The concept of svabhaavahetu is a major contribution of Dharmakiirti to Buddhist and Indian logic. In the case of a svabhaavahetu the invariable relation of pervasion between the probans and probandum is based on identity (or nondifference). This implies, according to our interpretation, the thesis that some general statements are true by virtue of meaning. Although such statements are true by virtue of meaning, they are not contentless. This is in disagreement with the view of a large number of recent philosophers who have held that statements true by virtue of meaning must be contentless. Besides giving an exposition of the svabhaavahetu, we have also explained its threefold classification, to which little attention has been paid by recent scholars.
Just as traditional European logic is dominated by the study of the categorical syllogism, so also is traditional Indian logic dominated by the study of nyaaya (literally: proof, deduction, argument, rule, and so forth). The nyaaya has a deductive core which has a substantial affinity with the categorical syllogism. In a nyaaya, the conclusion is typically of the form that something is characterized by something. The former is called the subject (pak.sa or dharmin) and is analogous to the minor term. The latter is called the probandum (saadhya) and is analogous to the major term. The conclusion is drawn on the basis that the subject is characterized by what is called the probans (saadhana or hetu), which is analogous to the middle term, and that the probans is pervaded (vyaapta) by the probandum. The following is a common example of a nyaaya: whatever is originated is noneternal: sound is originated; therefore, sound is noneternal. In this argument, sound, being originated and being noneternal, is respectively the subject, the probans, and the probandum. The statement "whatever is originated is non-eternal'' expresses the pervasion (vyaapti) of the probans by the probandum.
In an argument of this kind the relation of pervasion between the probans and the probandum is of crucial importance. It is no wonder, then, that Indian logicians through out the ages have devoted considerable attention to the analysis and explication of pervasion. In this regard Dharmakiirti and other Buddhist logicians are of the view that pervasion is an "invariable relation" (avinaabhaava-sambandha) and that an invariable relation may hold only under two conditions. The first condition is that of identity (taadatmya) or nondifference (abheda). In such a case the probans is said to be of the svabhaava variety. Thus, according to Dharmakiirti, svabhaava (literally, the entity itself) is one kind of probans (hetu) through which the probandum may be inferred.  In such a case the probans is related to the probandum by way of identity or nondifference. If it is known that the subject is characterized by the probans, it may be inferred that the subject is characterized by the probandum, for the probans is identical with or nondifferent from the probandum: the probans is tadaaman, that is, identical with that (the
probandum), or tadabhinna, that is, nondifferent from that (the probandum). This is also expressed as that the probans is tatsvabhaava, that is, that (the probandum) entity itself. For example, if it is known that something is a mango, it may be inferred that it is a fruit, for a mane is nothing but a fruit, (To an Aristotelian the relation between fruit and mango is that of genus and species; not so from Dharmakiirti's point of view. The precise nature of the relation of identity between mango and fruit is explained later.)
The second and only other condition under which there may be an invariable relation between the probans and the probandum is that of causation (tadutpatti). In such a case the probans is said to be of the kaarya or the effect variety.  In this case the probans is different from the probandum; still the latter may be inferred from the former, for causality admits of no exceptions. If it is known that the subject is possessed (in the specified sense) of the probans, which is an effect of the probandum, it may be inferred that the subject is possessed of the probandum. For example, if it is known that something possesses smoke, which is an effect of fire, it may be inferred that that thing possesses fire.
Dharmakiirti devoted more space to the explication of svabhaava as a probans than that of kaarya as a probans. He supplied slightly variant "definitions" (lak.sa.na) of the former in his different works although he did not supply any "definition" of the latter. His commentators felt that this was quite in order, for the notion of svabhaava was controversial.  This remark has proved to be prophetic: the interpretation of svabhaava as a probans has remained problematic even today with different scholars reaching diametrically opposed conclusions.
One important issue in this connection is the epistemic status of the statements enunciating the relation of identity between the probans and the probandum in such cases. In the argument already cited, 'all mangoes are fruits' is such a statement. Let us call such a statement a svabhaava statement. Is it analytic or synthetic? According to Stcherbatsky it is analytic in Kant's sense, for the predicate 'fruit' is contained in the subject 'mango', and further the connection of the predicate with the subject is though through identity.  Moreover, that Dharmakiirti was aware of this may be gathered from his claim that the predicate belongs to the "own essence" of the subject and can be inferred "from the existence of the subject alone," that is. the subject alone, without having recourse to experience. If Stcherbatsky is right, Dharmakiirti may be credited with anticipating, at least partially, the concept of analyticity.
Stcherbatsky's classical exposition of Dharmakiirti's view has been ably challenged by Steinkellner, according to whom such a statement, to Dharmakiirti's own mind, is synthetic.  For Dharmakiirti does not establish the necessity of such a statement referring to the terms themselves. To ascertain the truth one has to resort to an additional proof showing that the intersection of the probans and the complement of the probandum is empty. This latter proof, according to Steinkellner, is purely empirical, having recourse to perception directly, or indirectly via additional inferences. Further, the idea of the predicate being contained in the
subject does not fit with Dharmakiirti's apoha (differentiation) theory, according to which concepts signify exclusion rather than inclusion. Finally, the logical relation between the probans and the probandum is based on the factual identity of the reality corresponding to the two concepts. This reference to the real would have been superfluous if the relation of the two terms were thought to be analytic.
It is remarkable (Steinkellner fails to emphasize this) that Stcherbatsky was well aware of the points made by Steinkellner. Stcherbatsky did hold that the identity between the probans and the probandum in the case of a svabhaavahetu signifies real and existential identity.  He gave a detailed account of the apoha theory in terms of exclusion. He also would not deny the importance of additional indirect reasoning supporting a svabhaava statement if its truth is questioned. Still he held that a svabhaava statement is analytic. The question is: what is the correct understanding of a svabhaava statement?
To answer this question we shall first look at some of the crucial texts of Dharmakiirti on the svabhaavahetu. 
Svabhaava is the probans for that inferable character which follows upon (literally "which arises from") its (the probans') being alone. For example, this is a tree because of being a 'si^m`sapaa (a kind of tree). (NB 2.15, 2.16)
Everything which is real is noneternal, such as a pot, and so forth. This is a formula of svabhaava probans which is pure. (NB 3.9) Whatever is originated is noneternal. This is a svabhaava formula by way of introducing distinction from a character which is its being (nature). (NB 3.10)
Whatever is a product is noneternal. This is by way of invoking a distinction which incorporates something additional. (NB 3.11)
A product is that entity which is dependent on the operation of something else for the realization of its being. (NB 3.12)
As a matter of fact, there is identity between the two. (NB 3.18)
That which is not realized with the realization of that is different from the being of that. (NB 3.19)
An entity is the probans of that which is of such a nature that it follows upon the being (presence, existence) alone. (PV 3.2)
Svabhaava is the probans for that inferable character which follows from the mere being (existence, nature) of the character which is the probans. (HB 4.10)
The following of the inferable character from only the character which is the probans is the distinctive feature (lak.sa.na) of a svabhaava probans. (VN p.9)
It appears from the preceding that the basic account of the svabhaavahetu in the NB is only marginally different from those in the other works of Dharmakiirti. It should be noticed, however, that the account in the VN is somewhat simpler than that in the NB. While the VN simply speaks of the "character which is the probans," the NB speaks of the "being (existence, nature) of the probans." In this
respect the NB account is similar to that in the HB: the latter, too, speaks of the "being (existence, nature) of the character which is the probans."
The being (or nature) of a probans, however, should not be thought to be different from the probans, for Dharmakiirti does not subscribe to any such ontological position. In this respect Steinkellner's translation of svabhaava as "essential property" could be misleading unless one takes note of the warning supplied by Steinkellner. Svabhaava does not signify an essence or essential property either in a Platonic or in an Aristotelian sense. It does not also signify a common character shared by all the particulars of the same kind in the sense of a Vai`se.sika jaati or universal. Under the circumstances it seems better to avoid such a translation. It is also not proper to say that it "is a particular, necessary characteristic of a real thing without which that thing could not exist."  This implies an ontological distinction between a character (dharma) and the characterized (dharmin) to which Dharmakiirti is opposed. Svabhaava is avyatirikta dharma, as Karanakogomin says and Steinkellner notices.  But again Steinkellner's translation of avyatirikta dharma as "inseparable property" is not completely accurate. First avyatirikta does not mean "inseparable," but rather "nonadditional" or even "nondifferent." Second, dharma usually means "property" or "characteristic." But dharma is also used as virtually synonymous with bhava or vastu. The latter are almost a synonym for the English word "thing"; dharma should be taken in this sense in the present context. Avyatirikta dharma thus signifies something nonadditional or something nondifferent. In this context svabhaava should not be taken as a .sa.sthiitatpuru.sa compound, but as a karmadhaaraya compound: a close English equivalent of svabhaava then is "the thing itself," but it is best to retain the Sanskrit in such a case. However, though admittedly cumbersome, "the self probans'' and "the effect probans" could be offered as English substitutes of svabhaavahetu and kaaryahetu, respectively. It may then be said that the self probans is that probans which is nondifferent from the probandum. This is why it is sharply distinguished from the effect probans, for the latter is different from the probandum.
Our task is to gain a clear understanding of svabhaava as a logical notion and not of svabhaava as an ontological notion. Svabhaava signifies identity/nondifference (taadaatmya/abheda) between the probans and the probandum. We have to clarify identity/nondifference as a logical relation as distinguished from identity/nondifference as an ontological relation.
It may be useful to turn to some commentaries at this point. Commenting on Dharmakiirti's example, namely, this is a tree because of being a `si^m`sapaa, Dharmottara construes it as follows: "This is that to which 'tree' is applicable, because of being that to which '`si^m`sapaa' is applicable."  Durveka Mi`sra, the subcommentator on Dharmottara, understands it in a slightly different way: "That
to which `si^m`sapaa-ness is applicable is that to which tree-ness is applicable, for example, a previously encountered `si^m`sapaa to which tree-ness has been applied. This is that to which '`si^m`sapaa' is applicable."  The premise, namely, "that to which '`si^m`sapaa' is applicable is in every case that to which 'tree' is applicable," has been left understood in Dharmottara's construction. On the other hand, the conclusion, namely, this is that to which 'tree' is applicable, has been left understood in Durveka Mi`sra's presentation.
It appears that both Dharmottara and Durveka Mi`sra took the example to be primarily a linguistic exercise. The point of the inference is that the word "tree" has to apply to something to which the word "`si^m`sapaa" has been applied. In other words if something is called or named a `si^m`sapaa, it has also to be called or named a tree. This is because some of the ideas put together under the name "`si^m`sapaa" are the same as those put together under the name "tree.'' To quote Dharmottara again: "Tallness and so forth are not here the grounds for applying "tree"; rather the ground is `si^m`sapaa-ness as such, that is, possession of branches and so forth, which are included in '`si^m`sapaa'." 
Dharmottara's point, of course, is that calling something a `si^m`sapaa implies that it has branches and so forth. Having branches and so forth is also the grounds for calling something a tree. Hence, the grounds for calling something a `si^m`sapaa and the grounds for calling something a tree are partially identical. It is this kind of identity which gives us the clue to understanding identity/nondifference as a logical relation as involved in a svabhaava probans. The svabhaava probans is logically identical with or nondifferent from the probandum in the sense that the grounds for calling something by the name of the probandum are the same as some (or all, as we shall see) of those for calling something by the name of the probans.
For a fuller explanation of this logical relation, we now turn to the threefold classification of the svabhaava probans. In the first place is that which is called pure (`suddha) or unqualified (nirvi`se.san.a). "Everything real is noneternal'' has been given as an example of it. This kind is called pure because, as we have understood it, the relation between the probans and the probandum is that of pure and unqualified identity. This means that the grounds for calling something by the name of the probandum are exactly the same as those for calling something by the name of the probans. Dharmakiirti's example may seem to be odd to one not initiated into the Buddhist ontology; for the ideas associated with "real" are not, so far as non-Buddhist usage is concerned, exactly the same as those associated with "noneternal." But from the point of view of Dharmakiirti's own ontology the example is quite unexceptionable. According to the Buddhists in general, everything real is noneternal and vice versa; in fact, the Buddhists claim that "real" and "noneternal" are both extensionally and intensionally identical. If an author introduces a new concept and makes it definitionally equivalent to another concept, the relation between the two would be that of pure identity in the sense just given.
If this is so, why should there be a reference to a pot and so forth as supportive evidence for the generalization that everything real is noneternal? Does not such use of perceptual evidence show that the generalization is empirical? If "real" and "noneternal" have the same meaning, the truth of "everything real is noneternal" should be established independently of any experience.
It has to be remembered, however, that the reference to a pot and so forth occurs as a part of a step in a 'demonstration for others' (paraarthaanumaana). The Naiyaayikas, the principal philosophical opponents of the Buddhists, required that supportive examples should be supplied with a general premise in such a demonstration. If Dharmakiirti had not supplied the example, the demonstration would have been regarded as a failure by the Naiyaayikas.
Secondly, though such examples would satisfy the demand of the philosophical opponent, from the point of view of Dharmakiirti's own epistemology this does not amount to producing perceptual evidence for the given proposition. According to Dharmakiirti, any experience in which there is the intrusion of names or concepts is nonperceptual.  Further, names like "pot" do not refer to anything real, but are merely symbols introduced for the convenience of communication. Hence, the reference to a pot and so forth should be taken as inviting attention to linguistic usage, namely, that ideas associated with "pot" include the ideas associated with both "real'' and "noneternal." The implicit claim is that there is no name, the ideas associated which include the ideas associated with "real," that does not include those ideas associated with "noneternal." This is not incompatible with the thesis that the proposition holds by virtue of the meaning of the symbols occurring in it.
Thirdly, the empirical world of common sense is, according to Dharmakiirti, nothing but a linguistic construction. From this perspective a proposition which holds owing to the meaning of its symbols need not be void of content. The distinction between propositions which hold owing to the meaning of their symbols and those which are not so still remains crucial. Hence the need for the distinction between general propositions based on identity (taadaatmya) and those based on causation (tadutpatti); the truth of the former is based upon meaning. To quote Dharmottara: "What is real is necessarily noneternal.... [T]hus reality is necessarily ascribed to noneternality."  Given Dharmakiirti's view about the relation between language and reality, these remarks should be taken to indicate that the investigation is linguistic and/or semantic in spirit, though, from the same point of view, there is no bar to a linguistic/semantic investigation being applicable to the "constructed world."
That svabhaava statements are true by virtue of meaning is also borne out by looking at the two other kinds of svabhaava formula. In neither of these other two kinds is the relation between the probans and the probandum that of pure identity. Hence, the grounds for applying the names of the probans and the probandum are not wholly, but rather are partially, the same. The way in which the second and the third kinds differ from each other is the following. In the
second kind additional notions are not required for the connection in meaning between the probans and the probandum; however, the bridge supplied by additional notions is required in the third kind of case. The point may be explained with the help of the examples supplied by Dharmakiirti. "What is originated is noneternal" is an example of the second kind, while "what is a Product is noneternal" is an example of the third kind. No additional notion is necessary for the transition from "originated'' to "noneternal." However, by a product is meant, as Dharmakiirti says, that which is dependent on the operation of something else for the realization of being. With the help of the additional notion of dependence on something else and so forth, the transition from "product" to "noneternal" may be made.
Svabhaava statements are then either meaning postulates or consequences of meaning postulates with the proviso that in some cases the consequence may be indirect. They are not analytic in the sense of being tautologies or logical truths. They are also not analytic if an analytic statement is required to be empty of content. However, if a statement is called analytic in the broad sense of being a statement, which holds by virtue of the meaning of its symbols, and if in this sense an analytic statement need not be devoid of content, svabhaava statements may be regarded as analytic.  The important point is not to fight over the word "analytic," but to recognize svabhaava statements for what they are: their truth is not based on experiential evidence; rather their truth is based on meaning and may be ascertained through linguistic/semantic analysis; still they are not contentless, They have to be distinguished from general statements based on cansation (tadutpatti). The truth of the latter kind of statements is not based on meaning; an additional factor, namely, knowledge of causal connection, is required as the basis of their truth.
Let us now turn to the apoha (differentiation) theory of names. According to Dharmakiirti, to call something, for example, a mango does not signify that it possesses some positive character (which may be called mango-ness and) which it shares with all other mangoes. The primary significance of such a word is negative, namely, the exclusion of others (anyaapoha) . To call something a mango, then, signifies primarily that it is not a non-mango. Dharmakiirti denies that such a judgment that it is not a non-mango is equivalent to the affirmative judgment that it is a mango, or even that the former presupposes the latter. Similarly, to call something a fruit signifies primarily that it excludes non-fruits. Thus, the general statement, "all mangoes are fruits," should nor be construed to mean that all that possess mango-ness also possess fruit-ness; rather it should be construed negatively that all that excludes non-mangoes also excludes non-fruits. The fact that according to Dharmakiirti's apoha theory names signify exclusion does not conflict with the interpretation of svabhaava statements as being true by virtue of meaning. For the basis for saying that all mangoes are fruits is that the grounds for excluding non-mangoes includes such ideas as, say, possession of pulp and so
forth, which is the same as the grounds for the exclusion of non-fruits. This is why Dharmakiirti says: "In spite of the difference between the characters by virtue of the difference in the exclusion of others (the probans) is as a matter of fact necessarily (eva) identical with the probandum" (HE 4.111). Needless to say, possession of pulp and so forth, too, should be understood as signifying exclusion. Further, although the opoha theory is aimed against the theory of real universals and essentialistic ontologies, it is not against class inclusion as a logical relation. It is still permissible to say that the class of things excluding non-mangoes includes the class of things excluding non-fruits. Similarly, the idea of exclusion of non-mangoes may be said to include the idea of exclusion of non-fruits.
There remains finally the question whether the additional proof brought in in support of a svabhaava statement is purely empirical in character. For this let us look at such a proof in the Vaadanyaaya:
Everything which is so (real or originated) is noneternal, for example, a pot and so forth. Here a proof showing the falsity of the negation (viparyaya) may be produced to substantiate the generalization. Suppose it is not the case that everything which is real or originated is momentary. But that which is nonmomentary cannot be causally efficacious either successively or simultaneously; hence, because of lacking the characteristic of causal efficacy, it must be unreal. Indeed, that which is devoid of all efficacy is unreal. 
This proof is an indirect reasoning to support the statement that everything which is real or originated is noneternal. It starts with the contradictory assumption that some real or originated things are nonmomentary ("momentary" and "noneternal" being used synonymously in the present context). This amounts to assuming that some real or originated things are not noneternal. Then it is pointed out that it is already known that nothing is causally efficacious and further that everything which is not causally efficacious is unreal. Thus, it would follow that nothing nonmomentary is real. Then it is left to the reader to realize that this leads to something unacceptable, namely, that some real or originated things are unreal and, therefore, that the assumption is false. Thus, the reasoning shows that the denial of the svabhaava statement in conjunction with certain other well-established metaphysical theses implies something obviously untenable. It would be wrong to think that such a metaphysical thesis as "nothing non-momentary is causally efficacious" is an empirical truth. In our view such a statement, too, is true by virtue of the meaning of the symbols and is not contentless. Buddhist philosophers including Dharmakiirti have argued at length in support of such a thesis. The arguments are long and difficult and cannot be taken up for discussion here. But in our view those arguments too, are not empirical, but a priori and metaphysical.
We now look at another passage in the Hetubindu:
In the case of svabhaava probans the ascertainment of universal inclusion (anvaya) is as follows: Since the character which is the probandum is as a matter of fact the same as that (the probans), its following upon the mere presence of the character is established. This is a proof of the falsity of that the probans (belongs) to where the probandum is absent. Thus, that which is real is necessarily momentary, since causal efficacy is incompatible with nonmomentariness, the characteristic of being real, which is the same as that (causal efficacy), does not belong to what is nonmomentary. 
Here it is clearly asserted that the identity of the probans and the probandum establishes that the probandum would belong to that to which the probans belongs. The additional reasoning is similar to that in the immediately previous passage and, in our view, nonempirical and a priori in character. The important point here is that the identity of the probans and the probandum is said to be the basis of the ascertainment of pervasion between the former and the latter. This identity is, as explained, constituted by the fact that the grounds for calling something by the name of the probandum are wholly or partially the same as those for calling something by the name of the probans. This goes to show that svabhaava statements hold by virtue of meaning.
BL I, II - Theodore Stcherbatsky, Buddhist Logic. 2 vols. Reprint, New York, 1962.
HB - Dharmakiirti's Hetubindu. Tell I, Tibetan text and reconstructed Sanskrit text. Ed. E. Steinkellner. Wien, 1967.
NB - Dharmakiirti's Nyaayabindu.
NBTP - Dharmakiirti's Nyaayabindu with Dharmottara's .Tiikkaa and Durveka Mi`sra's Dharmottarapradiipa. Ed. D. Malvania. Patna, 1955.
PV - Dharmakiirti's Pramaa.naavarttikam. Ed. Dvarikadas Sastri. Varanasi: Bauddhabharati. 1969.
VN - "Dharmakiirti's Vaadanyaaya with `Saantarak.sita's Commentary." Ed. R. Saa^nk.rty^na. Journal of Bhandarkar Oriental Research Society 21 and 23.
1. BL I, pp. 248-251, and BL II, pp. 65-67.
3. NBTP 108.
4. BL I, pp. 251, 271.
5. E. Steinkellner, "On the Interpretation of the Svabhaavahetu.h, " Archiv fur Indische Philosophie (1974): 117-129.
6. BL II, p. 66.
7. Svabhaava.h svasattaamaatrabhaavini saadhyadharme hetu.h Yathaa v.rkos'ya^m `si^m`sapaatvaat. (NBTP 106)
Yat sat tat sarvam anityam, yathaa gha.taadir iti `suddhasya svabhaavaheto.h prayoga.h. (NBTP 156)
Yad utpattimat tad anityam iti svabhaavabhuutadharmabhedena svabhaavasya prayoga.h. (NBTP 157)
Yat k.rtaka.m tad anityam iti upaadhibhedena. Apek.sitaparavyaapaaro hi bhaava.h svabhaavanispattau k.rtaka iti. (NBTP 158-159)
Vastutas tayos taadaatmyam. Tanni.spattaav ani.spannasya tatsvabhaavatvaabhaavaat. (NBTP 162-163)
Hetu.h svabhaave bhaavo'pi bhaavamaatraanurodhini. (PV 251)
Saadhanadharmabhaavamaatraanyayini saadhyadharme svabhaavo hetu.h. (HB 39)
Saadhanadharmamaatraanvaya.h saadhyadharmasya svabhaavahetulak.sa.na^m siddha^m bhavati. (VN 9)
8. Steinkellner, "Interpretation," pp. 123-124.
9. Ibid., p. 123.
10. V.rk.savyahaarayogyo 'ya^m `si^m`sapaavyavahaarayogyatvaat. (NBTP 106)
11. Ya.h `si^m`sapaatvavyavahaarayogya.h sa v.rk.satvavyavahaarayogyah. Yathaa pravarttitav.rk.sat- vavyavaharaa puurvaadhigataa `si^m`sapaa. `Si^m`spaavyavahaarayogya`s caayam iti. (NBTP 107)
12. Noccatvaadi nimittantaaram iha v.rk.savyavahaarasya. Api tu `si^m`sapaatvamaatra^m nimittam--`si^m`sapaagatasaakhaadimatva^m nimittam ity artha.h. (NBTP 107)
13. BL II, pp. 14-24.
14. Yat sat tad anityam eva.... sattvam anityatve niyata^m khyaapita^m bhavati. (NBTP 157)
15. Most philosophers of this century hold that if a statement is analytic and true by virtue of meaning, it is factually empty. This is not in agreement with Dharmakiirti's point of view. The view that an analytic statement may have factual content is, however, upheld by some recent philosophers. See, for example, Hao Wang, Beyond Analytic Philosophy (Cambridge, Massachusetts: MIT Press, 1986), p. 14.
16. Ya`s caiva^m sa sarva.hanitya.h yathaa gha.taadir iti. Atra vyaaptisaadahana^m viparyaye baadhakapramaa.nopadar`sanam. Yadi na sarva^m sat k.rtaka^m vaa pratik.sa.navinaasi syaat ak.sa.nikasya kramayaugapadyaabhyaam arthakriyaa'ayogaad arthakriyaasaamarthayalak.sa.nam ato niv.rttam ity asad eva syaat. Sarvasaamarthyopaakhyaavirahala.k.sa.na^m hi nirupaakhyam iti.(VN 6-8)
17. Anvayani`scayo 'pi svabhaavahetau saadhyadharmasya vastutas tadbhaavataya saadhanadharmabhaavamaatraanubandhasiddhih. Sa saahyaviparyaye hetor baadakapramaanavrtti.h. Yathaa yat sat tat k.sa.nikam eva, ak.sa.nikatve `rthakriyaavirodhaat tallak.sa.nam vastutva^m hiiyate. (HB 37)