Chapter 8: Examining Definitions
The longest chapter of the treatise, and the one it has been building towards: the definition of valid cognition. It opens with the general account of definition, definiendum and illustration — proved by self-awareness, and pervading every object of knowledge — much of which is provoked by the refutation of the early Tibetan epistemologists. It then fixes the definition of pramāṇa itself, the basis and essence of its classification, the fixed number of its kinds, and its etymology, and shows how superimposition is eliminated.
Condensed from Khenpo Sodargye’s Chinese lecture transcripts on the Pramāṇayuktanidhi, and rendered into English. Root verses are given as the lectures give them; the prose beneath summarises the exposition. On the provenance and the limits of this rendering, see the series introduction.
Chapters: 1. Objects · 2. Cognition · 3. Universals and Particulars · 4. Proof and Exclusion · 5. Expressed and Expresser · 6. Connection · 7. Contradiction · 8. Definitions · 9. Perception · 10. Inference for Oneself · 11. Inference for Others
Ascertaining the Nature of the Valid Cognition that Knows [Objects]
Two parts: (1) positing the definition; (2) ascertaining the meaning of each instance. Positing the definition has three parts: the general account of definition, definiendum, and illustration; specifically ascertaining the definition of valid cognition; ascertaining the meaning indicated by that definition.
The chapter’s weight falls on the definition of valid cognition. The second chapter gave much space to invalid cognition; here the genuine article is set out in detail, since epistemology just is the discipline that states what valid cognition is. The debates about definition, definiendum, and illustration that open the chapter are largely provoked by refuting the early epistemologists; they are intricate but carry less soteriological weight than the later question of the Buddha as a valid being.
I. Positing the Definition
a. General Account of Definition, Definiendum, and Illustration
1. Establishing that the Three Phenomena Pervade All Objects of Knowledge
The three phenomena pervade all objects of knowledge; Therefore their positing is here explained.
Definition, definiendum, and illustration reach every object of knowledge — everything subsumed within saṃsāra and nirvāṇa, conditioned and unconditioned alike. Whatever can arise before the mind has all three, as space pervades the world.
The proof is self-awareness. When an object comes before the mind, the feature that marks it off from every other phenomenon is its definition; the ox’s hump and dewlap, borne by no other animal, or “understanding speech and meaning” for the human. The definiendum is the phenomenon graspable, unconfusedly, by convention and cognition together: on the basis of the definition people fix a name, and name plus the object held in mind, unmuddled, form the definiendum — “ox”. The illustration is what bears the feature, and since definition and definiendum are both universals, they need a representative concrete thing to stand on: the dappled ox one can point at.
Non-entities have all three too. “The son of a barren woman” raises an image bearing a feature distinct from other sons, and that image is its definition, its name the definiendum; the representative aspects can be posited by the exclusion-cognition, though not upon a particular as with the vase (function as definition, “vase” as definiendum, a golden vase as illustration).
The practical point: whatever field one studies, knowing a thing means knowing its definition, definiendum, and illustration. Without them a phenomenon is easily confounded with others; with them it can be expressed without difficulty.
2. Ascertaining the Nature of the Three Pervading Phenomena
Recognising the Essence
The definition is the means of knowing — a phenomenon of the object; The definiendum is what is to be known — a phenomenon of the mind; The illustration is what is to be known — the supporting phenomenon. These are the means of proof and what is proven.
The definition is the means of knowing, the establishing cause: through hump and dewlap one comes to know the ox. It is called a phenomenon of the object because, from the standpoint of conceptual grasping, it is neither the name nor the representative animal but a feature we take to sit on the external thing itself. It must not, however, be supposed to exist truly out there. If it did, then whoever saw hump and dewlap would necessarily know the animal to be an ox — even someone who had never met one — which is false. Even the definition is to be conjoined with the exclusion-cognition.
The definiendum is a phenomenon of the mind: the result known through the definition, imputed and not existing in the external object. So definition and definiendum stand as cause of knowing and result of knowing.
The illustration is what supports both: hump and dewlap must sit on some animal, and the name “ox” must refer to some dappled or black beast. Hence the three are related as proof and probandum, meaning knower and known — one asks what animal bears hump and dewlap, learns the name “ox”, then goes to the cattle-pen and sees the thing. Their relation of support and supported works like a seal and its impression: what is carved on the seal is known only through the design left on this wiki.
All three are positings of the exclusion-cognition; the early epistemologists’ failure to see this is the root of the debates that follow. Whether one calls the definition a phenomenon of the object and the definiendum a phenomenon of the mind is a matter of the standpoint of grasping — in fact one essence becomes many, and many become one, before exclusion.
The Definition of Each
General Statement
All three phenomena require their own reason; Without it, everything becomes confused.
Why are hump and dewlap called the definition and not “ox” or “dappled”? Each of the three needs its own uncommon reason for being posited as it is. Without such reasons, phenomena would become definition and definiendum indifferently and everything would run together.
The Reason for the Definition
Refuting the View that a Definition Needs No Definition
[Objection:] Some say the definition is a substantial entity, and so needs no [further] definition — to need one would be an infinite regress. Some say that in meaning it needs none, but for establishing the convention one is required. Others say it needs one in meaning too, and its definiendum is likewise [endless], as is the reason — endlessly.
Three positions of the early epistemologists.
[Early Epistemologists, first position] The definition is a substantial entity, and being substantial it fully indicates the essence; a second definition would start a regress that would leave the root definition itself unestablished. Sakya Paṇḍita’s tradition answers that a definition needs no further definition only once three applications are reached; before that, since every object of knowledge bears the three characteristics, one must ask whether the definition of the definition is an object of knowledge. If it is, it has all three; if it lacks a definition, it is not an object of knowledge at all — contradicting the opening thesis.
[Shawa Chözang and others] In meaning no definition is needed, but one is needed to establish the convention. Untenable: a definition that holds only in name and not in meaning is a will-o’-the-wisp.
[A third group] A definition needs a definition in meaning too, endlessly; and the definiendum likewise has a second definiendum, a third, without end; and so with reasons, since to show “sound is impermanent, because it is produced” correct one would need a second three-mode-complete inference, then a hundredth, without end.
[Reply:] If the establishing cause is undetermined yet the established effect is determined, Then the definition has become unnecessary; And if [the effect too is] undetermined, the endless [regress] collapses everything.
One dilemma disposes of all three positions. While your establishing cause — the definition — is still undetermined, is the established effect, the definiendum, determined or not?
If it is, the definition is pointless: one would have understood the ox entirely without understanding hump and dewlap, exactly as if impermanence were known while “produced” remained unknown, leaving the reason no function whatever.
If it is not, the endlessness thesis destroys itself. Granting an endless series of definitions, no final definition is ever reached; and if the definiendum cannot be known while the definition is undetermined, then on the opponents’ own terms the moment of knowing the definiendum never arrives.
[Objection:] The illustration of the reason establishes the meaning of the probandum, And so there is no fault of infinite regress.
[Early Epistemologists] No regress touches us, since one knows not only through the definition but through the illustration of the reasoning. In “sound is impermanent, because it is produced”, the definition of the reason is completeness in the three modes and its illustration is “produced” itself; the illustration suffices to establish the probandum.
[Reply:] If the three modes and the dewlap are undetermined, Then by determining merely “produced” and “dappled,” Impermanence and ox should also be known.
Then merely knowing “produced” would make impermanence fully clear, so that no one who knew “produced” could ever superimpose permanence on the pillar; and merely seeing a dappled animal would suffice to know it an ox, so that a Han Chinese who had never been to India and knew nothing of oxen would recognise one at sight. Neither follows. One must rely on the definition: this dappled animal is an ox, because it possesses hump and dewlap.
Khenpo Sodargye adds a note of respect: the early epistemologists’ writings are rich, and their contribution to the field very great. Their divergence lies in placing exclusion, particulars, definition, and illustration in the external object, which does not accord with Dharmakīrti’s intent — and it may be that they displayed themselves so deliberately, to give the later masters something to sharpen wisdom against.
The parallel question about the definiendum is then pressed. [Opponent] Does your definiendum possess a graspable object — that is, a definition? If not, it lacks a definition and so is not an object of knowledge. If it does, the first definiendum has become the definition of a second definiendum, that of a third, and so on: what is the name of “ox”, and the name of that name? The little monk is called Dongdong; has the name “Dongdong” a name? If not, no one can know his name; if so, Dongdongs multiply until they fill the heavens.
[Reply:] If so, branches would be likewise: without branches it is not a tree; If it has branches it is a tree — and so its definition too would be endless.
Sakya Paṇḍita answers by posing the same question on parallel ground. The definition of a tree is possessing branches and leaves. Take “sandalwood is a tree, because it possesses branches”: does the branch used as reason itself possess a branch? If not, the branch is not a tree, since it lacks the definition; if so, ask the same of the second branch, and the branches run on without end.
The framing matters. Before the exclusion-cognition an endless series can indeed be generated — one may ask the colour and shape of the horns of a rabbit’s horns, and nobody can forbid it. But in positing definition and definiendum one cannot cut loose from the thing judged in the external object; once name and meaning are established as inseparable, the purpose is accomplished and further division is idle.
[Objection:] Though the branch has no further branches, the branch itself establishes [the tree], Being of one body with the sandalwood tree; hence there is no infinite regress.
[Early Epistemologists] Neither fault touches us. In establishing the sandalwood by its branch, the branch has no further branch, yet it is not thereby a non-tree, because when it establishes, it is established as inseparable from the sandalwood; the branch is not split off as a separate body and set against the tree like fire against water.
The same holds for “the pillar is impermanent, because it is produced”. One should not ask whether “produced” is itself impermanent — if yes, reason and thesis collapse into one; if no, impermanence cannot be established through it. Pillar, impermanence, and produced are one in essence though separable before conceptual thought, and the argument is applied with the three joined.
This answer is precisely what Sakya Paṇḍita wanted, since it hands him his own reply.
[Reply:] Even though the definiendum has no second [definiendum], the definiendum’s own definition Establishes the essence of the knowable; so how would the definiendum become an infinite regress?
The opponents’ own method of establishment is accepted and turned back. The animal possessing the dewlap is named “ox”; “ox” has no second name, yet is not therefore unknowable, because the argument “this dappled animal is an ox, because it possesses the dewlap” establishes name and thing as inseparable. Once established, the “second name” of ox is simply ox. Divide in the mind as much as you please — the son of a barren woman may have a son of a barren woman for ever — but for establishing a phenomenon the work is already done.
Establishing Our Own Position
When the bond of name and meaning is understood, At that moment the convention is established.
“Name” is the definiendum, “meaning” the thing bearing it. When, before the exclusion-cognition, saying “ox” brings to mind the animal possessing the dewlap and that animal is replaced by the name “ox”, the relation is established and the convention with it. There a full stop is placed. Asking whether the definiendum of the definiendum has a definiendum is a foolish exercise with no purpose. The analogy: once one has thought of India, said India, and arrived in India, the aim is achieved; there is no need to go on looking for India before conceptual thought.
The reason and the definition, these two, Each have two kinds — general and particular; General and particular are completed in three formal arguments; After the fourth, none is needed.
Reason and definition each divide into general and particular. Generally, smoke is a correct reason because in dependence on it a probandum can be established, being complete in the three modes; particularly, smoke is a correct reason because fire on the mountain can be proven by it. Likewise hump and dewlap are the general definition because they eliminate direct contradiction and establish the meaning-isolate, and the particular definition because they do so with respect to the ox. The same double treatment applies to the definiendum (“ox” is a name argued for with the meaning-isolate as its reason) and to the illustration (“dappled” is the support of the definition) — six correct reasonings in all.
Three formal arguments exhaust the matter. First the root reason: sound is impermanent, because it is produced. Second, meeting the doubt whether the reason is genuine, the meaning-establishing argument: “produced” is a correct reason, because it is complete in the three modes. Third, meeting the doubt whether completeness in the three modes really counts as a correct reason, the convention-recalling argument: because the epistemologists at the outset named it so. A fourth is never needed; anyone who will not accept the third is simply dull, and there is nothing more to be done. The definition runs in parallel: this animal is an ox, because it possesses hump and dewlap → hump and dewlap are the ox’s definition, because they eliminate direct contradiction and establish the meaning-isolate → that phrase may be called the definition, because the epistemologists posited it so.
[Objection:] If the three-mode [reason] has no [further] three-mode [reason], it is not a reason; If it has one, there is an infinite regress.
[Certain masters] Take “there is fire on the mountain, because smoke is seen”. Does this three-mode reasoning possess a second three-mode reasoning? If not, it is a pseudo-reason, being incomplete in the three modes; if so, the series never ends.
[Reply:] The three modes are simply the phenomenon of smoke alone, Inseparable, with no other phenomenon.
As with branch and sandalwood: one does not establish by casting the reason off from what it establishes. The three modes and smoke are applied together and are inseparable; further, fire is established by smoke through their relation of non-occurrence-without, and smoke apart from fire serves as no reason at all. Only if smoke alone were taken as the subject of judgement would the regress threaten, and no reasoning works that way.
By this same reasoning is also removed The supposed fault that the definition becomes an infinite regress.
The same inseparability disposes of the parallel charge against the definition. Khenpo Sodargye adds a practical rule for debate: if the opponent argues from object-phenomena, do not skip across to mind-phenomena; if a single question mixes the two, separate them, and mark also what has been established by grasping universal and particular as fused.
Positing the Needed Definition
[Objection:] [If you say] the reason that establishes the definiendum [is the definition of the definition], Why is the definition of the reason not the same as this?
[Ngok Lotsāwa and other early epistemologists] If the ground that establishes the definiendum is the definition of the definition, then by parity the definition of the reason should be “the means of knowing the probandum”, since among subject, probandum, and reason the probandum is known through the reason.
Refuted: “produced” is a means of knowing, so on this account anyone who knew “produced” could not fail to know impermanence, and every phenomenon that serves as a means of knowing — pseudo-reasons included — would become a correct reason. If the account held, no superimposition about impermanence could survive in the continuum of one who knows “produced”.
Some hold that [the definition of the definition] is the completeness of the three phenomena.
[Shawa Chözang; also the Collected Topics tradition and some later Gelugpa masters] The definition of the definition is completeness in three conditions: that it be the general definition; that it not be other than its own definiendum; and that it be established upon the illustration. Hump and dewlap must be the ox’s definition, must fall within the scope of “ox”, and must be found on the dappled or black animal.
Because it would depend upon the definiendum, this is untenable.
The middle condition is fatal. If the definition can be posited only in dependence on its definiendum, then to know hump and dewlap one must first know “ox” — and to know “ox” its uncommon feature must already be clear. The definition is supposed to be the means of knowing and the definiendum what is known; make the definition depend on the definiendum and the definition has no function left.
[Our own:] It is the eliminating of direct contradiction and the establishing of the meaning-isolate.
“Eliminating direct contradiction” is casting off everything outside its own class — every animal and thing without hump and dewlap. “Establishing the meaning-isolate” means establishing before the exclusion-cognition an isolate that cannot be confused with the definiendum and illustration beneath it: from the angle of substance the three are one body within a single argument, but from the angle of isolate the definition’s isolate is not the definiendum’s and the definiendum’s is not the illustration’s. Without that, the definition of the definition would over-pervade the other two.
Applied: the ox has a definition, hump and dewlap; hump and dewlap have a definition, “the phenomenon that eliminates direct contradiction and establishes the meaning-isolate”; and there the series stops, since within one argument only a second application is required.
The Detailed Classification of the Faults it Eliminates
The Opponent’s View
[Objection:] The masters of the Land of Snow assert That the faults of a definition are subsumed under three.
[Tibetan early epistemologists headed by Shawa Chözang] The same three conditions restated in reverse, as faults to be avoided. First, “the own-isolate is not established as a substance” — as in “this animal is an ox, because it is an ox”, or “he is human, because he is human”, where the definition’s own essence never gets established. Second, “the meaning-isolate transferred elsewhere” — as in “this is a horse, because it possesses hump and dewlap”, where the definition cannot approach the definiendum at all; the auto-commentary notes that non-pervasion, over-pervasion, and impossibility are all folded in here. Third, “not abiding upon the illustration” — “this sheep is an ox, because it possesses hump and dewlap”.
Refuting that View
[Reply:] If you regard [the fault] as occurring when a formal argument is applied, then “the meaning-isolate is transferred elsewhere” cannot arise; If you deem it a fault when no [argument] is applied, then “not abiding in the illustration” becomes meaningless.
A dilemma: are these faults posited by way of a three-membered argument or without one?
If by way of an argument — “this animal (a non-ox) is a horse, because it possesses hump and dewlap” — then no transfer of the meaning-isolate can occur, because with such a subject the reason has no relation to the definition at all and the argument cannot be established in the first place.
If without an argument, then “the definition does not abide upon the illustration” loses all sense, since a definition is established upon its illustration only through an argument. With no argument at all, there is no knowing whether the thing is an ox.
Only the latter two of the three alleged faults are refuted; the first is left standing.
If these faults were tenable, Why did the crowning ornaments of the learned not assert them?
Had the definition of the definition really carried these three faults, Dignāga and Dharmakīrti would have stated them in the seven treatises. They did not, and the many learned have not adopted the scheme; constructing tenets out of one’s own conceptual thought when there is no need runs against correct reasoning.
Establishing Our Own
[Our own:] Non-pervasion, over-pervasion, and impossibility — These are the general faults of a definition. As the negations of the meaning of the term, these three alone, no other.
Dharmakīrti himself taught avoidance of non-pervasion and over-pervasion when giving the definition of valid cognition in the Pramāṇavārttika, and avoidance of impossibility elsewhere.
Non-pervasion: “this animal is an ox, because it possesses a dappled dewlap” — correct for the dappled ox, but it fails to reach the black ox.
Over-pervasion: “because it has a head”, “can walk”, “has four limbs” — other animals have these too. Note the trap in “because it possesses an ox’s head”: it looks uncommon to the ox alone, but nobody can know an ox’s head before knowing the ox, so this definition depends on its definiendum and repeats the fault charged against the early epistemologists.
Impossibility: “this animal is an ox, because it is a vase”, or “because it is human”.
A definition free of all three carries three negations: the negation of non-existence (the ox does not lack hump and dewlap), the negation of existence-elsewhere (only the ox has them), and the negation of possibility (no other animal could have them).
Removing Disputes
[Objection:] Whether you posit an illustration or not [for the non-pervasion fault]; And where the definition is impossible upon the [illustrating] thing, Does it negate the definiendum or not? These two disputes do not harm our position.
[Opponent, first dispute] In “this animal is an ox, because it possesses a dappled hump and dewlap”, do you posit an illustration or not? If you posit the dappled ox, the argument is simply correct, not non-pervading. If you posit a black ox, the fault is impossibility, not non-pervasion, since a black ox cannot bear a dappled dewlap. If you posit none, then anything at all — vase, elephant, lion — could be the subject, and with no subject fixed no fault can be constituted. So your non-pervasion fault is untenable. (Jamyang Loter Wangpo’s commentary records the opponent’s insistence that with no illustration there is simply no ground for a fault.)
[Reply] We do posit an illustration; without one, correctness cannot be judged at all, so the second horn need not be answered. But we do not judge by taking a single dappled ox as subject; we take the ox in general. Said of the class — “this kind of animal is an ox, because it possesses the dappled dewlap” — the definition covers only the dappled body and not the black or white, so it cannot yield the definiendum or illustration of oxen at large. Non-pervasion stands.
[Opponent, second dispute] Where the definition is impossible upon the illustrating thing — “hump and dewlap are an ox, because they are hump and dewlap” — does the argument negate its definiendum “ox” or not? If it negates it, then hump and dewlap have excluded “ox”, so either oxen lack hump and dewlap or hump and dewlap cannot indicate the ox. If it does not, then the impossible reason has failed to reach some definienda, and your impossibility fault has turned into a non-pervasion fault.
[Reply] Two answers. Hump and dewlap genuinely are the ox’s definition, so no impossibility attaches; and they pervade no other animal, so no non-pervasion attaches. Or, more directly: the argument does negate the name “ox”, because what it establishes is not the ox’s illustration — its subject is set on a mistaken basis. In a correct argument, “this dappled animal is an ox, because it possesses the hump”, nothing is negated, since hump and dewlap pervade all oxen.
Closing the topic: the definition’s own definition is “the phenomenon that eliminates direct contradiction and establishes the meaning-isolate”; the definition’s illustration is hump and dewlap themselves; the definition’s definiendum is the name “hump and dewlap”. On the second round these exist; a third is not needed.
b. The Reason for the Definiendum
Refuting Others
[Objection:] Some hold that definition and definiendum are one substantial entity. And [some] hold the definiendum’s illustration to be [mere] convention.
[Tsangnagpa and other early epistemologists] Two claims. First, that in the external essence definition and definiendum are one taste, one body, since they are established by the meaning-establishing valid cognition and so by a nature-reason, as in “sound is impermanent, because it is produced”. Second, that the illustration of the definiendum is convention.
[Reply:] Since the definiendum has the name for its essence, the sense-cognition that sees the definition Would also become a conceptual cognition; and [it would become] a meaning-establishing [valid cognition], and a real entity too.
The definiendum has the name for its essence, with no real content: “ox” is not the dappled animal but the name of that whole class. Hump and dewlap, by contrast, the eye can see, and so can it see the white, black, or dappled animal that illustrates them. Three faults follow from fusing the two:
- The sense-cognition seeing the definition would become a conceptual cognition, since the definiendum is known only before conceptual thought in dependence on the exclusion-consciousness, never by the eye-faculty.
- “This dappled animal is an ox, because it possesses a hump” would cease to be a convention-establishing valid cognition and become a meaning-establishing one, like “the pillar is impermanent, because it is produced”.
- The definiendum, a mind-phenomenon, would become a real phenomenon, being inseparable from the external hump and dewlap.
This does not conflict with the later verse in which Sakya Paṇḍita himself calls definiendum, definition, and illustration inseparable, since that inseparability is established before the exclusion-cognition, in order to set up a nature-reason.
[That] the definiendum’s illustration [is mere convention] is also untenable: the initial act of naming [is the illustration]; its application [thereafter] is convention.
A thing is named with its definition as reason, and that initial naming is the illustration of the definiendum; the name once in circulation is convention. Broadly a name may be called convention, but narrowly a name is certainly convention while convention is not necessarily a name.
Establishing Our Own Position
The cognition of convention that possesses a reason — This is the definition of the definiendum. The illustration is the [first] naming, given through a reason. Knowing this, one acts through the three gates [of body, speech, and mind], And thereby that convention is formed. It is a name dependent on meaning, And therefore it is imputed existence — Imputed because it is not established as a real entity.
The definition of the definiendum is the cognition that, in dependence on a reason such as hump and dewlap, fuses convention and meaning into one — the conventional name “ox” and the universal appearing before conceptual thought grasping the ox, joined.
The illustration of the definiendum is the first naming: the ancient namers deciding that the animal possessing hump and dewlap “should be called ox”. Thereafter people adopt and reject with the body, say “ox” with speech, negate and affirm with the mind, and that process is convention.
Hence convention is not necessarily a definiendum, nor is a name. A dull person nicknamed “Yak” bears that as a name, but his definiendum is “human”, not “yak”. A definiendum, though, should be convention. And since the definiendum is a name dependent on meaning — “meaning” being the reason, that is, the definition — it is imputed existence, not established as a real entity.
Removing Disputes
[Objection:] If definition and definiendum were not one entity, Then the nature-reason would be untenable.
[Early epistemologists] “This dappled animal is an ox, because it possesses hump and dewlap” should be a nature-reason, and a nature-reason requires reason and thesis to be inseparable.
[Reply:] When the convention and the meaning are mistaken to be one, And one thereby applies the convention, Since this is non-deceptive in worldly affairs, How does it contradict the inference established by common consent?
The conventional universal and the external meaning-universal are fused before the mind and the determined object is attained; applying the name on that basis is tenable, and the world does in fact establish the dappled animal as an ox by hump and dewlap. This is the inference of worldly renown, taught further in the tenth chapter.
c. The Reason for the Illustration
The support of the definition is the illustration; Its division is twofold — true and false.
The illustration’s definition is being the support of the definition, since every definition must be posited on something that bears the feature. The true illustration is the thing genuinely indicated in an argument — the dappled or black animal in “this dappled animal is an ox, because it possesses hump and dewlap”. The imputed (false) illustration is one used in name only: “this horse is an ox, because it possesses hump and dewlap”, where the illustration is inverted. The root text gives the illustration only two lines, the rest being easily understood by analogy.
Positing Each of the Three Phenomena
Three parts: recognising the own-isolate; the mode of connection; and positing what each indicates in connection with the definition.
Recognising the Own-Isolate
Refuting Others
[Objection:] Some hold that the own-isolate of the three phenomena Can appear as its own aspect [by its own power].
[Early epistemologists who did not separate exclusion from appearance] Each of the three has an isolate that appears before cognition by its own power, using no other aspect: the isolate of hump and dewlap, the isolate of the name “ox”, the isolate of the representative ox. Whether they stated this in their own treatises is disputed; Sakya Paṇḍita treats it as what their position forces on them, in the way an opponent’s view in Madhyamaka may be either what he says or what reasoning shows he must accept.
[Reply:] If so, the three phenomena Would become objects of non-conceptual [cognition].
The refuting argument: the three characteristics would become particulars, because they appear by their own power, like the particular of a vase — and then, depending on no exclusion-consciousness, they would show their own essences to the eye and ear. If the three sat on the external particular, whoever saw hump and dewlap would at once know the ox and, with it, its definiendum and illustration, since a single essence cannot have one part that appears and another that does not.
The auto-commentary’s distinction: “particular” covers both the genuine particular existing on the external thing, such as the pillar’s impermanence, and what is called a particular only from the angle of the grasped object of cognition, as the definition is here.
Establishing Our Own
Therefore the isolate of appearance is not the three phenomena; Before the exclusion-[cognition], the own-isolates of the three phenomena Arise — and there is [both] negation and establishment.
If the isolate of appearance were the definition, then the particular hump and dewlap now seen could indicate only this one ox before me, never all oxen of the three times, since a particular’s place, time, and aspect cannot be mixed. Likewise the definiendum: were it posited on a particular, the “ox” I utter would name only one, and the initial naming and the names of countless present and future oxen could not be obtained. And the illustration: were this dappled ox the illustration, every other dappled ox in every place and time would cease to support the definition.
So all three are posited before the exclusion-cognition, as vase, impermanence, and produced are. Posited there, they are three in number and inseparable in essence, yet may be separated as subject, thesis, and reason when an argument requires it; and both negation (“this animal is not an ox, because it lacks a hump”) and establishment (“this animal is an ox, because it possesses hump and dewlap”) can be carried out. This has real purpose, since beings with wrong knowledge cannot see the single as many nor the many as one, and the convention-establishing valid cognition is what reaches them.
The Mode of Connection
Actual Explanation of the Mode of Connection
Definition and definiendum are bound by identity of nature; The illustration is, for the most part, bound only temporarily.
Definition and definiendum stand in a relation of identical nature before the conceptual cognition, like impermanence and produced, or fire and the heat of fire: whatever has a definiendum has a definition and conversely, and the case of one existing without the other never occurs — only that some people fail to recognise it.
The illustration’s link is looser. Usually all three appear together, as in “this sandalwood is a tree, because it possesses branches and leaves”. But “dappled” is not required in every argument about the ox; black or white will serve, and when another is used the dappled illustration simply drops out while definition and definiendum remain.
The Valid Cognition that Ascertains Connection
Because [mind and convention] are mistakenly grasped as one entity And [thus] experienced, the connection of definition and definiendum is established.
The connection is established by self-aware valid cognition. The definition is the meaning-universal and the definiendum the convention-universal; these two are mistakenly grasped as one before the conceptual exclusion-cognition, and that grasping is experienced by self-awareness. That the relation is not tied to any particular set of words is clear from the fact that speakers of different languages replace “ox” with their own term while the uncommon feature and the uncommon name, and the relation between them, remain.
[Objection] If it is a mistake, how can it be correct? Two kinds of mistake must be distinguished. One is wholly unfitting — grasping a pillar where there is none — erring in both appearing object and determined object, and this is invalid cognition. The other, fusing name and meaning, errs in the appearing object but attains the determined object, and so is non-deceptive in worldly affairs. Conceptual thought is mistaken in essence, yet some conceptual thought is held correct for just this reason.
The connection of illustration and definiendum The learned one directly sees and recalls; But before the dull one it must be established By the convention-inference that recalls the name.
Sakya Paṇḍita’s Gateway to Learning distinguishes two senses of “learned”: one who unerringly comprehends the nature of all phenomena, worldly and supramundane — Mipham Rinpoche, Sakya Paṇḍita himself — and one merely unobstructed within his own field, like a physicist who cannot knead tsampa or herd yaks. Here “learned” means one well versed in definition and definiendum: on seeing hump and dewlap he recalls the connection and knows at once that this is an ox, needing nobody to introduce it, as one who knows what a vase is knows a golden or silver one at sight.
The dull one, though he knows the word “ox”, does not grasp the relation, and must be told: the dappled animal you saw is called an ox, because it possesses hump and dewlap; and if that fails, because the ancients named it so. If he still will not have it, nothing further can be done. The Zen story of the father and son fits: the son insisted he was dull, the father insisted he was clever, and at last the father, out of arguments, said “you are just a dunce” — thereby conceding the son’s point.
Positing What Each Indicates in Connection with the Definition
Classification of Formal Arguments
The three phenomena, each in two — general and particular — Belong to six formal arguments in all; The [non-other] isolate too in six arguments; [Even] where definition and illustration meet, there is no fault.
Two sets of six, twelve arguments in all: six affirmative, establishing each own-isolate generally and particularly, and six negative, showing that none of the three is either of the others.
Affirmative: hump and dewlap are the general definition, because they eliminate direct contradiction and establish the meaning-isolate; and the particular definition in this argument, because they do so with respect to the ox. The name “ox” is the general definiendum, and the particular definiendum here, because it is a name argued for with the meaning-isolate as reason. “Dappled” is the general illustration, being the support of the general definition, and the particular illustration here, since on it the hump and dewlap find support. (“General” is not to be misread: it means general within the ox-class, not that hump and dewlap are the definition of vase or human.)
Negative: the ox’s definition and definiendum are not the illustration, generally and particularly, because they are not the support of the definition; definition and illustration are not the definiendum, because they are not names argued for with the meaning-isolate as reason; illustration and name are not the definition, because they do not eliminate direct contradiction and establish the meaning-isolate.
Of the twelve, ten hold and two fail, both among the general negative arguments. “Dappled is not the general definition” is uncertain, since in “this animal is called a dappled animal, because it is dappled”, dappled functions as a definition. And “the name ‘ox’ and hump-and-dewlap are not the illustration” is uncertain, since anything can serve as an illustration in some argument — one may take the ox’s dewlap as illustration to prove something else. Both therefore carry the fault of non-pervasion and are incomplete in the three modes.
(The commentary appends a picture of the Indian ox bearing hump and dewlap — formerly called “yellow ox”, now zebu.)
Removing Disputes about the Mode of Indication
Removing Disputes about Indicating the Own-Isolate
[Objection:] The definition belonging to this argument — if it is negated upon the illustration and definiendum, Then it negates the dewlap that is the illustration; And if it is not negated, then there is over-pervasion.
[Opponent] Does the definition of the definition — “eliminating direct contradiction and establishing the meaning-isolate” — exist upon the definiendum “ox” and the illustration “dappled” or not? If not, it fails to pervade them, and since what it establishes is the definition itself, hump and dewlap would end up not existing on “ox” and “dappled” either; and since hump and dewlap are themselves the illustration of the definition of the definition, the definition would fail to pervade its own illustration. If it does exist there, then it is no longer solely the definition of the definition but also the definition of the definiendum and of the illustration — over-pervasion.
[Reply:] It pervades by way of the meaning-isolate but does not pervade By way of the own-isolate; hence there is no fault.
Each of the three characteristics divides into three — definition, definiendum, and illustration of each — giving nine phenomena. From the angle of meaning-isolate the nine are inseparable, one body, as pillar, impermanence, and produced are one in essence (without which no nature-reason could be set up). From the angle of own-isolate they are wholly distinct, as impermanence is not produced and produced is not pillar (without which the three-mode argument could not be laid out, one term as subject, one as thesis, one as reason).
So: by meaning-isolate the definition of the definition does pervade definiendum and illustration and all their parts; by own-isolate it is only the definition of the definition, and over-pervasion never arises. The distinction is worth holding on to, since it recurs constantly in debate.
Removing Disputes about Indicating the Non-Other-Isolate
[Objection: there are three disputes —] that of “definiendum-contradiction,” that of “negating the convention,” and that of “the direct contradiction of thing and name” — [but] there is no dispute [that harms us].
[Opponent, first: definiendum-contradiction] “Dappled” is the illustration and “hump and dewlap” the definition, so neither is the definiendum; call these “the non-definiendum conventions”. Are those two conventions definienda or not? If they are, you contradict your earlier claim that definition and illustration are not the definiendum. If they are not, then these names depend on no real establishing reason, and names could be assigned at random — a vase called an ox, an ox called human — with names and meanings wholly confused.
[Reply] From the angle of own-isolate they are certainly not the definiendum: dappled is a particular and hump and dewlap are real particulars in the external object, more than a mere name, whereas a name is a will-o’-the-wisp. From the angle of meaning-isolate the convention “dappled” and the convention “hump and dewlap” belong to the definiendum “ox” — in meaning they have no separate bodies. So neither alleged fault lands. Sakya Paṇḍita states this in the auto-commentary, and Gorampa remarks that once one of these reasonings is understood the rest are plain at a glance.
[Opponent, second: negating the convention] Upon hump and dewlap, is the general definition of the definiendum negated or not? If it is — if “the name argued for with the meaning-isolate as reason” does not exist there — then upon the definition the very essence that establishes a name is missing, the definition’s essence is gone, and the definiendum with it: convention is cut off. If it is not negated, the definition of the definiendum exists upon the definition too, and over-pervades.
[Reply] The same distinction. By meaning-isolate the definition of the definiendum exists upon illustration, definition, and definiendum alike; by own-isolate it exists only upon the definiendum and not at all upon the other two.
[Opponent, third: direct contradiction of thing and name] The illustration “dappled” is a real entity and the definiendum a universal, a non-entity. How can two directly contradictory phenomena become one body, as you say they do within an argument?
[Reply] We never claimed they fuse in the external object. Before the conceptual thought of the exclusion-consciousness, the grasped “dappled” and the name “ox” are joined and grasped together; real entity and non-entity, external particular and mental meaning-universal, are fused by way of mistake, and this carries none of the faults alleged.
Illustrating Parallel Arguments that Seem Contradictory
Therefore some [arguments] appear contradictory, Yet parallel [non-contradictory] arguments are also admissible.
Reasoning goes by meaning, not by surface wording. Two examples.
First: “in the argument establishing sound as permanent, ‘produced’ is a pseudo-reason, because it is incomplete in the three modes”; then “incompleteness in the three modes is a genuine reason, because it is complete in the three modes”. This reads as flat contradiction, but the second statement means that the reason “incomplete in the three modes”, used to establish “produced” as a pseudo-reason, is itself a sound reason within that argument.
Second: “the isolate of the object of knowledge is an object of knowledge, because it serves as an object of the mind”. Subject, thesis, and reason look identical, but the first term means the isolate of all worldly phenomena taken by scope, and the predicate means that such an isolate can be known by conceptual thought.
Positing their Mutual Determination
[Objection:] [Some say] definition and definiendum are one in the object; [some say] they are distinct entities; [some say] their numbers are unequal — all of these are mistaken.
Three views. [Some Tibetan masters] definition and definiendum are one taste, one body in the external object. [Hearer masters, especially the Vaibhāṣikas] they are wholly distinct entities — the vase on one side, and the four characteristics of conditioned phenomena (arising, abiding, decaying, ceasing) that show it impermanent on the other, as analysed in the Abhidharmakośa. [Others] one definition may have many definienda and one definiendum many definitions, so their numbers are unequal.
[Refuting the early epistemologists:] Real existence and imputed existence are fixed.
Hump and dewlap are a real thing; the definiendum merely indicates a name and is imputed. That division is fixed, so they cannot be one body in the external object — whatever their unity before conceptual thought.
[Refuting the Hearers:] Definiendum and definition are of one essence, equal in number.
“One essence” answers the Hearers: if the vase’s definiendum and its impermanence, abiding, and ceasing were wholly separate, the vase would possess impermanence and destruction and yet could not be established as impermanent.
“Equal in number” answers the third view: one definition, one definiendum, like a seal and its impression. Since the indicating phenomenon is one, what is indicated should be one.
Whatever mutual inversions [they assert], Our positing here is unharmed.
Disputes often turn on unnoticed shifts of sense — same sound, different meaning, or same meaning, different sound. If someone insists the vase is not conventional but ultimate, note that in the epistemological context “ultimate” means what performs a function, and on that reading the claim can be granted. Faced with “does the definition of the definition pervade the definiendum or not?”, ask back: the definiendum’s meaning-isolate or its own-isolate, its essence or its wording?
But the tool has limits. If someone says “you are a yak, because you possess the definition”, one should not simply agree, but ask from what angle the definition is meant. Equally, one must not answer “do you mean the universal’s meal or the illustration’s meal?” when called to eat, or when asked where one is going reply “my essence or my appearance?” — that is taking epistemology too far, and ordinary communication becomes impossible.
II. Specifically Ascertaining the Definition of Valid Cognition
Valid cognition means perception and inference: a correct cognition, and one who possesses it in its ultimate form is a valid being — Śākyamuni Buddha. The Pramāṇasiddhi chapter of the Pramāṇavārttika likewise gives the definition of valid cognition first and the valid being after; here the eighth chapter treats the definition and the ninth the valid being.
Four masters’ views are refuted in turn — Devendrabuddhi, Prajñākaragupta, Dharmottara, and the Kashmiri paṇḍita Deshé Gawa — before Sakya Paṇḍita’s own position, which accords with Mipham Rinpoche’s great commentary on the Pramāṇavārttika.
Refuting Others
[Objection:] “Non-deceptive” and “illuminating a hitherto unknown object” — [Some] call these synonyms; [some] state them separately.
Two candidate definitions run through Dignāga’s treatises: non-deceptive cognition, which does not deceive before perception or before inference; and illuminating a hitherto unknown object, where what was not seen or known before is now seen or known. Dharmakīrti in the Pramāṇavārttika gives both.
[Devendrabuddhi, with Śākyabuddhi] The two are synonyms — either may be used, they are merely different names. (Later Sakya masters, Gorampa among them, note that the fault arises because he takes them as two from the angle of meaning-isolate; taken from the angle of own-isolate, Sakya Paṇḍita’s own tradition could accept the claim.)
Khenpo Sodargye pauses over Devendrabuddhi’s stature: he received the Pramāṇavārttika from Dharmakīrti in person, yet his first commentary was thrown into the fire and his second into the water; only at the third, when he said a human life is short and he would not try again, did Dharmakīrti allow that the direct meaning was adequate. That commentary survives in Tibetan. If a master of that calibre struggled, the difficulty students meet is unsurprising — and Dharmakīrti’s closing verse in the Pramāṇavārttika says his wisdom returns into itself like rivers into the ocean, comprehended by no one.
[Prajñākaragupta, author of the Pramāṇālaṃkāra] State them separately. The definition of ultimate valid cognition is illuminating a hitherto unknown object alone, since non-deception requires determining some meaning, and a noble being’s self-aware wisdom, free of grasper and grasped, has no conceptual determination of white, red, or blue. The definition of conventional valid cognition is both, since there grasper and grasped have not dissolved.
[Reply, refuting Devendrabuddhi:] If the definition is posited as two, The definiendum would also have to become two entities.
Two definitions wholly separate in essence would require two definienda, and two definienda cannot appear in one mind. A definiendum is not a string of words but a single concept — “ox”, the animal that is neither vase nor pillar nor human. Nor may one deny a second definiendum while keeping two definitions, since definition and definiendum are equal in number: one definition with two definienda leaves one definiendum without a definition (non-pervasion), and two definitions with one definiendum makes the second superfluous, since one already does the work.
The apparent counter-example fails: that Chinese, Tibetans, and Indians name the pillar differently, or that one person calls it “Dharma-wheel”, does not give two definienda, since the different names express one external thing and only the concept of name-and-particular fused is the definiendum.
[Refuting Prajñākaragupta:] If one accepts two definitions and definienda, A single general definition becomes impossible.
Separately his account may pass, but ultimate and conventional valid cognition are both valid cognition, so a single general definition must pervade them, as “human” has a general definition pervading man and woman even though each has its own. On his construction, asked for the definition of valid cognition as such, one could produce nothing: the ultimate definition does not reach the conventional, nor the conventional the ultimate.
[Refuting Dharmottara:] This way of [supplementing the] removal of non-pervasion is untenable; And dividing [the definition and the function] into two is not [Dharmakīrti’s] intent.
[Dharmottara] The definition is non-deceptive cognition, understood through three non-deceptions. Non-deception in essence — the particular is attained, whether by perception or gradually by inference — which excludes the Cārvāka, who count only the grasped thing as valid cognition and reject inference and all hidden objects. Non-deception with respect to the object — the determined object is attained through analysing wisdom, which removes the fault of impossibility. Non-deception with respect to function — as fire’s burning or the impermanence of conditioned things, an uncommon capacity that removes over-pervasion. But non-pervasion, he holds, is not removed by these and must be supplied by other means; and “realising the true reality”, which he equates with illuminating a hitherto unknown object, is the function of valid cognition rather than its definition.
Two answers. If the non-pervasion fault is not eliminated within the words of the definition itself, the definition as posited is faulty, and eliminating it by other words elsewhere is no help. And the split between definition and function is found nowhere in Dignāga’s or Dharmakīrti’s treatises, so it does not reach Dharmakīrti’s ultimate view.
[Refuting the Brahmin masters:] Asserting “comprehension of the true reality” [as the definition] Possesses [the faults of] superfluity, non-pervasion, and impossibility.
[Śaṅkaranandana and Deshé Gawa, masters of Brahmin caste] Non-deception is not the definition; the definition is comprehension of the true reality, equivalent to illuminating a hitherto unknown object, with three distinctions: in essence, it eliminates the superimposition of the opposing side; with respect to object, it comprehends what was not understood before; in mode of apprehension, cognition and object are unconfused, so it does not err upon the object.
(An aside: “Brahmin” does not mean non-Buddhist. Among the Six Ornaments, the eighty great adepts, the Chan lineage gurus, and many vinaya masters, some were of royal and some of Brahmin caste; the claim in certain Chinese-Buddhist works, including a History of Indian Buddhism, that Tibetan Buddhism is Brahmanism does not stand.)
Three faults. Superfluity: if the third distinction already secures that apprehension does not err upon the object, the function of comprehending the object is complete and the second distinction adds nothing. Non-pervasion: eliminating the superimposition of the opposing side suits inference, which is conceptual, but not perception, which is non-conceptual and cannot eliminate superimposition at all — so the definition fails to reach perception. Impossibility: comprehending the object’s true reality exactly as it is lies beyond both. Perception merely displays the external aspect; inference proceeds by reasoning and cannot directly apprehend the true reality; full comprehension of that kind belongs to a Buddha’s wisdom.
Establishing Our Own Position
“Non-deceptive” and “illuminating a hitherto unknown object” Are the same, for both comprehend the particular; And both conventional valid cognition and ultimate valid cognition May employ these two.
Either may be used, since both comprehend the particular of things. This is not Devendrabuddhi’s position, because the two are not being posited as two definitions in the meaning-isolate; spoken from the angle of own-isolate, both fall within the single essence of valid cognition, so no second definiendum is generated.
Both apply at both levels. Conventional valid cognition — the perception and inference by which ordinary beings determine and attain the external particular — is non-deceptive with respect to object, agent, and function, and knows now what was not known before. The root wisdom-absorption of a noble being does not deceive with respect to object or mind, and a first-ground bodhisattva’s wisdom likewise knows later what was not known at first. Dharmakīrti’s own procedure in the Pramāṇavārttika is the same: he first posits non-deception, free of the three faults, and later adds that illuminating a hitherto unknown object may be posited too.
Among them, the non-deceptive has the three natures — Of function, of agent, and of the acted-upon.
Non-deception with respect to function: if the particular exists, its function exists; if the particular does not, neither does the function. Fire’s particular existing, its burning can occur.
Non-deception with respect to the agent: the cognition itself, perception or inference, does not deceive in determining existence or non-existence.
Non-deception with respect to the acted-upon: engaging a particular external object, that object is actually attained; engaging what has no external object at all, there is deception.
The middle one is decisive, and explains two exclusions made in the second chapter. Genuine investigative thought — believing water is in the old well, where water in fact is — has non-deception of function and of the acted-upon, yet is not valid cognition, because the agent neither saw by perception nor determined by genuinely effective reasoning. Subsequent cognition likewise knows its object, but lacks the non-deception of perception and inference in the agent, and so falls among invalid cognitions.
Removing Disputes
[Objection:] [Your definition] does not pervade the valid cognition of non-entities, And over-pervades upon defiled cognition.
[Opponent] If non-deception is posited upon the particular, then a valid cognition measuring a non-entity — the absence of the son of a barren woman, or selflessness seen by wisdom — would become invalid, since no particular is there; so your definition does not pervade. And if attaining the particular makes a cognition valid, then the eye afflicted by cataracts, though wrong about yellow, did see the conch’s shape and particles, and the traveller in a car, though wrong about the motion, did see trees and houses; so a portion of valid cognition would attach to defiled cognitions, and your definition over-pervades.
[Reply:] Because [we] analyse whether the particular exists or not, [it pervades]; And because [the cognition] is defiled, it is rightly said to deceive.
On the first: we never take a non-entity out on its own for analysis. As stated in the first chapter, non-entities are judged while observing real entities — in measuring the particular of a real entity one knows at the same time that the particular of the non-entity is absent.
On the second: where the sense-cognition is defiled, the object it reaches is mistaken through and through, with no correct portion. The shape seen is the shape of a yellow conch, not the correct shape of a white one; the trees seen are swaying trees, and no swaying exists in the object at all.
How Superimposition is Eliminated Thereby
How it is Ascertained
By the power of valid cognition, ascertainment is induced; The contrary class of ascertainment is superimposition. Its definition is “forsaking the truth”; Conceiving the other extreme, it is of two kinds — The conceiving of a mistaken cognition, and doubt.
Ascertainment is the wisdom that dispels doubt by the power of valid cognition. The auto-commentary divides it into valid ascertainment, which cuts off doubt about a non-deceptive meaning, and mistaken ascertainment, which does so about a deceptive one — as non-Buddhists hold with conviction that past and future lives do not exist. A valid ascertainment arising later overturns the mistaken one, as the ascertainment of impermanence overturns a former grasping at permanence. (The Beacon of Certainty’s ascertainment, which dispels darkness, may be read the same way.)
Superimposition is its contrary class: taking the existent as non-existent or the non-existent as existent, forsaking the truth and conceiving the other extreme — grasping phenomena as really existent, or denying past and future lives. All such wrong views reduce to the extremes of permanence and annihilation. It divides into two, the conceiving of a mistaken cognition (all worldly wrong views) and doubt (“do cause and effect exist or not?”).
Both divisions fall within conceptual thought, since ascertainment eliminates superimposition by a contrary wisdom arising in the continuum, whereas a non-conceptual mistake is eliminated only by cutting off its cause: knowing perfectly well that there are not two moons in the sky does not stop the eye from seeing two.
Khenpo Sodargye adds that subtle non-conceptual superimposition does exist, though Sakya Paṇḍita does not analyse it here, and the Pramāṇavārttika would be the place to study it. It is bound up with propensities. Attachment requires that the object already appear attractive to the eye-cognition, or the later conception could not grasp it as attractive; yet “good-looking” cannot exist in the object, since not everyone who sees the same person is attracted, what humans find beautiful the gods may find repulsive, and animals find each other attractive where humans see only ugliness. One who cultivates the contemplation of impurity sees the body as impure in the very group of eye-cognitions; one of strong desire sees it as pure. Conventionally, the Madhyamakāvatāra’s criterion holds — an unimpaired sense-cognition’s object is posited as conventional valid cognition — but pressed further, the appearing object leads to the Mind-Only position, and further still to the luminous mind of Dzogchen.
Removing Disputes
[Objection:] In the spontaneous [inference] of one who had not [previously] comprehended, The comprehension does not [yet] eliminate any superimposition.
[Opponent] Spontaneous inference — walking the road to Sêrtar, seeing smoke, and at once knowing there is fire — eliminates something, but not either of your two. It was not doubt, since no question “is there fire?” had arisen; and it was not mistaken superimposition, since no notion “there is no fire here” had arisen. Nothing at all had been thought.
[Reply:] Inference apprehends not by appearance But by exclusion of its object; [The mind] had grasped [the place] as like another that has no fire, And so [the eliminated superimposition] is held to be a mistaken cognition.
Inference is an exclusion-wisdom; were it to apprehend by appearance it would be perception. Though the traveller formed no explicit judgement, in his attitude the place stood as one like any other with no fire — a latent grasping — and that is what the spontaneous inference removed. It is therefore included, indirectly, in mistaken superimposition.
These two — ascertainment and superimposition — In terms of object, of essence, and of time, Are examined as to whether they are one or distinct; These three disputes do not harm [our position]. Ascertainment and superimposition grasp one object, But their modes of grasping differ — hence exclusion.
[First dispute, object] If their objects are distinct, they have no relation, and ascertainment can no more eliminate superimposition than daylight outside can lighten a dark basement. If their object is one, then either the object is true, in which case superimposition becomes errorless, or it is false, in which case ascertainment becomes false.
[Reply] From the angle of the grasped object they are one — both grasp the pillar, whether taking it as permanent or as impermanent. From the angle of the mode of apprehension they differ: one accords with the actual and meets no harm from valid cognition, the other does not; the first is ascertainment, the second superimposition.
[Second dispute, essence] If they are one essence, like fire and its heat, there can be no eliminator and eliminated, as light cannot eliminate itself. If wholly distinct, they are like superimposition in one continuum and ascertainment in another, so one could never remove the other — as the Buddha’s enlightenment did not remove the superimposition in Devadatta’s continuum.
[Reply] Within a single continuum they are one in essence, at different stages: first the notion that the pillar is permanent, later, through a guru’s guidance, the ascertainment of impermanence. This is said of the continuum, not of a single instant. Their own-isolates are wholly different.
[Third dispute, time] If simultaneous, two conceptions would arise at once in one continuum, and the Buddha states that a being cannot have two mind-continua. If not simultaneous, then like today’s noon sunlight and last night’s darkness, one cannot eliminate the other.
[Reply] They are not simultaneous, and elimination still holds. Superimposition arises first — taking a coloured rope for a snake — and by other causes and conditions ceases, ascertainment arising in its place; that sequence is what “eliminator and eliminated” means in convention. There is no war between them in which one side is destroyed. Two kinds of elimination should be distinguished: cutting the seed, as a first-ground bodhisattva’s wisdom destroys the seed of ordinary clinging; and merely checking the conception, as when, having studied epistemology, one can no longer hold the pillar permanent in this life. Only the latter is available to ordinary beings, and people call it eliminating superimposition.
III. Ascertaining the Meaning Indicated by the Definition
Four parts: the detailed classification of the definiendum, valid cognition; refuting mistaken conceptions about instances; the valid cognition that ascertains the definition upon the instance; and how negation and proof are carried out. The classification itself has four: the basis of classification, the essence of the classification, the fixed number, and the etymology.
The Basis of Classification
Refuting the Untenable Position
The “real [cognition that is yet] invalid” has already been refuted; And to divide the “real” [into two as the basis] is mistaken.
Perception and inference are divided from something, as man and woman are divided from “human”. The early epistemologists’ claim that subsequent cognition and investigative thought are invalid cognitions yet genuine cognitions has already been refuted: a genuine correct cognition is either perception or inference, and what is neither cannot behold the particular directly nor know a hidden object by three-mode-complete reasoning.
[Jinendrabuddhi] The basis is correct cognition, but correct cognition splits in two: the actively functioning kind — seeing fire burning something, water moistening something — which belongs to perception and so cannot serve as the basis, since being already determined it does not reach inference; and the kind that merely sees fire or water, which is the genuine basis.
Refuted: a basis of classification must be the general phenomenon. The object of knowledge is the basis for real entity and non-entity, and for the two truths; “human” is the basis for man and woman. On Jinendrabuddhi’s method one might as well make sense-perception the basis of the four perceptions, or inference by renown the basis of the three inferences — neither of which can be granted.
Establishing the Tenable Position
Therefore correct cognition is itself valid cognition, And this should rightly be taken as the basis of classification.
“Valid cognition” and “correct cognition” are different names of one thing, as “cool-rayed one” and “moon”. Whatever is correct cognition is perception or inference, and whatever is perception or inference is correct cognition: the four perceptions accord with the actual and behold their object-particulars clearly, and inference, established by three-mode-complete reasoning, deduces in accordance with the actual. Pseudo-perception and pseudo-inference are excluded. It is because the Dharma the Buddha taught is without error, whether by perception, inference, or scriptural valid cognition, that he is called a valid being and his words scriptural valid cognition.
The Essence of the Classification
The classification is into perception and inference; [We] examine whether these are one entity or distinct.
Perception measures the particular, inference the hidden essence. Four possible relations are canvassed: different isolates of one essence, like the produced and the impermanence of a vase; wholly different things, like vase and pillar; different names of one thing, like “cool-rayed one” and “moon”, or “Tathāgata” and “Sugata”; or a relation of real entity and non-entity that negates a single body, like an ox’s horn and a rabbit’s horn.
Both valid cognitions are real entities, And so they are not [related as] negating-a-single-entity; With respect to object — one or distinct — they are neither one entity nor distinct entities; Therefore in the essence of mind they are one, But with respect to the object they are distinct entities.
Both are real entities — perception a wisdom beholding the particular as it is, inference a wisdom understanding a hidden thing by correct deduction — so the fourth relation is out.
With respect to a single object they are not one, since perception knows the object-particular while inference’s direct object is the universal. With respect to distinct objects they are not simply distinct either, since both are cognition and neither can be parted from self-awareness when introspecting. Hence: one in the essence of mind, distinct with respect to object. In knowing fire by smoke, inference works by the universal of fire and perception by its particular.
[Objection, first] If they are distinct with respect to object, inference should not be a cognition, since from the angle of object it is not self-awareness.
[Objection, second] If from the angle of mind inference is one with self-awareness, and self-awareness is included in perception, inference is one with perception in general; but inference is distinct from sense-perception, which is also included in perception in general, so inference must also be distinct from perception in general — self-contradiction.
[Reply:] Inference is not [posited] from the object, and it is one With self-awareness; hence it is free of the disputed faults.
Inference is posited as a cognition from the angle of mind, not of object, so the first charge misfires. The second rests on a misunderstanding of general and particular: what is one with a particular within a general phenomenon is one with the general phenomenon — anything inseparable from aloeswood is inseparable from tree — but what is distinct from a particular need not be distinct from the general, since sandalwood is distinct from aloeswood and yet is a tree.
The Fixed Number
Because the objects of comprehension are two, valid cognitions too are two; Any other number has already been refuted.
The objects measured are particular and universal, and nothing else; so the measuring cognitions are perception and inference, and no more. This does not conflict with the first chapter’s statement that the object of comprehension is solely the particular, since that spoke of the ultimate object of comprehension, whereas this speaks from the angle of the mode of measuring.
Other schemes — Candrakīrti’s four valid cognitions in the Prasannapadā, Bhāvaviveka’s four, the non-Buddhists’ sound-arising valid cognition — are either subsumed under inference (as scriptural valid cognition and valid cognition by conviction are) or fall among pseudo-inferences. Dharmakīrti refuted the surplus one by one.
[Cārvāka] Perception alone; inference is not needed. But rejecting inference leaves every hidden matter undecidable, and in practice they argue that past and future lives and karmic cause and effect do not exist because they have not been seen — a pseudo-inference, and so an indirect admission of inference, which makes the position easy to refute. Mipham Rinpoche does so at length in his commentary on the Madhyamakālaṃkāra.
The Etymology
Names are [determined by] the etymological term and the conventional term; Their [combination of] etymology and convention has four kinds: Contradictory, non-contradictory, And sometimes contradictory, sometimes not.
A name has an etymological term when it can be explained upon the essence of the thing: “omniscient” for the Buddha, who does know all phenomena worldly and supramundane. It has a conventional term when it rests on usage alone — “vase”, “pillar”, “road”, named once by someone and kept up by custom, with no further reason. (Tibetan personal names mostly fall here: someone is called “Goddess” whether or not she resembles one.)
Four cases, following Jamyang Loter Wangpo’s commentary, whose division differs slightly from the auto-commentary’s:
- Convention without etymology: the “shepherd-bird” that herds no sheep, the “sesame-eating sparrow” that eats no sesame, a plain woman called “Goddess”, a man called “Auspicious” whose arrival brings nothing of the kind.
- Both: “omniscience”; “consciousness”, which does know the object; “water-born” for a lotus grown in water.
- Sometimes one, sometimes the other: a lotus growing on a mountain has the convention “water-born” but no etymology; frogs and tadpoles have the etymology but not the convention; a lotus in water has both; stone and wood have neither.
These names, [together with] the generic term, The learned divide into two kinds.
Names of things (personal and place names — Devadatta, Maitreya) and generic terms (tree, humankind, ox, fine horse).
Applied to valid cognition: perception is pratyakṣa, “in dependence on the faculty”. The six perceptions — five sense-perceptions and mental perception — have both terms, since all depend on a faculty and all apprehend their object-particular as it is. Yogic perception and self-aware perception have the conventional term but no etymology, since neither depends on a sense-faculty. The deluded sense-cognition seeing two moons has the etymology, since it does depend on the eye-faculty, but not the conventional term, since no particular “two moons” exists.
Inference is anumāna: having seen a correct reason and recalled the connection, knowing the deduced object by way of the universal. All genuine inference — scriptural valid cognition, inference by renown, factual inference — has both terms, since each is called inference and each knows its object through a correct reason.
Refuting Mistaken Conceptions about Instances
Instances of Perception
Taking What is Not Perception as Perception
[Objection:] Some hold that the [non-deluded] portion within delusions of place, time, and aspect Is perception.
[Prajñākaragupta and other early epistemologists] Each such delusion contains a genuine perceptual portion.
Place: seeing a jewel’s light through a door-crack, one takes the jewel to be at the door though it lies within the room — mistaken as to place, but the light seen is genuinely the jewel’s light. Or a person on a summit seen from the foot of the mountain appears to stand on another peak or in space — mistaken, yet the person was seen.
Time: dreaming at midnight of conversing with one’s children at midday — night taken for day is mistaken, but the children exist and the grasping of them is correct.
Aspect: a white conch taken for yellow, or roadside trees seen as moving — the yellow and the motion are wrong, but the shape of the conch and the trees themselves were seen.
[Reply:] Since there is no particular, it is not perception; But by the [appropriate] reason it can be determined as inference.
No particular was beheld in any of the three, so none is perception. The jewel’s light cannot stand for the jewel, whose particular was never perceived. What is dreamt has no particular of the daytime children at all; it is a deluded consciousness arising from a propensity left on the ālaya. And the yellow conch and swaying trees are produced by the sense-cognition’s own delusion — no tree runs backward, and a normal eye cannot see the shape of a yellow conch.
Each case can nonetheless yield knowledge by inference. Knowing the generated-generator relation of jewel and light, one reasons from light at the door to a jewel within, as one infers a butter-lamp or street-lamp from its light. Dreaming repeatedly of a relative shows a prior affection, by an effect-reason. And the yellow conch’s shape and weight can be inferred, as can the true stillness of the trees: I am in a moving car, so my sense-cognition is deluded; the trees stand still.
[Objection:] Some declare that the cognition apprehending colour Is also perception with respect to the tangible.
[Ngok Lotsāwa and other early epistemologists] Seeing a white conch, the eye-cognition also directly beholds its weight and smoothness — the object of the body-cognition — so perception of colour is at once perception of the tangible.
[Reply:] Through failing to understand the distinction of the [sense-]spheres, [They] mistake the reason-of-co-present-factors for perception.
The twelve sense-spheres of the Abhidharmakośa answer this: the outer six (form, sound, smell, taste, tangible, phenomena) correspond severally to the inner six (eye, ear, nose, tongue, body, mind). If the eye-faculty could behold the tangible, the distinct functions of the spheres would be pointless.
What is really at work is the reason-of-co-present-factors, one of the five effect-reasons: where several phenomena are gathered on one essence — colour, shape, taste — one deduces the others from one of them, as one deduces a sweet’s colour and shape from its taste. (An objection raised when studying the Pramāṇavārttika: after the sweet is eaten the taste lingers, so taste need not imply shape. Answer: so long as the taste is there its particles are there, invisible though they be.) Deducing the object of one sphere from that of another is inference, not perception.
Taking What is Perception as Non-Perception
[Objection:] Some say that because [these cognitions] are deluded in conception of object, time, and aspect, They are not perception.
[Certain masters] A vase outside the door taken to be inside; a person seen this afternoon taken to be the person seen yesterday; a red pillar taken for red woollen cloth, or the blue sky for blue cloth. Conception makes mischief in each, so the accompanying eye-cognitions are not perception.
[Reply:] The conception is deluded, yet it does not impede [the sense-cognition]; The sense-cognition is valid, just like the apprehension of blue.
In all three the particular was perceived directly; the delusion lies wholly in the conception, and a single sense-cognition cannot be both deluded and undeluded. Delusion of conception does not obstruct perception: one who has never studied impermanence sees blue cloth exactly as it is while thinking it will last for ever. If conception followed perception automatically, everyone who could see would already be an accomplished practitioner and the contemplation of impermanence would not need teaching.
Refuting the Claim that the Two Cases are Alike
[Objection:] Certain masters say the two cases are alike. [Reply:] They are not, because the objects are distinct entities.
[Certain early-epistemologist masters] If the sense-cognition that takes a pillar’s red for cloth-red is valid as regards red in general, then taking a white conch for yellow should be valid as regards the spiral shape.
Not so: the objects differ in kind. Seeing red cloth and later a red pillar, the eye beheld each particular correctly, and only the conception grasped both as one “red” — the conception’s object being blurred and universal, the particular belonging solely to the non-conceptual. But taking a white conch for yellow is caused by the sense-cognition’s own delusion, and a deluded sense-cognition has no undeluded portion. When the eye sees two moons there is no correct half; the early epistemologists’ insistence that one of the two moons must be correct comes of not distinguishing delusion from non-delusion in the sense-cognition — the same confusion that arose over appearance-without-ascertainment.
Conversely, undeluded sense-cognition with deluded conception does occur. When the non-Buddhist teacher looked at Śākyamuni Buddha, his eye-cognition saw the adorned form, but his karma turned the conception into wrong views about it.
Instances of Inference
Taking What is Inference as Non-Inference
[Objection:] Some say that to infer a jewel from the light of a jewel Is not inference.
[Certain early-epistemologist masters] Three ways all fail. Deducing the jewel from the jewel makes proof and probandum inseparable, like “you are human, because you are human”. Deducing it from the jewel’s light presupposes the jewel, for if it is hidden, how is the light known to be its light? And deducing it from light in general is uncertain, since the light may be sunlight or lamplight.
[Reply:] If so, then rice-sprout and rice-seed and the like — Most effect-reasons — would become absurd.
The same three horns would wreck every inference from effect to cause: seed from seed is inseparable, seed from sprout presupposes the seed, and seed from a sprout taken by itself is uncertain. Likewise fire from smoke. The opponent’s own answer — that one who observed sprout arising from seed knows their relation and can afterwards deduce the cause from the effect without seeing it, as one recognises a friend on the telephone by his voice alone — is exactly the reply. Having known that jewel and jewel-light are related, one sees the light and concludes: this is not lamplight but a jewel’s; a jewel is there. That is the recalling of connection already taught, and the inference is correct.
Taking What is Not Inference as Inference
[Objection:] Even though the reason and probandum are erroneous, In the portion of the external object [the inference] is still valid.
[Some non-Buddhist masters] Using the notion of the “part-possessor” — a whole complete with its parts — they argue: on the part-possessing mountain there is permanent fire, because there is ultimate smoke, like the stove made by Maheśvara. Reason and probandum are both absurd, they concede, yet since there is after all fire on the mountain and it is deduced by smoke, from the angle of the external object the inference holds.
[Reply:] Such a reason lacks the three modes, And the qualifier of the reason [used] to prove it cannot be established.
Nothing in it is established: the mountain is no part-possessor with respect to fire and smoke; fire is momentarily impermanent, not permanent; there is no ultimate-truth smoke. With none of the terms established the three modes cannot obtain. Further, the qualified phenomena, fire and smoke, are established only through the qualifiers “permanent” and “ultimate”, and those cannot be established at all — the argument is on a par with “the vase exists, because the vase of the son of a barren woman exists”. Better to ask plainly whether there is fire on the mountain and whether there is smoke.
The Valid Cognition that Ascertains the Definition upon the Instance
The Question
[Question:] When those valid cognitions are ascertained, Is this known by themselves or by another?
Valid cognition is non-deceptive; but what ascertains the non-deception? If the cognition itself, then no being would ever be confused about valid and invalid cognition, since anyone with an eye-cognition would behold its own essence. If another cognition, then that one needs a third, and so on without end.
Refuting Others’ Answer
Some assert self-ascertainment; Certain ones declare other-ascertainment. These two forsake reasoning; [we] assert what accords with reasoning.
[Gemarina and others] wholly self-determined. [Others] wholly determined by another’s power. Both forsake Dharmakīrti’s factual reasoning.
The Pramāṇavārttika says non-deception is known by self-awareness, which overturns any claim that it must be known by another’s power; and it also says that because beings’ faculties and delusions differ, some cannot know by their own power and must depend on another condition — as one who has seen the pillar but does not see its impermanence must be told, the pillar you saw is impermanent, because it is produced. Both cases hold, and fixing on either alone is untenable.
Establishing Our Own Answer
The two object-establishments, self-awareness, And inference are self-ascertained; The first [cognition], the inattentive one, And those with a deluding cause are other-ascertained. By [the cognition of] function, by familiarisation, And by inference, all can be ascertained.
Self-ascertaining valid cognition, needing no other aid, is of three kinds. The two object-establishments: by force of function, as when one sees a red thing burning firewood and knows at once, with no deluding cause, that it is fire; and by force of familiarisation, as when a grey thing in the sky is identified as smoke and not cloud, or a red stone as coral and not ruby, by someone long habituated to such things. Self-awareness, in which ascertainment necessarily arises upon introspection. And inference, since a three-mode-complete reason needs no second reason: smoke alone suffices to establish fire.
Other-ascertaining valid cognition is of three. The first cognition — seeing an ox, or coral, for the very first time, one sees it but cannot generate the ascertainment “this is an ox”. The inattentive — the eye sees an acquaintance in a crowd while the mind is fixed elsewhere, and no recognition arises. And those with a deluding cause — seeing a grey thing but unable, through obstruction, to tell smoke from mist.
These three are ascertained afterwards by the three aids named: function, as one who does not know the ox learns it by hump and dewlap, or identifies fire rather than red cloth by seeing it burn; familiarisation, whereby what conceptual thought has often met is gradually recognised (Mipham Rinpoche’s great commentary notes here that a woman’s body appears as an object of desire to one habituated to desire, as impure to one who cultivates the contemplation of impurity, and as food to certain beasts — one object, three functions, according to the force of familiarisation); and inference, as “the grey thing is smoke, not mist, because it is born of fire”.
Beyond both stands the Buddha’s omniscience, which is neither the self-power nor the other-power of conceptual thought, yet deceives in no respect and is therefore also valid cognition.
Removing Disputes
[Objection:] Analysing in terms of object and of essence, Other-ascertainment is inadmissible. [Reply:] [You] have confused together the functions Of perception and of conception.
[Opponent, by object] If the ascertained and ascertaining valid cognitions have one object, the second becomes subsequent cognition, the first having known it already. If their objects are not one, the second’s knowing cannot induce ascertainment in the first.
[Opponent, by essence] Does the ascertained valid cognition cut off superimposition about its object? If not, it is not valid cognition at all; if so, it is self-ascertaining, not other-ascertaining.
[Reply] From the angle of essence, or of the continuum of the determined object, the ascertaining and ascertained cognitions are one — the cognition that was inattentive and the cognition that later knows through familiarisation or inference are one in the mind’s essence; so the charge of distinct objects fails. From the angle of earlier and later substances in that continuum they differ, one having produced no ascertainment by its own power and the other having produced it; so the charge of a single object fails too.
The deeper error is the opponents’ conflation: perception, being non-conceptual and undeluded, is valid without needing to cut off superimposition at all; it is before the inferential cognition induced by perception that superimposition must be cut off. Holding that perception itself must cut off superimposition mixes the functions of the two.
How Negation and Proof are Carried Out
Positing Negation and Proof
The subject that appears as a real entity [is “appearance”]; Before the exclusion-[cognition] there is negation and proof; Within negation there are two modes — That which eliminates [an implied] establishment, and that which does not.
Objects are particular and universal, subjects non-conceptual and conceptual, and their modes appearance and exclusion. Appearance is the apprehension of a real particular; people call it establishment, but genuine negation and proof are posited only before the exclusion-cognition — “the pillar exists” is establishment, “the pillar does not exist” negation.
Negation is of two kinds. Non-implicative: “there is no vase here”, which negates the vase’s essence and establishes nothing. Implicative: “this thing is not a vase”, where the thing itself remains and something is indirectly established; or “the vase is not permanent”, which indirectly shows its impermanence. The Beacon of Certainty gives the same criterion — whether another phenomenon is drawn out by the negation.
All negation and proof operate upon language and conceptual thought, never on the material essence. Sound is simply impermanent; there is nothing there to negate or establish. What “negation” names is the correction of someone whose view has turned into another state — sound is impermanent, but he holds it permanent, and scripture and reasoning set him right. A permanent vase could be thrown out if it existed, but it is no different from the son of a barren woman, so there is nothing to remove.
How Objects are Cognised Thereby
Refuting Others
[Objection:] Some say the direct and the indirect Are also [perception’s] establishment and negation.
[Shawa Chözang and other early epistemologists, followed by some later Gelugpa masters] Perception has both a direct and an indirect cognition. Seeing a blue lotus directly, one also knows blue in general indirectly; seeing a red-robed monk, one knows red in general. Correspondingly on the side of negation: fire not existing, smoke is indirectly not seen. Both, they hold, fall within perception.
(Khenpo Sodargye notes the commentary he follows explains the verse in terms of the direct as establishment and the indirect as negation, whereas his own gloss above draws in Khedrup Je’s statement, on which perception and inference each have a direct and an indirect side.)
[Reply:] [Then ordinary] inference would become indirect cognition; And if not, it becomes a third valid cognition.
Knowing “smoke does not exist” from “fire does not exist” is a non-observation reason — specifically the non-observation of the cause, one of the four non-observations of the related (of cause, of nature, of effect, of pervader) — and so plainly inference. If the opponent accepts the parallel, commonly acknowledged inference collapses into indirect cognition. If he denies it, on the ground that his case involves no three-mode reasoning, then his indirect cognition is neither perception nor inference and becomes a third valid cognition — but the objects are only particular and universal, and Dharmakīrti posited no third.
In fact each supposed case is inference. Knowing blue in general from the blue lotus, or humankind from seeing a person, or a tree from a sandalwood, deduces the general through a particular and belongs to the nature-reason. Knowing that smoke is absent because fire is absent works through the cause-and-effect relation and is the non-observation of the cause.
Establishing Our Own
“Is the object,” “is the cognition,” “the object exists,” “the cognition exists” — By these, establishment is fourfold, two by two; Negation, the reverse, divides into four; And both negation and proof are confirmed by valid cognition.
All phenomena fall under object and cognition, and all valid cognition under negation and proof. Establishment divides by “is” and “exists”, each applied to object and to cognition: this is a red object; the eye-cognition grasping it is a red-apprehending cognition; this white Vajrasattva exists; the eye-cognition grasping Vajrasattva exists. Negation reverses each — “is not” and “does not exist”, again on object and on cognition. Four and four.
Both are settled by perception and inference alone. Negation is carried out by perception and by the inference that establishes non-implicative negation; establishment by the ascertainment drawn out by perception and by the inference of implicative negation. Perception can negate: that the pillar has no life is known, after seeing it, by the force of familiarisation, since no living pillar has ever entered one’s experience. The absence of a self of persons is known by yogic perception — not by sense or mental perception, which do not take it as object — though it can also be established by reasoning: within the aggregation of the five aggregates the self does not exist, like a place without a vase.
Capacity varies. A person of wisdom, seeing a place where the vase has been removed, ascertains at once that no vase is there; a duller one needs the non-observation reason. One who observes the pillar’s or the river’s changes knows impermanence directly; another must be brought to it through “produced”.
Removing Disputes
The so-called indirect cognition Intends [either] the function [of perception] or [spontaneous] inference.
[Opponent] Dharmakīrti’s treatises do speak of knowing a meaning by way of the indirect, so a perceptual indirect cognition must exist.
[Reply] The phrase intends one of two things. Either the function of perception: the fourth chapter of the Pramāṇavārttika says a stone has no mind, and this is known by the force of perception’s function rather than by perception itself. Or spontaneous inference: a thirsty person reaches for water without reasoning it out, and one long trained in impermanence thinks “the pillar is impermanent” on sight without invoking “produced” — the result of the reasoning appearing in an extremely brief time, though it remains inference. The traveller who sees smoke and thinks “there is fire here” without formulating the argument is the same case.
Saying “perceptual indirect cognition” as a mere form of words does little harm; accepting it as genuine perception does. Blue in general is a universal, and if a universal were perceived apart from the blue of the blue lotus, perception would have become a conceptual cognition — and no perception whose subject is conceptual is found anywhere in the epistemological literature.
Here ends the Eighth Chapter, Examining Definitions, of the Pramāṇayuktanidhi.
← 7. Examining Contradiction | All chapters | 9. Examining Perception →