Structure and Authorship of the KusumÀ·jali

This paper suggests that the classic of Indian theology, the Nyāya-kusumâñjali is in fact two texts: an earlier treatise in 65 ślokas, and Udayana’s (mostly prose) commentary on it. Internal evidence consists in: (a) the ślokas read as a continuous text; (b) there are extremely long prose passages without verses; (c) Udayana does not comment on his own verses, only on the ślokas; (d) the basic plan of the two texts are markedly different; (e) different content of some chapters: ch. 1 about karma vs. rituals to reach heaven, ch. 2 about creation vs. eternality of sound, and in ch. 5 Udayana doubles the arguments for God; (f) Udayana deals extensively with atoms and yogic perception and rejects the concept of śakti, in contrast to the verses; (g) there are a few manifest disagreements (on creatio continua and the sacredness of god-images). External evidence for the thesis: (a) there are mss. of the verses only; (b) there are many commentaries on the verses only; (c) these commentators—with a single exception—do not seem to take Udayana for the author of the verses; (d) the first commentary on Udayana names his own work a subcommentary; (e) in his other works, it is atypical of Udayana to insert ślokas in his text; (f) a legend of the Bhāduṙī Brahmins stating that Udayana “received” the Kusumâñjali.

commentary on them 1 . When preparing a paper 2 for a conference some four years ago, I noticed that the commentary does not always harmonise with the verses, and proposed that the author of the prose part (NKp) was different from the author of the kārikās (NKk). Now I am going to modify this thesis slightly. The NKk consists of 65 simple śloka verses and 8 verses in elaborate classical metres; I think that the latter belong actually to the commentary, and the work commented upon was the śloka-part only. This hypothetical composition in 65 ślokas I will abbreviate as NKś.
Udayana, Udayanâcārya or simply Ā cārya wrote the NKp around the start of the 11th century CE. 3 The universally accepted title of the work is Nyāya-kusumâñjali, although in the text itself, in the first and last verses, we find only two synonyms: Nyāya-prasūnâñjali (1.1d) and Nīti-kusumâñjali (5.20a), all meaning An offering of the flowers of logic. These variants are not significant, since they are necessitated by the metre-"Nyāya-kusumâñjali" does not fit either śārdūla-vikrīḍita or vasantatilakā, the metres of 1.1 and 5.20, respectively. More importantly, the colophon of the whole work calls it Nyāya-kusumâñjali-prakaraṇa (NKp), Treatise on the  The NKp-commentary Bodhanī also names it a 'treatise'. 5 The work is divided into five chapters that all the editors call stabakas, bunches of flowers, already in the colophons of the first four chapters of the NKp. This naming convention seems to be centuries later than the text itself, for it is absent from our earliest witness. There are many important old and modern commentaries on the NKp; the most helpful and by far the earliest extant commentary is Varadarāja's Kusumâñjali-bodhanī, Clarifying the Kusumâñjali. It was written about a century after the NKp, 6 while the next published commentary, Vardhamāna's Nyāya-kusumâñjali-prakāśa (Light on the Nyāya-kusumâñjali) followed only after another two centuries. 7 Neither Varadarāja's Bodhanī, nor the text of the NKp itself uses the term stabaka. The Bodhanī consistently calls the chapters paricchedas, sections, in all the five chapter colophons and also in the text itself. 8 A later, unpublished commentary by Vāmadhvaja (1561 CE) still uses this terminology. 9 This suggests that the colophons of the NKp chapters are not original; in fact one partial ms. does not have any chapter division at all. 10 The treatise consists of 73 verses (kārikās) and a 26 times longer prose discussion built around them. It may be seen as a very loose commentary on the kārikās. Therefore it would be logical to ask if it is really one work, or in fact two? And even if they are two, can they have the same author? Unfortunately it seems that these questions were never asked: modern scholarship just took it for granted that it is a single text 11 written by one person, Udayana.
In the NKp itself there is no obvious evidence either way. The prose does not refer to the author of the verses. Sometimes the kārikās are handled in a proper commentarial fashion: short introduction (avataraṇa), then the verse itself, followed by a full explanation of its terms and logic, e.g. 1.7. Other kārikās, e.g. 1.9., are very well integrated into the continuous flow of argument. In all cases except three, the ślokas are preceded by a clear delimiting expression like ucyate ('it is said', 26 times), iti cet, na 12 ('it is not so', 15 times), tathā hi or api ca ('for', 'further', 10 times). Unfortunately Udayana never uses an active verb form that would show if it is first or third person ('I say' vs. 'he says').
Considering internal and external evidence, in this paper I will suggest that the kārikās (or, more precisely, the kārikās in śloka metre) are in fact an independent composition, written by an earlier Nyāya author.

The kārikās as Stand-Alone Text
Reading the NKp one cannot fail to notice that it is not a miśra-prakaraṇa, a text written in prose and verse mixed, as it is sometimes stated. The kārikās make up a fairly continuous text, clearly written with the intention to be independently 8 At the start of chapters two and four. Atha dvitīyaṃ paricchedam ārabhamānas… "Now, he begins the second section…" (NKpB 1 59.3 = NKpB 2 209.2 = NKpB 3 233.1). Atha turīyaḥ paricchedaḥ. Sattve 'pi tasyâpramāṇatvād ity-asyâsiddhi-vyutpādanaṃ paricchedârthaḥ. "Now the fourth section. The purpose of the section is to develop a refutation to the thesis, »Even though He exists, He is no source of knowledge«" (NKpB 3 502.1-2). In the āryā verse closing the first chapter, it is called-no doubt, metri causa only-khaṇḍa, part. Udayana-gambhīra-bhāva-prakaṭana-nipuṇena Varadarājena | vyākhyāta eva gahanaḥ Kusumâñjaleḥ prathama-khaṇḍaḥ || "Varadarāja, skilful in bringing to light the deep purport of Udayana / has now explained the impenetrable first part of the Kusumâñjali" (NKpB 1 58.26-27 = NKpB 2 208.11 = NKpB 3 218.1-2). 9 The ms. is carefully described by DALAL (1937: 103-104), giving the colophons of chapters 1 and 5, in both cases called pariccheda. Dalal's information is somewhat analysed by RADHAKRISHNAN (1940). 10 COWELL (1864: x-xii) describes it in some detail, saying that "there is no sign of division between the two chapters,-the discussion of the second objection goes on in unbroken continuity with that of the first." 11 Actually MIŚRA (1968: 2) remarks without further details that "some people think that first the verse part of the Nyāya-kusumâñjali was written, but on account of its difficulty Ā cārya later wrote the commentary," kucha logo kī dhāraṇā hai ki pahale 'Nyāya-kusumāñjali' kā kārikā-bhāga likhā gayā, parantu isakī kaṭhinatā ke kāraṇa bāda me Ācārya ne vr tti bhī likhī. (Most of the anusvāras are dropped in this text.) 12 Or close synonyms, like abhidhīyate and etad api nâsti, na + Ablative etc. memorised. And in fact they do exist as an independent text, NKk, especially in Bengal. "The book Kusumâñjali has two forms: one consists of the kārikās only, the other is accompanied by the exposition written by Udayana Ā cārya. … The first, consisting of the kārikās only, is well-known primarily in Bengal." 13 It seems even more significant that there are many commentaries on the NKk, most of them 14 come again from Bengal. At least five NKk-commentaries have printed editions; the earliest one, NKkH by Haridāsa Bhaṫṫācārya (1530 CE) seems to be more popular than the NKp itself and has seven editions. The other four commentaries, by Rāmabhadra Sārvabhauma Bhaṫt˙ācārya (NKkR, 1580 CE), Jayarāma Nyāyapañcānana (NKkJ, 1620-1659, Gopīnātha Maunin (NKkG, 1650 CE), and Nārāyaṅa Tīrtha (NKkN, 1700 CE) have one edition each. The mere existence of these commentaries proves that their authors thought of the NKk as an independent work: it would seem quite unusual and disrespectful to the author to detach parts of his composition (without even mentioning this procedure) and comment only on them. Also our earliest witness, Varadarāja called his Bodhanī a subcommentary (ṭīkā) on the Nyāya-kusumâñjali, 15 suggesting that the NKk is the root-text and the NKp is a commentary on it.
From among the NKk-commentators only Gopīnātha Maunin thinks that the kārikās are the work of Udayana. He says in an introductory verse that he starts his exposition because "the sayings of Ā cārya are impenetrable, young people cannot comprehend them", and in his last words unambiguously states that "Gopīnātha prepared this exposition of the kārikās of Ā cārya." 16 The other four commentators are conspicuously silent on the question. The regular practice would be to pay homage to the author in one of the introductory verses. They don't do this, and as a sort of compensation NKkJ and NKkN mention the title of the work commented upon, Kusumâñjali-kārikāḥ. Further, NKkH and NKkN give separate colophons to the kārikās and to the commentary-and even in the kārikā-colophons no author is mentioned, although this is one of the main functions of a colophon. 17 So apparently these four commentators thought that the author of the NKk was unknown.
The way the NKk commentators quote Udayana's prose is not very informative.
(a) Haridāsa Bhat˙ṫācārya always clearly identifies the kārikās, most often (66 times out of 73) with an immediately preceding āha, 'he says'. When he quotes the NKp (NKkH pp. 3.9-4.6), nothing shows that a quotation starts; at the end an iti is ambiguous, for the quoted text itself ends with iti. Then he comments on many words of the quotation (p. 4.7-17) and within this section he refers to the text once with an āha. 24 (b) Rāmabhadra Sārvabhauma marks neither the kārikās nor the Udayanaquotations 25 in any systematic manner, although commentaries on the NKp are normally identified (see fn. 19). In a single case he says before a NKp-quotation: tathā cÂcāryaḥ, 'Ā cārya also says so' (NKkR p. 17.16). Frequently instead of himself commenting on the NKp, he quotes immediately after the text a subcommentary, mostly Ś aṅkaramiśra's Āmoda.
(c) Jayarāma Nyāyapañcānana, like Haridāsa, typically introduces the kārikās with āha. On the other hand, when he quotes the NKp, 26 there is nothing to show this (or, in a single case, a closing iti only, NKkJ p. 61.11).
(d) Gopīnātha Maunin, who does identify the author of the NKk with Udayana, quotes both similarly, i.e. preceded by āha. He quotes the NKp only once (NKkG pp. 8.29-9.8), ends the text with iti; no commentary added.
(e) Nārāyaṅa Tīrtha marks the kārikās clearly, most of the time with the usual āha. He does not quote the NKp at all.
To sum up: with the exception of Gopīnātha Maunin, the NKk-commentators never mention by name or describe otherwise the author of the kārikās. They introduce his text with a 'he says', 'he refutes' etc. Only Gopīnātha quotes the NKp similarly; the others simply include Udayana's text in theirs. In a single case Rāmabhadra names the NKp author as Ā cārya, the others never name him or refer to him in any way.
There is some evidence that at least Gopīnātha and Jayarāma extracted the verses themselves from a copy of the NKp where the kārikās were not numbered, for they each have three extra "kārikās" -in each case, verses quoted in NKp at that place. They have misidentified these illustrative quotations of Udayana as kārikās. 27 Further, all the commentators take the verses in artistic metres as kārikās, although -as we will see below-they belong to the NKp.
The picture emerging is fairly clear. Udayana's NKp was accepted by everyone as the authoritative commentary on the NKk and therefore the reliable source of the text of the kārikās; but before the 19th century, perhaps Gopīnātha Maunin was the only person who thought that Udayana was the author also of the verses.
There is an interesting legend mentioned by Narendrachandra VEDAN-TATIRTHA (1944: xxviii) that would corroborate our inference. "It is sometimes believed that Udayana did not compose N[yāya-]K[usumāñjali] himself but found it on a pilgrimage and circulated its content in Gauḋa (Bengal) Unfortunately Vedantatirtha's information seems to be not fully accurate. I could not find the line quoted in the three editions of the BhP consulted, although they have very detailed tables of contents, and BhP 3 has a full ślokaindex in vol. 3. These editions do not have a part or chapter entitled Bhaktimāhātmya. However, Rāmacandra DVIVEDIN (1992: [10]-[11]) also quotes four verses, first giving the source as Bhakti-māhātmyam ity-ākhye granthe, ('in the book called The greatness of devotion') then as Bhaviṣya-purāṇe triṃśattame pariśiṣṭâdhyāye ('The thirtieth appendix-chapter in the Bhaviṣya-purāṇa'). More than a century ago, Vindhyeśvarī-Prasāda DVIVEDIN (1916: 152-159) published a seemingly complete text in 85 ślokas, naming the source as Bhaviṣya-purāṇapariśiṣṭe Bhagavad-bhakta-māhātmye 30 adhyāye ('in the 30th chapter of The greatness of the devotees of God, appendix to the Bhaviṣya-purāṇa'). The colophon of the text reads Bhaviṣya-purāṇa-pariśiṣṭe Udayanâcārya-caritaṃ 28 The only major difference is that Nārāyaṅa Tīrtha excludes the last kārikā.-Interestingly in a ms. of Varadarāja's Bodhanī (the earliest commentary on the NKp) in the Sarasvatī Mahāl Library in Tanjore (No. 6532, and see also No. 6533), the last two verses are omitted (SASTRI 1931: 4728-4730). Unfortunately Mahaprabhulal Goswami, the editor of NKpB3 (the only edition containing also the last two chapters of the Bodhanī) did not utilise these mss. The mss. do not have a colophon at the end; the possibility of a lost last folio should be checked. Further, NKkH practically does not comment on the last two kārikās, and does not introduce them with his usual āha.
The printed editions give the appearance that several other verses are missing from some NKk commentaries: 1.3 (NKkH, NKkR); 3.14 (NKkJ); 5.5 (NKkG); 5.6 (NKkH). But in fact the verses are there, only not marked as kārikās, and sometimes even printed as prose text. 29 In the commentary on verse 5.3: etac ca sarvam Ātma-tattva-viveke nipuṇataram upapāditam, iti nêha pratanyate. "But all this has been more completely proven in the Ātma-tattva-viveka, therefore it is not elaborated here.  Now this text is in fact completely unrelated to the Bhaviṣya-purāṇa. Already AUFRECHT (1891) listed three mss. of the Bhagavad-bhakti-māhātmya and named its author, Candradatta of Mithilā. Quite recently Bābūrāma Ś arman critically edited the complete work in 150 chapters from three mss. In his edition the Udayanâcāryacarita is the 31st chapter . This is probably a relatively modern text, 30 but for our purposes it is irrelevant, since it does not contain the legend about the origin of the Kusumâñjali-actually the only work by Udayana mentioned at all is the Kiraṇâvalī.
Vedantatirtha's real source for the legend is probably BHATTACHARYA (1958: 5), who mentions the Bhaviṡya-Purāṅa immediately after discussing the legend in a footnote. (This may have caused the confusion, especially for someone quoting from memory.) According to him, there is a strong tradition in the Bhāduṙī family (Vārendra area in North-West Bengal) that Udayanâcārya, the author of the Kusumâñjali belonged to them. He quotes the śloka in full: sa evÔdayanâcāryaś cikāya Kusumāñjalim | tīrtha-paryaṭane labdhaṃ, tasmād Gauḍe pracāritam || "This Udayanâcārya collected the Kusumâñjali / that he acquired on a pilgrimage; then he popularised it in Bengal." He names his source as "a clever Vārendra scholar of the last century … (Laghubhārata Vol. III, p. 160-61)". The Laghubhārata is not available to me, but MONAHAN (1920: 15) says that it is "a genealogical work… published about 50 years ago", i.e. around 1870. He takes it entirely seriously as a historical source for the circumstances of the birth of Lakṡmaṅa Sena, only a century after Udayana; therefore it is not improbable that the legend about the Kusumâñjali preserves some truth. Bhattacharya, however, thinks that it is impossible, for according to him this Udayana of the Bhāduṙī family lived after Ballāla Sena, at least 150 years too late. Unfortunately on this point Bhattacharya gives no references. Rāmacandra DVIVEDIN (1992: 11), on the other hand, finds it unproblematic to quote about our Udayana from the Bhāduṙī genealogies (Bhāduḍī-vaṃśâvalyām). So without further research, especially into the Bhāduṙī and generally Vārendra Brahmin genealogies, it is impossible to say how old this legend is.

Structure and Style
The distribution of the kārikās in the NKp is summarized in the table below. The number of verses in each block, if more than one, is given in parentheses. The verses

Introduction
Introduction (2) Chapter intro Chapter intro Chapter intro Chapter intro Chapter intro (15) (2)  (2) Viewed thus, the chapter division seems quite irregular. The body of chapter two has only two ślokas, while chapter three has 21.
On the other hand, the length of the prose chapters appears normal. In the following table, the length of the prose is given in theoretical lines, where a 'line' means a hundred roman characters (without spaces).  (2) This huge difference between NKk and NKp in the proportions of the chapters is the result of the varying lengths of the commentaries on the ślokas. The average is 30 lines of prose per śloka, yet sometimes there is but one line of comment (on 1.4 and 5.1), or even less: nigada-vyākhyātam etat, 'this is self-explanatory', on 3.5. (This brief remark clearly suggests that it is a commentator speaking, not the author of the verse.) Conversely, we find some quite extensive discussions on a single śloka. The commentary on kārikā 2.1 is 229 lines long, so it makes up 12% of the whole prose! If Udayana were the author of the kārikās, he could have added a few more ślokas here. On the other hand, when a scholar comments on a fixed text (especially an older one), often he has some important new things to say for which there is no root-text. Then he inserts his sometimes lengthy discussion as an aside to a related topic.
As the chapter division fits well the NKp while for the NKk the length of the chapters is rather anomalous, we have to consider if this was really a part of the original Kusumâñjali, or is it an innovation of Udayana? A text of 73 verses (like the Sāṃkhya-kārikā) usually does not have any formal divisions. Here however, as seen in the first table, the chapters are clearly indicated by an introductory śloka and a concluding prayer in śārdūla-vikrīḍita metre. Now if we removed the kārikā-numbering from the text, it would be perfectly clear that the verses in artistic metres (printed in bold above) do not belong to the kārikās commented upon, but to the prose commentary. Otherwise the prose would not have the compulsory maṅgala-śloka, the benedictory verse invoking God at the very start of the text, nor the almost-compulsory concluding verse. All other published works of Udayana have both. 31 Further, all the true kārikās receive at least a few words of comment, and they are introduced with an ucyate or similar expression; while none of the artistic verses show either feature. They stand always at the extreme margin of a unit: no prose comes before the introductory verse, nor after the concluding prayer in each stabaka. Some of these verses are quite difficult (especially 1.1.) and they receive lengthy explanations from all commentators both of the NKp and the NKk. That Udayana does not comment on them shows that they were not part of the text he commented upon (the Kusumâñjali ślokas, NKś) but his own writing. (And conversely this shows that the NKś was not his own work.) The introductory ślokas are partly an illusion caused by accepting the current stabaka division. Verses 4.1 and 5.1 do in fact outline the discussion in the following chapter, and 2.2-3 prove a part of 2.1 (sarga-pralaya-sambhavāt, 'for the world is created and will dissolve'). On the other hand, 1.4 is not an introduction to the whole chapter, only to 1.5-9, and 3.1 is discussed only in 3.2-3. Also, there are other "introductory" verses in the NKk like 1.11 (saṃskāraḥ puṃsa evêṣṭaḥ, 'rituals influence only the soul') to 1.11-19, or 3.10 to 3.11-14 (upamāna, 'comparison' is not a separate kind of pramāṇa, 'valid source of knowledge').
The basic plan of the NKp and the NKk differ noticeably. Udayana is explicit about the structure of the NKp (and most commentators of the NKk follow him blindly). He says that although everybody accepts the existence of God in one sense or another, "summarily there are five objections: (1) There are no transcendental means to reach heaven; (2) These means can work otherwise [i.e., without God]; (3) There are proofs for His nonexistence; (4) Even if He exists, he is not a source of valid knowledge; (5) There is no proof for his existence." 32 The five chapters answer these five objections.
The NKk follows a more continuous flow of argument, an outline of which is as follows: Human destiny has non-physical causes, and rituals can influence them (1.4-9). The substrate of these karmic influences is the soul, which is the eternal subject of both actions and experiences (1.10-19). Ritual knowledge is not eternal, for its decay is observed; therefore it needs an author. That can only be God, who taught it to the first Brahmins after the creation of the world (2.1-3). It is impossible to disprove God (3.1-22), nor that He is the source of scripture (4.1-5). God's existence can be proven from the physical world needing an intelligent creator (5.1-5) and from scripture (5.6-17). 31 Of course the unfinished Kir does not have a concluding verse, and it is also not found in the incomplete edition of the NVTP. (On the other hand we know that "Udayana begins each chapter of the Pariśuddhi with a prayer verse", BHATTACHARYA 1958: 20.) NP does not have a maṅgala-śloka, and this corroborates our impression that it is only the last part of a (probably lost) larger work-in its extant form it is a commentary on the last chapter only of the Nyāya-sūtra. 32 Tad iha saṃkṣepataḥ pañcatayī vipratipattiḥ: alaukikasya paraloka-sādhanasyâbhāvāt; anyathâpi paraloka-sādhanânuṣṭhāna-sambhavāt; tad-abhāvâvedaka-pramāṇa-sad•bhāvāt; sattve 'pi tasyâpramāṇatvāt; tat-sādhaka-pramāṇâbhāvāc cêti. Immediately before 1. There are three conspicuous differences between the conceptions of the two works. For Udayana the first chapter is about rituals that get us to heaven, while in the NKś argument svarga ('heaven') does not even appear, it is about karmic influences on our lives. 33 In the NKś, chapter two is about the creation and dissolution of the world and of scripture with it; Udayana inserts here his extremely long (220 lines) refutation of the eternality of śabda ('sound' or 'word'). Finally, in the fifth stabaka Udayana doubles the arguments for the existence of God, saying that 5.1 can be interpreted in two different ways-in the NKś there is no hint about this possibility.
The style of the two works is again markedly different, the only similarity being that both are very difficult to read. In Udayana's case, the reason is that he presupposes only highly educated scholarly readers who are well versed both in the special Nyāya terminology and methods of argumentation and in other fields of classical scholarship like Pūrva Mīmāṁ sā. The style of the NKś recalls rather the kārikās of the old masters like Īśvarakr ṡṅa, Vasubandhu or Diṅnāga. In that age writing was probably not yet the central medium of education. 34 The students memorised the verses, then the teacher gave them the explanation. The kārikās were not meant to be read and understood; they contained all the relevant terms and concepts with perhaps a hint at their relation. The verse format helped the memorization and ensured to an extent the preservation of the key elements and their structure. With their help the students could later readily recall the essence of the oral commentary received decades ago.
Of course, matters of style are difficult to judge. What can be objectively said is that using verses in the body of the text is atypical of Udayana, and he prefers artistic metres to ślokas. In his other works I found only five internal verses (excluding quotations): three ślokas, two in artistic metres. 35 Among the fifteen introductory or concluding verses nine are in some artistic metre. 36 Among the artistic metres he uses most frequently śārdūla-vikrīḍita (four times). All this is perfectly consistent with our hypothesis: in the NKp, the introductory and concluding verses are in śārdūla-vikrīḍita, except the very last (vasanta-tilakā, also used in Udayana's other works). 33 In the NKp, alaukika paraloka-sādhana (see the previous footnote), in the NKk praty·ātma-niyamād bhukter / asti hetur alaukikaḥ || (1.4cd), "since experiences are determined for each self, there is a nonphysical cause". 34 This is, of course, speculation based on some work with those texts and some knowledge on how orthodox (Vedic) education works even to-day. I have no direct evidence about the teaching methods of the first millennium CE. 35 LĀ p. 78.21-29, śārdūla-vikrīḍita; Ā TV p. 529.2-5, vasanta-tilakā. The three saṃgraha-ślokas:  Three in śārdūla-vikrīḍita: Ā TV p. 1.1-4; NP p. 124.8-11; Kir p. 5.1-4. Two in mandâkrāntā (Ā TV p. 947.5-8; NP p. 124.12-15) and in mālinī (Ā TV p. 947.1-4; Kir p. 7.1-4); one in śikhariṇī (Ā TV p. 935.6-9, not a real concluding verse, as 26 lines of prose follow it before the end) and in vasanta-tilakā . The six ślokas: LM p. 1.11-12 = Kir p. 397.1-2 (introductory verse to chapter 2); LM p.

Ideas and Terminology
Since both texts belong to the Nyāya school, 37 we can hardly expect serious discrepancies here, still some differences can be observed. The general attitude of the NKś is entirely serious and focussed exclusively on proving the existence of God. The NKp seems to think that there is no real need to prove God, but it is an excellent topic around which many interesting discussions of Nyāya tenets can be woven. Probably it was written in a dominantly theistic environment; the verses in artistic metres read in fact as pure devotional poetry, Śiva-bhakti.
The theory of atoms is irrelevant for the NKś. Atoms are briefly mentioned once, in the course of proving that rituals affect only persons, not objects, for "their qualities are [based on] the differences of atoms produced by heating and the like." 38 Unsurprisingly Udayana-who is also a Vaiśeṡika author-writes a lot about atoms, the word occurs 39 times in comments on fifteen kārikās; 39 he even builds two of his arguments for the existence of God on the atomic theory.
The problem of yogi-pratyakṣa, the supposed ability of yogins to perceive things invisible for others is absent from the NKś; neither yoga nor yogins are mentioned in any way. On the other hand Udayana thinks that yogins can perceive God directly; he elaborates on the problem at some length (26 lines) and at a quite conspicuous place: this is the very end of the NKp prose. 40 Conversely, the concept of śakti ('power, ability') is important in the NKś, appearing in five verses, 41 whereas for Udayana it is a faulty Mīmāṁ sā concept. His commentary on kārikās 1.9-13 is essentially a substantial (120 lines) essay on why śakti does not exist. At the very end of this discussion-clearly in order to make room for the kārikās' frequent use of the word-he writes: "It is not the case that our school does not have even the category of śakti. -What is it then? -Being a cause." 42 Still he avoids the word, using it only to present opponents' views and to comment on the verses where it occurs. 43 A minor but noticeable difference in terminology is that the verses use for 'perception' both pratyakṣa and adhyakṣa (both twice), while Udayana practically always chooses pratyakṣa (71 times). Only when commenting on adhyakṣa in 1.16, he uses the same word thrice, and in a single case he freely selects this term-for alliteration's sake. 44 As CHEMPARATHY (1972: 43) remarks, "Another interesting, though less frequent, feature of his works is the use of alliteration usually rare in philosophical works." There are a few cases where Udayana clearly disagrees with the kārikās. Listing the arguments for God's existence, the NKś says: "The omniscient and eternal one can be proven from effect, joining, keeping, etc., …" 45 The idea is that the world is an effect, since it is not an unchanging eternal substance; therefore an intelligent being must be supposed who constructs, preserves and destroys it. That dhr ti (literally, 'holding, keeping') here means 'sustainment, preservation' is clear in itself, especially considering the obvious parallel of the Brahma-sūtra: Janmâdy asya yataḥ, "[Brahman is that,] from whom the birth etc. of this [world comes]", where janmâdy is explained by Ś aṅkara as "birth, continued existence and destruction." 46 The NKś elaborates on the list of arguments, and in verse 5.2 we read dhr ti-vināśayoḥ, "in preservation and destruction", which confirms the interpretation given above. In spite of this, Udayana explains dhr ti as 'holding in space': "The earth… is controlled by the will of its supporter, for, being heavy, it is not its nature to fly." 47 The verses 3.20-22 consider the problem whether an·upalabdhi, 'non-perception' is a separate source of valid knowledge. Udayana in the introduction to 3.20 suggests that it is superfluous: "It has been considered before that non-perception cannot disprove God and it does not go beyond perception etc." 48 (And in fact, in the NKp the problem has been profusely discussed earlier.) According to kārikā 1.12cd, "god-[images are sacred] because [the gods stay] near them, or because we recognise [the particular god in the image]." 49 Udayana accepts only the first explanation and rejects the second: "nearness is our view. For others, it is recognition". 50
For a critical edition, the unpublished commentaries of the NKk must be taken into consideration. Even when they are philosophically less interesting, as VEDANTATIRTHA (1944: xlix) suggests, 53 they may preserve old readings. Again it is impossible to say how many NKk commentaries there are, for many catalogues do not distinguish them from commentaries on the NKp. AUFRECHT (1891) lists eight commentaries on the Nyāya-kusumâñjali that are to my knowledge still unpublished. Their authors are Gadādhara, Candranārāyaṅa, Trilocanadeva, Raghudeva, Raghunātha, Rāmabhadra (son of Bhavanātha), Rudra Bhaṫt˙ācārya and Vāmadhvaja. It is clear that at least three among them are NKk-commentaries: those by Raghudeva Bhat˙ṫācārya, by Rāmabhadrâcārya Bhavanātha-putra (for both see STEIN 1894: 148) and by Rudra Bhaṫt˙ācārya (HALL 1859: 84). The other five are also probably NKk-commentaries (as perhaps most of the vyākhyās where the author is unknown), but we cannot be sure. The Nepalese German Manuscript Cataloguing Project also has at least two further NKkcommentaries, by Candraśekhara Sūri (NGMCP Acc.Nr. 5/4172 and 3/567) and by Ś rīrudranyāyavācaspati (NGMCP Acc.Nr. 4/1672 and 5/4170).
Since we saw that the (direct or indirect) source of the NKk text in the commentaries is Udayana's treatise, it must be carefully investigated: perhaps some other ślokas in the NKp do in fact belong to the NKś. The next step may be an attempt to interpret the verses on their own, counting with the quite real possibility that some ślokas are not in their original location. Then a comparison with other relevant works (especially Nyāya, Mīmāṁ sā and Buddhist) may suggest the historical position of the work.
Our results so far advise some caution with 'self-commentaries' in the Indian tradition. They are usually early and quite reliable commentaries, but often not in fact the work of the same author. I have arrived at this result with Nāgārjuna's Vigraha-vyāvartanī and with Vasubandhu's Viṃśikā and Abhidharma-kośa; 54 I also have suspicions about Diṅnāga's  Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this 51 One in Srinagar, Ś āradā script, 7 folios (YACH-KAUL n.d.: 168, accn.no. 1043-11); and one in Benares, Bengali script, 20 folios ([SCLB] 1911: 223, item 1437 = Nyāya Ç āstra no. 206). 52 One in Vrindaban, Bengali script, 11 folios (Gosvami 1976: III.158-159, Ser.No. 7653, Acc.No. 8071); one (possibly in Kathmandu), script unspecified, 12 folios (NGMCP, Acc.Nr. 5/6873). 53 With the exception of Raghudeva Nyāyālaṁ kāra's commentary, VEDANTATIRTHA (1944: xlix) 54 Such claims cannot be substantiated in a footnote; their main basis is always the different content. But there may be other traces like the commentator calling the author of the verses ācārya, 'teacher' in the Abhidharma-kośa (in the introduction to 1.3, AKBh 1 p. 2.17 = AKBh 2 p. 3.4). For the Viṃśikā see RUZSA-SZEGEDI (2015: 101-104, 116). 55 Here suspicions only, since I am not a Tibetologist and therefore with the Tibetan text I have to rely mostly on translations, and from the Sanskrit text of the commentary we have but fragments. Also I have studied the fifth chapter only. article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creativecommons.org/licenses/by/4.0/.
Funding Open access funding provided by Eötvös Loránd University.