The Pali Subodhlakra and Dain's Kvydara Author(s): J. C.

Wright Source: Bulletin of the School of Oriental and African Studies, University of London, Vol. 65, No. 2 (2002), pp. 323-341 Published by: Cambridge University Press on behalf of School of Oriental and African Studies Stable URL: http://www.jstor.org/stable/4145617 . Accessed: 15/11/2013 11:15
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Cambridge University Press and School of Oriental and African Studies are collaborating with JSTOR to digitize, preserve and extend access to Bulletin of the School of Oriental and African Studies, University of London.

http://www.jstor.org

This content downloaded from 163.1.62.81 on Fri, 15 Nov 2013 11:15:58 AM All use subject to JSTOR Terms and Conditions

The Pali

and Dandin's Subodhalanikara Kavyadarsa

Schoolof Oriental andAfricanStudies Subodhalankara, a Pali treatise on rhetoric in 367 verses, has been attributed, on the evidence of the colophons, to a twelfth-century author Safigharakkhita Mahasa-mi. A commentary, whose fourth-chapter colophon associates the title and name with itself rather than with the verses (Mahasami-namikayam has acquired the name Pordna-tTka. G. E. Fryer had Subodhalanikaratfkayam), edited the verses alone (in 'Pali studies. No. 1 ', JASB, 1875, 91ff.): his omission of the commentary and failure to note the close affinity of the combination of these two texts (as Karika and Vrtti, so to speak) with Dandin's Kavyadarsa, coupled with their reputation for being wholly dependent upon Sanskrit models, must account for the neglect of a work that nevertheless seems to have an important bearing on the earliest strata of Alamkara literature. In a welcome new edition, P. S. Jaini has included, besides the 'PoranatTka' which he takes to be Safigharakkhita Mahasami's auto-commentary, a so-called 'Abhinava-Tkia'which an oral tradition assigns to fifteenth-century Burma.' As he explains, the Abhinava is no independent work, but rather a more detailed and more helpful version of the same commentary. Jaini's texts substantially reproduce the somewhat erratic Burmese edition of 1964, with collation of an even rarer one of 1928, but without reference to Fryer's readings based on two Burmese manuscripts, and with no information to offer on the Ce of 1910, which was 'with (purdna)sannaya' according to Helmer Smith in CPD, I, 60*. G. P. Malalasekera imputed a 'TikaS'to Vacissara (Pali literature of Ceylon, 1928, 204), which Smith listed (loc. cit.) as 'Pordnatfkak', along with a 'NavatTka';but this does not guarantee that any commentary exists independent of the Vrtti in two recensions, Pordna and Abhinava, that has been edited from Burmese sources. Fryer's obtuse statement (1875, p. 106) 'I have met with no commentaries on the work. There is, however, a gloss (tfka), which is said to be scarce' obscures the fact that his Burmese manuscripts evidently included the Abhinava. That was 'the Tikad' from which, in his study of Vuttodaya ('Pali studies. No. 2', JASB, 1877, 410), he cited six specimens of Yamaka as 'rhyming', for the Pordna has only two examples; and his paraphrase of the verses agrees with the wording of the Abhinava.2 Subodhalanikaraand Dandin verses, the inspired kavi achieves his poetic According to the Subodhalan4kira effect (vimhaya) through the art of rhetoric, and not from the study of
'Padmanabh S. Jaini (ed.): Subodhalankdra.Porana-tkad (Mahdsami-tTkd)by Sanfgharakkhita Mahasami. Abhinava-ttka (Nissaya) (anonymous). xix, 315 pp. Oxford: The Pali Text Society, 2000. Reference will be to Jaini's verse numbers, mainly one less than Fryer's (since Fryer's v. 20 has become v. 19ef). 2 His rendering 'minor poems' for v. 6 kabba- reflects a gloss beginning muttakakulakadi- in the Abhinava (p. 19, line 22). The sense 'minor and major poetry' is clearer in the Porana, because it is not averse to a clumsy inclusion of the word mahikabbam within the definition of kabba... pajjamayam ... mahd[vik]yaripam (p. 15, line 9' muttakadivdkya-... -vdkyasamuddyasampannam mahakabbam ca': the edition reads 'mahdkavyaripamn mahakabbamr',but cf. avayavasabhavehi... mahavakyasabhavenaca in the Abhinava). Bulletin of SOAS, 65, 2 (2002), 323-341. ? School of Oriental and African Studies. Printed in the United Kingdom.

This content downloaded from 163.1.62.81 on Fri, 15 Nov 2013 11:15:58 AM All use subject to JSTOR Terms and Conditions

324

J. C. WRIGHT

traditional forms of kabba and nataka (v. 6f.).3 Approved heroic and romantic imagery are usually illustrated in terms applicable to Buddhist eulogy (Jaini, p. xvii); and a definition of Kavya genres is smuggled in only within the Vrtti. (The term Vrtti will be used to refer to the consensus of both Tikas.) Dandin's layout, a compendium of Alamkara with preface on diction and appendix on faults, agrees with the Pali definition of composition in terms of sadda, attha, dosa, in that order. In his own definition, the sequence is reversed: he applies an image of the 'body' of poetry first to faults (i, 7: as blemishes on its body), and then to content and form (i, 10: as its ornamented body).4 His phrase istarthavyavacchinnapaddvalTsuggests a more plausible reading (with typical enjambment) and significance (reflecting the moral, as well as linguistic and aesthetic, implications of attha) for the Pali definition in v. 8ab than its Vrtti provides: bandho ca nama sadd' atth[a]sahita dosavajjitd 'words with good sense and without fault constitute a composition'. This in v. 10b, as continuing bandho in agrees well enough with the use of bhdratT v. 8a. The Vrtti, though it professes to read 'saddatthd sahitd', glosses sahita as atthasahita in order to read into it the terms of the much more sophisticated definition of bandha in v. 13f.5 Indeed, its 'saddattha sahita' is awkward, whether construed as words and meanings in combination or in concatenation. The Subodhalarikaraproceeds to discuss faults and their avoidance in vv. 17-115 (dosa with reference to obscure wording, as in Yamaka, and to cacophony, as in alliterative verse); merits in 116-163 (guna with reference to lucidity and euphony), and finally figurative and emotive diction in 164-367 (atthalarikara and rasabhava). Its structure more readily explains, than is explained by, Dandin's bracketing of Alamkara (more than twice as much text for roughly the same set of figures) in between a prefaced listing of guna and appended listing of dosa. While he incorporates into guna a treatment of Anuprasa (a classification as unknown to the Pali as to Bharata), dosa is appended after an equally over-blown analysis of Yamaka. Where the Pali declines to illustrate perplexing Yamakas (v. 33 n'ekantamadhurani ti, upekkhiyanti), Dandin decides, for no obvious reason, to postpone the entire topic (i, 61cd tat tu naikantamadhuramatah pascad vidhasyate). Curiously, his juxtaposition of Yamaka with faults seems to imply no disrespect; and his use of the term duskara to introduce a set of even more artificial constraints is markedly contrived, as compared with the distinction between approved (icchita) and unduly obscure (accantadukkara)Yamaka in Pali.6
3 Dandin's version, i, 103f., differs crucially. Success in stereotyped genres (kavyasampad) demands formidable qualifications: pUirvavdsanagundnubandhi pratibhdnamadbhutam,.rutam bahu nirmalam, yatna, and divine favour. Pali vimhayakararm param seems to aim at the ultimate Adbhuta Rasa; Dandin's pratibhanam adbhutamconveys nothing very specific. 4 The Vrtti reads an image into v. 9, inappropriately since the verses introduce a bandhasarTra comparison of bandha with mukha in v. 10f. and with kafiinain v. 14. andhandma tamsajjitataddvali (*.'ibddrthalarmkarasajjita'abdrthasamuddyah) b5 gunalanikarasahabhavam gatia' samyutt-. Hence, for sahita in v.8, Abhinava, p. 25, line 9 '?anurfLpatthena similarly, it seems, Pordna, p. 22, line 14f. 'ektbhhta vuttalakkhan [atthi] ye/hi, te pubbacariyehi sahita vuccanti'. Cf. Kulluika's imputation of the meaning 'and sisters' to sahitdh in Manu 9.212, when he might just as well attribute it to bhratarah. 6 See also J. C. Wright, BSOAS, 59/1, 1996, 48-54 and 59f., re Dandin's debt to Buddhist literature for a Prahelika and definitions of Kavya. Also id., 'Pali dipam attano and attadfpa' in R. Tsuchida and A. Wezler (ed.), Hardnandalahar-,volume in honour of Professor Minoru Hara on his seventieth birthday, Reinbek, 2000, 488-92: re his debt to Pali poetry for types of Upama, Dipaka, and Rtipaka. [Ibid., p. 490, 13-16, re Bhinnapada Slesa, read: Dandin emulates Pali su-d-fpa(*-dvipa) with pra-dosa 'evening/enmity' (*pra-dvesa: Prakrit pad(d)esa, pa(d)osa, Pali padosa): his sambadhnan rdiijii pradosah is rather 'the enmity pertaining to the king' than 'a wicked man associating with a king '].

This content downloaded from 163.1.62.81 on Fri, 15 Nov 2013 11:15:58 AM All use subject to JSTOR Terms and Conditions

THE PALI SUBODHALANKARA

AND DANDIN'S

KAVYADARSA

325

Historians of Pali literature assert that the Pali texts are adaptations of Kavyddar'a,7 as the Vrtti's reference to a 'Dandi' as one of its own sources appears to imply. It is never at all obvious that this is so. The starting point of the analysis of dosa and guna in Pali is padadosa 'lexical fault'. It is the final item in Dandin's appendix on dosa, and the more obvious hypothesis seems to be conversion from a Middle Indo-Aryan source into Sanskrit. The Pali classification of faults (faults of wording, construction, and import) includes v. 34 virodhi padarm'contradiction in terms', and only this matrix accounts for the neuter gender of Dandin's virodhi (iii, 126), kificit ... dejsddivirodhi (iii, 164). To explain how the fault can be used to good effect, Dandin offers the same vague reference to artistry (iii, 179); but his word order, inverting the padas of v. 68 (to accommodate kavikau'aldt for kavikosalld) and preferring utkramya dosaganandm to atikkamma, is far dosasanikhyam less natural. In Dandin, virodhais used to refer (iii, 179) to the Dosa-cum-Guna desddivirodhi. The words virodha (ii, 333) and virodhin .lesa (ii, 315) also designate two barely distinguishable forms of paradox Alamkara. The Pali distinguished rather more clearly, both in terminology and in its examples, between antithesis (v. 323 virodhitd) and a punning paradox (v. 300 virodhisilesa), and between these figures and the contradiction (v. 68 virodha) involved in virodhipadam. Dandin's definition of the latter (iii, 162-178) as an absolute fault in poetry kavyesu), exemplified with nonsense stanzas, is followed immediately (varjyd.h (iii, 179-185) by specimens of its effective use in portents. In Pali, the approved omen verses were set in a mitigating context of Dosapariharavabodha(v. 70ff.) which is wanting in Dandin. The list of types of solecism was given separately in v. 34, and there we are spared any irrelevant nonsense: the verse has only a cross-reference (uddharanato phutam), which the Vrtti implements with examples taken from the approved verses. The Karika, v. 146, provides a Middle Indo-Aryan basis for Dandin's Kanti Guna: lokiyatthanatikkanta kanta sabbajanenapi, kanti ndma 'tivuttassa vutta sa pariharato. Not overstepping the bounds of reality and beloved by all: Kanti is so-called because it avoids the fault of over-statement. A Prakrit pun on kram- and kam- is explicit, and the form kanti confirms that 'charm' is primarily intended. Indeed the example that is given of the fault of crass overstatement (v. 56 ativuttam, explained as lokiyattham atikkantam) is singularly charmless: atisambadhamcIkasametissa thanajambhane.The illustration of Kanti in v. 95, a figurative mitigation of the same overstatement 'no room for it in the cosmos', makes an indirect reference to kanti 'moonlight' in the Riipaka image that equates the Buddha's fame with moonbeams: Munindacanda-sambhu-ta-yasorastmar-cinam, sakalo py ayam dkaso ndvakaso vijambhane. Dandin's definition also links the two etymologies (i, 85 sarvajagatkdntam laukikarthanatikramat);but since the pun and the 'moonlight' figure are both lost, one has no idea why he chooses to equate matter-of-factness with kanti 'charm'. He assigns i, 91, his expansion of v. 56cd, crass atyukti, to the Gaudas, and (in lieu of the Buddhist v. 95) foists upon the Vaidarbhas a less fulsome, but no more charming, compliment (i, 87 stanayor jrmbhamanayoh,
7

1960, vi. Similarly, H. N. Chatterjee, Comparativestudies in Pili and Sanskrit alanmkaras,

This content downloaded from 163.1.62.81 on Fri, 15 Nov 2013 11:15:58 AM All use subject to JSTOR Terms and Conditions

326

J. C. WRIGHT

avaka-o na parydptas tava bdhulatantare).The implication is possibly that he recognizes Magadhi and Maharashtri as the source of inspiration for notions of ugliness and beauty in the Pali, but avoids terms that specify a Prakrit linguistic medium.8 He misses the point that it is the Rupaka, identifying the Buddha's fame with moonbeams, that makes it possible to visualize its infinite extent. In the Pali, the separate enumeration of Kanti really serves only to boost the number of Gunas to a round number, for this sort of figurative measurability could well have been included among its specimen instances of Samadhi. These go beyond personification (which alone emerges from Dandin i, 93-100), by including four forms of imputed concretization: ruipa, rasa, drava, kathina. Like Kanti, Samadhi is defined with reference to a combination of metaphor with reality: this is implied by its prefix sam- (v. 150bc lokasTmanurodhato samma adhTyate). Dandin loses all cohesion. Separating figurative Samddhi from factual Arthavyakti, he juxtaposes Samadhi with Kanti, although their similarity is no longer apparent. His lotus metaphors, simple or complex (i, 94a kumuddni 96 padmdni ... vamantiva), lack the linking factor, the consistent use nimTlanti; of comparisons with moonlight as a link with reality. The Pali combined its lotus metaphor (v. 153 viniddd kumudint) with a full-moon Rilpaka; and to demonstrate that Samadhi can render even vulgarity highly attractive (v. 159): vaman'uggilanadyetamrgunavuttyapariccutaim atisundaram;aiman tu kimam vindati gammatamr.9 the example had toenails that seem to swallow the moonlight (candakanti) by spewing radiance (kanti). Dandin's resort to mere metaphor reflects a misunderstanding of the The Vrtti, followed by the editions, claims to enjambed reading to read etam guna' and identify guna arbitrarily as an alien 'appadhaina' etam.guna. the topic throughout Samadhi is the transfer of a natural property, although property 'mukhyavatthuno pasiddhaguno' (p. 145, line 6f.). As in the case of v. 8 sadd' atthasahiti (above), the Abhinava conflates the two readings. It construes etam implausibly with vaman'uggilanddi(hitherto unmentioned); but construed with guna' in its 'idani imesu eva it has also a gloss on etamrn [p]dnidhammidisu' [Ed. vani': but Samadhi comprises paninam dhammo,etc.: p. 146ff.]. In deference to the caesura, etam has been glossed separately as an adverb ('iddni'); but, although the syntax is wrong, it is rightly explained as a The term etamguna is reference to the natural property ('painidhammaddsu'). indeed essential, as a reference back to v. 150 afifiadhammo(tato ai-iiattha ... adh-yate), the property to be transferred. Unlike the Vrtti, Dandin excises the seemingly superfluous etam, takes gunavutti to be *gaunt vrttih, and replaces pariccu- with a purely rhetorical notion of 'descent' (... vantadigaunavrttivyavigahate). pairayam ... gramyakaksram Importantly, however, apariccuta stressed the advantage of 'retaining' a

8 Non-Buddhistic illustrations in the Karika can be reminiscent of Maharashtri verse (n. 18, below), and later taste might well condemn as charmless and wilfully obscure the archaic Magadhi ballads, specimens of which survive in the Jain Manipaticarita (Ludwig Alsdorf, 'Zwei Proben Ernst Waldschmidt der Volksdichtung aus dem alten Magadha' in Beitrage zur Indienfjorschung, zum 80. Geburtstaggewidmet, 1977, 17ff.; Kl. Schr., 1998, 777ff.). 9 Ed. uggira? (as in v. 162); Fryer and Vrtti (1928 ed.) uggila?. Ed. etam guna?. Fryer ?vutyap, for *vutt'ap?as a compound ('tato apariccutam'); Ed. has ?vuttydp?, Abhinava possibly ?vuttyap?, sandhi "vuttya 'p, after Dhvanyaloka 1.17. Read *gamya? (<gramya, beside with external gamma <gamya)? Fryer and 1928 ed. gamma?, Ed. gamma.

This content downloaded from 163.1.62.81 on Fri, 15 Nov 2013 11:15:58 AM All use subject to JSTOR Terms and Conditions

THE PALI

SUBODHALANVKA4RA

AND DANDIN'S

KAVYA4DARSA

327

literal diffusion of radiance, as justification for the daring conceit candakanti pivanti 'the toenails absorb the moonbeams'. The compounding of etam- is confirmed by v. 161: acittakattukam rucyam icc e[t]amgunakammakam, sacittakattukam p' etamrgunakammamn yad' uttamam. [Ed. p' etamrn ?kammatam(for ?kammdtam?, Fryer, etc. ?kammakam); iccevam.; yad, Vrtti yadi]. 'Thus (the spewing verbs) with etamguna (spewable) guna?; are pleasing when with inanimate subject (viz., the toenails that spew object radiance), but they are at their best when even with animate subject (and figurative object: viz., the Buddha who spews forth Rasa in v. 162 uggiranto va sa sneharasarm Jinavarojane, bhasanto madhuramdhammarn...).10 The Vrtti in pada c as resuming etam vaman' (although it would has to treat its etamrn as a conjunction then properly belong in the first hemistich) and its iccevamrn the stanza with v. 159f. a caesura follows the predicate Since, however, linking rucyam and the pa-da-initialicc is stressed in any case, such an evam would be redundant and can have been a device to get rid of one of the apparently unaccountable three successive occurrences of etamrn before guna?. The proximity of ativutta (Dandin's atyukti) in v. 146, sabhava-vahka-vutti in v. 165, and atisayavutti in v. 173 implies here *etarngunokti, rather than a *gunavrtti which indeed invites alteration to Dandin's gaunavrtti. Those with *-ukti are parsed as genitive Tatpurusas; and the Abhinava's gloss (guno 'appadhdno' vutti) makes sense only on the assumption that a tassa before vutti has been lost in deference to the implied *gaunt vrttih of Dandin and the Pordna (guna 'padhanetard' vutti 'payogarupa ')." Transitive sub-glosses, amukhyavisaye payuj- and visayantarapariggaha-, seem to favour *gunokti rather more than *gunavrtti. As against Dandin's gaunavrtti, the Pali form -gunavutti is confirmed by gunavrtti in Dhvanyaloka. The Buddhist 'supreme' form of metaphor (v. 161f. uttamarn),ignored by Dandin, must have inspired Anandavardhana's adaptation of v. 159 -gunavuttyapariccutamrn into a definition of the limitations of metaphor (DhA 1.17).12 His tatra na skhaladgatih 'incurring no loss thereby' recalls the Pali 'tato apariccutam' (skhal- = cyu- 'to fall, be lost '): mukhyam vrttirnparityajya gunavrttya.arthadarsanam yad uddiSyaphalam, tatra Babdonaiva skhaladgatih. 'When to achieve a purpose a word forsakes its primary application and conveys meaning by metaphorical application, it does not lose (its primary sense) in the process.' The prosody, with enjambed hemistichs (yad arthadarsanam ... tatra) as in v. 159, baffled its own Vrtti (into construing yat, as yadi, separately from arthadarsanam and tatra, and tatra iabdo naiva skhaladgatih apparently as 'then its use would be faulty diction, which is not the case': 'tada tasya [sabdasya] prayoge dustataiva sydt, na caivam') and the Locana (into dropping the enjambment altogether, construing yat phalam with tatra,
10 ... kam na janam. Ed. va sasneha-, Vrtti va so sneha-, Fryer va senaha- (sic). Are sanmppnaye uggir- and sa features of the quotation, as opposed to uggil- (n. 9) and so elsewhere? ' Analysis as *gunena vartate 'functions indirectly' and *gunena vartayati seems to underlie the instrumental Tatpurusa glossing of gunavrttya ('gunasamuddyavrtteh ') in DhA Locana 1.1, either as 'gunadvdrena vartanam' or as 'tair [samipyadibhir dharmaih] upayair [Kaumudf comm.: nimittabhitaih] vrttiryasya/yatra sah'. This passage seems to imply that the concept of prayojana in metaphor rests on the alleged instrumental in gunavrtti. 12 This supports a claim that the first twenty verses of DhA Karika are an adaptation of Buddhist ideas (cf. S. Paranavitana, 'The Dhvanyaloka in fifteenth-century Ceylon', JAOS, 1971, 131ff.): at least an adaptation of Buddhist wording is evident. There must be a suspicion that the name 'Abhinavatfka'was coined in misguided homage to Abhinavagupta.

This content downloaded from 163.1.62.81 on Fri, 15 Nov 2013 11:15:58 AM All use subject to JSTOR Terms and Conditions

328

J. C.

WRIGHT

and "abdonaiva skhaladgatih as 'an unimpeded verbal operation, hence not Laksana but Dhvani'). The damage done to the argument is made explicit in the Locana (cf. the comments of Ingalls et al., tr., 1990, 188). The 'supreme' example of metaphor in v. 162 (*udgirann iva rasamjane, bhdsan dharmam), the truncated reading gunakammaka (p. 152, line 23 'appadhanakammavanta '), and the ambiguity of Pali jane (where the Vrtti, evidently an accusative plural 'sakalasattanikaye '), presumably gave wrongly, implies rise to the terminology whereby verbs of speaking, with dharma as direct object, can be gunakarmaka 'with indirect object'. This yields a definition of the indirect object in the context of the Akathita Karaka (Panini 1.4.51). Its source in poetics tends to confirm that in the relevant Slokavdrttika verse the words gunena ... acaritam kavind can be quite naturally construed as 'used by an author as indirect object', i.e., a transitive pendant to *gunena vartate 'functions indirectly' (n. 11, above):13 # upayoganimittamapi~rvavidhau, duhiyacirudhiprachibhiksicinidm tad akTrtitam ca sacate # acaritam kavind. bruvisisigunena yat 'The (factor) "which has not been mentioned" is that which, used by an author as the indirect object of (the verbs) duh-, yac-, rudh-,prach-, bhiks-, ciand bru-, sas-, conforms itself, failing applicability of previous rules, to the would direct object.' Dependence of gunena ... acaritam on confirm an early date for the basic Pali material. etam.gunavuttiand Dandin Subodhalafikara-t-ika The Pali Vrtti helps to explain Dandin's arbitrary exclusion of lyric poetry anuktah) and drama (muktakamkulakam ... iti tadriah sargabandhaiigarupatvad in order to but It uses similar impart the informawording, (anyatra vistarah). tion about Kavya in general that the verses pointedly withheld: p. 19, line 22ff. kabbam nama muttakakulakddivdkyavasena ca avayava-sabhavehi ... -dassitamahdvakya-sabhavena ca titthati and Bhdratadindtyasattha-... bhavam.'4The idea of defining full-scale Kavya as a concatenation of sentences or stanzas is more satisfying than its reversal by Dandin (i, 13), who claims that the definition of individual stanzas is implicit in that of Mahakavya. Again Dandin's readings can readily be seen as an inevitable result of the adaptation from the Pali into Sanskrit. The Vrtti's definition of Mahakavya (p. 15ff.), compared with which Dandin's appears garbled, leads up to its explanation in verse of the term 'canto': nativitthinnasaggehipiyavuttasusandhibhi sabbatthabhinnasagganteh'upetamrn 'comprising moderate-sized cantos, i.e., separate individual episodes (*sarganta, like suitranta, vrttanta?) of the whole narrative (*sarvartha), suitably
13 As suggested in BSOAS, 1997, 151. The stumbling-block is the doggerel versification and split compound gunena beside oddsigunena. The technical term upayoganimitta must refer 'cifmirn as the proper 'occasion for applying' the term karman; sacate cannot be to the direct object construed as sambadhyate with instrumental gunena; nor can acaritam kavind mean uktam sutrakarena (Kadika)or'enjoined' (MW). Thus 1.4.51 need not be a pointless interpolation (Joshi and Roodbergen, Karakahnika, 1975, 174ff.). The KaLikadoes not equate gunena with pradhanena (ibid., 183): it says 'the guna of bri- and .ds- is an ancillary factor; the pradhana object is dharma, etc.; when (the former is) connected with the latter, it is akTrtita'. 14 Ed. Bharato, 1928 ed. Bharat?. The Abhinava reads -natasattha- (for *natyasattha-?). 15 Ed. sabbattha bhinna?, Abhinava 'sabbasmim paricchede bhinnasaggd pariyosana yesam (despite having glossed sagga as 'pariccheda'); Porana ignores this (or derives from it the Ed. ?rasussavamr (p. 15), ?udayalakkhanam(p. 20), for ?ratussaexplanation (p. 15), must be oversights. (p. 20), ai"iamafiiasanigaha). vamr ?udayavannanam

This content downloaded from 163.1.62.81 on Fri, 15 Nov 2013 11:15:58 AM All use subject to JSTOR Terms and Conditions

THE PALI SUBODHALA NKARA

AND DANDIN'S

KAVYADARSA

329

linked with pleasing metres'. Such casuistry does not affect the probability that saggabandha would originally amount to 'sustained composition' as opposed to muttakadi. The Vrtti specifies Indavajiradi metres for the coda, which is valid for Anustubh texts like Ramayana (and Subodhadlaikara itself, which ends sections with Upajati and Vasantatilaka), not for the Kavyas of AMvaghosa. Dandin's fragmented vague reference to euphony and cohesion (sravyavrttaih susandhibhih) will reflect classical Kavya's generalization of piyavutta Indavajiradithroughout the canto. His equally fragmented sarvatra bhinnavrttantaih,which leaves commentators undecided between 'episodes' and (improbably) 'metres' for vrttanta, does no justice to the *sarvdrtha-that neatly complements *ndtivistfrrnasarga, or to *sarganta as a designation of sagga once it had come to mean pariccheda. His prosy ?ratotsavaih ... ?udayavarnanaih,vipralambhairvivahaidca kumdrodayavarnanaih ... alamkrtam lacks the elegance of "rat'ussavam ?udayavannanam, vippalambhavivdhehikumarodayavaddhihi ... alanikatam (Abhinava: ussavo yattha'). To maintain 'uyydnasalilakTld-madhupanaratikT~l-sanikhato prosody, his Sanskrit has to resort to rata for abstract rati and an awkward repetition of udayavarnanaihin lieu of *udayavrddhibhyam. When the Abhinava, deeming the Buddhist example in v. 326 to be misleading, chooses to reinstate a secular specimen of Parivrtti Alamkara from its source (ten' eva Dandiyam), the participial construction of Dandin's Sanskrit dadatd bhujenatava ... hrtam tesadm (ii, 356 iastrapraharam yasah) seems pedestrian beside the elegant gerund of the Pali (-ppaharanam datvd); and it is Sanskrit prosody that would compel him to substitute circrjitam hrtam tesam yasah for the more graphic ciramcit' ibhato tesam yaso. The so-called Pordna-tfka is briefer, failing to gloss the Vrtti's specimen verses and illustrating, for example, only the first two of the sixteen Pahelikas. Jaini considers (p. xiii) that it is an auto-commentary. Whereas, in the definition of' poetry' in the Abhinava, the word viracita refers us to textbooks of prosody it is used in the Pordna to refer, (p. 28 chandasi nidit.thavuttavisesaviracitam), in the apologetic manner of Dandin, to the Karikakara's own treatise (p. 24 amhehi yeva viracitaVuttodaydkhyechandasi nirupita ... ten' ev' ettha tesam adassanarm).Such manipulation of the wording implies embroidery in the Pordna, a Vrttikara putting words into the mouth of the Karikakara, for it is clear that the Pordna cannot aspire to the status of auto-commentary. It fails to recognize the distich syntax of v. 338f., while the Abhinava retains a correct gloss, despite superficial corruption: kavind 'nibaddha ti sambandho' and nibandhd (read nibaddhd) 'vuttagunopeta-kavind bandhita'. For v. 156 Sabbatthasiddha mahaguna, its gloss '?siddhassa ... mahaguna', instead of the c.lakaputapeyyd required vocative (Abhinava:'he Tathagata tava ... gund '), is an elementary blunder. A Vrtti is in any case unlikely to be an auto-commentary, when it defines genres of poetry and drama which the Karika considers to be negligible as models, and exemplifies varieties of Yamaka which the Karika expressly leaves to the imagination (v. 31) or omits as tedious (v. 33). An instance where both Tikais have lost the correct understanding of the Pali that Dandin attests is v. 189: yadi kirmci bhave 'mbhojarn locana[b] bhamuvibbhamanm, dharetummukhasobham tam tavet' esa 'bbhutopama. [Ed. locanam taveti abbhut?,Dandin ity asdv adbhut?,Abhinava 'iti tdisT';1928 ed. tava te abbhut? reflects rather Fryer's reading.] Here Fryer had rightly followed the Abhinava reading locana-bbhamu-(at least to the extent of printing

This content downloaded from 163.1.62.81 on Fri, 15 Nov 2013 11:15:58 AM All use subject to JSTOR Terms and Conditions

330

J. C. WRIGHT

locanambhamu- as a compound): 'if any lotus could flirt with its eyes and brows, it would achieve the beauty of your face'. Dandin's pedestrian Sanskrit # subhru vibhrantalocanamwith its verseversion yadi kimrcidbhavet padmamn filling vocative demonstrates the superiority of the Pali text; but dharetum mukhasobhamis rightly construed by Dandin as having dharetu in a conditional sense (ii, 24 mukhairiyam dhattaim),whereas the Abhinava and Pordna resort to 'dharetum samattho hoti' and nisijjhate'. date than Abhinava. In its definition of The Pordna shows signs of later 'dharetum. Mahakavya saggabandho mahdkabbam# uccate tv assa lakkhanam, # panamo vatthuniddeso Aassa pi ca tammukharn. the Abhinava sensibly explains saggabandha as a plain Tatpurusa (p. 20, line 15: paricchedavisesehi bandho), in keeping with the Karika's use of bandha. The Porina takes it to be an adjectival Upapada compound (p. 15, line 13: paricchedavisesehi bandhyate karTyati). Yet for nibandha (v. 9), which would appropriately be defined thus as a Krt derivative, it produces muttakadthi... ... nirantaram bandho, which must be a truncated version of the muttakdadThi katva bandho of the Abhinava. Its gloss '... saggabandho ti pavuccati. tassa tu lakkhanam ...' amounts to a mistranslation of uccate tv assa, implying the influence of Dandin's ambivalent reading (i, 14)

# ucyate tasya laksanam sargabandho mahakaivyam just as its reading niddeso vi (instead of astsd pi ca) as the last item in the definition of' Mukha' agrees with the series of alternatives, with a prosodically inferior weak caesura, to which Dandin had to resort in order to accommodate Sanskrit Ldis:
namaskriyd vastunirde'o vdpi tanmukham. afr vi more in keeping with Abhinava (p. 20) has assa ... tu vuccate and on the Karika-text. The Pordna has no direct gloss (n. 15, above), 'saggantehi astsanam, no doubt because Dandin had dropped ?saggantehi in favour of ?vrttantaih. in v. 2: Both Tikas give a complicated analysis of the term suddhamagadhika santi santo puratand-, Ramasammadyalarikara suddhamagadhikana te. tathdpi tu valanljenti The Pordna sees the whole word as a Karmadharaya compound meaning 'latter-day monks and novices' (yatipotd), and tries to justify this by explaining first mdgadhika as referring to those who speak or learn Magadhi language, and then suddha as pure (visuddha) or purist, i.e. uninfluenced by (asammissa va aparicitattd) Sanskrit, etc. These people, it holds, cannot appreciate the Alamkaras of Ramasamman, etc., once beautiful, but now flawed (malaggahita) because they are in Sanskrit and Prakrit (sakkatadibhdsitakalusiya-). The Abhinava seems to be on firmer ground. It uses the idea of epigonic unfamiliarity with these languages (without the complication of latter-day Sanskrit hybridization to which the Pormnarefers) to justify a gloss kevalamagadhika 'monoglot Pali-speakers'. This reasonable notion is followed by another that starts with the full explanation of mdgadhikaas 'those who speak or learn Magadhi', as found in the Porana, but with possibly the more idiomatic reading te [saddd] etesam atthi (versus santi in the Porcna); then it notes a choice between Taddhita derivation from suddhamagadhaand a Karmadharaya suddha-magadhika.Either way, they would be unable to appreciate the excellent

This content downloaded from 163.1.62.81 on Fri, 15 Nov 2013 11:15:58 AM All use subject to JSTOR Terms and Conditions

THE PALI SUBODHALANKARA

AND DANDIN'S KA4VYA-DARSA

331

(vififilhi pasatthatta sobhand), but linguistically unrefined (bhdsdyapuratanatta aparisuddha)material of Ramasamman, etc. The Pordna agrees verbatim only on the gloss assigned to mdgadhika. It may be revealing its youth, not only by adding the notion of hybrid Sanskritization, but also by dropping the gloss kevalamagadhika, possibly on the basis that twelfth-century monks would not be monoglot Pali-speakers. It elaborates, however, upon the Karmadharaya explanation, leaving only a trace of the alternative Taddhita explanation (n. 33, below). The Vrtti uses the Karika-'sexplanation of the term Alamkarasastra (v. 12 gantho ... alanikarappakasako)in order to claim that Ramasammadyalanikara santo puratand could designate the treatises of pubbacariya, as well as their 'fine old Alamkaras'. Its postulation of a Sastra called *Ramasamma-Alanikdra is echoed in the equally elliptically named Rdmasarma-Acyutottara, to which Bhamaha ascribes a definition of obscure Yamaka (ii, 19): yamakavyapadefinT, nadndhdtvarthagambhIra prahelika sa hy uditd Rdmadarmacyutottare. This is close enough to the bracketing together of Pahelika and kilitthapada Yamaka in Pali to make it doubtful whether Bhamaha knew anything more of Ramasamman than is preserved in the Karika. Like the Pordna which reads a negative judgement on Ramasamman into v. 2 (despite its tathapi), Bhamaha's interpreters implausibly understand uditai to imply 'the obscure Yamakas perpetrated in Ramasarman's Acyutottara',16 and an attribution Ramasarmanah has been appended to Bhamaha's specimen of faulty Upama (ii, 58). Unlike the Pordna, the Karika does not specify Ramasamman's language, and the Abhinava mentions Sanskrit only as something that is excluded by the term suddhamdgadhika.It is unlikely that the Karika could have been referring to the canons of classical Sanskrit poetry as obsolete (puratand), or that the Abhinava would describe Sanskrit, which the Vrtti introduces at p. 28ff. as devatdbhdsa,as unrefined (aparisuddha).They could very well regard the canons of Jain Maharashtri literature as santo purdtand, sound but outmoded; but there must be a suspicion that it would be its forerunner, Magadhi secular poetry and 'Ardhamagadhi', that the term suddhamagadhikd (as a Taddhita had originally served to disown. derivative from *Auddhamagadha) There is thus no reason to doubt the attribution of the Pordna to a twelfthcentury Mahasami. The attribution of the Karika verses to the same period in ... paricchedo) the editions (Sanigharakkhita-mahiiasaimi-viracite Subodhalnaikdre to the Karika text. Such a could rest on a transfer of the title from the Vrtti transfer is implied by the 1928 edition's final colophon Alanikdrappakaranam samattamn (Jaini, p. 304) and by the Tika colophons: 'Subodhailaikaranissaye Subodhalari... paricchedo' in the Abhinava, and 'Mahassdminimikiyamrn kiratTkayam ... paricchedo' in the fourth chapter of the Pordna (its third chapter has the conflated reading Mahasamindmikiyam 'Subodhalanikare samattam, the 1964 Beside Alanikarappakaranamr Subodhalanikdrat-kdyamn'). version pakaranam sampunnam, Subodhalarikaratika sattkam looks like a 'restoration'. samatta (p. 305)Subodhilarikdram.
16 The title, ostensibly Vaisnava, makes sense as 'above Indraloka, supreme' (with the Middle Indo-Aryan connotation of Acyuta), and as an allusion to the Pali discussion of the supreme form of the ultimate of the most delightful type (v. 159 (v. 161 uttamamr) ?apariccutarnatisundaramrn) guna Samadhi, a topic which, as has been seen, can have influenced the formulation of anukta Dhivani as well as akathita Karaka theory. Thus the title Acyutottara implies at least a belief that the Pali owed its stance on guna to Ramasamman.

This content downloaded from 163.1.62.81 on Fri, 15 Nov 2013 11:15:58 AM All use subject to JSTOR Terms and Conditions

332

J. C. WRIGHT

The yamakavannana section and Dandin The Karika text, unwilling to illustrate faults with verses on Buddhist themes, retained traditional examples, such as v. 45f. kaftie kdmayamanam mam na kamayasi. kim nv idam?'7and yd bhavato ... for inelegance and v. 56 atisambadham dkisam etissa thanajambhanefor overstatement. For inelegance, even the counter-example, inelegance mitigated by subtle innuendo, is not Buddhist but recalls the canons of Maharashtri verse: v. 92 dunoti Kcmacandalo so mam ... sukhf pi kim upekkhase.'8 It gave some simple Buddhist instances of approved lucid Yamaka (v. 74 yamake ... icchite) in vv. 27-30, but only a secular example of tolerated Pahelika in v. 73.19 This could have some bearing on the Vrtti's non-committal attitude towards Pahelika (pubbacariyehi niddittha, versus uddigyategatih in Dandin).20 An explanation for the equivocal status that Dandin bestows upon Yamaka and Prahelika, placing them in between alamkdra and dosa, emerges only from the Pali, where Yamaka is treated (within its basic overall scheme of faults, mitigations, and merits) under dosa as properly exhibiting the merit of 'clarity', but liable to the fault of 'obscurity', of both form and meaning. The Karika illustrates the distinction with approved simple three-syllable 'unsplit geminations' (e.g. kusaldkusald sabbe), but with only a warning against accantadukkara (' unduly obscure') Yamaka.21 The Karika text defines Yamaka 'gemination' in v. 26: dvuttanekavannajamn,22 abyapetabyapetanf-am yamakam taft ca pdddnam ddimajjhantagocaram.
17 The Vrtti takes kim nv idam to be so lacking in poetic beguilement as to be boorish (milakkhukanamin the Abhinava, coupled with an explicit denial that there is any impropriety in 'girl, you do not love me ...'). It is grimya in the sense of lacking nagarika sophistication, as is shown by the counter-example v. 92. The key phrase is virtually lost in the Sanskrit, and Dandin has rightly (cf. n. 18) been credited with a merely prudish objection to i, 63 kanye kamayamanam mdm tvam na kdmayase katham [var. na tvam]. 18 The Vrtti takes boorishness to be avoided by sophisticated innuendo: sukh- conveys 'whose love is requited' and hence 'unfaithful, neglectful'. Dandin's inferior version in i, 64 (kdmam Kandarpacdnddlo...), allowing kdm- to appear only in the adverb kamam, appears to make mealymouthed preciosity a prerequisite of rasavaha poetry. In the Pordna, for akando read pracando (canddlo pracando va asahyopatap'vahattd), rather than ananigo as Jaini tentatively suggests: the Abhinava has canddlena akdrunena. The Pordna seems to take kdmam, kandappa, avaha from Dandin's version and to work them all into its commentary. kam dlinigati nu no 'Whom, being embraced by his beloved, does 19piya sukh' alirigitarm happiness not embrace?' The Vrtti specifies the Prakritism piyd for piydya. The conundrum will entail a more humorous sense, with nominative piya and compounded sukhdlitigitam. 20 Failing any definition in the Pali, Dandin uses for Prahelikf (iii, 97 kridagosthTvinodesu $ tajj-iairdkirnamantrane)the Pali mitigation for the fault of ambiguity: v. 111 kTla-adi-.He seems to have read *krTdad-dau into the Abhinava'sgloss' sambddhasammantrane k-ldsambddhasammantana-adi-', thus encouraging the Pordna, with 'adisaddena aikinnasammantandadim sariganhdti', to give the latter an implausible Dvandva interpretation. ' The Porana, with 'sukhena karTyanti payujjanti' for sukara, implies that dukkara might be a worthy challenge. Bhamaha views duskara rhyme-schemes as potentially desirable (ii, 16 duskaram sadhu tadriam), and he castigates gambhTra,vyikhydgamya Yamaka as a quite separate issue. The dpad in MBh.); and the gloss 'dussadhiya' Kfriki's accantadukkarais 'unduly hard' (like duskara. in the Abhinava could imply 'hard to construe'. Dandin exemplifies only simple initial rhyme as sukara, an idea that can be derived from the Vrtti's presentation. His duskara end-rhyme iii, 41 can indeed be distinguished from his sukara pada-initial rhyme iii, 29 by its formal ambiguity, which has led to severe textual corruption (separation of saccarita from pramatta in the editions yields no good sense: following the Tibetan source in all other respects, we should rather read tava priya, saccaritapramatta,yd / vibhiusanamr dharyam iharm.umat tayd, ... na me phalamn kimrncana kantimattayd). 22 Jaini's edition reads abyapetarm and byapetai and iam separately (not as a simple misprint, but somehow in collusion with a wrong reading abyapeta'mbyapetai ti in the Porana): but the parsing of it as 'samaharadvando' in the Abhinava;the readings abyapeteccadi and abyapetabyapetan ti elsewhere in the Porana; and Dandin's reading avyapetavyapetatma, all confirm the compound. Fryer's avyapetam vyapetaf c' aim' (with an awkward sandhi at the caesura) will reflect the analysis with thrice ca in the Abhinava.

This content downloaded from 163.1.62.81 on Fri, 15 Nov 2013 11:15:58 AM All use subject to JSTOR Terms and Conditions

THE PALI SUBODHALAriKARA AND DANDIN'S KAVYADARSA

333

According to the Vrtti it is threefold: 'unsplit, split, and another', viz. another avyapetavyapeta type which is both unsplit and split (ubhayamissakattd). Dandin's term avyapetavyapetatma is simpler, but its application is more contrived: he uses it twice, first to convey the prior two options (iii, 1) and then to convey the third (iii, 33 avyapetavyapetatma vikalpo 'pi). 'Other' is exemplified (p. 48f.) as multi-pada split recurrence of unsplit geminations, rhyming ABBA: = samalam sam ramnjaye, n alamkattu.m sucira.m-suc[j] alam samalambhi yo. suciram suci-ramgam tam sam [Ed. -suci]. While the Vrtti has here, as for simple unsplit and split, a trisyllabic gemination, Dandin illustrates various possible rhyme-schemes with disyllables (kalam kalam ..., etc.).23 Thus any notion of rhyme is relegated in Pali, along with Mahakavya and Natya, to the Vrtti. There is no guarantee that the Karika intended 'split gemination' to be a reference to multi-pada rhyming (as Dandin's examples imply), given that one single occurrence of sujandsujandsabbe within a pada constitutes an unsplit Yamaka, and split Yamaka within the pada is a concomitant feature in all Sanskrit sources, even Bharata (Malayamaka: Halt balt half mad/). Its term vyapeta, even when glossed 'vannantarehi vyavahita' (Porana: 'vannantaravyavahita'), conveys the idea of pada-internal repetition better than that of rhyme. In the Vrtti's example of dukkara vyapeta (p. 53: the compounds are misprinted on p. 51) na bhasurd te pi sura vibhisita # tatha 'surd bhuri surd-pardjita, sabhasu raja pi tatha surdjito# yatha surdjanti surd-vinissata. the term dukkara imposes decorative AAAA rhyming recurrence upon the basic pada-medial split Yamaka surd te pi surd. Otherwise the term pddamajjha vyapeta would be ambiguous, as between pada-internal repetition of 'surd' or multi-pada repetition of 'surd ... surd'. Mere multi-pada occurrence of different split geminates would, on the analogy of unsplit Yamaka, remain within the scope of sukara. The Vrtti's one example of simple split gemination (p. 48f.) piyena vacasa sabbe #pi yena 'ppiyabhdnino, paddnate jino kasi #so dhammo hantu vo 'ppiyam. is notably different from the Sanskrit examples (Dandin's madhurena- rhyming AAAA with madhurena, etc.). The Abhinava's defence of the caesura (pi functions as api: 'all, even those who spoke ill') is undermined by the enclitic pi in its own gloss 'appiyabhdninopi sabbe ', which does not rule out the more obvious sense 'all his opponents', balancing 'your opponent' in the main clause. Since in any case pi can only be enclitic (CPD, I, 289ff.) and the rules for Yati (v. 50ff.) require a combined vowel to follow the caesura in external sandhi, the line remains faulty, unless treated as a hemistich without regular
caesura.24

Such a hemistich-internal rhyme would be a close approximation in Aksaracchandas to the internal rhyming at a variable caesura that occurs in Ganacchandas (cf. Sbhnen, op. cit., p. 501). A Giti verse that exhibits pada-final assonance in Bharata (xvi, 65):
23 The terms 'gemination' and 'rhyme' will be used as in Renate Sihnen[-Thieme], 'On the concept and presentation of yamaka in early Indian poetic theory', BSOAS, 58, 1995, 496. 24A sixth-syllable caesura recurs in v. 247ab na santdpapaham, n' ev' i-cchitadam .... (The piyena ... verse is misprinted and wrongly labelled 'abyapeta' in both Tikas.)

This content downloaded from 163.1.62.81 on Fri, 15 Nov 2013 11:15:58 AM All use subject to JSTOR Terms and Conditions

334

J. C. WRIGHT

lokandmprabhavisnur 0 daityendragadainipa-tanasahisnuh, # bhagavan asuravaramathana[krd] Visnuh. jayati suradaityajisnur is unmetrical] has been treated as spurious, since most manu[GOS ed. ?karT scripts present an Aksaracchandas verse with four -mandalam rhymes.25The feebleness of the latter and of Dandin's punning madhur ena-/madhurena, coupled with the contrived nature of the Pali Anustubh specimen, suggest that all are attempts to adapt Gana rhyme to Aksara metre. Since the Pali seems thus deliberately to have avoided treating either unsplit or split gemination as a matter of separate rhyming padas, it is strange that the third category amina, unexplained in the Karika text, was illustrated in the Vrtti by combining unsplit geminations in an ABBA rhyme-scheme (samalam sam alam- ...).26 It then had to illustrate the 'obscure' varieties (v. 32) with distinctive AAAA schemes. Thereupon Dandin includes AAAA schemes within his basic definitions, and goes on to compile catalogues, first of split and unsplit-cum-split pada-initial rhyme-schemes (iii, 20-29 and 34-36), and then of duskara types (39-50) which now comprise AAAA schemes other than simple pada-initial rhymes. Bhamaha reverts to individual specimens and (although nothing in his definitions prescribes this) goes beyond Dandin by restricting his examples to four-pada schemes. Bhamaha's Yamaka has thus lost all trace of the simple sujandsujandsabbe of v. 27; but his LatilyaAnuprasa (drstim drstisukham)restores a niche for it, certainly for such as v. 29c vandand vandand-mdna-bhijaneand p. 50, line 1 manohara hara.27 The gulf that separates the Vrtti's complex ai-fla from sujandsujandand pi yena ... pi yena is explicable on the basis that the term, necessarily implying multi-pada rhyme, is intrusive in the Karika. Dandin has only a verse-filling -atmd, which could imply Pali *abyapetabyapetattamif, as seems possible, the phrase afitiam abyapetabyapetam originally represented a piece of casuistry on the part of the Vrtti. That af-liais intrusive in the Karika's inscrutable expression abyapetabyapetaifiiam is shown by the fact that there it is inherently ambiguous. It clashes with more natural uses of the same word in v. 31, where a~fiianiyamakdni means only 'other than verse- and pada-initial avyapeta ', and in v. 74, where denotes 'unduly obscure' Yamaka as a reference back to the distinction ann"iam made in v. 32 between 'simple' and 'unduly obscure'. It is unlikely that aniiia existed in v. 26 as a third primary variety when vv. 31 and 74 were framed. The Vrtti will have derived its notion of afifia as 'unsplit-cum-split' from the fact that samalam sam alam- exemplifies ai~iiain v. 31 in the sense of' other than padadi avyapeta'. The Abhinava also imposes upon that example the 'unduly obscure' sense of afifla in v. 74, by attributing to it an exaggerated kilitthapada difficulty. Distinct from raiijaye suciram, the suciram-suctof pada b is glossed as *Auci-rasmi-ruci('sobhanaramsind ruciyutto yo Jino'), as if postulating an elision of -r- in -ramsi-r-uci. Dandin conflates this composite notion of 'otherness' with the term duskara (which stems from v. 32, the Vrtti's repository of AAAA rhyme), by illustrating the latter primarily as The Vrtti substitutes dukkara 'other than merely pada-initial', i.e., as an,-ia.
25 M. R. Kavi (GOS ed., 1934) found the Gana verse only in ' Bha', which presumably (GOS ed., 1926, Preface, p. 9) represents a group of Telugu and Tamil MSS. He did not specify whether it occurs in place of the -mandalam verse. It has remained in square brackets ever since; but *manthakrd and the corruption -mathanakarT would explain its replacement. ' 26 Unsplit-cum-split' might more naturally have referred to the split gemination of an unsplit gemination within the pada (mayamayc- ... -mayamayd) that recurs rhyming AAAA in Dandin iii, 48: but no such monster is found in the Pali. 27 The Latiya criterion 'sinngleich+ silbengleich' (H. Brinkhaus, BISt., 1989, 14) chooses to ignore the Yamaka-based syntactic variation which Bhamaha's examples show to be essential.

This content downloaded from 163.1.62.81 on Fri, 15 Nov 2013 11:15:58 AM All use subject to JSTOR Terms and Conditions

THE PALI

SUBODHALA1NKARA

AND DANDIN'S

KA VYADARSA

335

rhyming for accantadukkara obscurity; but Dandin goes on to use the label duskara as an excuse for all manner of word games. On this basis, the definition of Yamaka in Agnipurana (KSS 343.11ff.) may reflect the original intention of the Karika and Vrtti.28It explicitly defines both unsplit and split as occurring within one pada (padasya.cddimadhydntagocaram), and then creates a third category apara, which is either unsplit or split. This is indicated by its stating that apara encompasses two sets of adimadhydnta positional varieties, six in all (cf. S6hnen, p. 509). Since apara involves rhyme by definition (pfirvena cet padena ... tulyah), mere multi-pada occurrence remains within the scope of avyapeta and vyapeta. Dandin's device of duplicating avyapetavyapetatma could have been inspired, and ostensibly justified, by the apara of AP. The category a/ida is superfluous and incompatible even in the Vrtti. There, multi-pada occurrence is covered by terms like catupdda,and rhyme by quoting the rhyme-scheme.29The effect of the third category is that multi-pada geminations, when rhyming, cease to be vyapeta and avyapeta as in the AP, and become vyapeta and avyapetavyapeta. The Abhinava identifies the unsplit geminations (samalam sam alam-, etc.) and the two rhyming geminations that properly justify the latter; but it could readily be understood to be referring to the example's ABBA rhyme-scheme, one that is itself 'unsplit-cum-split'. It may be this that encouraged Dandin to illustrate the whole range of possible rhyme-schemes, also for vyapeta. The notion of relating the term avyapetavyapeta to rhyme-scheme seems to complicate the exegesis of tan-ca padanam adimafihantagocaram in the basic definition of Yamaka. The Karika text simply illustrated initial position within the padas, the option of multi-pada occurrence being catered for by the plurality of the term paddnam. The Vrtti takes the phrase to refer not only to the three positions in any one of the four padas but also to seven basic possibilities in each pada (adding adimajjha, majjhanta,adyanta, together with adimajjhanta or sabba, i.e. repetition of the whole pada). A third inference computes stanza configurations as ultimately innumerable (aneka), although that point is made separately in the Karika text in v. 32 (accantabahavo), probably with reference to non-rhyming occurrences. The Abhinava adduces ten possibilities (of one-pada and two-pada occurrence) for both unsplit and split, clearly stating that split gemination also occurs within one pada, like unsplit (p. 43, line 33 tatha vyapeta pi dasa). In this computation, it fails (like the postulated original Karika text) to envisage an unsplit-cum-split category or any distinction between multi-pada occurrence and rhyme. It ignores even three- and four-pada occurrence, and instead refers vaguely to canto configurations to make up the numbers. For his definition of Yamaka, Dandin adopts the wording of v. 26, as referring literally to the three positions within any pa-da(iii, 1). In an additional defining verse (iii, 2)
28 The AP definition'anekavarnavrttih' ca vyapetam ca' iti yv ... yamakams& 'avyapetam dvidha ... sodhataddaparamappearsto be quotingthe Abhinava, vuttehi p. 43, line 11punappunam and line 20 abyapetam ca byapetam anekavannehi ca. Cf. Sdhnen,p. 511ff.,visualizingthe course of events"shouldhe [Dandin]be earlierthan AP ': but the primitiveschemeof AP could hardly have arisen from a conflationof the Abhinavaand Dandin. The specificationof differencein in the AP wouldfollowfromits treating Yamakaas an Alamkilra, meaning(bhinnarthapratipadikd) no longerin the contextof lexicaldosa. 29 ABBA is described as pathamacatuttha-dutiyatatiya and AAAA as ekaripavyapeta. vyapeta, We may readcatupada with the 1928edition(Jaini,p. 51), or *catuppadda, for catukkapada (p. 51, line 7); also for catutthapada (p. 46, line 15). The Poranahas padacatuttha (p. 42, line 26) in lieu of padacatukka (p. 43, line 23, and pp. 51-3).

This content downloaded from 163.1.62.81 on Fri, 15 Nov 2013 11:15:58 AM All use subject to JSTOR Terms and Conditions

336

J. C. WRIGHT

vikalpanah, ekadvitricatuspadayamakanam adi-madhya-anta-madhycnta-madhyddy-ddyanta-sarvatah. he adapts the phraseology of the Vrtti's second computation, although the words 'in one, two, three, or four padas' draw attention to the anomaly that two of his three varieties do not in practice occur 'in one pada'. His second hemistich echoes the Vrtti's seven possibilities within each pada; but, as may be seen from the example verses, he applies the wording to its third computation of stanza configurations. Apart from its duplicating what is already implied by the plural padindm in iii, 1, and emphasizing the existence of vyapeta in one pada, this second definition is so artificially compressed that it was misconstrued in B6htlingk's translation. As in Bharata's Giti verse, the element of true rhyme is strong in split gemination: cf. bhasura/-tha 'sura and pi surd/-ri sura-/-ti sura- in the dukkara vyapeta example quoted above. The Abhinava has another specimen with recurrences of internal true rhyme ... -hitdsa sitasad ... satdas vatasd. Such internal true rhymes are never carried through all four Aksaracchandas pddas,30 possibly with a deliberate intention of distancing even this form of Yamaka from true rhyme. At any rate, Dandin seems in iii, 43 to imitate the slight asymmetry of the Pali (where -tha surd- in pada c interrupts pi sura, -ri surd-, -ti surd- in abd) by offering one random -atena among seven -itena ...) fails rhymes. His version of the Pali na bhasurdstanza (iii, 40 sabhasu raijan to sustain the true rhymes; and where Bhatti (x, 17) offers four internal true rhymes, they are entirely different in each pada. Like Bharata in the Gana verse, Dandin and Bhatti have widely spaced true internal rhymes at the caesura and pada-end (S6hnen, p. 501), and Bhamaha's specimen involves endrhyme, not internal rhyme (ii, 15 with the same foible: three -a1asandhrhymes The Pali tended in the opposite direction, towards narrowly plus one -rsdsasanah). like -hitasa sitasI. geminations split pada-internal Plain Amredita gemination, which does not occur in the Pali or in Dandin, etc., appeared in Bharata's Yamaka as a decorative adjunct to, or substitute for, multi-pada rhyme. The concentration in the Pali Karika on initial unsplit gemination, which Dandin develops into a complete systematic exposition, could reflect a desire to differ from Bharata, where pride of place is given to pada-final assonance, such as has been attested in the epics. Bharata's version is noted for a discrepancy between the definitions and the illustrations offered (S6hnen, pp. 499 and 507). The definitions lead us to expect basically assonant syllables, along with an appendix that covers wordgemination, pada-gemination, and sustained consonantal alliteration. Their reference to monosyllabic assonance is a significant archaism (S6hnen, p. 497),3i as compared with the illustrative verses, where the concept is reduced to unedifying jingles that resemble a compromise between the intended assonances and the sophisticated classical Yamaka. The absence in Bharata of padamedial assonance is possibly another sign of age, since it implies an interest in
30 Although the Abhinava has 'honti' for hontu (p. 53, line 22), yielding *hont' atasa hata-sa and perfect true rhymes in each pada, the modal reading hontu 'tasa is convincing. 3 KM ed., xvi, 66 padiin"m ante sy;t samam aksaram; 78 adau padasya ... samdve"ah samaksarah, 84 yatraikam vyalijanambhavet, tan malayamakam. Even if samam aksaram is merely a prosodic substitute for an expression like 75 aksaraih samau, there is significance in the insistence on .abda 'sounds', on aksara referring to pada-final and phda-initial syllables, and on vyaiijana. The main discrepancy in Bharata seems to be between the first four verses (defining and subdividing Yamaka) and the individual definitions. The latter are clearly at a loss to reconcile their examples with the basic terminology. They miss the hemistich-final rhyme in Samdasta (significant as it completes the paddnta section); and their pointless restriction of Amredita word-gemination to pada-final position probably indicates a desire to pair it with Samdasta (wrongly construed as pada-initial gemination) instead of with Caturvyavasita pada-gemination.

This content downloaded from 163.1.62.81 on Fri, 15 Nov 2013 11:15:58 AM All use subject to JSTOR Terms and Conditions

THE PALI SUBODHALANKARA

AND DANDIN'S

KAVYADARSA

337

end-rhyme and initial Stabreim which are a more normal preoccupation of poets than gemination. Punning seems to occur only of necessity when the rhymes and geminates exceed a basic 2-3 syllables; and since the word yamaka makes no reference to meaning, it will have been coined to describe Bharata's system of assonance and gemination. The theory of rhetoric distanced itself from Bharata also by keeping assonance separate from alliteration. A deliberate poeticization of each of them endowed Yamaka with an element of essential decorative word-play and alliteration with emotive qualities. Play on words, irrelevant to Bharata, is implied by the division into coherent and incoherent varieties (v. 24f. silittha-padasandhi and kilitthapada-dosa)that accompanies the Pali definition. Dandin drops this link with dosa, and therewith any implication of word-play in the definition, although he makes incidental use of the Karika's concluding epithets -dukkara and n' ekantamadhura. Some slight semantic difference appears in unsplit gemination, as in v. 28 kusalakusald sabbe ... (a Dvandva, though printed as kusald 'kusald in the edition) and v. 30 sugato Sugato ... (unnecessarily taken, but only optionally in the Abhinava,to be a pun *sugadah Sugatah); and also in the Vrtti's examples of pada-internal rhyme. Unlike Dandin, the Vrtti is prepared to waive this characteristic semantic differentiation when geminations are rhymed: thus surd'liquor' recurs in padas b and d of one verse, and manam 'mind' recurs in all padas, and manam manam recurs as *mano mandk in most padas, of another (p. 51).32 This further confirms that multi-pada rhyming is an optional decorative adjunct (reversing the appearance in Bharata of Amredita gemination as an adjunct to basic rhyming). The section rasabhavabodha and Dandin In the Karika text, the fifth and final section on nine Rasabhavas blossoms forth with integral verses in Sragdhara, Gana, and Matra metres, and it lacks the expected closing stanza in Indavajiradi metre. This section is surely an accretion, for its promulgation of a ninth Bhava sama is not consistent with the text's initial use of the eighth Bhava vimhaya to identify what is lacking in untutored poetry (n. 3, above). That sama (*Aama)has been superadded, as a means of adapting Kavya theory for Buddhist purposes, appears from its tacitly duplicating the item sama (*frama) that appears in the list of Vyabhicarins in v. 344. The manuscript colophons also imply that the chapter is spurious. The preceding 'Atthalanikara'chapter carries the more plausible, and hence probably basic, version of the Pordna colophon (Mahasamindmikayam The 'Rasabhava' chapter ends in the 1928 edition SubodhalaikadratTkdyam). with Alaiikarappakaranam The final Rasavant Alamkara may also be an accretion, attracted by the adjacent Bhava Alamrnkara, since the preceding samattam.. AMisAlamkara (v. 334 ndtho patu lokam) is an appropriately auspicious final item. Apart from a brief and belated explanation of sihigaradiRasa (vv. 352-55), the topic of this final section is Rasabhava. Abstract rati appears in v. 356 (clearly *sa has to be read for so in 357c) on a par with ruddo raso, etc. Thus the Vrtti is hardly justified in distinguishing a Sthayin from a Rasa in v. 366b
32 The Abhinava seems most implausibly, and doubtless corruptly, to read an Amredita *manomanah into one occurrence. Pada-internal vyapeta survives in Dandin only in his absurd sura ... sura specimen (iii, 40), clearly modelled on the Vrtti's specimen, but now with doggedly different semantic applications for each of the eight occurrences of surd; and in iii, 47, where widely spaced rave-... -rave is incidental to a Samdasta stanza.

This content downloaded from 163.1.62.81 on Fri, 15 Nov 2013 11:15:58 AM All use subject to JSTOR Terms and Conditions

338

J. C. WRIGHT

vimhayo 'yam raso 'bbhuto:that distinction is made explicit only for the ninth Rasa. The chapter can be seen as a pendant to the purely metaphorical Rasavant Alamkara with which the previous chapter ends (v. 336: the Buddha's feet caress the lotus-face of Earth), emphasizing the distinction between it and Rasa proper. As in Dhvanyaloka, the Rasavant figure is a mere implication of Rasa (v. 335 rasappatitijanakam)or of a semblance of Rasa (Vrtti: rasabhasa). Another cardinal principle of Dhvanydloka appears also as a significant starting point in the Pali Vrtti (p. 7f.): pana kifici vatthu# atth' eva vanTsu patTyamdnam mahdkavTnam, # abhati ldvanyam iv' ariganasu. yam tamrpasiddhdvayavatirittam is required and, despite [Ed. tam pasiddha?:but the specification vanipasiddha? the Vrtti's 'yam kif-ci vatthu', an enjambment of yam alone is impossible.] Again the application in Pali is simpler. Traditional literature is not enjoyable because the subtler implications of its obsolete language (atthassa upatthambhakabhito bhavatthalesopi) can no longer be understood.33The Sanskrit context (DhA 1.4) required the insertion of anyad eva punar anyad eva V vastv asti vdnTsu pratTyamdnamr mahdkavTnam and the resulting emphasis on vastu instead of vanisu impugns *tatprasiddha= vant . The DhA Vrttirenders the ensuing yat tat as kim api 'je ne sais quoi', more in line with the 'yam kif-ci vatthu' and 'bhdvatthaleso' of the Abhinava than with any attested Sanskrit usage. The Karika illustrates Rasavant, and the Abhinava illustrates even the Rasabha-vas, with individual Buddhist stanzas. Dandin adopts this means of incorporating the theory of dramatic evocation (ii, 285 rasavattvamsamarpaya-) into Alamkara~astra, but by assigning to Rasavant Alamkara the full connotation of Pali Rasabhava. He expects us to compress entire dramatic effects into one isolated Rasavant stanza (ii, 281 ripabdhulyayogena), whereas what the Abhinava had offered for Rasabhava was a set of Kavya verses on one sustained narrative theme, Vidhurapanditajataka (Jaini, p. 292, note), such as would accompany a Campfi or a dramatization of the text. Fragments of the Pali treatment of Rasabhava subsist in Dandin. He attributes his version of the piyatara figure (v. 262) to Vidura and couples it as Preyas with Rasavant. His Bhavika Guna, in the sense of sustained artistic unity (ii, 364ff.), replaces Bhava Alamkara (v. 329f.) and Rasavant as the culmination of the Alamkara section.34 The Abhinava shows that Dandin's gloss on Bhavika 'kaver abhiprdyahkdvyesv asiddhi samsthitah' is achieved by reading the Pali rubric bhava-adhippiy[o] 'definition of Bhava' into the definition of thdyTbhava (v. 341bc yo bhdvo na tirohito sTlenatitthati). For, when glossing the definition of bhava (v. 340), it had (optionally) read the relevant into pada b bhdvayanti rase yato, with rase rubric, as bhdva-adhippdya[mr], construed for the purpose as a locative (' tasmimrn yeva rasavisaye kavino, lokasabhavam visayam katvd, ... adhippdyam bhdvayanti'). It made no attempt to
33 The Abhinava uses the ambiguity of 'suddhamdgadhika' as a case in point. The Porana seems to misunderstand this, dropping the Taddhita alternative, expanding patttivisesayogato bhavatthaleso into a statement that Ramasamman wrote in Sanskrit (Rimasammadinam verse. Sakkatadibhavato ti ayam ettha saddattho), and abruptly tacking on the patTyamanam 34 The Pali Bhava Alamkara concerns praise (vanna) expressed with an argumentative particle (namadi) that conveys an opinion (bhavabodhana).The definition pavuccate yarmnamadi will refer to particles such as nama 'sarmbhavye';the example has nanu. The Abhinava renders pavuccate namadi yena kenaci vannenaas 'nouns, etc., are used, denying in a hesitant manner' (patisedhetva va no va), and attaches 'ti sambhavane' equally oddly to api. The Pordna simply ascribes the sense 'denial' (nisedhanarilpena)to yena kenaci vannena.

This content downloaded from 163.1.62.81 on Fri, 15 Nov 2013 11:15:58 AM All use subject to JSTOR Terms and Conditions

THE PALI SUBODHALANKARA

AND DANDIN'S

KAVYADARSA

339

read either bhava-adhippcyaor thdyTbhava-adhippaya into v. 341; but the Pordna has substituted lokathitim for lokasabhavam, thereby imposing Dandin's conflated treatment of v. 341 (' Bhava' as a sthdyin abhipriya) upon the Vrtti's gloss on v. 340 (' Bhava '). Buddhist RasabhaSva with nine Rasas would hold little attraction for Dandin. In v. 367 samo metta-dayd-mod'-adisambhavo, thdyTbhavo bhdvaddhi tadukkamso santo santanisevito. it is clear that samo and santo have the connotation of sama and sinta, so as to provide another auspicious closing stanza in place of the Asis verse. The terminology is, however, again probably Middle Indo-Aryan. The sama that already existed as the Vyabhicarin rama could, with sramana and tapahkranta in mind, be punningly combined with upacama and promoted to the status of a supreme Rasabhava based on mettadaydmodddias Vibhavas. The glossing of samo as 'santaguno' at v. 342 (as well as 'upasamo' at v. 367), and of v. 367 santanisevito as 'sddhiihisevito ', would favour *sranta, as more appropriate to the notion of seva than *Sdnta (while *sant is irrelevant to the etymology). Thus santa would be a portmanteau word like *k(r)antiguna, *d(v)Tpa, and the like, and could not be Sanskritized. An implication of srdnta appears in the specimen verse (Abhinava, p. 304): pi ... yakkhe vicalayati, no cali Tsakam santim gato va Vidhuromadhurd [p]i bhaiv[d] an improbable Bahuvrihi with -avibhavo, rests on garbled [Ed. madhuravibhavo, *madhura vibhava, as is shown by the gloss 'madhurd bhdva api madhurasabhavatopi']. The Abhinavatakes the phrase *sdntimgata iva to be an explicit mention of santarasa, denoting upasama ('santarasassa pakatibhfit[e] uddese dassit[e] navamo thdyibhavo dassito hoti' [Ed. ?bhiIto uddese dassito]), and claims that it implies all requisite Vibhavas, etc., by simarthya. This feeble effort may explain Dandin's willingness to subsume Rasabhava within Rasavant; but it does no justice to the verse. The phrase gato va being a purely incidental Alamrnkara, the key phrase would be madhura santim, (sva)bhdva, implying the mettadaydmodadi qualities that underlie *srintaguna-cumidntarasa. Subodhalafikara and Vuttodaya It likewise cannot be taken on trust that the Vuttodaya prosody that is also ascribed to Safigharakkhita is 'wholly based on works dealing with Sanskrit prosody, the terms of which it has borrowed and adapted' and 'in some instances, whole sentences are incorporated from Pifigala and other authorities with no more alterations than are necessary when Sanskrit is turned into Pali' (G. P. Malalasekera, PLC, 198; similarly, J. W. de Jong, in the preface to R. Siddhartha, Samgharakkhita's Vuttodaya, ed.2, 1981; and K. R. Norman, Pdli literature, 1983, 168). The Vuttodaya passage that Malalasekera adduced was 'the last ten verses of the first chapter Pada-ccheda (on caesura), which are taken directly from Halayudha'. The first chapter of Vuttodaya deals with types of metre, and defines Yati only incidentally (v. 9 'padacchedo yat[F] bhave' [Editions:yati bbhave]). The fact that Halayudha alludes to samudrendriyarasddisu while glossing 6.1 yatir vicchedahpresumably reflects not only the fact that samudda-usu-rasa-ddihappens to be defined in the same Vuttodaya verse as Yati, but also the ambiguity of *samudresu-rasadisuthat precluded direct transcription from Pali into Sanskrit.

This content downloaded from 163.1.62.81 on Fri, 15 Nov 2013 11:15:58 AM All use subject to JSTOR Terms and Conditions

340

J. C. WRIGHT

Helmer Smith, in Saddaniti, Iv, 1, 1949, 1148, described the text as based on Kedarabhatta's Vrttaratndkara.Taking an example at random, however, Vuttodaya,v. 73ff., has fourteen twelve-syllable metres where the Sanskrit text defines 24 (3.45ff.). The first thirteen recur in the same sequence in the Sanskrit; and of these the first three read: vadanti Vamsatthamidamja-td ja-rd ja-rau ja-tau tu Vamrnastham udtritamr sa-y-Indavamsa khalu yattha ta ja-ra ta-ta-jai ra-samnyutaih sydd Indravam.d idha Totakam ambudhi-sehimitam iha Totakam ambudhi-saihprathitam. This is not 'no more alterations than are necessary when Sanskrit is turned into Pali', but more obviously those that are required when Pali is turned into Sanskrit. The Sanskrit rejects (or misreads) banal vadanti... idam at the expense of rational word-order; its rejection of hiatus after sd forces it into a clumsy construction; and the loss of a syllable in -sehi engenders prathita, since pramita was obsolete in the relevant sense. There is no observable reason why the Pali would drop Sragvini, Jaladharamala, and Prabha, while retaining Puta, Priyamvada, and Ujjvala; or indeed why Pali should be dependent on Sanskrit texts for metres which must have developed within the freer confines of Middle Indo-Aryan prosody. Fryer was nearer the mark in merely listing some of the many instances where Kedara's definitions differ from those of Vuttodayain name, scansion, or caesura. The question of the direction of borrowing simply did not occur to Smith. His description of the Pali text as based on Kedara followed on from his reporting the text's self-description lokiyacchandanissitamas 'base sur la metrique sanskrite lokiyacchandas'. A Ramasarman is named as one of Kedara's sources for prosody (according and also as a compiler of Unadisiftra in an Anustubh to BhdvarthadTpika), version (Aufrecht, Cat. Bodl., vii, no. 399). Perhaps the appearance of the names of Ramasarman and Dandin in both literatures is evidence of some measure of awareness of genetic links between Pali and Sanskrit in poetics, prosody and grammar. While both Tikas claim to draw on 'Dandi' and other pubbacariya authors (p. 270),35 the Abhinava seems to cite ten' eva Dandiyamr as the name of a work (p. 262, possibly implying *Ddnfdland referring to the subject-matter of the example in question satthappaharanamdatva ...). Thus the Vrtti seems to have no more to contribute on the subject of Dandi than on the subject of Ramasammadyalainkara. The Abhinava, i.e. the poetics manual 'Subodhalarikaranissaya',is sometimes conflated with readings used by Dandin, and is probably a fair reflection of the Vrtti that inspired the composition of Kavyadarsa. This Vrtti had somewhat traduced the pious intention of the Karika text 'Alatikarappakarana ', by introducing material that had been deliberately excluded. The Por1na, or 'Subodhalarikara-ttka',is no auto-commentary, but rather an abbreviated and Sanskritized re-working of the Nissaya. The firmest tradition is that which associates Saligharakkhita and other medieval scholars with various 'Porcna'

35 The Abhinava has DandiBhaddapdnadi Li.e. Dandin and *Bhadrapdnin, with the attested name Pdnin?). Jaini suggests *Bhaddana for '(Ananda)vardhana', but this is hardly tenable.

This content downloaded from 163.1.62.81 on Fri, 15 Nov 2013 11:15:58 AM All use subject to JSTOR Terms and Conditions

THE PALI SUBODHALANKARA

AND DANDIN'S

KAVYADARSA

341

commentaries. Despite other traditions that ascribe the Alanikarappakarana and Vuttodaya to the same author in twelfth-century Ceylon, the indications are that Buddhists had turned to a Prakrit secular tradition for guidance long before Sanskrit poetics and prosody flourished at the hands of Dandin, Anandavardhana, Halayudha and Kedarabhatta.

This content downloaded from 163.1.62.81 on Fri, 15 Nov 2013 11:15:58 AM All use subject to JSTOR Terms and Conditions