Buddhist Logic before Dinnaga
(Asanga, Vasubandhu, Tarka-sastras)
By professor guiseppe tucci
Journal of the Royal Asiatic Society of
Great Britain and Ireland
1929.07, p 451-488
We must admit that very little is known about the
first development of Indian logic and particularly
about Buddhist logic before Dinnaga. If we take the
best manuals of Indian logic now available, such as
those by Suali, Vidyabhusana, Keith, or the most
comprehensive Histories of Indian philosophy like
those of Dasgupta and Radhakrishna we shall easily
recognize that the data contained therein are far
from being satisfactory; more than that, they are
also very often wrong. In fact, almost the only
source from which their statements are derived is the
book by Sugiura,(1) who certainly had the merit of
giving the first account of Indian logic as preserved
in Chinese sources, but, being himself absolutely
without knowledge of orthodox nyaya and of Sanscrit,
is in his statements and in his translations very
On the other hand, it is evident that a better
knowledge of the logical schools before Dinnaga might
settle many a vexed question, including those of the
originality of Dinnaga himself, his indebtedness to
previous masters, and the relation between his theory
of the syllogism and that expounded in the
Unfortunately the largest part of the texts on
logic anterior to Dinnaga seems to be lost.
We have, it is true, two fragments preserved in
Chinese; one is the so-called Upaya (?)-hrdaya--not
1. Hindu Logic as preserved in China and Japan,
2. On the other hand, a great deal of information can
be gathered from Ui's book on the Vaisesika
philosophy. Cf. also his Studies on Indian
Philosophy, 印度哲學研究 Tokyo. The classical book
of Stcherbatsky, Erkenntnistheorie und Logik nach
der Lehre der spateren Buddhisten, deals chiefly
with Dharmakirti's thought.
hrdaya as suggested by Nanjio and accepted by Bagchi;
the other is a fragmentary treatise in three
chapters, attributed by some catalogues of the canon
to Vasubandhu. By Nanjio, Ui and Bagchi it is called
Tarka-sastra. The first was translated by Ki Kya
Ye(1) ; the second by Paramartha. Although the
statement of Takakusu that all the works translated
by Paramartha are anterior to A.D. 500 is too
dogmatical, we are at any rate confronted here with a
fairly ancient book.(2)
So far as the first text is concerned, there are
no grounds either for affirming or for denying its
attribution to Nagarjuna; but there is no doubt that
it represents fairly ancient theories which are very
nearly akin to those contained in the
These two treatises have been retranslated into
Sanscrit by me and will shortly be published in the
Baroda Sanscrit Series. As they are certainly the
most ancient fragments of the Vivada-sastras that we
possess, their bearigg upon the problem of the
relations between pure heuristic and Later Nyaya
doctrines is very great. But we should like to have
other texts of indubitable authorship, in order to
fix a terminus a quo and to ascertain which school
must be credited with an original contribution to
Fortunately, such texts have been preserved. We
may divide them into three categories, (a) Chinese
sources; (b) Tibetan sources; (c) Sanscrit sources.
The first category includes the translations of
the following books:--[l] Yogacarya-bhumi-sastra, 瑜
伽師地論, by Asanga. Of this monumental work we have
also the Tibetan translation (Bstan agyur, mdo dsi,
1. On Ki Kia Ye (fifth century A.D.) see Chavannes,
Cinq cent contes, iii, n. 1, Demieville, BEFEO.
xxiv, 1924, pp. 65-6, n. 4. We know from the K'ai
yuan shih kiao lu, 開元釋教錄, that before Ki Kia
Ye another translation of this work had been made
by Buddihabhadra of the Eastern Tsin. Cf. Bagchi,
Canon Bouddhque en Chine, p. 346.
2. BEFEO. 1904, p. 3.
3. See Ui's Studies in Indian Phil., vol. ii, p. 428.
Catalogue du fonds tibetain, vol. iii, p. 378). 
Prakaranarya-vaca-sastra, 顯揚聖教論. The chapter
concerning our subject corresponds almost verbatim
with the preceding. Mahayanabhidharma-sangiti-sastra,
大乘阿毗達磨集論. The theories expounded
in this work differ very often from those contained
in the two preceding texts. Its doctrines are
explained in the commentary written upon it by
Sthiramati(1) and called 
磨雜集論. Then we can collect a great deal of
information from the commentaries written by K'uei
Chi, the disciple of Yuan Chwang. I have used the
Commentary on the Nyayapravesa, which has been partly
translated by me in a previous study(2) which may
complete in some way the statements contained in the
present paper. I am aware of the fact that K'uei Chi
wrote also a commentary upon the
Yogacarya-bhumi-sastra, called 瑜伽師地論略纂, and
another on the Abhidharma-samyukta-sangiti(3); but I
could not here in Indin get copies of these two
texts. On the other hand, I have used the commentary
of Shen T'ai, another disciple of Yuan Chwang, on the
Nyaya-mukha.(4) The second category is represented by
the Pramana-samuccaya-vrtti, by Dinnaga(5) It
contains, as we shall see, much precious information
about the logical activity of the schools that
1. On Sthiramati see Peri in BEFEO. 1911, 348 and
2. Notes on the Nyaya-pravesa in Bollettino della
Scuola di Studi Orientali, 1928.
3. This is called 大乘阿毗達摩雜集論述記, usually
quoted under the abridged form 對法論疏.
4. The Nyaya-mukha (not Nyaya-tarka-dvara-sastra; see
JRAS. 1928, p. 7) has been translated into English
by me and compared with the corresponding portions
of the Pramana-samuccaya. It will shortly be
published in the Materialien zur Kunde des
Buddhismus of Professor Walleser.
5. The Pramana-samuccaya is preserved in Tibetan,
together with two translations of the vrtti of
Dinnaga himself (Bstan agyur, mdo, ce, Cordier, p.
434). I have used the copy of the University of
Calcutta, which has been kindly put at my disposal
by the authorities. This copy belongs to the
In the third category we may include those
quotations and allusions which can be found in
Uddyotakara's Nyaya-varttika and Vacaspati Misra's
It is evident that the second and the third
category can supply us only with fragments, while in
the first we are confronted with complete texts or
commentaries, which, through the intermediacy of Yuan
Chwang, are likely to go back to a tradition of
exegesis current in the Indian monasteries at the
time of the travels of the great pilgrim. In any
case, by combining all these references, we can
attain a better knowledge of Indias logic before
Dinnaga than we have had up to the present. We shall
begin by studying the Chinese translations which
belong to the so-called Yogacara school started by
Asanga and developed by Vasubandhu. The teaching of
this school, in its dogmatical structure, seems to be
more related to the Sautrantika doctrines than to
the ontological theories expounded in the Lankavatara
or in the Mahayana-sraddhotpada.
The contents of the chapters that are of interest
to us were made known by Sugiura, and after him by
Vidyabhusana, who based himself upon the resume given
by the Japanese scholar. But even this summary is far
from correct or complete. Moreover, there is in the
books referred to many a detail which has been passed
unnoticed by the Japanese and Indian scholars. Even
the attribution of the Yogacarya-bhumi-sastra to
Maitreya is wrong, although it is generally acceptcd
and repeated by Indologists. This fact is rather
important because, accepting the attribution of the
text to Maitreya, we should be compelled to admit
ipso facto an earlier date for it; but there is no
doubt that it is by Asanga and represents perhaps one
of the last and most complete products of the
wonderful activity of the great master.
In the exposition that follows we shall indicate
by A the group Yogacarya-bhumi-sastrra and
Prakaranarya-vaca and by B the group Sangiti and
The first classification that we meet is that
concerning "speech", vakya, 論, smra.ba. There are
seven heads, viz.:--
I. vakya in itself, 論體性, smra.ba.
II. The place where speech is made, the parisat, 論處
III.The basis or the support of speech, 論所依,
smra. bai.gzi, vakya-mula or vakyasraya.
IV. Adornment of speech, 論莊嚴, smra.bai.rgyan,
V. The defeat of a speech, i.e. in argument, 論墮負,
VI. That which derives or comess forth from a
speech, 論出離, smra.ba.las byun.ba.,
VII. Those characteristics which are the causes of a
speech being appreciated (by the hearers), 論多所
作法, smra.ba.gces.spres.la.dgos.pai.c'os. rnams.
We shall later discuss the third point, which has
the main interest for us, and give here a mere
summary of the various subdivisions of the other six
items, as they have not the same bearing on the
history of logical theories in India.
I. vakya. This can be of six kinds--
(a) Vakya in itself.
(b) Excellent words, that is words with which the
world is pleased.
(c) Disputation-words, which are uttered when two
men engaged in a discussion maintain quite different
opinions about a particular object or a particular
thesis. It is worthy of notice that, while 13 simply
states that it consists in holding opposed views, A
insists at length upon the various causes of the
dispute. It asserts that these are to be found in the
abhinivesa, "attachment," of the creatures belonging
to the kama-dhatu or in the criticism that human
beings are inclined to express about the sinful deeds
of body, mind and speech of others, or in the
discussion of the various drstis, e.g.
those of eternity or of uccheda, at a time when the
disputants are not yet free from passion.
(d) Rebuke-words (apavada-vakya), 毀謗論 or (B)
毀論, ts'ig.nan.pa.smra.ba. It includes unpleasant
words or the teaching of false theories.
(e) Accordant speech, 順正論, mt'un.par.smra.ba.:
any speech which is in accordance with the dharma and
aiming at producing a right knowledge in tile mind
of the hearers.
(f) Teaching, 教導論, gdams.par.smra.ba.
The first two items can be either good or bad,
and therefore it is necessary to distinguish
them according to circumstances; the next two are
always bad and therefore must be avoided. The last
two are always good and therefore must be practised.
II. Place where a speech is made--
(a) before a king;
(b) before a governor;
(c) in a great assembly;
(d) before sramanas who are well versed in the
(e) before Brahmans;
(f) before those who like to hear the dharma.
IV. Adornment of speech. Its fundamental aspects
are five according to A, but six according to B.
A. I and II B. Sangiti and Sthiramati
(a) Perfect knowledge of Id.
one's own as well as of
another's system, 善自他宗
ses . pa (sva-para-siddhanta-
(b) Perfection of the phrase, Id.
語句圓滿 ts'ig.sbyor. b
p'un.sum.ts'ogs. pa. A phrase
is perfect when it is possessed
of five good characteristics.
That is to say, it must be:-
(1) devoid of any rustic No mention of these five
expression. subdivisions. The perfection
(2) easy. of the phrase consists in
(3) evident. avoiding mistakes through
(4) coherent. knowledge of the sabda-sastra
(5) having a good meaning and the vyutpatti-sastra.
(c) 無畏, mi.ajigs.pa., Id.
abhirutva, fearlessness. Even
if one finds himself among a
parisat numerous or hostile,
he must be sure of himself.
(d) 敦肅, brtan.pa, dhirata, Id.
(e) Speech possessed of Id.
those characteristics that will
be esteemed and attractive,
應供, no.mi.bzlog.pa. Adds: 辯才=pratibhana,
when sentences flow
At this point A gives a list of twenty-seven
prasamsa-gunas, which are the ornaments, as it were,
of an excellent speech:-
(1) high estimation by hearers;
(2) belief and acceptance by hearers;
(3) absence of fear;
(4) knowledge of the mistakes in the thesis of
(5) knowledge of the superiority of one's own
(6) absence of abhinivesa;
(7) not to be partial towards one's own system;
(8) not to renounce one's own law and rules;
(9) to understand quickly what has been said by
(10) to grasp quickly what has been said by the
(11) to explain quickly what has been said by the
(12) the power of captivating the assembly with
gifts of speech;
(13) to be able to rejoice those who like
(14) the power of expressing in the best way the
meaning of the arguments;
(15) no trace of depression in the body, while
(16) no depression in mind while discussing;
(17) no stammering;
(18) to maintain always presence of mind
(19) no bodily fatigue to be shown;
(20) memory always functioning;
(21) mind uninjured;
(22) no pain or impediment in the throat;
(23) expressiveness of the voice;
(24) restraint of one's own mind in order to
(25) to comply with the other's mind in order to
avoid his wrath;
(26) to act in such a way that the adversary may
be persuaded in his own mind;
(27) to be considered everywhere as a great
V. Nigraha-sthanas. These can be of three fundamental
(a) vacana-sannyasa, 拾言, brjod.pa.gton.ba;
(b) when the speaker perceives that his words
have been refuted with success by the opponent and
therefore tries to avoid further discussion, 言屈,
(c) erroneous speech, vacana-dosa, 言過,
Vacana-sannyasa consists in confessing one's own
defeat and in acknowledging that the thesis of the
adversary is right. According to group I it can be of
thirteen kinds; e.g. my thesis is wrong, your thesis
is right, etc.
Vacanabhibhava occurs when a speaker, realizing
that his arguments are wrong, tries to avoid the
discussion, saying that he has something else to do,
or brings into the discussion
new arguments not connected with previous ones, or
looks irritated, angry, conceited, or reveals some
defect or fault in the adversary which the latter
does not like to have disclosed, or looks offended or
shows impatience or distrust, or has nothing to reply
and therefore keeps silence, or looks abashed and
trembling or bends his head or appears as if he were
deprived of the faculty of thinking and specking.
Vacana-dosa can be of nine kinds--
(a) to speak at random;
(b) violent expressions, suggested by anger,
(c) obscurity of expression, when the speaker
cannot be understood either by the assembly or by the
(d) lack of proportion, when the expression is
either defective or excessive (adhikya-nyunatva);
(e) meaningless, 非義相應, don.dan.ldan.pa.ma.
yin.pa, vyartha. It is of ten kinds:--
(1) anarthaka, 無義, dgos.pa.med;
(2) aparthaka, 違義, don.pa.med.pa;
(3) yukti-hani, 損理, rigs.pa.las.nams.pa;
(4) sadhya-sama, 與所成等, bsgrub.par.bya.ba.
(5) jati, 招集過難, ltag.gc'od.pa;
(6) arthanupalabdhi, 不得義利, don.mi.dmigs.pa;
(7) asambaddha, 義無次序, don.dan.mi.abrel.ba.;
(8) aniscita, 義不決定, ma.nes.pa;
(9) siddha-sadhya, when the probandum is already
proved, 成立已成, sgrub.pa.yan.sgrub.par.bya.ba.yin.pa;
(10) a speech according to illogical or wrong
Sthiramati knows only the first five of these
nigraha-sthanas, and he considers the other five as
mere explanations of them (1< 6, 2 < 7, 3 < 8, 4 < 9,
5 < 10);
(f) aprapta-kala, when the various arguments are
not brought forward in order;
(g) aniscita (or aniyata), when someone either
attacks an argument that he has already established
as his thesis or establishes as a thesis an argument
that he has already attacked or suddenly changes his
(i) lack of cohesion.
VI. That which derives or comes forth from a
This is threefold, consisting of (a)
guna-dosa-pariksa, 觀察德失, yon.tan.dan.nes.pa.brtag;
(b) parisat-pariksa, 眾會, ak'or.brtag.pa;(c)
pandityapanditya- pariksa, 善不善, mk'as.mi.mk'as.
The first consists in examining whether the
discussion undertaken will be of some use or not to
the speaker and to the hearers. If one knows that no
good result is to be expected from the discussion,
he must avoid it.
The second consists in ascertaining whether the
parisat is impartial, learned, strictly honest. If
this be not the case, the discussion must be avoided.
The third consists in examining whether one has
the knowledge and the ability necessary to carry on
the discussion satisfactorily, If an aspirant
acknowledges that he is not possessed of the
requisite and indispensable qualities, he must
renounce the disputation.
VII. The characteristics which cause a, speech to
be appreciated by the hearers are (a) knowledge of
one's own and opposing systems, (b) absence of fear,
(c) promptitude of intelligence: (a)
sva-para-mata-jnana, 善自他宗, bdag.
dan.p'a.rol.gyi.gzun.lugs.ses.pa; (b) abhiruta, 無畏,
mi. ajigs; (c) pratibhana, 辯才, spobs.pa.
Now we shall study the section dedicated to the
third item, that is, to the basis or support of a
speech. In a discussion we can distinguish two
elements, which are respectively called (a) the
probandum, sadhya, 所成義, bsgrub.par.bya. bai.don,
and (6) the proof, sadhana, 能成, sgrub.pa. The
probandum is twofold, that is to say, we may prove
either a subject (lit. an entity, svabhava, 自性,
no.bo.nid) or an
attribute (lit. a quality, visesa, 差別, bye.brag).
In the first case I can affirm or deny the existence
of something, that is, I can say that it is or is
not. In the second I may affirm or deny that a given
quality belongs or not to the subject. In this way
according to the example given by Sthiramati a sadhya
can be of either of the following types:-
(a) "the atman is, is not."
(b) "the atman is all-pervading" or "sound is non-
The proof, or sadhana, consists of eight terms,
although the list and the definition of these vary
remarkably in the various texts that represent our
(1) pratijna, proposition, Id.
(2) hetu, reason, 辯因, Id.
(3) drstanta, example, 引 Id.
(4) sadharmya, homogeneity, Application, 合.
同類, mt'un. Pa.
(5) vaidharmya, heterogeneity, 異類, mi.mt'un.pa.
(6) pratyaksa, direct per- Id.
ception, 現量 ,mnon. sum.
(7) anumana, inference, 比 Id.
(8) agama, authority, 至 Id.
(1) "Proposition," pratijna.
A. Pratijna consists in maintaining as one's own
thesis a particular point of view concerning the
twofold probandum already referred to. It is either
based on the sastra, or is the result of an
independent intuition (pratibha), or has been
heard from somebody else. And it is designed either
to maintain one's particular point of view, or to
show the mistake in another's argument, or to subdue
the other's pride, etc.
B. Pratijna is the argument that the vadin
accepts of his own free will, as that which must be
proved (以所應成 自所許義 = *sadhyatvena svayam
anujnato 'rthah); and it must be expressed to others
in such a way that they can understand. Sthiramati
explains how the various elements of the definition
are necessary; "that which must be proved," because
what is already proved is not a thesis; " accepts of
his own free will," because what is said by another
is not a pratijna; "to others," in order to show that
it takes place where there are a vadin and a
prati-vadin; "expressed by words," because what is
expressed by mere signs (ingita) of the body is not a
pratijna; "in such a way that they can understand it"
because a proposition the meaning of which is not
clear cannot be called a pratijna.
 "Reason," hetu.
A. Hetu is meant to prove the probandum, and it
shows forth that logical reason which is derived from
the "example", "homogeneity, " "heterogeneity, "
"direct perception," "inference," and "authority".
B. When an object (artha) to be proved is not yet
evident, the reason consists in the indication of
those characteristics which will make it known, and
which rest upon its perceptibility or
non-perceptibility by direct perception and so on.
Perceptibility and non-perceptibility concern either
the essence (自體, svabhava) or the form (相貌,
 "Example," drstanta.
A. This also is designed to proved the probandum;
it consists in adducing those same dharmas which are
inherent in a reason and which are accepted by common
belief, general knowledge, etc.
B. It consists in expressing the relation between
seen (drsta-anta) anil what is not yet seen (未川,
 A. "Homogeneity, " that is similarity of
characteristics (相貌, rtags, nimitta); similarity of
essence (自體, no.bo.nid, svabhava); similarity of
action (業, las, karma); similarity of attributes (法
, c'os, dharma); similarity of cause and effect (因果
, rgyu.dan.abras, karya-karana). It is worthy of
notice that according to the Chinese translation the
last four are subdivisions of the first item.
B. "Application" is a logical rule, rightly
expressed, which adduces other facts belonging to the
same class or genus in order to prove the attribute
(of the subject).
 A. "Heterogeneity,'' reciprocal diversity. It
has four aspects, which are the opposites of those
referred to under  (or five according to the
B. "Conclusion." This consists in affirming that
certitude has been reached.
Here Sthiramati gives the following example of a
syllogism:- Suppose that a Buddhist wants to maintain
against an atma-vadin that the atman does not exist.
He will argue in this way:--
 pratijna: all dharmas are anatman.
 reason: because, if we assume (prajnapti)
that the atman is in the skandhas, we fall into a
1. Four cases are possible: --(a) the atman has the
characteristics of the skandhas; (b) it is in the
skandhas; (c) it is in another place; (d) it is
assumed without any relation to the skandhas.
(a)As the skandhas are not autonomous, but dependent
on causes and conditions and subject to birth and
destruction, the same implication would be
necessary as far as the atman is concerned; but
this is contradictory to the common definition of
(b)As the skandhas which are the basis (所依,
asraya or adhara) are non-eternal, the atman which
rests upon them (能依, adheya) must be
(c)In this case the atman would be without cause
and therefore without function (無用, niskriya).
(d)In this case the atman would be isolated and
free; no need therefore to strive for its
 "Example," namely those which we make when we
assume that in the present the past is still
 "Application": as the atman has been refuted,
the other attributes also, such as eternity, etc.,
are to be declared non-existent.
 "Conclusion": therefore the five skandhas are
anatman and non-eternal.
 "Direct perception." A. This has three
characteristics, that is:-
(a) it is evident, 非不現見, lhog.tu.ma.gyur.pa.,
(b) devoid of imagination, 非思構所成; but Tib. mnon.
par. brtags. zin. yin. pa. ma. yin. pa. dan
(reading doubtful). brtags. par. bya. ba. yan. ma.
yin. pai. mnon. sum. gyi. ts'ad. ma. =
(c) devoid of error, 非錯亂所見, ma. ak'rul. pa.,
(a)It derives from the senses when they are
uninjured, and it precedes manaskara. It depends
upon (a) production of homogeneous perception, 同
類生, mt'un. pa. skyes. pa; (b) production of
heterogeneous perception, mi. mt'un.skyes.pa; but
Tibetan, followed by Yogacarya-bhumi-sastra,
yan.dag.par. adas.pa. skyes.pa,
samatikranta-utpada; (c) proximity, 不極違,
(a) When the indriyas belonging to the sphere of
kama (kamavacara) perceive (lit. are born in)
objects belonging to the same sphere.
(b) When the senses belonging to a superior bhumi
perceive objects belonging to a superior bhumi.
(c) Obstructions which must be absent in order to
have a direct perception are of four kinds, (i)
obstruction which derives from covering, as
through darkness or ignorance; (ii) obstruction
which derives from being hidden, as through the
force of some mantra, etc.; (iii) obstruction
1. The same arguments must be repeated here mutatis mutandis.
from being overpowered, 映障所礙, zil.gyis.gnon.pa,
abhibhava, as the small by the great, etc.; (iv)
obstruction which derives from bewilderment, moha,
such as magic power, maya, sleep, taimirika, etc.
(b) The second term also is twofold; first of all
it includes the perception of objects which results
as soon as these come in contact with us. So, e.g.,
when a doctor gives a medicine to a patient, through
the colour, the smell, the taste, etc., he has a
direct perception of the medicine. On the other hand,
the virtues which are inherent in the medicine can
only be imagined until the disease is over. They are
no more imagined when one knows that the patient has
recovered, The term refers also to the adhimukti or
realization of a particular element, e.g. water, in
another element, e.g. earth, in the process of
(c) abhranta means absence of seven kinds of
errors; these errors are the following:--
(a) samjna-bhranti, to think that an object is this
when it is not this, atasmin tad eva; e.g. to take
a mirage, marici for water.
(b) sankhya-bhranti; e.g. to see the complex in the
elementary, as happens to the taimirika, who sees
two moons instead of one.
(c) akara-bhranti, to suppose that an object has a
certain form when it has not; e.g. to see a wheel
in a turning fire.
(d) varna-bhranti; as in the case of someone
suffering from kamala, 迦末羅, mig.ser.gyi.nad.
(e) karma-bharanti, to attribute an action to
something which in fact is not so acting; e.g. the
appearance of movement in trees when one runs very
(f) drsti-bhranti, to persist in the errors already
referred to and to think that they correspond to
(g)citta-bhranti, to rejoice in these errors.
All these varieties of perception can be reduced
to the four following:--
loka-pratyaksa, including in fact the two preceding;
suddha-pratyaksa, which can be laukika, as well as
B. Perception is the very thing, rightly
perceived, devoid of error. "The very thing" and
"rightly" are intended to express the right
perception of the rupa, etc., through the eyes and to
indicate that a pot, etc., that, according to common
belief, is the object of perception, is, in fact, not
the object of perception, as it is only a
conventional assumption 假, "perceived," is meant to
indicate that in the act of perception all the causes
of obstruction must be absent; "devoid of error"
excludes false and erroneous perceptions, as that of
a marici, etc.
A. It consists in the discrimination of an object
through imagination. It is of five kinds (cf. above,
p. 463,  A):- nimitta-anumana; as to infer fire
from smoke; it depends
on the fact that the relation between the two was
sva-bhava-anumana; as to infer unperceived existence
from a present perceived existence or from one
part of an entity to deduce the unperceived part,
e.g. to infer the past from the present or a car
from a single portion of it, as a wheel.
karma-anumana, from an action to infer the basis or
the support of it; e.g., when we see an object
from afar, if it is motionless, we infer that it
is a tree; if it moves, we infer that it is a man.
dharma-anumana; when we know that many dharmas are
inter-related, from the perception of some we
infer the existence of the others. From birth we
infer death, etc.
karya-karana-anumana, inference of notions which are
related as cause and effect.
B. "Inference" is any conviction besides that
derived from direct perception; as, when me have
already seen an
object and now see only a part of it, we infer the
A. It includes the teachings of the wise or the
doctrines that have been heard from them or are in
accordance with them. It is of three kinds: (a) it is
included in the holy words; or (b) it represents the
opposite (pratipaksa) of the passions; or (c) it is
not contradictory to the characteristics of the law.
B. It is not contradictory to the other two
At the end of this chapter A adds the following
notes on the syllogism in general:--
If someone asks why we have to formulate the
proposition when we want to establish the argument
assumed by us, the reply is that this proposition is
meant to show the argument that we wish to prove. The
"reason" shows, on the other hand, that that logical
and sure evidence which is based upon a manifest fact
is not absent in the object to be proved. The
"example" indicates that evident object in which this
logical reason is seen to be present. The other five
elements of a syllogism are meant to express
contradiction and noncontradiction with the "reason"
and the "example". This contradiction consists in two
kinds of fallacies, aniscita, uncertain, 不決定,
ma.nes.pa, and sadhya-sama, identical with the
probandum, 同所成, bsgrub.par.bya.ba.dan.adra. ba.
The aviruddha, on the other hand, is certain,
niscita, aikantika, 決定, gcig.tu.nes.pa, and
different from the sadhya, 異所成,
These are the contents of the logical chipters of
the Yogacara works as preserved in Chinese and partly
in Tibetan. It is quite evident that they have a
mixed character; purely logical doctrines are
inserted in dogmatical discussions, and generally the
are treated in such a way as to testify that
hetu-vidya did not yet fit quite well into the
general scheme of the doctrine. Even those sections
that deal with mere vivada-rules, e.g. those
dedicated to the nigraha-sthanas, have a far less
systematic character than in the Caraka-samhita or in
the Upaya-hrdaya; many of the items which come under
that group have in fact very little to do with logic.
The theory of the nigraha-sthanas itself is not based
on the classification of the possible wrong
formulations of a syllogism. A comparison with the
list of the nigrahasthanas given in the
Nyaya-sutras and in the Caraka-samhita will prove
useful in establishing the relation between the
NYAYA-SUTRAS CARAKA-SAMHITA(1) A AND B
(V, ii, 1).
pratijna-hani id. 
pratijnantara = III, g
pratijna-virodha viruddha (13, cf.
pratijna-sannyasa --I, vacana-
hetv-antara id.  included in II
arthantara id.  included in II
nirarthaka id. (10, vyartha, cf. under III, e
avijnatartha =III, c, obscurity
aparthaka id. (11; cf. under under III, e
aprapta-kala kalatita  III, f
nyuna id.  =III, c
adhika id.  III, c
1. The numbers in brackets show the serial order that
the various Nigraha-sthanas have in the actual
list of the Caraka-samhita.
NYAYA-SUTRAS CARAKA-SAMHITA A AND B
(V, ii, 1).
punar-ukta id. 
ananubhasana included in II
ajnana id. 
apratibha included in II
viksepa vacanabhibhava II
matanujna id. (5, abhyanujna)
paryanuyojyo- id. (3, anuyojyasya-
niranuyojyanuyoga id. (2, ananuyoj-
ahetu  samsaya-sama
It is quite evident that we are confronted in A
with unsystematic and perhaps archaic theories of the
nigraha-sthanas, the classification of which seems to
have been suggested more by extrinsic reasons
concerning the behaviour of the disputants than by
analysis of the intrinsic errors of a speech.
Moreover, we do not find any trace of technical
terminology. An argument is considered as wrong
chiefly because it does not convey any meaning.
Therefore it receives the general designation of
"vyartha", meaningless. The ten varieties of this can
be reduced to five only, as rightly suggested. by
Sthiramati himself. But this list of five has nothing
in conmon with the five hetv-abhasas of the
Nyayasutras, except the sadhya-sama. The anarthaka,
which happens when there is arthanupalabdhi, is the
anartha-nigraha-sthana of the N.S. and Caraka. The
aparthaka is the same as that of Caraka and N.S. Jati
is simply enunciated.
Let us pass now to the most interesting section
and consider first of all the question of the
pramanas. Nagarjuna knows only four pramanas,
pratyaksa, anumana, upamana,
and agama; these are referred to in the Upaya-hrdaya
and are refuted as self-contradictory in the
Asanga in his treatises reduces the means of
knowledge to three only, pratyaksa, anumana and
agama,(2) and it is quite evident that in B even
agama is authoritative, according to him, only in so
far as it is based on the first two pramanas. We must
study these means of knowledge separately.
Pratyaksa, according to A, must be aparoksa,(3)
unmixed with imagination, nirvikalpa, and devoid of
error, abhranta, or avyabhicari.(4)
The first two items of aparoksa have some bearing
upon the study of the dogmatics of Buddhist
mysticism, but not so much upon the history of these
doctrines, with which we are dealing here. The other
two, avyavadhana and anatidurata, are more
interesting to us, as they represent a classification
of the various cases in which, owing to some
hindrance, the direct perception of an object cannot
be produced. The question of the paroksa was
discussed very early in Indian speculation. Patanjali
and Caraka have already a list of the various
avaranas; then the complete series of the eight
impediments that obstruct perception can be found in
Vasu(bandhu)'s commentary on tile Sata-sastra of
Aryadeva(5) and they perfectly agree with the list
given in the Sankhya treatises (Sankhya-karika, 7).
The following scheme will show the analogies which
our text presents with the other schools and at the
same time its peculiarities.
1. For the Upaya-hrdaya and the Vigraha I can refer
to my forth coming translation in the Baroda
2. Three pramanas can be found also in the Commentary
of Sthiramati upon the Trimsaka-karika of
Vasubandhu, p. 26.
3. Or apariksita; this expression is, in fact, in the
Caraka-samhita, Sutrasthana, xi, 8.
4. The two terms are almost synonymous, and the
Chinese as well as the Tibetan can be translated
in both ways.
5. For the list given in the Sata-sustra see my
translation of this text in Studi e Materiali di
Storia delle Religioni, 1925.
On the avaranas, according to Patanjali and Caraka,
see Strauss, Mahabhasya ad Panini, 4, 1, 3, in Aus
Indiens Kultur, Festgabe Richard von Garbe, 1927,
Comm. on the Sankhya
Patanjali. Caraka. S(ata) S(astra). texts. A.
ati-sannikarsa id. id. (2) as in S.S. --
all- viprakarsa id. id. (1) as in S.S. durata
murty-antara- avarana vyavadhana(6) id.
vyavadhana 1, 2
indriya-daur- karana-daur- indriya-ghata id.
ati-pramada mano-'nava- mano 'nav. id. 4, mola
samanabhi- samanabhihara id.
abhibhava abhibhava id. 3,
ati-sauksmya sauksmya id. abhibhava
The next necessary quality of pratyaksa is
according to A abhranta or avyabhicari; that is, it
must be devoid of error. These errors can be of five
kinds; in fact, it is evident that the two other
errors given by Asanga in the supplementary list of
the seven bhrantis, I mean citta-bhranti and
drstibhranti have more a dogmatical than a logical
bearing and belong rather to inference than to direct
perception. The samjna-bhranti, defined as consisting
in believing atasmin tad, corresponds to the
savyabhicari, as understood by Vatsyayana in
commenting on N.S., I, 1, 4. It is rather interesting
to note that the other varieties of bhranti were
accepted by Dharmakirti, as we can infer from the
examples given for each of them. Thus the
sankhya-bhranti (example, timira, as in Dharmak.)
corresponds to the indriya-gata-vibhrama-karana of
N.B.T.; the nimitta-bhranti (example, alata-cakra, as
in N.B.T.; N.B. asu-bhramana) corresponds to the
visaya-gata-vibhrama-karana; karma-bhranti (example,
a moving tree, as in N.B.T.; nau-yana of the N.B.)
corresponds to the bahyasraya-sthita-vibhrama-karana
of the N.B.T.; varna-bhranti (kamala) corresponds to
the samksobha of the N.B., that is, to the
adhyatma-gata-vibhrama-karana of the N.B.T. We must
not discuss here whether Asanga, was right in
assuming that samjna-bhranti is a separate class (1);
1. In fact, it is clear that all the various bhrantis
consist in assuming atasmin tad.
must insist upon this analogy between Asanga and
Dharmakirti. We know that Dinnaga, does not add the
attribute abhranta to his definition of pratyaksa and
that in Iris Pramana-samuccaya-vrtti he attacked the
epithet avyabhicari given by the Naiyayikas. On the
other hand, Dharmakirti defines pratyaksa not merely
as kalpanapodha, but as kalpanapodham abhrantam. This
addition is not an innovation introduced by him, but
due to his acceptance of the old theory of the
Sautrantikas. This fact is not only proved by our.
tests, but also is clearly pointed out by Mallivadin
in his Tippani (p. 19, Stcherbatsky ed.).
But Asanga adds another division of pratyaksa in
four items, that is:-
(b) manah-pr. ;
(c) laukika-pr., which includes these two;
(d) suddha-pr., pure, which can be eitherr laukika
This classification of direct perception is also
worthy of notice, because it shows some points of
contact with the fourfold pratyaksa which we find in
Dinnaga as well as in Dharmakirti. In fact, it is
easy to recognize that the first two items and the
last correspond respectively to the rupendriya,
manah, and yogi-pratyaksa of the Pramana-samuccaya,
Nyaya-mukha, Nyaya-bindu, etc. It is difficult to see
what the third item is meant to represent, but it
seems that it has nothing to do with the
sva-samvedana-prattajsa, which is very likely to have
been an innovation due to Dinnaga and depending on
his epistemological theories.
But, as it is evident from the texts, Asanga knew
another definition of direct perception, namely that
which we find in the Sangiti; here the pratyaksa is
the very thing rightly perceived and devoid of error.
The Chinese 自正明了 無迷亂義 presupposes an original
like this, svayam samyak-pratito 'bhranto 'rthah. The
mentary by Sthiramati clearly indicates that this
definition is meant to distinguish the exact
perception from the fictitious; we cannot speak of
having a perception of a pot. In fact, when me see a
pot we cannot say that the knowledge that we have of
the pot is direct perception, as this is confined to
rupa, etc., that is, to the dharmas from which it
results. We find, therefore, here the same definition
of direct perception which was formulated by
Vasubandhu. In fact, we-know from Uddyotakara, that
this master gave the following definition of
pratyaksa:--"tato 'rthad vijnanam" (Nyaya-varttika,
Benares ed., p. 40). That in this and in many other
places Uddyotakara quotes verbatim from the works of
Vasubandhu, and chiefly from his Vada-vidhi, is
proved by the refutation that Dinnaga writes of that
same definition, which he attributes to the
Pramana-samuccaya, chap. i, fol. 3a :-
don. de. las. skyes. rnam. par.ses.
P.S.Vrtti,(1) a, fol. 16 :-
don. de .las. skyes. pai. rnam. ses.
mnon. sum. yin. zes. bya. bai. adir.
P.S.V., b, fol. 79b:--
don. de . las. skyes. pal. rnam. par. ses. pa .mnon.
sum. yin. no. zes. bya. ba. adir .....
It is important to see how one and the same
author is trying to define perception in two different
The fact is that according to the Sutras or even
to the Abhidharma literature there is hardly any
place for Pratyaksa, as it is understood in the other
schools. It was relatively easy for the Vaisesika or
the Nyaya, both being realistic systems, to formulate
a theory of perception, but it was
1. As I said before, we have two translations of the
vrtti of the Pramana-samuccaya, which do not
always agree and seem to be very often defective.
This fact increases the difficulty of the test,
which is one of the most abstruse.
not so easy to introduce this doctrine into a system
which is chiefly based on the dharma theory and in
which there was only question of particular moments
of internal vijnanas, each corresponding to its
analogous external ayatana or dhatu. The definition
as given in the Sangiti and strictly related to that
of Vasubandhu is more in accordance with the
traditional dogmatics; the second is far more
elaborate and it is of the highest interest, as wve
already find there the terms which will be accepted
by Dinnaga (kalpanapodha) and by Dharmakirti
(abhranta), showing therefore the first noticeable
attempt towards the later and more organic
development of Buddhist logic.
If we pass now to anumana, or inference, we must
point out that no explicit mention is to be found
either in A or in B of the distinction between the
svarthanumana and the pararthanumana, which is
expounded in the Pramana-samuccaya, but which was
certainly anterior to Dinnaga. In fact, the
pararthanumana was known to the Tarka-sastras, as we
shall see later on. But the distinction is implied in
Asanga. Although, in accordance with traditional
dialectic, the syllogism comes first in his works,
anumana, included in the list of the pramanas,
represents the subjective means through which we can
apprehend an object or a truth, quite independently
of that verbal formulation which is inherent in a
syllogism, and consists in the evident and valid
conclusion that our mind can draw from some facts
previously ascertained by direct experience.
To the two definitions contained in A and B we
may add that of the Vada-vidhi, referred to and
criticized by Dinnaga:-
P.S., ii, fol. 9a:-
de . la. med .na. mi. abyun .ba.
ran. rig. rnam. pas. adod. ce. na.
P.S.V., chap. ii, a, fol. 34b :-
rtsod . pa . sgrub . pa . nas . ni . med . na . mi .
abyun . bai . don .mt'on. ba. de. rig .pa. rjes. su. dpag.
P.S.V., b, chap. ii, fol. 116b:-
rtsod. sgrub. par. med(1). na. mi. abyun. bai. don.
mt'on. ba. de. rig. pa .ni. rjes. su. dpag. pa'o.
Now in this sentence we can easily recognize the
definition of anumana quoted and refuted by
Uddyotakara in his Nyaya-varttika, p. 54, apare tu
bruvate nantariyakartha-darsanam tad-vido 'numanam.
It is therefore evident that here also we are
confronted with another fragment from Vasubandhu;
consequently the attribution of this definition to
Dinnaga himself, as suggested by Randle, cannot be
accepted. Anumana presents the five fundamental
aspects(2) which we shall find in the homogeneous
example; that is, we may have nimitta-anum.,
bhava-anum., karma-anum., dharma-anum.,
karya-karana-anum. Only two items of this fivefold
classification can be seen in the list of the
Vaisesika-Sutras (karya, karana, samyogi, virodhi,
samavayi, V.S. ix, ii, 1), while in Dharmakirti we
have, as is known, only anupalabdhi, svabhava, and
karya. But, as me should expect, the section which is
largely developed is that dealing with the syllogism.
This is divided into two parts, a probandum, sadhya,
and a proof, sadhana. The proof is said to be
eight-fold; but the eight members are in fact reduced
to five only, as the last three are nothing but the
pramanas already referred to.
The first thing that we must point out is that
the probandum is considered as separate from the
syllogism itself; it is not the pratijna or
proposition. This probandum can be of two kinds,
either an "essence" or a "quality", svabhava or
visesa. In the first case the mere existence or
non-existence of the subject can be predicated; e.g.
"the atman is", "the atman is not." In the second
case the probandum is a particular predicate which
must be proved as belonging or not belonging to the
subject, e.g. "the atman is all-pervading", "the
atman is not all-perrvading." This notion of the
sadhya is ---------------- 1. Xyl., byed. 2. Randle,
Fragments from Dinnaga, p. 21.
common to both A and B; but, if we consider the five
avayavas, which constitute the syllogism, the
difference between the two groups of texts is
greater. Pratijna, hetu, drstanta occur in both
groups, although there is some difference as regards
the various terms. But the last two terms are
enunciated in a quite different way. While in B we
find the same terms as in the Nyaya-sutras, which
occur also under another name in Prasastapada, in A
we have only "homogeneity" and "heterogeneity", which
are nothing but two different aspects of the drstanta
itself, as K'uei Chi already recognized.
This fact is worthy of notice, because it shows
that, while in the first instance Asanga followed the
ancient scheme, as handed down in the various
Tarka-sastras, or Vivada- sastras, in his greater
work he acknowledges that the last two members are
superfluous, thus practically reducing the syllogism
to three members only, as it is proved by the
additional notes with which he concludes the section
that we are studying. If we were to follow the
explanation of Sthiramati, we should be compelled to
admit that a three- membered syllogism is also
expounded in the Sangiti. But I do not think that his
interpretation is exact. Although the definition
given by the Sangiti is not perfectly clear, it seems
that upanaya consists for Asanga, in referring to the
subject the analogous facts ascertained by the
example, in order to prove the attribute expounded in
the proposition. Sthiramati lived long after Asanga,
when Buddhis logic, chiefly through the speculations
of the Tarka-sastras of Vasubandhu and Dinnaga, had
reached a well-developed and advanced stage. At that
time the syllogism was generally considered to be
composed of three terms only; so that in order to
bring the Sangiti into accordance with the new
theories without altering the textual reading of the
book, Sthiramati, who according to the Chinese
sources was well versed in logic (BEFEO. 1911, p.
379), tried to give the terms another meaning. In
fact, the syllogism that he gives as an
instance is really composed of three members only.
The other two are meant to express that other
attributes, proved by the same reason, can be
predicated of the subject. That the reason "because
it is a product" can prove the noncternity as well as
the absence of atman is accepted by Dinnaga also and
Therefore I am inclined to think that in the
Sangiti we have, in fact, the traditional type of
syllogism of five members, which Sthiramati, in his
commentary, endeavours to explain in accordance with
the new theories. If it be so, A would represent the
first text in which we find an attempt to decrease
the members of the syllogism.
We can represent the theories held by Asanga
concerning the syllogism in the following way:--
sadhyu sound "non-eternal"
pratijna "sound is non-eternal"
sadhana hetu "because it is a product"
drstanta homogeneity "as a pot"
heterogeneity, "as the ether"
sadhya as before
Be that its it may, the fact remains that we do
not find in Asanga any trace of the theory of the
threefold aspect of a "reason'', the trairupya, which
certainly representsthestartingating point of the
new logic. At least we have no grounds either for
affirming or denying that Asanga must be credited
with this innovation, which at any rate is very far
from that perfection of elaboration which is the
chief merit of Dinnaga's logic. At any rate, we know
that some of the Tarka-sastras expounded a
five-membered syllogism, while in the Chinese sources
this reduction of the syllogism to three members
is generally attributed to Vasubandhu, a statement
that is supported by Vacaspati Misra himself.(l) For
Vasubandhu the three avayavas are pratijna, hetu and
drstanta. We shall see subsequently his definition of
the "reason". So far as the pratijna is concerned, we
know from Uddyotakara that the definition proposed by
the Vada-vidhi was sadhyabhidhanam pratijna (N.V.
117). This definition might at the first glance
appear similar to that given by the Naiyayikas; but
this is not the case, as the word sadhya has in the
definition of the Vada-vidhi a different and peculiar
meaning; here, in fact, sadhya is understood as
paksa-dharma, where paksa is the object to be proved
in the course of tile discussion. This can be
gathered from the full definition quoted in the
Pramana-samuccaya-vrtti, chap. iii, fol. 45b,
rtsod.pa.sgrub. par.ni. bsgrub. byar. brjod. pa.tsam.
dam. bca'. bar. agyur. ba. ma. yin. gji. adi. ltar.
p'yogs. bsgrub. bya. yin. no. p'yogs. de. ci. cig.
rnam. par. dpyad. pal. adod.pai. don. te. P.S.V. b,
127b: rtsod.pa. bsgrub. par. ni. bsgrub. bya. brjod.
pa. tsam. dam.bca'. ba. ma. yin. gyi. 'on. kyan.
p'yogs. kyi. c'os. bsgrub. bya'o. p'yogs. gan. yin.
pa. rnam. par. dpyad. par. adod. pal. don. p'yogs.
yin.te. This means that the definition
sadhyabhidhana-matram has not the same meaning as in
the Nyaya-sutras and therefore is not subject to the
refutation that Dinnaga made of the N.S. Sadhya is
said to have here a technical sense. The original of
the sentence, which is evidently composed of two
fragments put together by Dinnaga, in order to show
how the Vada-vidhi interpreted the definition, can
easily be restored into Sanscrit: Vada-vidhau
sadhyabhidhana-matram pratijna na bhavati, api tu
paksadharmah sadhyam.pakso vicaranayam isto 'rtho.
The restoration is obvious, as the second part of the
definition is also to be found in the Nyaya-varttika
(p. 106) in a place where Uddyotakara refutes the
Buddhist theories concerning the paksa. In this way
we have also identified
1. N.V.T., p. 288 (Benares ed.), atra Vasubandhuna
pratijnadayas trayo 'vayava dur-vihita
Aksapadalaksanenety uktam; cf. N.V., p. 136.
another of the manifold doctrines criticized by
Uddyotakara in the course of his work without giving
the name of their author.
No allusion, so far as I know, call be found in
the Nyaya-varttika to the theory of the "example"
expounded in the Vada-vidhi; but, fortunately, the
definition of the drstanta given in that book has
been preserved by Dinnaga; and from this it appears
that according to Vasubandhu the example is the
expression of the relation between the reason and the
sadhya. P.S.V. a, chap. iv, fol. 70b: rtsod.pa.sgrub,
stc. P.S.V.b, fol. 154a:
rtsod.pa.sgrub.par.ni.de.dag.abrel. bar. bstan. pa.
gan. yin. pa. adi. brjod. pa. dpe. yin. te.
But what about the " reason "? Must the
tri-laksana theory of the hetu be really attributed
to Dinnaga or is it an innovation of Prasastapada?
Or are there proofs through which we can safely
assume that it was anterior to both? Our sources show
beyond any doubt that the tri-laksana theory was
known to the Buddhist schools before Dinnaga.
First of all, we gather both from K'uei-Chi and
from Shen- T'ai(1) that the theory of the vi-paksa
was known to the ancient masters, who held two
different opinions about it, which were not accepted
by Dinnaga. Some thought that the vi-paksa is that
which excludes the sa-paksa as well as the paksa; so
in the syllogism "sound is non-eternal, because it is
a product, like a pot" the vi-paksa "ether" excludes
the contrary of the non-eternal as well as of the
pot. On the other hand, other logicians said that the
vi-paksa is everything except the non-eternal, while
for Dinnaga, as is known, vi-paksa is yatra pakso na
vidyate. We find here the same terms and elements
which are peculiar to the definition of the reason as
given by Dinnaga or in the versus memoriales quoted
by Prasastapada. Moreover, the actual
1. K'uei-Chi, chap. iii, Shen-T'ai, chap. ii. Even
for the Tarka-sastra preserved in Chinese (see
above, pp,. 452 sqq.) the third laksana of the
hetu is vi-paksa-vyavrtti.
definition of the hetu contained in the Vada-vidhi
and refuted by Dinnaga confirms the Chinese sources.
In fact, in that book the reason is said to consist
in the enunciation of that dharma which is
non-existent where there is no attribute analogous to
P.S.V. a, chap. iii, fol. 57b:
rtsod.pa.sgrub.pa.nas.de. mt'un. med. la. med. pa.
yi. c'os. bstan. rtags. zes. pa.
P.S.V. b, fol. 138a: rtsod.pa. bsgrub.par. ni.
de. lta. bui. med. na. mi. abyun. bai. c'os. ne.
bar. bstan. pa. ni. gtan. ts'igs. so.
Dinnaga objects to the formal exactness of the
definition; but it seems that even for Vasubandhu the
paksa-dharmata, vi-pakse sattva, sa-pakse sattva were
the three fundamental characteristics of the reason.
And, as we shall see later on, me have another text
almost certainly anterior to Dinnaga in which the
three laksana-theory is clearly expounded.
We must now consider the various theories
concerning logical errors. Asanga, in the concluding
portion of A, reduces all the possible logical errors
to the contradictory, which contains two sub-groups,
inconclusive, aniscila or anaikantika and
sadhya-sama. There is no trace of this theory in B,
where allusion to logical mistakes can be found
eventually in the section dedicated to the
On the other hand, the Vada-vidhi knows the same
list of the hetv-abhasas as is accepted by Dinnaga,
that is, asiddha, aniscita, and viruddha. The
definition of these errors, if we are to follow the
statement of the Pramana-samuccaya, was not given in
the Vada-vidhi; but they were only enunciated and
specified through a corresponding example.(1)
1 P.S. V. a, chap. iii, 63b: de. la. re. zig. rtsod.
pa. sgrub. pa. nas. ma grub. pa. dan. ma. nes. pa.
dan. agal. bai. don. ni. gtan. ts'igs. ltar. snan.
ba'o. zes. zer. ro. de. la. ma. grub. pa. la.
sogs. pa. ni. dper. brjod. nas. mts'an.nid. ma
.yin. te. dper. na.mig. gi[s]. gzun. bya. yin.
par. p'yir. mi. rtag. go. zes. bya. ba. ma.
grub. pa. dan. lus. can. ma. yin. pai. p'yir. rtag
. go. zes. bya. ba. ma. nes. pa. dan. bye. brag.
rnams. kyi. dban. po. las. byun. bai. p'yir. mi.
rtag. go. zes. bya. ba. agal. ba. gcig. dan. gran.
can. pai. rgyu. la. abras. bu. yod. pa. yin. te.
yod. pa. skye. pai. p'yir. ro. zes. bya. ba. agal.
ba. gnis. pa'o.
Asiddha: "sound is non-eternal, because it is
perceived by the eye."
Aniscita: "sound is eternal, because it is
The viruddha can be of two kinds, (a) if a
Vaisesika maintains that sound is eternal because it
can be perceived by the senses; (b) if a Sankhya says
that the effect is preexistent in the cause, because
it is born. This means that according to the
Vada-vidhi, as is in fact proved by another passage
of the Pramana-samuccaya, the viruddha-hetvabhasa is
either pratijna-viruddha or siddhanta-viruddha, a
theory that is refuted by Dinnaga in the Nyaya-mukha
as well as in the Pramana-samuccaya.
Thus, gathering and comparing the various
fragments and quotations scattered in the sources
still available, we can supply, in a certain way at
least, the loss of the original texts and attain a
better knowledge of logical theories accepted or
formulated by Buddhist writers before Dinnaga.
The first result of these investigations is that
long before Dinnaga logic, which as tarka or
hetu-vidya was blamed and condemned by the ancient
schools, was accepted at least as a subsidiary
science by the Buddhist doctors and developed
P.S.V. b, fol. 146a: re. zig. rtsod. pa. bsgrub.
par. ni. ma. grub. pa. dan. ma. nes. pa. dan. agal.
ba. ni. don. nid. gtan. ts'igs. kyi.skyon. yin. te.
de. la. ma. grub. pa. la. sogs. pa. rnams. kyi.
mts'an. nid. ma. bsad. par. dpe. rnams. bsad. pa. ni.
dper. na. ma. grub. pa. ni. sgra. mi. rtag. ste. mig.
gis. gzun. bai. p'yir. zes. bya. ba. dan. ma. nes. pa
ni. lus. can. ma. yin. pai. p'yir. rtag. go. zes.
bya. ba. lta. bu'o. bye. brag. par. rnams. kyi. dban.
pos. gzun. bar. bya. yin. pai. p'yir. mi. rtag. go.
zes. bya. bai. agal. ba. gcig. dan. grans. can. gyi.
abras. bu. rgyu. la. yod. pa. yin. te. skye. bai.
p'yir. ro. zes. pa. ni. agal. ba. gnis. pa. yin. no.
P.S. V. a, 46, chap. iii, b: rtsod. par. sgrub.
par. ni. adi. agal. bai. gtan. ts'igs. ltar. snan. ba
nid. kyi. k'on. du. bsdus. te. mts'an.nid. de. lta.
bu. las. ni. agal. ba. dan. ldan. min. de. ni. agal.
bar. rnam. pa. gnis. su. bstan. te. dam. bca'. bai.
don. dan.. agal. ba. dan. grub. pal. dan agal.ba'o.
P.S. V. b, chap. iii, fol. 129a: rtsod. pa. sgrub
pal. yan. adi. agal. bai. glan. ts'igs. nid. du.
adus. pa. yin. gyi. dei. mts'an. nid. k'o. nas. de.
ni. agal. ldan. nin. der. ni. agal. ba. rnam. pa.
gnis, bstan. te. dam. bca'. ba. dan. agal. ba. dan.
grub. pa. mt'a'. dan. agal. ba'o.
on independent lines. Great masters such as Asanga
and Vasubandhu, and perhaps many others whose names
are lost, perfected the ancient rules of discussion,
katha or vivada. Asanga was, as far as we can guess,
the first to introduce hetu-vidya in his dogmatical
The growth of the great philosophical systems,
the codification, so to say, of the sutras, the
blossoming of a large dogmatical literature, devoted
to commenting upon them, involved the sects in many
discussions and struggles, through which not only
were vivada and its rules perfected, but mere
heuristic began to leave the place to logic and
epistemology, an achievement for which Dinnaga was
Even for Vasubandhu logic was Still a section of
vada; and, in fact, all the books written by him on
this ropic seem to have had the title vada. According
to Shen T'ai some of his works were:--
 論心, Lun hsin, Vada-hrdaya, a title which
reminds us very much of the *Upaya-hrdaya. The
restoration as Vada-kausala, proposed by
Vidyabhusana, is untenable.
 論式, Lun shih, which is the
Rtsod.pa.sgrub.pa of the Tibetan sources and the
Vada-vidhi of Uddyotakara, the fragments of which we
have collected in this paper.
 論軌, Lun kuei. This Vidyabhusana restores
arbitrarily as Vada-marga. In a previous paper I had
no definite suggestion to advance.(l) But now I think
that more precision is possible: 式 shih and 軌 kuei
are synonyms in Chinese; therefore we have to suppose
that even in the Sanscrit original two synonyms were
used, vidhana conveying the same meaning as vidhi,
just as shih is equivalent to kuei. The Sanscrit
sources confirm this hypothesis; in fact, a
Vada-vidhana-tika is quoted in the Nyaya-varttika, p.
117, yad api Vada-vidhana-tikayam sadhayatiti
sabdasyatiti svayam parena ca tulyatvat svayam iti
visesanam sadhayatiti kilayam sabdah prayojye
prayoktari ca tulya-rupo bhavatiti... We
1. See my Notes on the fragments from Dinnaga, JRAS.
find here expressed the same theory as already met
with in that passage of the Samyukta-sangiti in which
Sthiramati, who, as is known, was a follower of
Vasubhandhu, is commenting upon the definition of the
pratijna contained in the Sangiti. Therefore I think
that we call safely restore the Chinese 論軌 as
Another conclusion that seems to follow from the
material collected is that tile question of a mutual
borrowing between Dinnaga and Prasastapada must be
dropped. The fact that the theory of the three
laksanas of a reason was known before Dinnaga rather
implies that each master took it, although perhaps
developing and formulating it in a better and more
organic way, form some other, previous, school of
Vada-sastra which, in this respect at least, held
different views from those expounded in the
Nyaya-sutras. That this is really the case is proved
by the fact that we have another text anterior to
Dinnaga in which the tri-laksana theory is clearly
enunciated. This text is the Tarka-sastra,(1) which,
if we are to judge from the Chinese sources, enjoyed
a very great authority not only in India, but also in
Central Asia and in China. Dharmagupta studied that
book while residing in Kucha. Paramartha translated
it into Chinese and commented upon it.(2) We do not
know its author; but it is evident that the present
redaction of the text, as it has been handed down to
us, was written by some Buddhist. Now in the second
section of this book, dedicated to the jatis, under
the item sadharmya-khandana or sadharmya-sama, we
read the following sentence 相離, which translated
into Sanscrit runs thus:-- asmabhis tri-laksano hetuh
pratisthapitah; tad yatha paksa-dharmah
sapaksa-sattvam vipaksa-vyavrttih. References to the
same doctrine can be found in other passages of the
1. On the 如實論 see Ui's Studies in Indian
Philosophy, vol. i, 222.
2. The commentary written by Paramartha was called 如
實論疏. Cf. BEFEO. 1911, p. 351, n. It is lost.
Although the text contains a list of
nigraha-sthanas which is almost identical with that
in the Nyaya-sutras, it presents also some very
precise similarities to views accepted by Vasubandhu.
We know from Dinnaga that the theory of the jatis, as
expounded by Vasubandhu in his Vada-vidhi, was
different from that expounded by Aksapada. Vasubandhu
divided all the possible cases of jatis into three
groups, viparita, abhuta, viruddha, and in each of
these were comprehended various sub-groups, which
have been quoted by Dinnaga in the following way:--
P.S.V. a, chap. vi, fol. 94a: rlsod. pa. sgrub.
par.ni. p'yin. ci.log.dan. yan. dag. pa. ma. yin. pa.
. nid. dan. agal. ba. nid. rnams. kyis. lan. skyon.
brjod. pa. yin. no. zes. brjod. do. de.la. p'yin. ci.
log. ni. c'os. mt'un. pa. dan. c'os. mi. mt'un. pa.
dan. rnam. par. rtogs. pa. dan. bye. brag. med. pa.
dan. p'rad. pa. dan. ma. p'rad. pa. rnams. la. ni.
gtan. ts'igs. dmigs. sin.. abras. bu. mts'uns. pa. la
sogs. pa. ni. t'e. ts'om. du. brjod. do.
P.S.V. b, fol. 177a: rtsod. pa. sgrub. par. ni.
p'yin. ci. log. dan. yan. dag. pa. ma. yin. pa. dan.
agal. ba. rnams. ni. lan. gyi. skyon. zes. bsad. pa.
yin. no. de. la. p'yin. ci. log. pa.ni. c'os. mt'un.
pa. dan. mi. mt'un. pa. dan. rnam. par. rtog. pa. dan
. k'yad. par. med. dan. gtan. ts'igs. dan. p'rad. pa.
dan. dmigs. pa. dan. t'e. ts'om. dan. ma. brjod. pa.
dan. abras. bu. mts'uns. pa. la. sogs. pa'o.
P.S.V. a, fol. 95a: yan. dag. pa. ma. yin. pa. ni
t'al. bar. agyur. ba. dan. don.kyis. go. bar.
mts'uns. pa. la. sogs. pa'o.
95b: agal. ba. ni. ma. skyes. pa. dan. rtag. par
.mts'uns. pa. la. sogs. pa'o.
P.S.P. b, fol. 178b: t'al. ba. dan. don. gyis. go.
ba mts'uns. pa. la. sogs. pa. ni. yan. dag. pa.
Ibid.: ma. skyes. pa. dan. rtag. pa. mts'uns.pa.
la. sogs. pa. agal. ba. yin. pa.
The text, especially in tile passage concerning
the viparita, does not seem to be quite correct; but
with the help of the commentary of Dinnaga we can
make a list of the jatis accepted by Vasubandhu
which is analogous to that found in the fragment of
the Tarka-sastra preserved in Chinese, as is proved
by the following scheme:--
sadharmya-sama 1 id.
vaidharmya-sama 2 id.
vikalpa-sama 3 id.
avisesa-sama 4 id.
ahetu(?)-sama 5 id.
prapty-aprapti-sama 6 id.
upalabdhi-sama 7 id.
samsaya-sama 7 id.
avarnya-sama 7 id.
karya-sama 7 id.
arthapatti, etc. id., plus: prati-drstanta
nitya id., plus: svartha-viruddha.
As I have said before, we do not know anything
about the author of this book, or its age; but we may
presume that it was anterior to Dinnaga. It may be
also that this Tarka-sastra, or a redaction of it,
was existent already in the time of Vatsyayana. There
is in Nyaya-sutras, ii, 2.3, an allusion to some
logicians who denied any validity to arthapatti, as
being inconclusive. Vatsyayana, commenting upon this
sutra, writes, asatsu meghesu vrstir na bhavatiti
satsu bhavatity etad arthad apadyate, satsv api
caikada na bhavati seyam arthapattir apramanam iti.
Now we have in the second
section of the Tarka-sastra referred to the following
passage:- 可顯物者。有二種有義至有非義至。 義至者若有
雨必有雲。 若有雲則不定或有雨或無雨, which can be
translated into Sanscrit thus:-- yad abhivyaktam
dvi-vidham, arthapattir anarthapattis ca.yadi vrstir
bhavati tada meghenapi bhavitavyam, meghe saty api tu
kadacid vrstir bhavati, kadacin na bhavatity
anaikantikata. The correspondence is almost perfect;
so we should be inclined to think that Vatsyayana and
even the final redactor of the Nyaya-sutras knew, if
not this same text, then another of those
Tarka-sastras which seem to have existed long before
Dinnaga and in which the criticism of arthapatti was
already formulated. That we can speak of
Tarka-sastras and not of a single Tarka-sastra is
proved by two references to them which can be found
in the Pramana-samuccaya-vrtti. In both cases
Dinnaga uses the plural; moreover, the second
fragment clearly shows a doctrine of the syllogism
quite different from that contained in the text
translated into Chinese. The first quotation is to be
found at the beginning of the third chapter
dedicated to the pararthanumana:-
P.S. V. a, chap. iii, 44: gal. te. rjes. su.
dpag. par. bya. ba. ston. pa. ni. dam. bca'. ba.
brjod. par. bya. ba. ste. rtog. gel bstan. bcos.
rnams. su. gzan. gyi. don. rjes. su. dpag. pa. dgon.
pa. de. ji. ltar. yin. ze. na.
P.S.V. b, 126b: hon. te. rtog. gei. bstan. bcos.
rnams. su. gzan. gyi. don. rjes. su. dpag.pa. la.
rjes. su. dpag. par. bya. ba. ston. pa. dam.
bca'.ba.bkod.pa.gan.yin.pa, which perhaps corresponds
to an original like this:--yady anumeyabhidhanam
pratijnavacanam iti Tarka-sastresu pararthanumanam.
The other quotation occurs just at the beginning of
the fourth chapter, in which Dinnaga expounds his
doctrine of the "example ":-
P.S.V. a, chap. iv, 66b: rtog. ge. pai. bstan.
bcos. rnams. su. ni. p'yogs.kyi.c'os.nid. tsam
.gtan.ts'igs.kyi.sbyor. ba. yin. no. zes. grags. te.
dper. na. adi. byas. pai. pyir. zes. pas. sgra. mi.
rtag. par. go. bar. byed.pa. lta. bu'o.
P.S.V.b, fol. 149a: rtog. gei. bstan. bcos.
rnams. su. sbyor. ba. la. gtan.ts'igs. zes. bya. bas.
p'yogs. kyi. c'os. tsam. nid. bstan. pa. yin. te.
dper. na. byas. pai. p'yir. zes. bya.ba. adir. sgrai.
zes. bya. ba. rtogs. pa. yin. no.
These two translations do not perfectly agree,
but their meaning is clear. According to the
Tarka-sastras the indissoluble connection between the
major and the middle term in the syllogism is
expressed by the paksa-dharma, and therefore the
"example" is not necessary, that is, the middle term
as residing in the subject of the inference is
sufficient to prove the probandum. The theory, which
we find else-where in the later development of Indian
logic, is not accepted by Dinnaga, who thought the
example absolutely necessary to express the other two
laksanas of the "reason". The same theory is also
referred to and criticized by the Jaina
Nyayavatara,(1) which calls it the theory of the
antar-vyapti. It was certainly not accepted by
Vasubandhu, as Vidya-bhusana thought, but it was at
any rate anterior to Dinnaga, as is sufficiently
proved by the above reference, which shows how far
logical speculations must have advanced even before
the advent of the great Buddhist thinker. This very
important development of logical schools in the
period between Asanga and Dinnaga,of which we have
unfortunately some fragments only, must change our
ideas of the authorship of the various theories which
we find in the texts handed down to us, and also Of
the relation between the various authors. We must
1. Nyayavatara 20: antar-vyaptyaiva sadhyasya siddher
bahirudahrtih vyartha syat tad
a-sad-bhave 'py evam nyaya-vido
that is, a syllogism like this, "on the hill there is
fire, because there is smoke," is perfectly valid, as
there is an inner indissoluble connection between
the major and the middle term and therefore the
example " as in the kitchen" (bahir-vyapti) is not
necessary. This theory cannot be attributed to
Vasubandhu, as suggested by Vidyabhusana, History of
Indian Logic, p. 268, n. (and in nis edition of tile
Nyayavatara, Calcutta, 1909, p.l7). That Vasubandhu
formulated the syllogism in three members is proved
by what we already said and by the clear statements
of K'uei-hi and Vacaspati Misra.
acknowledge that perhaps the treatises which still
remain are but a small part of all which was written
regarding this subject by some generations of
thinkers. The similarity that we can find between
this and that author does not imply a mutual
borrowing, but can be quite well explained as due to
the fact that either writer was following some
previous authoritative text or original.
Owing to distance from England I was not supplied
with proofs of my article, in which, consequently,
there are a few misprints and mistakes.
p. 451, l. 3: read Giuseppe.
p. 454: The Yogacarya-bhumi-sastra has been
proved by Ui to be by Maitreya: Studies in Indian
Philosophy (in Japanese), i, p. 359, and Zeitschrift
fur Indologie und Iranistik, Band 6,
Heft 2. So, while at first Asanga followed his
guru's views, he then altered his opinions.
p. 453, l. 2: The second Chinese character should
be 揚. ibid., l. 4: The fourth Chinese character
should be 毗; so also in l. 8.
p. 458, l. 23: under item(a): gton. ba.
p. 459, under item (9): siddha-sadhya: corr.
"when the probandum is already proved."
p. 461, l. 15: under (1): dam.bca'..ba.
p. 464, l. 24: t'ag. rin. ba.
p. 466, l. 10: The Chinese character must be read
after "conventional assumption."
p. 479: Even this definition of the drstanta is
in Uddyotakara; see my article on the "Vada-vidhi,"
Indian Historical Quarterly, vol. iv, p. 634.
p. 484, ll.26-7: t'al. bar. Instead of K'uei Chi
and Shen T'ai read K'uei-chi, Shen-t'ai.