐inglybyTillemans(1990:I.20–21).30Theinterpretationoftheablativeas“because,”“since,”andthelike,towhichIsubscribe,isfavouredbythemajorityofscholars.

2.4.2.OnthemeaningofsāmānyaAlsoregardingthemeaningofsāmānyatherearetwocamps.Onetakesitas“similar”or“equal,”etc.,theotherunderstandsitas“ingeneral”or“generally.”ThelattertranslationhasbeenproposedbyYaita(1987:6,8) and van Bijlert (1989: 124), and it is also the one preferred byDunne. Dunne (2004: 363–364 n. 10) argues for this interpretationagainst theunderstandingofŚākyabuddhi inorder toavoidacontra‐diction with PV 1.218. He does so even at the risk of attributing toDharmakīrti a form of fallacious reasoning that Dharmakīrti himselfreferstoasśeṣavadanumāna–andwhich,aswewillseelater,alsoper‐tains toDignāga’sdefinitionandDharmakīrti’s interpretationof it, as 29Itgoeswithoutsayingthat,disregardingthecontext,tasyainPSVaswellasinPSṬ

easily could be understood as referring toāptavacana. In that case, however,wehavetoassumethatthequestionkathaṃtarhitasyānumānatvamremainscomple‐telyignored.

30This refutation holds good also for van Bijlertʼs interpretation “in so far” (1989:122).

96 HELMUT KRASSER wellastoŚākyabuddhiʼscomments–byapparentlytakingtheterm“ingeneral” in the stronger senseof “universally,without exception.”Hewrites, “On this interpretation, the argument is that, since the state‐mentsofaparticularauthorhavebeenobservedtobetrustworthyinterms of observable objects, this general trustworthinessmay be ex‐tended to unobservable objects.” This understanding leads Dunne tothefollowingtranslationofPVSV109,7–9(onk.216,seeabovep.90):

Thesekindsofstatementsbyacredibleperson–those[describedbyDig‐nāga]andthose[accordinglydelineatedabove]–aregenerallytrustwor‐thy.Hence, not observing those statements tobemisleading, one infersthe trustworthiness of a cognition of those statementsʼ object, eventhough it is not knowable by perception and empirical inference.31(Dunne2004:364)

This translation is fine, given that sāmānyātmeans “generally.”Dhar‐makīrtiherespeaksofasāmānyaintermsofavisaṃvādaoftwokindsofstatements:tasyacāsyaivambhūtasyāptavādasya.32Thefirstonere‐ferred to by tasya is, according to Karṇakagomin, theāgama said byDignāgatobeanumāna.Thesecondoneaddressedbyasyaivambhūta‐syaistheonethat,inaccordancewithwhathasbeensaidink.214,hasbeenfoundtobecoherent(sambaddha)andtoteachaproperhumanaim(puruṣārtha)aswellasasuitablemeansforobtainingit(anuguṇo‐pāya),whichthenhasbeentestedandascertainedtobetrustworthyasdescribed ink.215withPVSV.33Theāgama at stake inDignāgacon‐cernsheaven(svarga),whichisnotaccessibletoourknowledge.Thus, 31PVSV 109,7–9: tasya cāsyaivambhūtasyāptavādasyāvisaṃvādasāmānyād adṛṣṭa‐vyabhicārasyapratyakṣānumānāgamyeʼpyarthepratipattestadāśrayatvāttadanya‐pratipattivadavisaṃvādoʼnumīyate.

32Hereonewouldexpectanotherca:evambhūtasyaca.But this isnotsupportedbythecommentaries.PVSVTibD322b6–7/P479a7(dedaṅʼdideltaburgyurpaʼiñespazadpaʼitshig)suggeststhereading:tasyāsyacaivambhūtasyaortasyāsyaivambhūta‐syaca.Thetexteitherneedstobecorrectedorwehavetosupposethatca ismis‐placed(bhinnakrama).

33PVSVṬ394,9–10:tasyāgamasyācāryadignāgenanirdiṣṭānumānabhāvasya.asyetyasmābhisasambaddhānuguṇopāyambityādināvicāritasya.

a asmābhiḥ refers to Dharmakīrti; cf., e.g., the comment on atrocyate (PVSV39,30):atrasāmānyalakṣaṇeʼsmābhirucyate(PVSVṬ174,21).bsambaddhānuguṇopāyamem.(PV1.214):sambandhādanuguṇopāyamPVSVṬ,PVSVṬms142a1.

LOGIC IN A REL IG IOUS CONTEXT 97

Dharmakīrti would be saying that “these kinds of statements by acredibleperson–those[thatspeakaboutheaven]andthose[accord‐inglydelineatedabove]–aregenerallytrustworthy.”Thisimpliesthat“statements speaking about heaven are generally trustworthy,”whereby “generally trustworthy” means “checked by the threefoldtest.”Inordertoavoidthisproblematicimplicationonecanassume,ofcourse, that Dharmakīrti with tasya does not mean anything specialandonlywants to informus thatDignāgaconsiderseven thatpartofāgamaheisn’tspecificallyconcernedwithtobetrustworthy.

Bethatasitmaybe,itisinanycasenotclearwhygeneraltrustwor‐thinessjustifiesDignāgaʼsholdingtheāgamathatspeaksofheaventobeanumāna.Thisleadsustothequestion,OnwhatbasiscanDignāga,according to Dharmakīrti, claim āgama to be anumāna, even whensāmānyaisunderstoodas“similar”or“equal,”etc.,intermsofitsbeingavisaṃvāda?First,however,onemustanswerthequestion,Similartowhat?Itseemstherearetwopossibilities:(a)eitherleaveitopen,asIhavedoneinthetranslationofPVSV108,1–2(seeabove,p.84),or(b)takeitspecificallyas“similartoinference.”Thosewhohaveoptedforthe first alternative areHayes (1988: 238), “insofar as they have thecommoncharacterofnotbeing false,”Tillemans(1990:20),“becauseauthoritative words are similar in not belying” (similarly, Tillemans1993:10,11),andEltschinger (2007a:219), “est semblableen fiabil‐ité.”Yetthesamescholarshavealsoonoccasionoptedforthesecondalternative: Hayes (1980: 252), “insofar as they have (in) common(withinferencethe)characterofnotbeingfalse,”Tillemans(1986:32),“becauseauthoritativespeech(āptavāda)issimilar[toaninference]inbeinginfallible,”and(1990:22),“theyaresimilartonormalinferencebecause they tooarenon‐belying,”andEltschinger (2007a:224), “estsemblable[àlʼinférence]en iabilité.”

Thus,thestatementofacrediblepersonwithregardtotranscend‐entobjectssuchasheaveniseithersaidtobeinferenceonaccountofthe general trustworthiness of the words of this credible person re‐gardingperceptibleandinferablematters,oronaccountoftheirbeingsimilar tonormal inference innotbeingbelying.Both interpretationsarefarfrombeingself‐evident,noristhereanycommentarialsupportfor them,as faras I can tell.Thereare,however, clearstatementsbythecommentatorsthatDignāgainPS2.5abintendedwhatDharmakīrtiwould later call a svabhāvahetu and that anumāna should be under‐

98 HELMUT KRASSER stoodinthesenseoflogicalreason(liṅga,hetu).34ThisdidnotescapeDunne, who provides the relevant material in the footnotes to histranslation.Thecommentatorsalsotellusthattheequalityorsimilar‐ityisbetweenthestatementsofacrediblepersonconcerningmattersthat can be checked by perception and inference and transcendentmatters suchasheaven thatwecannot check.Already the commentsbyŚākyabuddhi(followedbyKarṇakagomin)ontheopeningquestionatPVSV108,1–2arequitetelling.

Anystatementofacrediblepersonisnon‐belying,like[astatement]suchas“Allconditionedthingsaremomentary.”Andthis isastatementwithregardtoatranscendentobject.Thus,thistooisnon‐belying.Becausethestatementofacrediblepersonisnon‐belying(avisaṃvāditvāt,[PVSVṬ,no equivalent in PVṬ]) as its character of being non‐belying is thesame in the way [explained] (evam), it is inference. Thus (iti), theteacherDignāgasaidthatscriptureisinferencewithregardtotheexter‐nalobject.35(AlsotranslatedinDunne2004:361n.2)

Śākyabuddhi’s and Karṇakagomin’s comments on PV 1.216 are alsoquiteclearonthepointthatthecharacterofbeingnon‐belyinginthecaseofacrediblepersonʼsstatementconcerningtranscendentmattersisequaltoorthesameasthispersonʼsstatementconcerningmattersthatcanbechecked.Anditisforthisreasonthatāptavādaservesasan

34For āptavacana as a kāryahetu see Tillemans’ explanation on PV 4.92 (Tillemans

2000:126ff.). 35PVṬ JeD242b3–5/P285b6–286a1:gal tephyirolgyidṅospo lasgra tshadmañidyodpama yinna/ ʼonkyaṅgaṅ ʼdi yid ches tshignimi sluba spyi las (PVSV108,1)tegaṅdaṅgaṅñespazadpaʼitshigdedaṅdenimislubayinte/dpernaʼdusbyasthamscadniskadcigmaźesbyabalasogspaltabuʼo//śintulkogtugyurpaʼidonlañesparzadpaʼitshigʼdiyaṅyinno//debasnaʼdiyaṅmislubayinnoźesbyabadeltanañespazadpaʼitshignimislubaspyilasrjessudpagpañidyinpadeltarna/slobdpongyi(read:gyis)luṅniphyirolgyidonlarjessudpagpañiddugsuṅspayinno//≈PVSVṬ390,15–19:yadibāhyevastuniśabdasyanāstiprāmāṇyam,yattarhīdam āptavādāvisaṃvādasāmānyāt. yo ya āptavādaḥ, so ʼvisaṃvādī, yathākṣaṇikāḥ sarve saṃskārā ityādikaḥ. āptavādaś cāyam atyantaparokṣe ʼpy arthe.tasmādayamapyavisaṃvādītyevamāptavādasyāvisaṃvādasāmānyādavisaṃvā‐ditvādanumānatetyāgamasyabāhyeʼrtheʼnumānatvamuktamācāryadignāgena.

LOGIC IN A REL IG IOUS CONTEXT 99

inferentialmark36onaccountofwhichwecaninferthatourcognitiondoesnotbelie.

“Becausethatcrediblepersonʼsspeechisthesameinitstrustworthiness(avisaṃvādasāmānyāt). That is, just as the credible personʼs speech istrustworthy with regard to an object that can be determined [throughperceptionandordinaryinference],likewiseitistrustworthywithregardtoanextremelyremoteobjectalso,preciselybecauseitisthespeechofacredibleperson.”(Dunne2004:364n.10)ItisthereforethattheteacherDignāgatermedaninferenceacognitionarisingfromaninferentialmarkcharacterizedas[this]personʼscrediblespeechthat[suchacognition]isnotbelyingwithregardtoarealstateofaffairs(donlamisluba...bloni,*arthāvisaṃvādabuddheḥ).37

This interpretation of Śākyabuddhi and Karṇakagomin is in accordwithwhatDharmakīrtihimselfexplainedinPVSV109,8–11whencom‐mentingonPV1.216(seeabovep.89).Thus,Icanseenoharminac‐ceptingthisinterpretationandtranslatingPS2.5abfromtheviewpointofDharmakīrtiasfollows:

Because the characterof beingnon‐belyingof a statementof a crediblepersonis thesame[in thecaseof transcendentobjectsas inthecaseofobjectsthatareaccessibletous,it]isaninference[inthesenseofanin‐ferentialmark].

sāmānyātinPV1.216mightbeunderstoodaccordingly.Itgoeswithoutsaying that for the timebeingsuchanunderstanding is justifiedonly 36Jinendrabuddhi(PSṬms63b5–6,referredtoabove,p.94)alsoexplainsthestatement

ofacrediblepersontobealogicalreason:āptavacanaṃgṛhītvāvisaṃvādādhigama‐hetuṃtasyānumānatvamuktam.

37PVṬJeD245a7–b1/P289b5–7:ñespazadpaʼitshigdenimislubarmtshuṅspaʼiphyirte/ciltarmṅonsumdaṅrjessudpagpasmislubaʼidonyoṅssugcodparnuspalañespazadpaʼitshigmislubadeltarśintulkogtugyurpa(read:pala)yaṅyinte/ñespazadpaʼitshigñidyinpaʼiphyirro//debasnadonlamislubañespazadpaʼitshiggimtshanñidcangyirtagslasbyuṅpaʼiblonirjessudpagpañidduslobdpongyisbrjoddo//≈PVSVṬ393,25–28:tasyāptavādasyāvisaṃvādasāmānyāt.ayathā śakyaparicchede ʼrthe āptavādasyāvisaṃvādaḥ, tathātyantaparokṣe ʼpi, āpta‐vādatvādeva.tataścāptavādalakṣaṇāl liṅgādutpannāyāavisaṃvādabuddheranu‐mānatācāryadignāgenābhihitā.

atasyāptāvādasyāvisaṃvādasāmānyātPVSVṬms141b6(PVṬ):tasyāstāvadasyā‐visaṃvādātsāmānyātPVSVṬ.

100 HELMUT KRASSER forDharmakīrti.WhetherthisinterpretationisfaithfultoDignāgaʼsun‐derstandingstillremainstobeinvestigated.

2.5.PV1.217–asecondinterpretationofPS2.5ab

TheaboveinterpretationprobablyisthebestDharmakīrticouldcomeup with without contradicting Dignāgaʼs wording, although I cannotseeanythinginDignāgaʼsphrasingthatwouldsuggestthatascripturemustbetestedinitsentirety,asproposedbyDharmakīrti.Perhapsbe‐cause this is hardly possible Dharmakīrti provides uswith a second,less demanding interpretation ofDignāgaʼsāgama definition, accord‐ingtowhichaperson’sreliabilitywithregardtotranscendentobjectsisinferredfromhisknowledgeofthemainreligiousteachingsthathepresents,suchastheFourNoble[ʼs]Truths.38

heyopādeyatattvasyasopāyasyaprasiddhitaḥ/pradhānārthāvisaṃvādādanumānaṃparatravā//PV1.217

heyopādeyatadupāyānāṃtadupadiṣṭānāmavaiparītyamavisaṃvādaḥ.ya‐thācatūrṇām āryasatyānāṃvakṣyamāṇanītyā. tasyāsyapuruṣārthopayo‐gino ’bhiyogārhasyāvisaṃvādādviṣayāntare ’pitathātvopagamonavipra‐lambhāya, anuparodhān niṣprayojanavitathābhidhānavaiphalyāc ca vak‐tuḥ.PVSV109,15–19

Or[scripture]isinferencewithregardtotheother[domain]duetoitsbeingnotbelyingwithregardtotheprincipalpoints[i.e.theFourNo‐ble(ʼs)Truths],becausethenature(tattva)ofwhatistobeabandonedand what is to be realized together with their means is wellestablished[byit].

38IconsiderthemethodproposedinPV1.217asanalternativetothethreefoldanaly‐

sis,i.e.,thesecondpossibilityproposedbyDunne(2004:241):“Thistesteithersup‐plementsorperhapsreplaces…thethreefoldanalysis.”YaitaandTillemansconsi‐deritdifferently:“Onthebasisofcredibilityofāgamaexaminedlikethat,Dharma‐kīrti proves the non‐disagreement of theāgama evenwith regard to inaccessiblethingtalkedofbytheauthor,inthefollowingtwoways(v.216–7)”(Yaita1987:3);“(e)v.217elaboratesuponaspectsofv.216:whenthescriptureisnon‐belyingonimportant rationally accessible things it should also be so on the inaccessiblethings”(Tillemans1999a:400=1999b:42).

LOGIC IN A REL IG IOUS CONTEXT 101

Reliabilityconsists inthefactthatwhatistobeobtainedandwhatistobe avoided [togetherwith] theirmeans,which has been taught by that[credibleperson],correspondstoreality(avaiparītya); liketheFourNo‐ble(ʼs)Truthsinthewayitwillbeexplained[inthePramāṇasiddhi‐chap‐ter]. Because that very (tasyāsya) [thing that has been taught], whichservesthehumangoal,[andhence]issuitabletobeingpracticed,isrelia‐ble,theassumptionthatthisissoalsointhecaseoftheother,transcend‐entrealmmaynotleadtooneʼsdeception.[Andthisisfortworeasons:]1)becausethereisnocounterevidence(anuparodha),and2)becauseitispointlessforaspeakertomakefalsestatementswithoutapurpose.

2.6.ConclusionofthisexcursusonPS2.5ab

WhatwehaveseensofarcertainlydoesnotsubstantiatemyclaimthatforDharmakīrtiscriptureisnotameansofvalidcognition.Onthecon‐trary,welearnthatthestatementsofacrediblepersonandinferencesbasedonascripturethathaseitherbeenwellcheckedorexpoundsthemainreligiousteachings,arereliableevenwithregardtorealstatesofaffairsandnotjustthespeaker’sintention.Thus–toinvertacommonsaying–whatyougetiswhathehasseen.

However,beforerefutingotherwaysofaccountingforthereliabilityof credible persons offered by opposing schools (PV 1.218ff.), Dhar‐makīrti concludes his interpretation of Dignāgaʼs de inition with thefollowingwords:

tadetadagatyobhayathāpyanumānatvamāgamasyopavarṇitam–varamāgamātpravṛttāvevaṃpravṛttiriti.nakhalvevamanumānamanapāyam,anāntarīyakatvādartheṣuśabdānāmitiniveditametat.PVSV109,20–22Thus, thisscripturehasbeenexplainedinbothwaystobe inferenceforwantof[anyother]possibility(agatyā),[havinginmind:]“Giventhatonehastoproceedonaccountofscriptureitisstillbetter(varam)toproceedinsuchaway.”However, inference insuchaway indeed isnotwithoutproblems (na ... anapāyam), for words are not invariably concomitantwith[their]objects.This[wealready]haveexplained[inPV1.213].

Herewelearn,ifmyinterpretationofagatyā iscorrect,thatscripturecan only be said to beanumāna (in the sense of an inferentialmark[liṅga]orlogicalreason[hetu]),becausethereisnobetterwaytode‐fineitandbecauseitisstillbettertoproceedonthebasisofachecked

102 HELMUT KRASSER or“purified”āgamathanjustblindfaith,e.g.,intheVeda.But,inwhich‐everwaythisdefinitionistobeunderstood,toinferthereliabilityofacredible personʼs statement with regard to transcendent matters isproblematic. Thus, although Dignāgaʼs de inition is in contradictionwith Dharmakīrtiʼs theory thatwords are not connected to their ob‐jects, Dignāga nevertheless did the best he could under the circum‐stances that we require scripture to live our lives and, thus, ideallyshould rely on one that is optimally reliable. What kind of problemDharmakīrti had inmind,whenhe saysna ...anapāyam, he does notsharewithushere.

3.OTHERPV(SV)PASSAGESONĀGAMAAlas, it is onlymuch later that Dharmakīrti tells us specificallywhatkindofproblemhehadinmind.Thiskindof inferencewhereonein‐ferscorrectnesswithregardtotranscendentmattersfromthecorrect‐nessofstatementsinonepartisafallaciousinferencecalledaśeṣavad‐anumāna.Moreover,weknowthatDignāgawasawareof this typeoffallacy, because inhisPramāṇasamuccaya he refutes the śeṣavadanu‐mānaoftheNaiyāyikas.Dharmakīrtirefers,withoutactuallysayingso,toPS2.28bwithVṛttiwhereDignāgarefutestheNaiyāyikasusingthesameexamplesasDharmakīrti.Themaindifferencebetweenthesetwokindsof śeṣavadanumāna applied by theopponent, in this case aMī‐māṃsaka,andbytheBuddhist isthattheBuddhistscriptureis“puri‐fied”byanexaminationandthusitmightsometimesbereliable.Thus,inthiscasetheprobabilitythatitistrueisgreater,butthereisnocer‐tainty.

But another [theorist], renouncing the definition of scripture as [thatwhichis]authorless[becauseofasimilarcriticism,and]wishingtoprovetheauthorityoftheVedainanotherway,says,“[Those]Vedicsentencesforwhich there isnocognition [on thepartofBuddhistsasbeing true]are[in fact] true,because theyarepartof the[same]Veda, just like thesentence‘Fireistheremedyforcold,ʼetc.”...... This kind of inferencewasdeclared to benot [really] a proof by themaster [Dignāga] himself in pointing out the deviating character of the

LOGIC IN A REL IG IOUS CONTEXT 103

Naiyāyikasʼśeṣavat‐inference[inPS(V)2.28b],39likethe[so‐called]proofthatfruit[onehasnottasted]hasthesametaste[asfruitonehastasted]becauseithasthesamecolour,andthe[so‐called]proofthatricegrainsonehasnotobservedarecooked,likethosewhichonehasobserved,be‐causetheyareinone[andthesame]vessel.Andthemannerinwhichthis[śeṣavat‐inference]isnotaproofhas[already]beenstatedearlier[inPV1.14with PVSV]. And [true,]we have stated this definition of scripture[too].However,this[isjustifiedonly]if,foreveryobjectcapableofbeingexamined, there iscorrectness(viśuddhi)ofpositiveandnegativeasser‐tionsbyappropriatemeansof valid cognition. [And]even if there isnonecessary relationbetweenwords and [their]meanings [,whichwouldensurethevalidityofscripture],itisbetterthata[person]actin[astateof] doubt [when it comes tomatters relating toworldly prosperity andsalvation];40for[scripture]mayoccasionallybereliableinthiscase.41

WhytherecanbenocertaintyDharmakīrtihasalreadystatedbefore,wherehealsoreferredbacktothedefinitionofscriptureIhavedealtwithatthebeginningofthispaperandtohisremarkthatsuchanin‐ferenceisnotwithoutproblems.

39PSVK115b2–4: lhag ldan layaṅdpebyedna//(=PS2.28b:śeṣavatyapicedvatiḥ)lhagmadaṅldanpalayaṅgaltedpebyedpayinna/lhagmṅonsumdaṅmtshuṅspaʼiyulgaṅlayodpadeʼiśespanilhagmadaṅʼdrabaʼoźesbyabaʼdiyaṅʼkhrulpaʼiphyirtshadmamayinte/gzugsmtshuṅspasrolasogspagdonmizabarmtshuṅsparʼgyurbanimayinno//deltarnalhagmadaṅldanpalayaṅdpermirigspayinno//.Seeabove,p.62andn.126.

40Theideathatoneactsinregardtotranscendentobjectsfromastateofdoubtoncethe scripture is purified is also propounded by Śāntarakṣita and Kamalaśīla; seeMcClintock2010:322–324.

41PVSV 173,16–174,1:anyas tvapauruṣeyam āgamalakṣaṇaṃparityajyānyathāprā‐māṇyaṃvedasyasādhayitukāmaḥprāha–avitathānivedavākyāniyatrāpratipattiḥ,vedaikadeśatvāt, yathāgnir himasya bheṣajam ityādi vākyam iti. ... svayam īdṛśamācāryeṇānumānaṃ naiyāyikaśeṣavadanumānavyabhicāram udbhāvayatā tulyarū‐patayāphalānāṃtulyarasasādhanavadekasthālyantargamāddṛṣṭavadadṛṣṭataṇḍu‐lapākasādhanavac cāsādhanam uktam. tadasādhanatvanyāyaś ca pūrvam evoktaḥ.uktaṃ cedam āgamalakṣaṇam asmābhiḥ. tat tu sarvasya śakyavicārasya viṣayasyayathāsvaṃpramāṇenavidhipratiṣedhaviśuddhaunāntarīyakatvābhāveʼpiśabdānāmartheṣuvaraṃsaṃśayitasyavṛttiḥ,tatrakadācidavisaṃvādasambhavāt…Onthispassage,seeabove,pp.62–64.

104 HELMUT KRASSER Objection: Isnʼt it the case that such a thing as the arrangement of theworld,eventhoughitisnotanobject[accessible]toreason,isknown[byyouBuddhists] from thestatementofapersonwhichmustbeassumed[tobetrue,onthebasisofhisreliabilityinregardtootherthings]?[An‐swer:]No,because [wehave]no confidence [in suchaperson]. It is in‐deednotthecasethat,since[apersonhasbeenobserved]nottoerrwithrespecttoacertain[matter],everything[thatpersonsays]islikethat[i.e.true, and this for tworeasons: first], becauseoneobserves that [peoplewhoareknowntobereliable inregardtoacertainthingdo in fact]err[in regard to other things]; and [second,] because a concomitance be‐tween the [verbal] activityof this [allegedly superiorperson]and relia‐bility is not established. Beyond that, the [aforementioned] definition ofscripturehasbeenaccepted for lackof [anyother]recourse.There is [in‐deed]noascertainment [of supersensible things] from [scripture thusde‐fined,and]thisisthereasonwhy[wehave]alsostated[above]thatscrip‐tureisnotameansofvalidcognition.42

Theseinferencesaremainlyuncertain,becauseonecannotestablishapervasionbetweenscriptureanditsreliability,andonlysuchaperva‐sion could guarantee the certainty of the inference. Therefore, Dhar‐makīrtirepeatsthatthisdefinitionwasassumedonlyinwantofabet‐terpossibilityandthatthereisnocertaintyfromscripture.Andthisisthereasonwhyhealsosaidthatāgamaisnotapramāṇa.

InthelastchapterofhisPVDharmakīrtireferstothesepassages.prāmāṇyamāgamānāṃcaprāgevavinivāritam/PV4.101abNow,ithadalreadybeenrefutedearlierthatscriptureswerepramāṇas.(Tillemans2000:141)

Sofarwehaveseenthatbyreadingbitsandpiecesfromdifferentdis‐cussionsofāgamainthePVSVtogethertheimpressionthatwegainedfromDharmakīrtiʼs interpretationofDignāgaʼsde initionofscripture,namelythatitisarealpramāṇawithregardtoexternalobjectssuchas 42PVSV167,23–168,2:nanukaścil lokasanniveśādirayuktiviṣayo ʼpisambhāvanīyapu‐ruṣavacanādarthaḥpratipadyate.na,apratyayāt.nahikvacidaskhalita iti sarvaṃtathā, vyabhicāradarśanāt, tatpravṛtter avisaṃvādena vyāptyasiddheś ca. agatyācedamāgamalakṣaṇamiṣṭam.nātoniścayaḥ.tannapramāṇamāgamaityapyuktam.Onthispassage,seealsoabove,pp.42–44.ThepassagebeginningwithnahikvacitisalsotranslatedinTillemans1999a:400;seealsoKataoka2011b:256n.10.

LOGIC IN A REL IG IOUS CONTEXT 105

heaven,hasnowchangedandithasbecomeclearthatāgama isnotapramāṇaatall.Itisonlyundercertaincircumstances,namelywhenithasbeencompletelycheckedandpurified,thattheremightbeachanceofproceedingsuccessfullybasedonāgama.Orelse,whenanāgamaorāptavāda expounds themain teachings, aswehave seen in PV1.217withPVSV.WhatDharmakīrtiformulatesinPV1.217andhiscommen‐tary thereon serves as a leitmotif for his second chapter entitledPramāṇasiddhi,whichisentirelydevotedtotheproofthattheBuddhaispramāṇa.43Butinordertoavoidthemistakeofemployingaśeṣavad‐anumāna, the Buddha is proved to be pramāṇa onlymetaphorically.Justasapramāṇaquacognitionisnotbelying(pramāṇamavisaṃvādijñānamPV2.1ab1)andmakesknownahithertounknownobject(ajñā‐tārthaprakāśovāPV2.5c), inthesamewaytheBuddhaisnotbelyingwithregardtoandmakesknowntheFourNoble(ʼs)Truths,whichhi‐thertowereunknowntous.44 43PV2.32(cf.PV1.217):heyopādeyatattvasyasābhyupāyasya(:hānyupāyasyaMiyasa‐

kaed.)vedakaḥ/yaḥpramāṇamasāviṣṭonatusarvasyavedakaḥ//.“Hewhomakesknownthenatureofwhatistobeobtainedandwhatistobeavoided,togetherwiththeirmeans,isassumedtobeapramāṇa,butnothewhomakeseverythingknown.”PV 2.145–146ab: tāyaḥ svadṛṣṭamārgoktir vaiphalyād vakti nānṛtam/ dayālutvātparārthañcasarvārambhābhiyogataḥ//tataḥpramāṇam,tāyovācatuḥsatyaprakā‐śanam/.“‘Protection’[i.e.,causingpeopletocross,]isstatingthepathseenbyhim‐self.Hedoesnottellalie,(1)becauseitispointless,(2)becauseheiscompassion‐ate, and (3) because he perseveres in all his undertakings for the sake of others.Therefore [he is] a pramāṇa. Or, protection is illuminating the four Noble(’s)Truths.”Ontheseone‐and‐a‐halfverses,see,e.g.,Kataoka2011b:258–259withn.20.

44SeeKrasser2001.ItgoeswithoutsayingthatIdonotagreewithFrancowhocon‐cludeshisverylearnedchapter“TheFrameworkandProof‐StrategyofthePramā‐ṇasiddhi‐Chapter”bycomparingDharmakīrtiʼsstrategywiththatofPakṣilasvāminVātsyāyana.Dharmakīrtiwaswell aware of the fact thatDignāgaʼswayof under‐standingāgamawassimilartothatofPakṣilasvāminandthatbothoperatedwithafalseśeṣavadanumāna,amistakethatDharmakīrtiavoided.Franco(1997:39–40):“Toconclude,Ihavetriedtoshowthattherearestrongresemblancesandparallel‐ismsbetweenDharmakīrtiʼsproofof thevalidityof theBuddhistāgama,which isbasedontheBuddhaʼsbeingpramāṇabhūta,whichisbasedinitsturnontheesta‐blishment of the four noble truths, and Vātsyāyanaʼs proof of the validity of theVeda,whichisbasedonitscompositionbyāptas,whosestatusisbasedinitsturnon the effectiveness of the statements of the Āyurveda, themantras, etc. IwouldevengosofarastosaythatalthoughstructurallythePramāṇasiddhi‐chapteriscon‐struedasacommentaryonthefiveepithetsoftheBuddha,theinterpretationofthe

106 HELMUT KRASSER 4.MOREONPVSV108,1–109,22

Thesetwoaccountsofāgama,accordingtowhich,ontheonehand,itcannotbecountedasapramāṇaatallandaccordingtowhich,ontheother,ityieldscorrectknowledgeevenofexternalobjects,seemcom‐pletely incompatible tome.Yetwe find them in theworkof oneandthesameauthor.Thus,towhichviewdidhereallysubscribe?

Thedepictionof scriptureashaving thespecial statusasan infer‐ence,butaninferencethatisnotwithoutproblems,leadsDunnetoaninterestingdescriptionofDharmakīrtiʼsapproachtoit.

Asignificantfeatureofthisargumentisthetensionthatitevidencesbe‐tweenanappealtooneʼsownempiricalknowledgeasopposedtotrust‐ing in anotherʼs (i.e., the Buddhaʼs) transempirical knowledge.... Never‐theless, despite this claimʼs empiricist tones, Dharmakīrti is apparentlytroubled by his own appeal to the Buddhaʼs transempirical knowledgeandconcomitantextraordinaryqualities.Howelsecanweexplainthefactthat he immediately backpedals by denying that scriptural inference isreallyaninferenceatall(6)?Moreover,hethenproceedstorejectexplic‐itlyanyappealtocredibility(7).(Dunne2004:241f.)

Tillemans, on the other hand,who iswell aware of the passages ad‐duced above in §3, considers scripture as depicted in the PV(SV)108,2–109,19 to be an “exception.” In his summary of PV 1.216 hewrites,

(d) v. 216: Dignāgaʼs point in saying that authoritative words were aninference was that when an authorityʼs words (=scripture) have beenfoundtobenon‐belyingonrationallydecidablematters,thenwearejus‐tified to understand radically inaccessiblematters based on that scrip‐ture.Inthisparticularcase,contrarytowhatwassaidinv.213,wedoin‐fersomethingmorethanjustthespeakerʼsintentionfromhiswords:wealso infer that thestateofaffairsobtains.Thisone ‘exception’ tov.213mustbeallowedbecauseotherwisetherewouldbenoway(agatyā)forustocometoknowradicallyinaccessiblethings.(Tillemans1999a:399–400=1999b:41–42)

epithets and the general strategy employed to prove the validity of the BuddhistteachingsowemoretoVātsyāyanathantoDignāga.”OnPakṣilasvāmin,seebelow,p.110n.49.

LOGIC IN A REL IG IOUS CONTEXT 107

Andinhisintroductiontothesamepaper,whichsummarizesDharma‐kīrtiʼspositiononāgama,weread,

ItiswellknownthatDharmakīrticommenteduponthephraseāptavādā‐visaṃvādasāmānyād anumānatā in Dignāga as showing that quotationsfromauthoritiescouldbeusedtoprovecertainpropositionsinferentially.Inparticular,useofscriptureorscripturaltradition(āgama)wassuppos‐edlynotaseparatemeansofvalidcognition(pramāṇa),butwassaidtobeaninferencebecausesimilartootherinferencesinreliablyrepresent‐ingitsobject,i.e.in‘notbelying’(avisaṃvāda).Thisthenisāgamāśritānu‐māna,‘scripturallybasedinference’.Ofcourse,notjustanyscripturecanbeusedasabasisforsuchaninference,norcansuchinferencespertaintoanyandallobjects.Themaincriterionseparatingreliablefrombogusscripturesisthethreefoldanalysiswherebyitisdeterminedthat...

There follows a description of the threefold analysis. ThenTillemansgoesontosay,

...Now,ifthescripturepassesthistripletest,itisfittobeusedin‘scrip‐turallybasedinferences’,butwiththeall‐importantstipulationthatsuchinferencesareonlytobeusedinthecaseofradicallyinaccessiblematters(atyantaparokṣa), ones which are not in the domain of observation orobjectivereasoning,butareonlyaccessibleoncewehaveaccepted(abhy‐upagata) scripture. In short,āgamāśritānumānaworkswhere objectiveinference and observation leave off. ... This interpretation of Dhar‐makīrtiʼs account of scripturally based inferences, whichwe shall termfor short ‘inference‐like‐any‐other’, isnot justahypotheticalpossibility.(Tillemans1999a:395–396;notreprintedin1999b)

ThequestionsIwanttoaddresshereare: WasDharmakīrti“troubledbyhisownappealtotheBuddhaʼstrans‐

empiricalknowledge”? Does Dharmakīrti hold the position that this “one ‘exceptionʼ to v.

213 must be allowed because otherwise there would be no way(agatyā)forustocometoknowradicallyinaccessiblethings”?

108 HELMUT KRASSER Is this “inference‐like‐any‐other”45 more than “just a hypothetical

possibility”andcanit“beusedtoprovecertainpropositionsinferen‐tially”?

Letmebeginwith the secondquestionby examining themeaning ofagatyā.

4.1.Onagatyā

agatyāinthecontextoftheāgamadiscussionhasbeenusedbyDhar‐makīrtithreetimes:onceinPV1.216,onceinPVSV109,19,andonceinPVSV168,1. Ithasbeen translatedvariouslyas follows (thepassagesaregivenapproximatelyinchronologicalorder):46PV1.216(seeabovep.88)

“for[otherwise]therewouldbenoway[toknowsuchobjects]”(Til‐lemans 1986: 32, 1990: I.25, 1993: 11; and similarly Dunne 2004:364)

“for there is no [other] possibility (agatyā) [to explain it]” (Yaita1987:8)

“since there is no [other] possibility [but to also regard the latterkindofstatementasaninference...]”(vanBijlert1989:124)

“[...mais cela,Dignāgane lʼaditquʼ]en raisonde lʼimpossibilité [oùlʼonsetrouvesinon]dʼaccéder[auxobjetsradicalementimpercepti‐bles]”(Eltschinger2007a:224)

PVSV109,19: tadetadagatyobhayathāpyanumānatvam āgamasyopa‐varṇitam,varamāgamātpravṛttāvevaṃpravṛttiriti

“Thus[thefactthat]āgamaisanumānahasbeendescribedin[differ‐ent]twoways[inv.216&217],owingtoimpossibility[ofexplaininginanyothermannertheāgamaʼsauthoritativeness,i.e.byconsider‐ingourordinaryconception:]‘When[someact]isgoingtobedone,it

45Ifoneacceptstheaccountofsāmānyātasproposedabovein§2.4, thenTillemansʼ

expression “inference‐like‐any‐other” for this kind of inference might not be thebestchoice.

46SeealsoMcClintock2010:259.

LOGIC IN A REL IG IOUS CONTEXT 109

iscertainly(evam)bettertoactonthebasisofāgamaʼ.”(Yaita1987:10)

“Thisfactthatscriptureisaninferenceisassertedinbothcases[i.e.invv.216and217]becauseofthelackofany[other]way.Ifoneen‐gagesoneselfonthebasisofscripture, it isbettertoengageoneselfinthisfashion[onthebasisofacorrectscriptureratherthanonthebasisofonewhichbelies].”(Tillemans1999a:400)

“[Maisdechacune]desdeuxmanières[dontonlʼafait,cenʼestquʼ]enraisondelʼimpossiblitédʼaccéder[sinonauxobjetsradicalementim‐perceptibles que nous avons] exposé le caractère inférentiel delʼEcriture,[jugeant]quetantquʼàagiràpartirdʼuneEcriture,mieuxvautagirainsi.”(Eltschinger2007a:227)

“Thus, in both cases, that the scripture is a valid inference is ex‐plainedbecausethereisno[other]way.”(Kataoka2011b:256)47

PVSV168,1–2:agatyācedamāgamalakṣaṇamiṣṭam “Now,weaccept thisdefining characterof scripture for lackof any

[other]way.”(Tillemans1999a:400) “Andthisdefinitionofscripture[asaninference]isaccepted[byus]

becausethereisnootherway.”(Kataoka2011b:256n.10)OfthesepossibilitiesIprefertheoneaccordingtowhichthisdefinitionwasgivenoracceptedforlackofanyotherway.Iunderstandagatyāasreferring back to the initial explanation I proposed in §2.1 of whyDignāga gave a definition at all, namely, because otherwise activitywith regard to transcendentmatters is notpossible. Thus, hehadnoother choice than to give this definition, knowing that, though it isfaulty,itisthebestdefinitionavailable.Buthegave,ofcourse,thebestone that is possible. Thus, I prefer agatyā to be read with the pastparticiples in these phrases: agatyābhihitā (k. 216), agatyā ... upa‐

47KataokaisoftheopinionthatagatyāshouldexcludesuchpossibilitiesastheBud‐

dhaʼsbeingaliarorhisteachinghavingnoaimandsoforth,byrelatingittotheMī‐māṃsakaproofofManuʼsreliability.

110 HELMUT KRASSER varṇitam(PVSV109,19–20),andagatyā ...iṣṭam(PVSV168,1–2).Thisisalsotheapproachtakenbythecommentators.48

Ifweunderstandagatyāas“for[otherwise]therewouldbenoway[toknowsuchobjects]”or somethingsimilar, thenwewouldhave tosupplythephrase“knowledgeoftranscendentmatters,”which,thoughimpliedinhisexplanationofk.216(arthāvisaṃvādādanumānamPVSV109,11),Dharmakīrtitriestoavoid.Byintroducingagatyā,understoodin thisway,Dharmakīrti isbetterable todown‐playtheobviouscon‐tradictionbetweenhis attitude towards scriptureandDignāgaʼsde i‐nitionandtoarriveatthe“solution”thatitisstillbettertoacttowardstranscendent matters on account of doubt, for sometimes it mightwork(tatrakadācidavisaṃvādasambhavāt,cf.aboven.41).Moreover,inthiswayDharmakīrticanavoidcommittingaśeṣavadanumānahim‐self.49

IalsodonotfollowKataokaʼshypothesis,thoughitistempting,thatagatyā excludes suchpossibilities as theBuddhaʼs being a liar or histeaching having no aim and so forth. Here, the context is to explainaway as much as possible the contradiction between Dignāga andDharmakīrti,notatalltoprovethattheBuddhaisapramāṇa.

Returningtothequestionunderconsideration,namelywhetherthis“one ‘exceptionʼ to v. 213 must be allowed because otherwise therewouldbenoway(agatyā)forustocometoknowradicallyinaccessiblethings”(Tillemans),IcannotseethatscriptureasdefinedbyDignāgainthewaywehaveseen,sothatitamountstoaśeṣavadanumāna,would 48See, e.g., Śākyabuddhi and Karṇakagomin on PV 1.216. PVṬ Je D245b2/P289b8–

290a1:deyaṅgoskabsmeddernampagźangyisśintulkogtugyurpala ʼjugpamedpaʼiphyirro//aʼjugpayodnadeltarʼjugpaʼoa[D:’jugparbya’oP]//≈PVSVṬ393,30–394,8: sā cāgatyābhihitānyenaprakāreṇātyantaparokṣepravṛttyasambha‐vāt,asatyāṃpravṛttauvaramevaṃpravṛttaitia.

acf.PVSV108,5–6:satipravartitavyevaramevaṃpravṛttaiti,andPVSV109,20–21:varaṃpravṛttāvāgamādevaṃpravṛttiriti.

49Thiskindofśeṣavadanumānawe find impliedbyPakṣilasvāmin.NBh97,2onNSū2.1.68: teṣāṃ khalu vai prāṇabhṛtāṃ svayam anavabudhyamānānāṃ nānyad upa‐deśād avabodhakāraṇam asti. “Now, when these [ordinary] living beings do notknowforthemselves[throughtheirownpowerofcognitionthosethingsthatshouldbeavoidedorobtained],thenthereisnoothercauseof[correct]knowledge[leftforthem] than the [valid] instruction [givenby the expert speaker].”TranslationvanBijlert1989:161.

LOGIC IN A REL IG IOUS CONTEXT 111

constitute an exception to verbal knowledge for Dharmakīrti inguaranteeing knowledge of or access to external objects such asheaven.50 It isonlywhenitcomestoreligiouspracticethat,actingonthebasisofapurifiedscripture,theprobabilityofsuccessishigher,butthereisnocertaintyatall.

Connectedtothisproblemisalsothequestionwhethersuchanin‐ferencebasedonscriptureismorethan“justahypotheticalpossibility”and whether it can “be used to prove certain propositions inferen‐tially.”

4.2.Onāgamāpekṣānumāna,āgamāśraya‐anumāna

FromthedescriptionbyTillemanscitedabove(pp.106–107)welearnthat“ifthescripturepassesthistripletest,itisfittobeusedin‘scrip‐turallybased inferences,ʼ” also calledāgamāśritānumāna. If I amcor‐rect inassumingthatwhatTillemansmeansbyāgamāśritānumāna iswhatDharmakīrti callsāgamāpekṣānumāna (PVSV109,1),āgamāśra‐yamanumānam (NB3.114,PVin3128,4),51orāgamāpekṣamanumā‐nam(PV4.48=PVin3.9;seeTillemans2000:78),52thenImustconfessIhavenotbeenabletofindanypassageinDharmakīrti,wherehecallsan inference based on a purified scripture that has passed the tripletest a “scripturally based inference.” In PV 1 and PVSV Dharmakīrtiusesscripturallybasedinferencesonlyinthecontextoftestingascrip‐ 50AsnicelyshownbyTillemans(1993:Intro.§§2–4,1999a:396–399),laterTibetans

suchasTsoṅkhapaorŚākyamchogldantooksuchaninferencebasedonscripturetobeafull‐fledgedprobativeinferenceinwhichthehetufulfilsthetrairūpyacondi‐tion. See, e.g., Śākyamchog ldanʼs account in the translationofTillemans (1999a:398n.5):“Asforprovingthe[reasonʼsthree]characteristics,thepakṣadharma(tā)isestablishedbyperception,forthisreasonispresentedtoanopponentwhoper‐ceptually observes the statement, ‘From giving comes wealth and frommorality,happinessʼ [i.e. he sees that the statement is indeed present in the text of theRatnāvalī].As for theproof of thepervasion (vyāpti), there are the following two[subsections]:provinganvayaandprovingvyatireka.Wenowtakeupthefirst[i.e.anvaya]. Take as the subject the Ratnāvalī; it is non‐belying with regard to thepropositionswhichitteaches,becauseitisascripture[judged]immaculatethroughthethree[kinds]ofanalysis.”ForIndianforerunners,seeEltschinger2007a:107.

51ForthetextandatranslationofNB3.114,seeTillemans2000:95n.336. 52InPVin2.42thiskindofinferenceisalsoreferredtoasāgamārthāśrayāyuktiḥ,see

above,p.55n.97.

112 HELMUT KRASSER ture.Hedoessowhenfirstexplainingthetripletest inhiscommentsonPV1.215.

āgamāpekṣānumāne ʼpi, yathā rāgādirūpaṃ tatprabhavaṃ cādharmamabhyupagamyatatprahāṇāyasnānāgnihotrāderanupadeśaḥ.PVSV109,1–3Alsointhecaseofascripturallybasedinference53[thereshouldbenoin‐validation(abādhana).Non‐invalidationis]asfollows(yathā):Havingac‐cepted(abhyupagamya)thatdemerit(adharma)consistsin[defilements]suchasdesire(rāgādirūpa)and[theactions(karman)]bornofthem(tat‐

53ThetextofKarṇakagomin(PVSVṬ393,13–17)explainingthelocative(āgamāpekṣā‐numāne ʼpi)seemstobeproblematic:yathātmādīnām(/)ādiśabdātpradhāneśva‐rādiparigrahaḥ/nahyeṣāṃkiñcilliṅgamastiyenānumeyāḥsyuḥ/etadapipratipā‐dayiṣyati/viśuddheviṣayadvaye( ) ʼtyantaparokṣecāgamaviṣayepaurvāparyaviro‐dhenayasmin cintāṃpravartayati tasminn āgamāpekṣamanumānamapi/abā‐dhanam itiprakṛtam/.ātman, etc., inātmādīnām (PVSV109,1) constitute exam‐plesofentitiesthatshouldnotbetaughttobe inferable inatreatise.Bythewordādi,accordingtoKarṇakagomin,primordialmatter(pradhāna)oftheSāṅkhyasandacreatorgod(īśvara)areincluded,becausethereisnoinferentialmarkonthebasisofwhichtheycouldbeinferred.Theproblemnowis,whetherviśuddheviṣayadvayeshouldbereadwithetadapipratipādayiṣyatiorwiththenextpartbeginningwithatyantaparokṣe.Theeditorindicatedviśuddheviṣayadvayeasanunidentifiedquota‐tionandreaditwithatyantaparokṣe.Inthatcaseviśuddheviṣayadvayehastobeun‐derstoodtogetherwithatyantaparokṣeastheobjectoftheinvestigationbymeansofthescripturallybasedinference,aswasdonebyDunne(2004:363n.9).Dunne,whoalsoprovidesatranslationofŚākyabuddhiʼsexplanation,realizedthatthiscon‐stitutes a problem and thus introduced his translation with the statement thatKarṇakagomin“offersa somewhat lesshelpful comment.”Thus, Ipropose to readviśuddheviṣayadvayewithetadapipratipādayiṣyatiandtocorrect thetexttoetadapipratipādayiṣyati–viśuddheviṣayadvaya iti (or: ityādinā).viśuddheviṣayadvayewouldthenbeareferencetoPV4.50whereDharmakīrtiwillexplainthatonecanadducepassages froma scriptureas inferentialmarksonlyafteronehaspurifiedthatscriptureintermsofperceptibleandempiricallyinferableobjectsaccordingtothemethod explained in the comments on PV 1.214, but not before – this is themeaning ofviśuddheviṣayadvaye, etc. (seeTillemans2000:80).And as the oppo‐nentscanproveātman,pradhāna,etc.,onlywiththehelpofinferentialmarks(liṅ‐ga)thataretaughtintheirtreatises,theyhaveno liṅgasavailableaslongastheirtreatiseshavenotpassedthetest.Thetreatises,however,willnotpassthetestastheyteachātman,pradhāna,etc.,tobeinferablewithouttherebeingaliṅgaavaila‐ble.Thus,DharmakīrtiinPV4.50willexplainthatthereisnoliṅgawhatsoever,onaccountofwhichātman,etc.,couldbeinferred(nahyeṣāṃkiñcilliṅgamastiyenā‐numeyāḥsyuḥ.etadapipratipādayiṣyati).

LOGIC IN A REL IG IOUS CONTEXT 113

prabhava),54 [a treatise] cannot prescribe [practices] such as the Agni‐hotraandablutions(snāna)asmeansofeliminatingit.

Later,thisabhyupagamaispresupposedinthecommentsonPV1.333(intheversetheformulationāgamāpekṣenānumānenaisused).55

anyadapipratyakṣānumānābhyāṃ aprasiddhaviparyayam āgamāśrayeṇacānumānenabādhitamagnihotrādeḥpāpaśodhanasāmarthyādikam.PVSV174,21–2356a prasiddha em. (PVSVṬ 612,7) : prasiddhi PVSV : pratisiddha PVSVms39b8.[TheVedastates]yetother[things]whicharecontrarytowhatisestab‐lishedbyperception and inference andwhich are negatedby inferencebasedonscripture,suchasthecapacityoftheAgnihotraand[ablutions]topurifyoneofsin,etc.57

Thus,ifIhavenotoverlookedapassageinwhichDharmakīrtiusestheterms āgamāśritānumāna, āgamāpekṣānumāna, and the like for aninferencethatisbasedonapurifiedscripture,āgamāpekṣānumanaforDharmakīrti isnotthesameastheinferencethatiscalledbylaterTi‐betans“inferencebasedonauthority”(yidchespaʼirjesdpag,seeTille‐mans1993:12).

Comingbacktothesecondandthirdquestionsposedabove(seepp.107–108),Isinferencebasedonapurifiedscripturemorethan“justahypothetical possibility,” as Tillemans alleges, and can it “be used toprovecertainpropositionsinferentially”?Iwouldanswer“no”toboth.For Dharmakīrti such an inference is a śeṣavadanumāna, and thus Icannotseehowitshouldbemorethan“justahypotheticalpossibility.”If such inferences nevertheless are used as in the Tibetan tradition,thenthe“certainpropositions”thatcanbeinferredbasedonapurifiedscripturewouldbeanyofthepropositionsofthatscriptureregardingtranscendentmatters. Once the scripture has been established to be 54Cf. PVSVṬ 393,19: tatprabhavaṃ rāgādisamutthāpitaṃ kāyavākkarma cādhar‐mamabhyupagamya.

55ThisabhyupagamaisalsopresupposedinPV4.107;seeTillemans2000:150–152. 56ForŚākyabuddhiʼsexplanation,seeabove,pp.69–70n.142. 57Thisāgama isattributedbyKamalaśīla toKapila, the founderof theSāṅkhya,and

others;seeMcClintock2010:319–320.

114 HELMUT KRASSER purifiedbythetripletest,thenanystatementofthatscriptureregard‐ingtranscendentmattershastobenon‐belying(avisaṃvāda),becauseitisastatementofthatscripture.Thelogicalreasonisasvabhāvahetu,asintheinferencesofŚākyabuddhireferredtoabove(p.99),atleastinthevariouspassagesdiscussedbyTillemans.

Nowwe can consider the first questionposed above (see p. 107).WasDharmakīrti “troubledbyhisownappeal to theBuddhaʼs trans‐empiricalknowledge”?Aswehaveseensofar,DharmakīrtihasaclearconceptofāgamaandofDignāgaʼsde inition:āgamaisnotapramāṇaandtherecanbenocertaintyfromitsstatementsregardingtranscen‐dentmatters;Dignāgaʼsde initionisfaulty,constitutingaśeṣavadanu‐māna.Thus,IthinkDharmakīrtiwasnottroubledbyhisownappealtothe Buddhaʼs transcendent knowledge, but he was troubled by Ku‐mārila. For Dignāgaʼs de inition came under ire in the Ślokavārttika,whereKumārilainthebroadercontextofhisrefutationofomnisciencehasasection(ŚVcodanā121–132),whichistitled“Denialofekadeśa‐saṃvādānumāna” by Kataoka (2011a: 348). Already the introductoryone‐and‐a‐half versesmake it clear that the argument at stake is thesameastheonepropoundedbyDignāgainhisdefinitionofāgama inPS2.5ab.

Even amanwho (yo ʼpi), after having seen [i.e., ascertained] (dṛṣṭvā) apersontobereliable(satyavāditām)withregardto[perceptible]objectsthathave connectionswith the sense‐faculties and soon (indriyādisam‐bandhaviṣaye), supposes (kalpayet) [that a statement of that personmust]also(api)[betrue]withregardtoan[imperceptibleandreligious]matter thatonecanonlybelieve (śraddheye ʼrthe), because it isa state‐ment of the sameperson (tadvacanatvena), he too (tenāpi)wouldhaveproved (sādhitā syāt) validity (pramāṇatā) throughdependence [on so‐methingexternal](pāratantryeṇa).58(Kataoka2011a:348–354)

AlthoughKumārilareferstoobjectsthatareconnectedwiththesense‐faculties (indriyādisambandhaviṣaye), I think there is little risk in in‐terpretingthisasareferencetotheobjectsofperceptionandempiricalinference. This is at least the way Kumārilaʼs commentators under‐ 58ŚVcodanā121–122ab:yo ʼpīndriyārthasambandhaviṣayesatyavāditām/dṛṣṭvātad‐vacanatvenaśraddheyeʼrtheʼpikalpayet//tenāpipāratantryeṇabādhitāsyātpramā‐ṇatā/.

LOGIC IN A REL IG IOUS CONTEXT 115

stand it, alluding to the Buddhaʼs teaching of momentariness (kṣaṇi‐katva)asthebasisforestablishingthathisteachingis(generally)true:Umbeka–pramāṇāntaragocarārthapratipādakehikṣaṇikādivākye;Su‐caritamiśra – indriyādisambandhaviṣaye hi jñānamātrakṣaṇikatvādau;Pārthasārathimiśra – asmadādipramāṇagocarārthaṃ kṣaṇikaṃ sarva‐saṃskṛtamityādibuddhavākyam(seeKataoka2011a:348–349n.374).ThisisexactlytheexamplethatisusedbyŚākyabuddhiandKarṇaka‐gominwhenexplainingPS2.5abaspresentedinPVSV108,1(dpernaʼdus byas thams cadni skad cigma źes bya ba la sogspa lta buʼo//;yathā kṣaṇikāḥ sarve saṃskārā ityādikaḥ; see above, n. 35). In thecontinuationofhisargumentKumārilaʼstonebecomesquitesarcasticasheshowswhatelseonecouldprovebysuchanargument.

Thisstatementofmine(vacomama)“theBuddha,etc.(buddhādīnām)arenotomniscient”(asarvajñamiti)istrue,becauseitisstatedbyme(mad‐uktatvāt), just like(yathaiva)[mystatements]“fire ishot”(agniruṣṇaḥ)and(api)“[fireis]bright”(bhāsvaraiti).59,60And(ca)itisaperceivedfact(pratyakṣam)thatIuttered(maduktatvam)[thissentence],whereasyouhavetoprove(tvayāsādhyā)thathe[i.e.,theBuddhaandsoon]stated(taduktatā)[theseteachings].Therefore(tena)mine (madīyaḥ) should be (syāt) a [correct] reason (hetuḥ), whereasyours (tava) isnotestablishedbecause it isdoubtful (saṃdigdhāsiddha‐tā).61(Kataoka2011a:364)

Ithinkthatbeingfacedwithsuchmockingcriticismandknowingthattheback‐boneofhisownapohatheoryiscompletelycontrarytowhatDignāgaproposedwithhisāgamadefinition,itfitsthecontextbetterifweassumeDharmakīrtitobetroubledbyKumārilaratherthanbyhisown appeal to the Buddhaʼs transempirical knowledge. Dharmakīrticouldnotbutbitethebullet.Andhediditquiteskilfully.

59OnthepossibilityofaMīmāṃsakausingsuchakindofśeṣavadanumāna toderive

thevalidityoftheVedainallitspartsfromastatementsuchas“Fireistheremedyforcold,”seeabovepp.102–103withn.41.

60ŚVcodanā 130:buddhādīnāmasarvajñam iti satyaṃvacomama/maduktatvādya‐thaivāgniruṣṇobhāsvaraityapi//.

61ŚVcodanā131:pratyakṣaṃcamaduktatvaṃtvayāsādhyātaduktatā/tenaheturma‐dīyaḥsyātsandigdhāsiddhatātava//.

116 HELMUT KRASSER 5.SUMMARY

Puttingnowtogetherthesevariousbitsandpiecesofinformationscat‐teredthroughoutthePV(SV),wecansketchthefollowingpicture.

BeingconfrontedwiththefactthatDignāgaʼsde initionofāgamaaswell as the cognition that arises from it as an anumāna is not onlyfaultybutalsoincompatiblewithhisapohatheory,theonlypossibilityleft forDharmakīrtiwastoresort todamagecontrol.Buthetriednotonly tominimize thedamage asmuch aspossiblebut also tobenefitfromit.

Firstof all, thisdefinitionofDignāga isnot tobeunderstoodonalogical level.62 ItwaspresentedbyDignāgaonlybecausehumans likeuscannotlivewithoutreligion.Thisisduetothefactthatcredibleper‐sonsreportdisastrousconsequencesofbadkarmaaswellasthepossi‐bility of final release. Since we do not see anything that contradictstheirreports,wearebetterofffollowingtheiradvicewhenweact.ItisonlyunderthesecircumstancesthatDignāgagavethisaccount;63beinghimself a specialist in logic (nyāyavid, PV 1.331), he knew very wellthatsuchadefinition,logicallyspeaking,isasdeficientastheonepro‐posedbytheNaiyāyikas.64Thus,Dignāgasetthestandardforscriptureashighaspossibleonarationallevel–theentirebodyofāgamahasto 62PV1.216(anumānatā...agatyābhihitā),seeabove,p.88;PVSV109,19–20(agatyā...anumānatvam...upavarṇitam),seeabove,p.101;andPVSV168,1–2(agatyācedamāgamalakṣaṇamiṣṭam),seeabove,p.104n.42.Seealso§4.1.

63PVSV108,2–6;seeabove,p.85n.11. 64PVSV173,22–25(onPV1.331);seeabove,p.103n.41.SeealsoTillemans1999a:

401:“KarṇakagominandŚākyabuddhiadv.216areclearontheimplications:scrip‐turalinferenceisaninferencebecauseofthethoughtofpeople(pumsoʼbhiprāyava‐śāt)whowanttoengagethemselves(pravṛttikāma)onthespiritualpath: it isnotaninferenceobjectively(vastutas).”Tillemanscites,inn.11,Karṇakagomin(PVSVṬ394,20–22): kiṃ tarhīṣṭasya pratyakṣānumānāgamyasyārthasyānantaroktena nyā‐yenāvisaṃvādādanumānamapipravṛttikāmasyapuṃsoʼbhiprāyavaśāt/vastutastvananumānam, śabdānām arthaiḥ saha sambandhābhāvāt/. Already in their com‐mentsonnāyam(PVSV108,2),etc.,ŚākyabuddhiandKarṇakagominhadexplainedthatDignāgadidnotcallāgamaaninferenceintermsofrealvalidity(bhāvikaṃprā‐māṇyam).PVṬD242b5–6/P286a2–3:ʼdiskadduslobdpongyisdṅossutshadmañiddubrjodpasluṅrjessudpagpañiddugsuṅspanimayingyi/ʼonkyaṅskyesbuʼiʼjugpa la ltosnasbstanpayinno//=PVSVṬ390,21–22:nācāryeṇabhāvikaṃprāmā‐ṇyaṃkathayatānumānatvamāgamasyoktam,apitupuruṣapravṛttimapekṣya.

LOGIC IN A REL IG IOUS CONTEXT 117

passthetripletest.65Ifitdoes,theprobabilitythatonewillbesuccess‐ful in attaining one’s religious goals ismuch higher.66 But this tripletest isnotonlyhelpful in religiousmatters, it also isaneffective toolwhendealingwiththescripturesoftheopposingschools.It isexactlythis triple test that isconsistentlyappliedbyDharmakīrti throughouthisworks in order to refute the scripturesof his opponents; namely,they have to pass the test (1) in terms of perceptible objects, (2) intermsofobjectsthatareinferableempirically,andthen,finally,(3)intermsofinternalcontradictionswhichcanbecheckedwiththehelpofscripturallybasedinferences.Andbecauseāgama isnotanindepend‐entpramāṇa,itcannotbeusedaslongasithasnotpassedthefirsttwostepsofthetripletest.Thus,manyofthecentraldoctrinesofthescrip‐turesofopposingschools,suchasapermanentsoul(ātman),primor‐dialmatter(pradhāna/prakṛti),oracreatorgod(īśvara),whichinfactcanonlybeprovenbymeansofscripture(ifatall),failtopassthese‐cond step of the test, i.e., ātman and so on, which are taught to beinferable,infactcannotbeestablishedbyempiricalinference.67

The price for escaping Kumārilaʼs critique, for keeping his theorythatthereisnorelationbetweenwordsandexternalobjectsasstatedinPV(SV)1.213–oneof thecornerstones inDharmakīrtiʼsrefutationofthevalidityoftheVeda–andfordraggingdownallscripturestotherealmwhere theycanbe testedbyperceptionorempirical inference,was very high. Dharmakīrti had to abandon the validity of all scrip‐tures,includingthatoftheBuddhists.Thisisquiteconsistentwiththerestofhis logicalsystemwhichbuildsonpervasion(vyāpti).There isnowaywhatsoevertoestablishapervasionbetweentheactofspeak‐ingofacrediblepersonandhistellingthetruth.68Inaccordancewiththis theory Dharmakīrti established a different method to prove theauthoritativenessof theBuddhainthePramāṇasiddhichapter,basinghimselfonempiricalinferenceandthusavoidingaśeṣavadanumāna.69 65PV1.214–215withPVSV;seeabove§2.2. 66PVSV174,1:tatrakadācidavisaṃvādasambhavāt;seeabove,p.103n.41. 67SeePV4.2and4.48–52(especially4.48and50)inTillemans2000:11,78–83;see

alsoabove,p.112n.53. 68PVSV167,26–168,1:tatpravṛtteravisaṃvādenavyāptyasiddheśca;seeabove,p.104

n.42. 69Seeabove,pp.105–106andnn.43and44.

118 HELMUT KRASSER While Dharmakīrtiʼs courageous reform of Dignāgaʼs theory of

scripture,whichhewas forcedtoundertakenotonlybyhisownsys‐tembutalsobyKumārila,turnedouttobequitesuccessfulwithintheframeworkofhisownsystem,itcouldverywellhavepleasedKumārilaandothernon‐Buddhiststoseetheirworstcriticgivingupthevalidityofthebuddhavacana.ItwasalsoverymuchwelcomedbyhisBuddhistadversaryBhāviveka,whoimmediatelysenthisgreetings:

atrocyatepramāṇaṃnaḥsarvaṃtāthāgataṃvacaḥ/āptopadeśaprāmāṇyādbhadrohipratipadyate//nāgamāntarasandigdhaviparyastamatiḥparaḥ/tasmāttatpratipattyarthaṃtanmṛgyoyuktimannayaḥ//MHK5.8–9Tothis[pūrvapakṣaoftheYogācāra]wereply:AllthewordsoftheTathā‐gataareauthoritative(pramāṇa)forus,becausetheteachingsofarelia‐blepersonare authoritative.A goodoneputs [these] intopractice.Theotherone,whosemindisindoubtandconfusedbyotherscriptures,doesnot.Thereforethepathofreasoning(yuktimannaya)shouldbefollowedbyhiminordertoputtheseintopractice.70

And itmightwellbe thecase thatBhāvivekawasnot theonlyonetoaccuse Dharmakīrti of not being a good (bhadra) Buddhist by beingtaken inbyKumārila’sargumentsagainstscripture, theargumentsoftheopponent(āgamāntara),and,inhisconfusion,thinkingthemtobehis own arguments. There might have been other Buddhists as wellwhothoughthehadthrownthebabyoutwiththebathwater.

70For the interpretationof these two verses and the commentary on the latter, see

Section1.3.1.1,“WhoistheopponentinMHK5.8–9?”inKrasser2012:545–546.Forthetextanddifferenttranslations,seeHoornaert2000:78and90,andEckel2008:225–227and394–395.