13. CHAPTER TEN MAHIMABHATTA' S VYAKTIVIVEKA

CHAPTER TEN MAHIMABHATTA' S VYAKTIVIVEKA

The most elaborate and strident, and also the last, major

attack on Anandavardhana's dhvani-theory is that offered by

Mahimabhatta in his Vyaktiviveka ("An Analysis of

Suqgestion"). Like Mukulabhatta before him, Mahimabhatta

challenges Anandavardhana's claims for the existence of

suggestion (vyañjana) as a mode of expression unique to poetry

and necessarily distinct from those semantic processes dealt

with by earlier philosophers and language theorists. For

Mahimabhatta, what Anandavardhana would call suggestion (and,

indeed, non-literal expression generally) can be satifactorily

explained only as a kind of inference (a criticism which, as

we have seen, Anandavardhana himself anticipated and strove to

refute -- cf. above, Chapter Five, pp. 239ff) .

Mahimabhatta knows and quotes both Kuntaka's

Vakroktijīvita and Abhinavagupta's Locana;1 Mammata (c. 1100

AD) is clearly familiar with his work.2 Thus, the

Vyaktiviveka must have been composed some time during the

1Mahimabhatta, Vyaktiviveka, ed. Revaprasad Dvivedi, Kashi Sanskrit Series 121, Varanasi 1981 [third ed.], pp. 96, 142, 284.

2At the end of the fifth ullasa of his Kavyaprakasa, Mammata responds briefly to Mahimabhatta's argument for the identification of suggestion with inference -- cf. Kāvyaprakāsa (ed. N.S. Venkatanāthacarya, Oriental Research Institute Series 120, University of Mysore, 1974) Vol. I, pp. 474-81. Mammata's examination of poetic flaws in the seventh ullāsa often draws very heavily on Mahimabhatta's treatment of the same topics -- cf., for example, ibid :; Vol. II, pp. 50-61 [compare Vyaktiviveka, op. cit., pp. 196ff.], and pp. 105-117 [compare Vyaktiviveka, op. cit., pp. 287ff].

471

472

eleventh century. The Vyaktiviveka is divided into three

chapters called vimarsas. The first vimarsa offers a detailed

critique of Anandavardhana's definition of dhvani and of the

theory of suggestion. The second vimarsa consists of an

extensive analysis of aesthetic flaws (dosas) -- the relation

of this seeming digression to the overall project of the work

will be considered below (pp. 493ff). The third vimarsa

examines individually Anandavardhana's examples of dhvani,

demonstrating in each case that the meanings Ānandavardhana

describes as suggested can be arrived at only through a

process of inference.3

In developing his theory of poetic inference, and in many

other areas besides, Mahimabhatta relies very heavily on the

ideas of the great seventh-century Buddhist logician

DharmakIrti. In his edition of the Vyaktiviveka Revaprasad

Dvivedi identifies one citation from Dharmakīrti's

Pramānavarttika,4 and C. Rajendran finds two more,5 but the

full extent of Mahimabhatta's indebtedness to Dharmakirti has

not been noted. His work is littered with quotations from all

of Dharmakirti's major works, and the views so quoted are

3A fairly good general survey of Mahimabhatta's views and arguments is C. Rajendran's A Study of Mahimabhatta's Vyaktiviveka (Published by Dr. C Rajendran, Feroke, 1991); more detailed examinations of selected topics dealt with in the Vyaktiviveka may be found in Amiya Kumar Chakravorty's Studies in Mahimabhatta (Calcutta University, 1975).

4Vyaktiviveka, ed. R. Dvivedi (Kashi Sanskrit Series 121, Varanasi, 1982 [third ed.]), p. 77.

5Rajendran, op. cit., p. 48.

473

invariably treated as absolutely authoritative.6 It is not

clear whether or not Mahimabhatta was himself a Buddhist,7 but

he was, in any case, a thoroughgoing follower of DharmakIrti

in matters of logic.

The Vyaktiviveka has not received anything like the

attention it deserves from modern scholars of Alamkārasastra,

probably at least in part because of its author's explicit

hostility to the widely favored dhvani theory of

Ānandavardhana. While it is generally conceded that

6The citations I have so far identified are as follows -- Pramānavārttika 2.57 (Vyaktiviveka, op. cit., p. 76), 3.26 (p. 80), 3.27 (p. 69), and 3.191-2 (p. 469); Nyayabindu 2.3 (p. 81), 2.34 (p. 465), 2.38 (p. 466), 2.39 (p. 53), and 3.1 (p. 67); he quotes from the Pramanaviniscaya on p. 81 (cf. Ernst Steinkellner, "New Sanskrit Fragments of Pramanaviniscaya, First Chapter," Wiener Zeitschrift für die Kunde Südasiens XVI, 1972, p. "Dharmakirti's 201), on p. 76 (cf. Ernst Steinkellner Pramanaviniscayah, Zweites Kapitel: Svarthanumanam, " Österreichische Akademie der Wissenschaften, Philosophisch-Historische Klasse Sitzungsberichte 287, Band 4, 1973, p. 25), and p. 316 (cf. E.M. Stern, "Additional Fragments of Pramanaviniścaya I-II," Wiener Zeitschrift für die Kunde Südasiens XXXV, 1991, p. 153); he quotes from the Vadanyaya on p. 333 (cf. Vadanyaya, Bauddha BharatI Series 8, Varanasi, 1972, p. 106) and p. 43 (cf. ibid., p. 109).

7K.C. Pandey finds indications of Kashmir Śaiva affiliation in Mahimabhatta's work (cf. his "Kashmir Saiva tendencies of Mahimabhatta," Bharatiya Vidya, Vol XI, pp. 187ff; also Rajendran, op. cit., pp. 47-51). His commentator Ruyyaka, however, certainly believes Mahimabhatta to be a Buddhist; he frequently accounts for Mahimabhatta's remarks by pointing out that they represent the "Buddhist view" (saugatadrs, saugataprakriya) on a particular question (cf. Vyaktivivėka, op. cit, pp. 48, 279, 339, 376), and often cites DharmakIrti's works himself in explaining Mahimabhatta's

3.1], standpoint (ibid., p. 48 [cf. Svopajñavrtti on Pramānavārttika p. 62 [Pramanavarttika 3.263], Bhattagopala, and 339 [Pramānavārttika 4.128]. a pre-fifteenth p.

century commentator on Mammata's Kāvyaprakāsa, describes Mahimabhatta as a Buddhist (cf. Rajendran, p. 48).

474

Mahimabhatta's work displays great logical acumen and

polemical skill, the significance of his contribution to

Sanskrit poetic theory is often downplayed. According to S.K.

De, for example,

Mahimabhatta's work is undoubtedly scholastic argumentation, a masterpiece of exhibiting much fastidious criticism and great learning of a miscellaneous kind; but its avowed object is polemical and it does not pretend to set up a new system ... A fine product of a scholastic age, he cannot yet look beyond the pettiness of immediate issues; and whatever might be the value of his peculiar proposition, he hardly ever adds to its limited interest any independent treatment of the larger problems of Poetics.8

Krishnamoorthy observes that, "In depth of perception,

clearness of exposition and vastness of erudition,

Mahimabhatta very nearly approaches Anandavardhana himself

The Vyaktiviveka excels in the art of subtle dialectics and

hyper-critical polemics and has hardly had a parallel in the

whole range of Sanskrit criticism. "9 Nevertheless, he labels

Mahimabhatta "a mere carping critic. "10

This dismissive attitude toward Mahimabhatta's work

underrates his continuing importance in the Alamkāra

tradition. While it is true that his identification of dhvani

with inference finds no followers among subsequent

Ālamkārikas, his arguments remain an important reference-point

8De, History of Sanskrit Poetics, op. cit., Vol. II, pp. 199-200.

9K. Krishnamoorthy, The Dhvanyaloka and its Critics (Kavyalaya Publishers, Mysore, 1968), p. 296.

10Ibid., p. 249.

475

for later dhvani-theorists, and the need to respond to them

effectively is keenly felt even by much later Alamkārikas.11

His elaborate analysis of poetic flaws is extremely

influential, and large portions of it are adopted wholesale by

later authors. Despite the ultimate triumph of the dhvani-

theory against which he struggles, Mahimabhatta clearly

remains a tremendously important figure in the tradition of

Alamkāraśāstra.

Suggestion and Inference

As already noted, the central thesis of the Vyaktiviveka

is that what Anandavardhana calls suggestion is nothing but a

kind of inference; in every case of "suggestion" the directly

expressed meaning of a passage, in concert with relevant

contextual factors, serves as the inferential mark (hetu or

linga) which allows us to infer some additional, unstated

meaning. It is important to stress that, apart from this one

major point of contention, there is very little fundamental

difference between Mahimabhatta's understanding of the nature

of poetry and Anandavardhana's. Like Ānandavardhana,

Mahimabhatta asserts the crucial role of unstated meaning in

poetic language, accepts three types of unstated meaning

(vastu, alamkara, and rasa), adopts a teleological, rasa-

centered model of literary aesthetics, and takes the ability

of any feature of a work to contribute to the work's overall

11Cf., for example, Jagannatha Panditaraja's (seventeenth century) Rasagangadhara (ed. Kedāranātha Ojha, Sarasvatī Bhavana Granthamala 113, Varanasi, 1977) , Vol. I, pp. 47-49.

476

rasa as the sole criterion by which to judge its aesthetic

value. Mahimabhatta himself explicitly points out the limited

scope of his quarrel with Anandavardhana: "Suggestion is the

life-breath of dhvani; thus this alone is analyzed here.

Regarding the rest, there is for the most part no disagreement

[between Änandavardhana and myself] . "12 Indeed, as we shall

see, Mahimabhatta not only accepts Anandavardhana's rasa-

centered poetic teleology but, like Bhattanāyaka and

Abhinavagupta, carries its logic even farther than

Ānandavardhana was prepared to do by insisting on rasa-

directedness as a necessary feature of all poetry.

Mahimabhatta's argument against the existence of

suggestion as something distinct from inference rests most

fundamentally on his Dharmakirtian epistemological principles.

For Dharmakirti there are only two kinds of knowable entities:

unconceptualized sense-particulars (svalaksana) , and

generalized concepts (sāmānyalaksana, kalpanā) -- and,

correspondingly, only two means of valid knowledge: perception

(pratyaksa), which apprehends sense-particulars, and inference

(anumana), which is the source of all generalized conceptual

knowledge. Thus all valid knowledge which is not immediately

and purely perceptual can only be the product of a process of

inference; all other means of valid knowledge which have been

proposed by rival epistemologists (e.g. upamāna

["comparison"], sabda ["verbal testimony"], or arthapatti

12Vyaktiviveka, op. cit., p. 511. Cf. also p. 111.

477

["necessary presumption"]) must, insofar as they are valid, be

included within the scope of inference, as they all deal with

generalized conceptual knowledge.

Mahimabhatta reprises this classic argument in order to

show that Anandavardhana's suggestion can be nothing but a

kind of inference:

For the understanding of one thing from another is not possible without inference; and [other means of knowledge such as] upamana, etc. are included in that [inference]. As they say, "When one thing is seen, the conception [kalpana] of another is [generally] not appropriate, because overextension would result. 13 But this will be possible if there is an invariable connection of this [other thing with the thing percieved] For this [percieved thing], its nature being invariably connected in this way, will make one aware of another which is established to be thus [invariably connected]. And the cognition of a thing by means of a general concept connected with it is inference -- thus there are only two means of knowledge. "14 And the understanding of another [unstated] meaning from an expressed meaning is not possible without the recollection of an invariable connection [between the twol, because otherwise there

13I.e. upon percieving any entity, one might cognize any other (unpercieved) thing. There would be no restriction at all on one's knowledge: any or all possible conclusions could legitimately be drawn from any perceptual awareness whatsoever.

14T.e. perception and inference (the latter being the only legitimate means by which knowledge of what is not immediately percieved may be arrived at). The quotation is from Dharmakirti's Pramanaviniscaya (cf. Steinkellner 1972, op. cit., p. 201). Unfortunately, this quotation has become somewhat garbled: In the [percieved thing] ... " sentence beginning "For this the word meaning "its nature being invariably connected" (pratibaddhasvabhavah) is omitted and the word "this" (sa) is replaced by "not" (na), distorting the meaning to the point of unintelligibility; the passage is quoted correctly (with minor variants) in Bhasarvajña's Nyayabhūsana (Varanasi: Saddarsanaprakasana-pratisthāna, 1968, p. 381)' and closely Vacaspatimisra's Nyayavārttikatāparyatīkā (Varanasi: Kashi Sanskrit Series 24, paraphrased in

1925, p. 158).

478

would arise the dilemma that any meaning at all could be understood from this [expressed meaning] .15

We can understand from an utterance a meaning not explicitly

stated only if this meaning is necessarily associated with the

meaning that is directly stated (otherwise there will be no

restriction on the meanings which may so be understood and no

definite meaning will emerge from any utterance) ; deducing one

thing from another on the basis of necessary association is

nothing but inference. Thus Anandavardhana's dhvani (along

with all forms of non-literal expression) can only be a kind

of inference.

To undermine Anandavardhana's claim that dhvani is based

not on inference but on vyañjana ("suggestion",

"manifestation" -- cf. above, Chapter Five, p. 204) ,

Mahimabhatta subjects to detailed analysis the notion of

vyañjana as it has been applied outside the realm of poetry

and concludes that the process of non-literal expression

Anandavardhana describes as vyañjana cannot properly be so

designated. Mahimabhatta offers the following definition of

vyañjanā: 16 "The appearance Of a thing revealed

[prakāsamana], whether real or not, together with its revealer

[prakāsaka], when this [revealer] does not depend [for its

functioning] on the recollection of some connection [with the

15Vyaktiviveka, op. cit., pp. 81-83.

16The term Mahimabhatta uses here is actually abhivyakti; throughout the Vyaktiviveka he uses the cognate terms vyakti, abhivyakti, vyañjanā Ānandavardhana's vyanjana. and interchangeably to refer to

479

thing revealed]. #17

This definition covers a variety of phenomena. One type

is the manifestation, due to some cause, of something which

had previously existed only in a latent, imperceptible form,

e.g. the appearance of curds in milk. 18 Another is the

revelation of some object by the light of a lamp; when a lamp

is lit in a darkened room the lamp itself and any nearby

objects previously obscured by darkness will become visible

simultaneously. It is this second type which Anandavardhana

sees as operative in poetry containing dhvani; he explicitly

says that the relationship between suggestive words and

meanings and the suggested senses they convey is analogous to

that between a lamp and a pot which is illumined by it.19

In addition to these phenomena which meet his definition

of vyañjana, however, Mahimabhatta examines others which have

sometimes been described as types of vyañjana but which do not

fulfill this definition -- those in which the supposed

suggestive factor depends on a remembered connection with the

thing it reveals; for example, the awareness of a cow produced

either by a picture or other representation of a cow or by the

17Ibid., p. 80.

18At least according to the Sāmkhya theory satkaryavada, which holds that before they become apparent all Of

effects (e.g. the curds) pre-exist latently in their causes (e.g. the milk); Mahimabhatta himself notes that some would regard this as the production of curds from milk, rather than the "manifestation" of curds which already existed, latently, in the milk (ibid., p. 80).

19Dhvanyaloka, op. cit., pp. 420-21.

480

word "cow". In such cases, the desired awareness can come

about only if one recalls the previously experienced

association of a particular arrangement of shapes and colors

or a particular set of sounds with the object being

represented or denoted. For the reasons already discussed,

Mahimabhatta regards any situation in which an awareness

arises in dependence on such a remembered connection as a case

of inference. Thus such phenomena cannot properly be regarded

as cases of vyanjana, and Mahimabhatta accordingly formulates

his definition so as to exclude them.

However, Anandavardhana himself does not regard poetic

suggestion as functioning like these latter cases of supposed

vyañjana. As already noted, he sees the vyañjana which takes

place in poetic expression as analogous to the case of the

lamp and the pot, which Mahimabhatta himself accepts as a type

of vyañjana. Mahimabhatta argues that the analogy cannot

hold: when the light of a lamp reveals both itself and a

nearby pot, the two become apparent simultaneously, while such

simultaneity (which, as we have seen, he regards as essential

feature of true vyañjana) is not possible between directly

expressed and unstated meanings. As Mahimabhatta notes,

Anandavardhana himself argues that a hearer's awareness of

directly expressed and suggested meanings is necessarily

sequential, even if this sequence is sometimes (as in rasa-

481

dhvani) too brief to be noticeable.20 Our understanding of

the unstated meaning of an expression cannot arise

simultaneously with our understanding of the expressed

meaning, because the former depends on the latter as its

cause; only after we have understood the expressed meaning can

any additional, unstated meaning be recognized.

Even if one were to amend the stated definition of

vyañjanā to remove the requirement for simultaneity,

difficulties would remain. Altering the definition in this

way would allow one to designate the process by which

expressed meanings give rise to unexpressed ones as vyañjanā,

but it would also, by the same token, require that all cases

of inference be so designated.21 To take the stock example

of inference: when one sees smoke rising from a mountain and

(subsequently) concludes that there must be fire there, the

smoke would have to be said to "suggest" or "manifest" the

fire. This would run counter to all established usage and

cognitive theory, and would totally eliminate any possibility

of distinguishing between inference and vyañjana.

The use of the term "suggested" (vyangya) by

Ānandavardhana and others to describe certain poetic meanings

which are not directly expressed is acceptable, but only if

(contra Ānandavardhana) one understands the term to be

20 Tbid., p. 85. Cf. Dhvanyaloka, op. cit., pp. 404-414 and above, Chapter Four, pp. 134-136.

21Ibid., p. 86.

482

employed metaphorically.22 The inference of an unstated

meaning from a stated one may sometimes (i.e. in the case of

rasa) come about so rapidly that our awareness of the two

appears to be simultaneous, and in this respect resembles a

genuine case of vyañjana. This apparent simultaneity may lead

us to describe the second, unstated meaning as "suggested"

(vyangya) . But, as Anandavardhana himself argues, this

simultaneity can only be apparent; it is for this reason that

he describes rasa-dhvani as "that in which the sequence

[between the expressed and suggested meanings] is not noticed"

(asamlaksyakrama). Therefore, the characterization of the

unstated meaning in such cases as "suggested" can be accepted

only if it is understood metaphorically. And even this

metaphorical application of the term "suggested" will be

acceptable only in the case of Anandavardhana's rasa-dhvani;

in vastu- and alamkära-dhvani there is not even apparent

simultaneity between the expressed and inferred senses, and

there is therefore no basis for any such metaphorical

usage.23

Mahimabhatta's contention that all such unstated meanings

are inferred, rather than "suggested" can be sustained only if

he can show that some valid inferential mark (liñga) is

present wherever the understanding of such an additional

22Ibid., pp. 59ff, 78, 113.

23Ibid., p. 61.

483

meaning occurs.24 Here, once again, Mahimabhatta finds

fodder for his argument in Anandavardhana's own analysis of

suggested meaning. In his own argument for the existence of

suggestion as a semantic function distinct from direct

expression, Anandavardhana makes the following observation:

The expressive capacity of a particular word is its fixed essence [ātmā], since it is seen to be invariably connected with that word beginning from the time that we learn it. But this [suggestiveness] is not fixed, since it depends on particular conditions. For this will be understood through the limitation imposed by factors such as context, and otherwise [i.e. in the absence of such factors] will not be understood. 25

Mahimabhatta quotes this passage with approval. Our

understanding of any additional meaning beyond what is

directly expressed by a word depends not on the word itself

(otherwise it would be understood whenever the word was used)

but on a complex set of contextual factors in connection with

the directly expressed meaning of the word (and is thus

24Later critics of Mahimabhatta's views, beginning with Mammata, make this the focal point of their attacks. In true cases of suggestion, it is argued, there is no necessary association (avinabhava) between the expressed and suggested senses. The relation between the two will be variable (anaikantika), and the expressed sense cannot therefore be considered a valid inferential mark with respect to this second unstated meaning. Sensitive readers will understand the poet's intended meaning even though it is not deducible by strict canons of logic from the directly expressed meaning of a verse (cf. Mammata, Kāvyaprakasa, op. cit., Vol. I, pp. 474- 81) . Kunjunni Raja, reading the later responses of the tradition back into the Dhvanyaloka, misleadingly presents this argument as Anandavardhana's own (Indian Theories of Meaning, Madras: Adyar Library and Research Centre, 1963, p. 291) .

25Dhvanyaloka, Vyaktiviveka, op. cit, p. 148. Also cf. above, Chapter Five, op. cit., pp. 436-437; quoted in

p. 212.

484

understood only occasionally and under certain specifiable

conditions) .

It is this entire set (samagri) of contextual and

semantic factors which serves as the inferential mark that

will allow us to understand a particular unstated meaning in

any given case.26 If this set of factors is sufficiently

determinative to allow us to understand a particular unstated

meaning when and only when a poet means for us to do so, it

will be necessarily connected with this intended meaning and

may thus be regarded as a valid inferential mark; if it is not

sufficiently determinative, the intended meaning of the verse

in question will be unclear (and the verse will hence be

faulty). Thus, no verse (or any other expression) will be

able to effectively convey any meaning not directly expressed

unless its directly expressed meaning, together with any

relevant contextual factors, constitutes a valid inferential

mark necessarily associated with that unstated meaning. In

the third vimarsa of the Vyaktiviveka27 Mahimabhatta examines

in great detail many of Anandavardhana's examples of dhvani,

attempting in each case to specify the complete set of textual

and other factors which allows a reader to infer the unstated

meaning which Anandavardhana believes to be "suggested".

Mahimabhatta's claims for the inclusion of

26Vyaktiviveka, op. cit., pp. 146-149; cf. also pp. 176, 504, and 510-511.

27Ibid., pp. 463ff.

485

Anandavardhana's dhvani within the scope of inference, as they

form the central thesis of his work, have attracted the most

attention from modern scholars of Alamkarasastra. It has been

far less frequently28 noted that he also argues for the

inclusion Of dhvani within the scope of figurative

signification.29 He mocks Anandavardhana's arguments for the

differentiation between dhvani and figurative signification,

incorporating into his own work the verses of the Dhvanyaloka

in which Anandavardhana argues for this distinction but

altering them slightly such that their meaning is inverted and

they present an argument for the identity of dhvani and

figurative signification.30 What Anandavardhana calls dhvani

is, in fact, nothing but figurative usage.

Mahimabhatta's critique of dhvani is thus actually quite

similar to that already offered by Mukulabhatta; both hold

that the process by which we understand unstated meanings in

cases of Anandavardhana's dhvani and the semantic and

contextual conditions required for this understanding are no

different from those found in cases of figurative

signification. The only fundamental difference in their

analyses stems from the old argument between the Mimāmsakas

28Though occasionally; cf. Krishnamoorthy, The Dhvanyaloka and its Critics, op. cit., pp. 286-87 and Rajendran, op. cit., pp. 95-96.

29Cf. Vyaktiviveka, op. cit., pp. 114ff, 174, 178.

30Ibid., pp. 123-4; cf. Dhvanyaloka 1.14 (op. cit., pp. 141-2), 1.16 (p. 146), 1.17 (p. 148), and 1.18 (p. 151).

486

and the logicians (both Buddhist and Naiyayika) over the

existence of arthapatti (necessary presumption) as a means of

valid knowledge distinct from inference.31 For Mukulabhatta,

as for the Mimamsakas, our understanding of figurative usage

depends on a kind of arthapatti: when a word or phrase,

understood literally, would render an expression incoherent or

unintelligible, we are forced to seek for some figurative

interpretation which will make the expression as a whole

intelligible.

However, Mahimabhatta, as we have already seen, follows

the Buddhist logicians in arguing that arthapatti, along with

all other proposed means of knowledge apart from direct

perception, must necessarily, insofar as they are valid means

of knowledge, be subsumed within the domain of inference.

Thus, for him, figurative signification itself, because of its

basis in arthapatti, can be satisfactorily explained only as

a kind of inference.32 Nevertheless, these two critics of

the dhvani-theory are united on several key points -- that

words can convey meanings other than their literal ones only

under certain specifiable conditions, that these conditions

are none other than those already dealt with by theorists of

non-poetic language under the framework of figurative

signification, and that there is consequently no justification

for positing a new mode of expression, vyañjana, in order to

31Cf. above, Chapter Three, p. 91.

32Vyaktiviveka, p. 118.

487

explain the semantics of poetic language -- and it is these

points which form the core of the anti-dhvani position.

Mahimabhatta follows Mukulabhatta in insisting that poetic

language is fundamentally no different from extra-poetic

language and operates according to the same principles.

Indeed he develops this idea much more fully than

Mukulabhatta; as we shall see, Mahimabhatta's detailed

exposition of poetic defects is shaped primarily by his

attempt to rigorously apply within the realm of poetry

standards of coherence and acceptable usage first developed in

the analysis of non-poetic language.

The Essentiality of Rasa-predominance

As already noted, Mahimabhatta, despite his determined

assault on Anandavardhana's theory of poetic semantics, adopts

quite enthusiastically many other major elements of

Anandavardhana's approach to literary analysis, particularly

his focus on rasa as the overall goal toward which all other

elements of a poem ought (at least ideally) to be directed.

For Mahimabhatta, as for Bhattanayaka and Abhinavagupta, rasa

is not merely a highly desirable element to be introduced in

poetry or one generally to be found in the best poems; it is

the necessary and constitutive goal of all poetry. For him,

poetry is poetry only insofar as it effectively conveys rasa.

"That activity of a poet, " he says, "which consists of the

joining together of vibhavas, etc., and which is necessarily

associated with the manifestation of rasa, is called 'poetry'

488

(kavya) ";33 and elsewhere, "The composition of vibhavas, etc.

alone is the activity of the poet, nothing else. "34

Mahimabhatta's understanding of the general nature of

poetic flaws is developed in accordance with this

unambiguously rasa-centered account of the function of poetry.

For Mahimabhatta, a flaw in a poetic composition is any

element which obstructs in any way the development of rasa.

"The common characteristic of these [flaws] is that they

produce some obstacle to the understanding of the intended

[vivaksita] rasa, etc. 35 Thus the unobstructed

communication of rasa to an audience is for him the sole

aesthetic value; the acceptibility of any element of any poem

can be measured only in terms of its capacity to further this

ultimate goal.

This insistence on rasa as the sole raison d'etre of any

poem, as Mahimabhatta recognizes, totally undermines

Ānandavardhana's tripartite classification of poetry.

Anandavardhana, it will be recalled, allows for three

varieties of poetry: dhvani-kavya (in which suggested meaning

is predominant), gunībhūtavyangya-kāvya (in which suggested

meaning is present, but plays a subordinate role), and citra-

33Vyaktiviveka, p. 101.

34Ibid., p. 142.

35Ibid., p. 182; cf. also p. 255. The "etc." refers to other rasa-like elements, e.g bhava, rasabhasa, bhavabhasa; cf. above, Chapter Two, pp. 52-58.

489

kāvya (in which there is no suggested meaning whatsoever) .36

But if, as Mahimabhatta argues, rasa -- a type of "suggested"

(really inferred) meaning -- is necessarily and by definition

the ultimate goal of all poetry, then all poetry whatsoever

will belong to the first of Anandavardhana's three categories.

For this reason, Mahimabhatta attacks Anandavardhana's

description of dhvani as a type (visesa) of poetry.37 It is

not a "type", it is poetry itself; any poetry which does not

belong to this "type" will not, properly speaking, be poetry

at all. Thus it is not possible to classify poetry on the

basis of the presence or absence, or the predominance or

subordination, of "suggested" meaning.

Apart from his insistence that all poetry have rasa as

its ultimate goal (and therefore belong to the first. of

Ānandavardhana's three types), Mahimabhatta finds altogether

objectionable Anandavardhana's division of suggested/inferred

meanings into predominant and subordinate varieties. For one

thing, he contends that there can be no motive for attempting

to draw such a distinction. Unless there is a general

distinction in beauty between verses in which

suggested/inferred meaning is subordinate and those in which

it predominates (and Anandavardhana himself does not claim

that there is one), there is no point in drawing any

categorical distinction between the two. Mahimabhatta

36Cf. above, Chapter Six, pp. 271ff.

37Vyaktiviveka, op. cit., pp. 98-107.

490

compares Anandavardhana's own examples of predominant and

subordinate suggestion of vastu, alamkara, and rasa, and

argues that the latter are no less beautiful than the

former. 38 It is the presence of suggested/inferred meaning,

not its (supposed) predominance or subordination, which is the

crucial determinant of poetic beauty. Thus the distinction

between predominant and subordinate suggestion is not

aesthetically significant and Anandavardhana's categorical

division of dhvani from gunībhūtavyangya-kavya consequently

serves no useful function.

Beyond this, however, Mahimabhatta argues that, in fact,

unstated meanings are never subordinate to the expressed

meanings from which they are inferred.39 When an expressed

meaning causes us to understand a further, unstated meaning,

it is serving as a means to an end and is therefore

necessarily and by definition subordinate to this end. In

such cases, the expressed meaning acts as an inferential mark

(linga) and, like any other such mark, is important not for

its own sake but for the further information (i.e. the

"suggested" meaning) it allows us to deduce.

This is true even in cases where Anandavardhana claims

that the directly expressed meaning of a verse is predominant:

That predominance which is said to belong to this [expressed meaning] in some cases, e.g. in [alamkāras] such as samasokti, can exist only with respect to its

38Ibid., pp. 161-168.

39Ibid., pp. 9-14.

491

contextuality [präkaranikatva] not with respect to the understood meaning [to which it gives rise]. example: "The moon, becoming red/passionate, touched the For

face of the night, with its twinkling stars/trembling eyes, so that her [the night's] cloak of darkness was condensed/bunched up [samasta] and fell away in the east/in front unnoticed by her, because of her own redness/passion." -- Here [according to Anandavardhana] the expressed meaning alone is predominant and is only accompanied by the understood meaning, "since night and the moon alone, on which the behavior of female and male protagonists have been superimposed, are the meaning of the sentence." But, because this [expressed meaning] is an inferential mark with respect to that [understood meaning], there is no departure from subordination.40

In calling the expressed meaning predominant in cases such as

these, Anandavardhana is really confusing two different kinds

of predominance. In the verse quoted the directly expressed

meaning is the contextually relevant one -- the actual

referents of the verse are the night and the moon, which the

poet only implicitly identifies with a pair of lovers -- and

may, in this sense, be called "predominant". But

nevertheless, the point of the verse is precisely to

communicate this second, implicit meaning (without which it

would nothing but a banal account of an everyday event). The

explicit description of the night and the moon is crafted in

such a way as to allow us to infer the identification of these

two with a pair of lovers; it thus serves as a means to this

end, i.e. as an inferential mark, and is therefore, despite

its immediate contextual relevance, subordinate to the

unstated meaning we infer from it.

40Tbid., p. 11. Cf. also Dhvanyaloka, Kashi ed., pp. 109- 11 and above, Chapter Four, pp. 160-161.

492

The inferable meaning a verse or a passage conveys,

whatever its relation to the narrative context or the literal

referentiality of the work in question, will always be

aesthetically predominant. It is the presence of inferable

meaning generally,41 and of rasa in particular, which is the

fundamental basis of poetic beauty -- indeed, which is itself

constitutive of poetry. It is thus never possible to speak of

such meaning as subordinate to the directly expressed meaning

which gives rise to it. The communication of inferable

meaning, especially rasa, is the whole point of poetry, and

such meaning is thus necessarily predominant.

We can see that there develops among the Alamkārikas of

tenth and eleventh century Kashmir a general consensus

regarding the essentiality of rasa-directedness in all

poetry. 42 It is a consensus which cuts across the lines of

the major aesthetic controversy of the period: Bhattanayaka

and Mahimabhatta, two of the most outspoken critics of

Anandavardhana's dhvani-theory, and Abhinavagupta, its most

celebrated defender, are at one in this matter. In defending

his teleological approach to literary analysis, Ānandavardhana

41Tbid., p. 160. Like Abhinavagupta (cf. above, Chapter Nine, pp. 454-455), Mahimabhatta allows the "suggestion" of vastus and alamkaras a major role in poetry, but only insofar as it is understood that they must always culminate in the communication of rasa (by themselves serving as vibhāvas, etc.) -- cf. Vyaktiviveka, p. 104. For both, it is ultimately always rasa and rasa alone that is primary.

42The one noteworthy exception being Kuntaka, for whom rasa is only one of several possible and equally valid aesthetic objectives -- cf. above, Chapter Eight, pp. 423-426.

493

tried to show that the best of existing poetry is, for the

most part, rasa-directed but, mindful of the diversity

prevailing in literary practice, allowed for the existence of

poetry in which suggested meaning (including rasa) is

subordinate or even wholly absent. These later authors make

no such allowance; for them all poetry must, by definition,

have the communication of rasa as its ultimate objective.

They are, in the end, more adamant in their defense of

Anandavardhana's rasa-centered poetic teleology than was

Anandavardhana himself.

Mahimabhatta's Analysis of Poetic Flaws

The second vimarsa of the Vyaktiviveka appears to be

nothing but an immense digression from the stated objective of

his work, namely the proof that what Anandavardhana calls

suggestion is nothing other than inference. The first vimarsa

offers a detailed critique of Anandavardhana's definition of

dhvani, and of the notion of suggestion generally,

demonstrating that dhvani can be explained only as a kind of

inference. The third gives a detailed examination of

Anandavardhana's examples of dhvani, showing in each case the

inferential process by which the allegedly "suggested" meaning

is understood. The second vimarsa (which comprises more than

half of the entire Vyaktiviveka) consists of a detailed and

sophisticated analysis of poetic flaws. The matters discussed

therein have, for the most part, no obvious bearing on

Mahimabhatta's critique of dhvani. It is almost as if a

494

second, unrelated work were inserted into the middle of his

attack on Anandavardhana. Mahimabhatta himself acknowledges

the digressive character of this section of his work; at the

conclusion of the second vimarsa he says, apparently referring

to the entire discussion of poetic flaws, "enough of this

extensive treatment of matters unrelated to the topic at

hand. "43

Whatever Mahimabhatta's motive may have been in including

this elaborate discussion of poetic flaws in work avowedly

devoted to the destruction of the Anandavardhana's dhvani-

theory, his analysis of these flaws is connected in important

ways with his view of the nature and function of poetry as

developed in the other portions of his work, and is worthy of

far more serious consideration than it has generally received.

As already noted, Mahimabhatta's understanding of the

43"alam aprastutavastuvistarena"; ibid., p. 262. Later interpreters of the Vyaktiviveka have, not surprisingly, had a hard time trying to discern the purpose of this enormous digression. After completing his analysis of aesthetic flaws, Mahimabhatta briefly (pp. 456-461) examines the first verse of Anandavardhana's Dhvanlyaloka and finds there several of the flaws he has just discussed; it has often been suggested, though Mahimabhatta gives no indication to this effect, that this critique of Anandavardhana's verse is itself the sole raison d'etre of the second vimarsa. Ruyyaka says as much in his opening commentary on the second vimarsa, and, presumably following his lead, several modern scholars of Alamkārasāstra have taken it as fact that this is the sole object of the

as second vimarsa (which they therefore rather cavalierly dismiss of minimal significance -- Dhvanyaloka and its Critics, Krishnamoorthy, op. cit., p. 271 and Gerow, cf. The

Indian Poetics, op. cit., p. 269). The argument is not really tenable, as only a few of the many types of aesthetic flaw dealt with by Mahimabhatta are found in Anandavardhana's verse; the vast preponderance of what is discussed in the second vimarsa has no bearing on the critique of this verse.

495

general nature of poetic flaws is tied to his claim that rasa

must be the ultimate goal of all poetry. Since the object of

all poetic language is to communicate rasa as effectively as

possible, any element of a poem which limits or impairs, in

any way, the poem's ability to convey rasa will be regarded as

a flaw to be avoided by the poet. Mahimabhatta divides these

flaws into two general classes: those which are intrinsic

(antaranga) and those which are extrinsic (bahiranga) .4 The

first are factors which directly impair the intended rasa,

e.g. the mention of vibhävas or other aesthetic factors which

contradict this rasa. Mahimabhatta states that flaws of this

type have been dealt with by previous authors, 45 and

therefore declines to examine them in detail. To the second

class belong those grammatical or semantic features which

obscure the poet's intended meaning and thereby indirectly

impede the communication of rasa, and it is these which

Mahimabhatta analyzes in detail.

Mahimabhatta divides these "extrinsic" flaws into five

major classes: vidheyavimarsa ("non-consideration of what is

predicated"), prakrama-bheda ("breaking of an initial

pattern"), krama-bheda ("breaking of sequence"), paunaruktya

("redundancy"), and vacyavacana ("not stating what should be

44Ibid., pp. 179-182.

45Apparently referring particularly the to discussion of rasa-virodhins to Anandavardhana himself, (factors contradictory to rasa) in Dhvanyaloka 3.17-28 (Kashi ed., pp. 360-397). Again we see that in many areas Anandavardhana's views are treated as authoritative even by his critics.

496

stated"). The first, vidheyavimarsa, occurs when a poet fails

to make clear the subject/predicate relation (vidhi/anuvāda

bhava46) obtaining within a verse. Mahimabhatta discusses

three varieties of vidheyavimarsa, all of which he finds to be

exemplified in a single verse:

That excitement which every lion has with respect to a wretched elephant or a piece of cloud is only a bit of playfulness determined by its nature. So, where can one who is non-excited even in the presence of the elephants of the directions47 and the massing of the doomsday clouds, the lion of [the goddess] Ambika,48 find a high degree of exhilaration?49

The first type of vidheyavimarsa Mahimabhatta deals with

involves the inappropriate use of term negation in preference

to sentential negation or vice versa.50 In the above verse

the use of negative compound "non-excited" (a-samrabdhavān) is

inappropriate. The denial of any excitement on the lion's

part ought to be the predicate (vidhi) of the entire clause;

46Cf. above, Chapter Three, pp. 86ff.

47The divine elephants who are thought to hold up the sky,

directions. one of which is positioned at each of the four cardinal

48I.e. that lion who acts as the vehicle (vähana) of the goddess.

49Ibid., p. 184. samrambhah karikītameghasakaloddesena simhasya yah sarvasyaiva sa jātimātraniyato hevākalesah kila/ ityāsādviradaksayambudaghatābandhe 'pyasamrabdhavān yo 'sau kutra camatkrter atisayam yātvambikākesarī//

The authorship of the verse is unknown, but it is also cited by Kuntaka (Vakroktijīvita, Krishnamoorthy's ed., p. 16) .

50Ibid., pp. 185-195.

497

the excitedness itself is only the subject (anuvada) whose

absence is predicated.51 But here the negation is relegated

to a subordinate role, serving as a qualifying element within

a nominal compound. What the clause as stated predicates is

not the absence of excitedness but rather the presence of non-

excitedness. The poet ought to have said "not excited" (na

samrabdhavan) -- when the negation is expressed by a separate

word, rather than a subordinate element of a compound, it can

be taken as the predicate of the clause, as the poet

apparently meant it to be taken.

Obviously Mahimabhatta doesn't mean to claim that the

poet's intended meaning is unintelligible. Here, and

elsewhere in his analysis of flaws, his method rests on the

assumption that it is possible to see what a poet meant to say

even when the poet has expressed himself poorly. Context and

other factors will allow us, on reflection, to understand the

intended meaning. Mahimabhatta's "flaws" do not create an

absolute barrier to understanding; they constitute an obstacle

to the process of comprehension (and thereby indirectly

detract from rasa), but this obstacle is not insurmountable.

If it were, aesthetic criticism of the kind Mahimabhatta is

practicing here would not be possible.

The second type of vidheyavimarsa stems from the failure

to properly connect relative and correlative clauses.52

51Tbid., p. 188.

52Ibid., pp. 196ff.

498

According to Sanskrit grammatical theory the relative pronoun

yat ("who, which") and the correlative pronoun tat ("that")

are inevitably associated with one another; whenever one of

these words is used it creates an expectation for the other

which must be satisfied if the expression in which it occurs

is to be satisfactorily understood. Again, this is a question

of subject/predicate relations.

"When one begins by the use of either yat or tat, it is appropriate to conclude by use of the other, which will refer back to that, since these words as well are desired to have subjective [anuvādya] and predicative [vidheya] senses [respectively] and since between these two there is always a mutual expectancy. "53

Therefore every correlative pronoun must be connected with a

relative pronoun (either expressed or implied), and vice

versa.

In the verse quoted above, the relative pronoun (" ... who

is non-excited ... ") is used without a corresponding

correlative.54 After a careful analysis of the conditions

under which either yat or tat can be implied rather than used

explicitly (pp. 196-204), Mahimabhatta concludes that such

53Ibid., p. 196.

5*It might be argued that the demonstrative asau ("the" or "that") Mahimabhatta argues (pp. 206-209) that this view is untenable; here takes the place of the correlative.

asau and tat cannot be seen as synonymous, and therefore the former cannot, like the latter, satisfy the expectation generated by the use of the word yat. syntactic If they were truly synonymous, then asau would likewise have to be connected with a relative whenever it is used, and the words asau and tat could never be used together in the same clause (as they would be redundant) -- it can be shown by numerous examples that neither of these conditions holds in practice.

499

conditions to not hold here. Therefore the poet's failure to

use a form of tat in the clause, "So where can ... the lion of

[the goddess] Ambika find a high degree of exhilaration,"

constitutes an aesthetic flaw. The correlation between the

two clauses is not clearly expressed, and the vidhi/anuvada

bhava between them is therefore obscured.

The third type of vidheyavimarsa involves the

inappropriate use of compounds. As a general rule,

Mahimabhatta argues, elements of an expression which are to be

predicated ought not to be used as subordinate parts of a

nominal or adverbial compound.

All these compounds contain two elements, a qualifier [visesana] and a qualificand [visesya]; however, when the qualifying portion, because it produces aesthetic appeal in the sentence meaning by lending some special quality to its referent [asraya], is intended as primary, and [thereby] attains the status of predicate, and the other [i.e. the qualificand] is subordinated becoming, as it were, the subject of predication, then it is not an appropriate case for the use of a compound. For, in a compound, the relation of primacy and subordination between these two is eclipsed.55

The incorporation of qualifier and qualificand in a compound

obscures the primary/subordinate relation between these two,

and the vidhi/anuvada bhava which depends upon it. The

obscuring of this relation ultimately spoils the ability of

the expression to convey its intended rasa, and it is for this

reason that it is to be considered a poetic flaw.56 Having

articulated this general principle, Mahimabhatta proceeds to

55Ibid., p. 229; cf. also pp. 264-265.

56Ibid., p. 255.

500

give an exhaustive analysis of its application, examining all

types of compound allowed for by Sanskrit grammar, citing

numerous examples from existing literature in which poets have

used these compounds inappropriately and juxtaposing them with

counterexamples in which poets have appropriately refrained

from the use of compounds where the clear expression of

vidhi/anuvāda bhāva demands it.57

In the verse quoted above, the phrase "the lion of [the

goddess] Ambika" is expressed by a single compound (ambika-

kesari). By the standard just articulated, the joining of

ambika and kesari in a compound is a defect in the verse.58

The lion's association with Ambika is not merely an

indicidental identifying mark; it is precisely this divine

association that sets this lion above all others and accounts

for its failure to be moved even by extraordinary sources of

excitement. Thus the word "Ambika" is predicative and is of

primary importance. Compounding it with the word "lion" masks

the intended vidhi/anuvada bhava between the two words and

thus constitutes an aesthetic flaw.59 Mahimabhatta composes

a revised version of the same verse which eliminates all three

of these types of vidheyāvimarsa.60

The second class of aesthetic flaw Mahimabhatta deals

57Ibid., pp. 230-276.

58Tbid., pp. 277ff.

59Ibid., p. 280.

60Ibid., p. 282.

501

with he calls prakrama-bheda ("breaking of an initial

pattern") .61 Like vidheyavimarsa, this flaw too consists of

the damaging of subject/predicate relations within an

expression and culminates in the impairment of rasa.62 It

occurs when a poet begins by using a particular lexical item

or grammatical construction and, later in the same verse or

passage, employs a different lexical item or grammatical

construction to refer to the same entity or serve the same

syntactic role. Such deviation produces a kind of "stumbling"

(pariskhalana) in the hearer's process of comprehension.

Pattern breaking in lexical choice is illustrated as follows:

"I have promised that I will never speak to you again if I do not bring you [to see your lover] ... 63 because of my no longer talking with you, my strength And,

will vanish and I will no longer be able to sustain my life. #64

Here, although "speak" [bhasati] and "talk" [lapati] both, distinction, conversing, when this thing is initially expressed by use without have the meaning of

of the word "speak", then it is appropriate to continue by using the very same [word], and no other. accepted that this kind of verbal appropriateness, called For it is

"not breaking an initial pattern", consists of a sort of

61Ibid., pp. 287-322.

62Ibid., pp. 287, 310-311.

63The speaker, acting as messenger for a friend's rejected lover, is trying to convince her friend to give up her anger and return to him.

64satatam anabhibhāsanam mayā te paripanitam bhavatīm anānayantyā/ ...

gatadhrtir avalambitum batāsūn analam analapanad aham bhavatyah/ (Magha, Sisupalavadha 7.9ab and 7.10ab)

502

vidhi/anuvāda bhāva. 65

The second of the two sentences provides further information

related to the same non-speaking mentioned in the first. The

use of a different word for speaking in the second sentence

obscures, if only very briefly, the intended identity between

the two. Even though there is ultimately no difference in

meaning between "speaking" and "talking", the terminological

shift produces the impression of a distinction where none is

intended. Such lexical substitutions are therefore to be

avoided.

The same principle applies in the selection of

grammatical forms. For example:

One woman, scattered with dust, with the beauty of her moonlike face obscured, resembled the sky; others, their beauty gone and their hearts agitated, burned within like the surrounding regions [of the earth]; some, like the winds, wandered in all directions; some [were] trembling like the earth; in this way, when the kings set out, the

follow.66 women revealed in advance that ill-fortune which was to

65Ibid., p. 288.

66 Tbid., pp. 299-300 kācit kīrna rajobhir divam anuvidadhe bhinnavaktrendulaksmīr aśrīkāh kascid antar diśa iva dadhire dåham udbhrantasattvah/ bhremur vatya ivānyāh pratipadam aparā bhūmivat kampamānāh

śaśamsuh/ / prasthāne pārthivanām asivam iti puro bhāvi nāryah

(Māgha, Śisupālavadha 15.102)

The verse describes the distress of the soldiers' wives as the army of Sisupala and his allies sets out to do battle with Krsna. The darkening of the sky, the appearance of forest-fires in surrounding areas, the blowing of strong winds, and the trembling of the earth are all omens of ill- fortune; the actions of these women resemble these omens, and

503

Here, the actions of the first three groups of women are

expressed by finite verbs in the perfect tense; that of the

fourth, those who were "trembling like the earth", is

expressed by the present participle kampamana. The use of the

participial form rather than a perfect tense verb breaks the

pattern established by the first three clauses. The shift to

a different grammatical form gives the appearance of

semantic distinction even though none is intended by the

poet.67 Semantic parallelism demands morphological

parallelism. Similar examples are adduced to illustrate the

importance of maintaining established patterns in the use of

pronouns, case, gender, number, etc.

Mahimabhatta's third aesthetic flaw is krama-bheda

("breaking of sequence").8 This flaw is not, like the first

two, linked with the Mimamsa concept of vidhi/anuvada bhava.

It does, however, hinge on the strict application of extra-

poetic restrictions on language use within the domain of

poetry. Since in Sanskrit so much grammatical information is

conveyed by inflection, word order tends to be relatively

free. There are however many words and grammatical

thus they themselves foretell disaster for the departing army.

67Where such a distiction is intended, a shift of grammatical form will not constitute a flaw -- indeed it is a positive requirement. For instance, when one of a series of stated actions is primary and the others are subordinate, the primary action should be expressed by a finite verb and the others by participial or other Vyaktiviveka, pp. 269-272. nominal forms; cf.

68Ibid., pp. 323ff.

504

constructions which demand a restricted word order. For

example, many indeclinable particles must be used immediately

adjacent to the words they qualify. As Mahimabhatta explains:

[Particles] such as ca ["and"] are modifiers ... which must not be separated from the words they qualify. Thus these [particles] are able to make a qualification only of those words to which they are immediately proximate. The occasional use of these that is made out of sequence is unacceptable. For, when these words are inappropriately positioned, they exercise their influence on some other, unintended meaning, and the confusion of the contextually appropriate meaning follows from this. Or, if, even though they are out of sequence, a connection with their intended meanings is deduced, then the comprehension of the contextually appropriate meaning will be obstructed, and consequently the savoring of rasa which depends on this will likewise be obstructed.69

Poets, largely due to metrical considerations, often

overlook restrictions of this kind in composing their verses,

and it is such lapses that Mahimabhatta criticizes under the

heading of krama-bheda. For example,

[Two things have now become lamentable by seeking union with the skull-bearer (i.e. Śiva) :] Both that splendid digit of the moon, and you [i.e. Parvati], a lotus to the eyes of this world. 70

Here the word "and" (ca) should imediately follow the word

"you" (tvam), as it correlates this with the "digit" mentioned

earlier; however, the poet has placed it after the word

69Ibid., p. 156. first vimarsa, but These remarks are made earlier, in the they allude to Mahimabhatta's later discussion of krama-bheda.

70Ibid., p. 328. [dvayam socanIyatām kapālinah/] gatam samprati samagamaprarthanaya

kala ca sā kāntimatī kalāvatas tvam asya lokasya ca netrakaumudi// (Kālidāsa, Kumārasambhava 5.71)

505

"world" (lokasya), thus violating the sequential restriction

on the use of ca and producing, at least momentarily, a

meaning which is unintended.71 So, despite its frequent

occurrence in actual practice, the loosening or ignoring of

the ordinary rules of word order in the context of poetry

cannot be regarded as acceptable. Violations of sequential

rules in poetry produce confusion, just as they would in

everyday usage, and consequently impair a poem's capacity to

accomplish its ultimate purpose, the conveying of rasa.

The fourth type of aesthetic flaw examined with by

Mahimabhatta is paunaruktya ("redundancy") .72 Mahimabhatta

is certainly not the first literary theorist to deal with the

problem of redundancy in poetry. Several of the early

Ālamkārikas (e.g. Bhāmaha 4.12-16, Dandin 3.135-138, Vāmana

2.2.11-19) mention a poetic defect they call ekärtha ("having

a single meaning"), in which a word or phrase expresses a

meaning already stated and hence serves no useful purpose.

What distinguishes Mahimabhatta from the earlier Alamkāra

tradition in this matter is not the insistence that poets must

avoid redundant expressions, but rather the tremendous rigor

with which he applies this critical principle. Redundancy,

for him, is not limited to mere verbal repetition or the

71Most of the types of krama-bheda similarly involve the misplacement of indeclinable particles, but other sequential rules are dealt with as well; e.g, one should not refer to something by a demonstrative pronoun without having first referred to it by a noun. Cf. ibid, pp. 323-326.

72Ibid., pp. 333-387.

506

juxtaposition of synonymous expressions: it applies to all

cases in which a poet states explicitly what could otherwise

have been understood implicitly.

Mahimabhatta, again citing plentiful examples from

existing literary works, demonstrates numerous situations in

which the use of particular stems, suffixes, words, or even

whole sentences adds nothing at all to the meanings of the

verses or passages in which they occur. Many common

expressions are, by this standard, redundant. E.g. one should

never call a person "one who has a confused mind"

(mudhacetana, mudhadhi), as one can simply call him or her

"confused" (mudha) -- it goes without saying that confusion is

always a property of the mind and nothing else;33 one should

not speak of a "feeling of suffering" (duhkhanubhava) -- it is

understood that suffering is, by its very nature, a

feeling.74

By the rigorous application of this kind of analysis,

Mahimabhatta is able to show that certain passages consist

almost entirely of such semantically useless filler. For

example:

[who] cuts the necks of his evil enemies with the extremely fierce blows with the front edge of the drawn, sharpened sword [he] held in [his] hand. "75

73Ibid., p. 373.

"4Ibid., p. 375.

75karakalitanisātotkhātakhadgāgradhārā- drdhataravinipatacchinnadustarikanthah/

507

Here nine words are worthless. The four qualifers of "sword" [i.e. "hand" (kara), "held" (kalita), "sharpened" (nisata), and "drawn" (utkhata) ] are redundant, since they are by nature invariably associated with this [i.e. a sword] whenever it is used. And the mention of the "edge" [dhārā] qualified by "front" [agra], is redundant, since whenever the sword is meant as an instrument, the understanding of this ["front edge"] will be arrived at merely through appropriateness. "blows" [vinipāta], qualified by And the "extremely fierce" [drdhatara], which are mentioned here as the means [of cuttingl, and the evilness of the enemies as well, are redundant, since they are understood by implication. Therefore, " ... [who] cuts the necks of enemies with a sword" [khadga-chinna-ari-kantha] -- these four words alone are the essence here; everything else is largely worthless and culminates only in the filling out the meter, not in any distinction of meaning.76

The use of any word or words whose omission would make no

difference in the final meaning understood from a verse is

redundant and hence flawed.

Kashmiri Ālamkārikas since Ānandavardhana's time, both

pro- and anti-dhvani, had generally emphasized the central

significance Of unstated meaning in great poetry.

Mahimabhatta erects this general preference for unstated

meaning into an absolute rule: wherever a meaning can be

understood implicitly it is redundant to state it outright.

This rule applies even in the selection of alamkāras; one

figure should be chosen over another if it implies more and

says less explicitly. Where possible, one should employ the

figure rupaka ("metaphorical identification") rather than

upama ("simile") ; the former figure implies similitude between

76Tbid., p. 377.

508

the things identified, without stating it outright.7 By the

same token, the figure samasokti ("condensed expression") 78

is to be preferred to either upamā or rūpaka.79 In samāsokti

even the standard of comparison (upamana) is not stated

explicitly; the adjectives or other qualifiers applied to one

thing are crafted so as to imply its similarity to another.

In constructing alamkaras, as in all other aspects of

composition, the poet must strive to maximize the implicit

content of his work.

The fifth and last type of poetic flaw examined by

Mahimabhatta is what he calls vacya-avacana ("not saying what

should be said") .80 As its name would suggest, this is

something of a generic, catch-all category -- one which would

seem to overlap, at least partly, with the previous four.

Under the heading of väcya-avacana Mahimabhatta deals with a

series of disparate topics too numerous to discuss here

individually.81 However, one particular variety of vacya-

avacana -- the one to which Mahimabhatta himself devotes the

most attention -- is of particular interest, as it leads into

77Ibid., pp. 350ff.

78Cf. above, Chapter Four, pp. 160-161.

"9Ibid., pp. 358ff

80Ibid., pp. 387-455. According to Mahimabhatta, this class of flaws includes, by implication, examples of avācya- vacana ("saying what should not be said") -- cf. pp. 436ff.

81E.g. failure to use pronouns where appropriate (pp. 387ff), errors in choice of alamkaras (pp. 391ff), improper formulation of similes (pp. 426ff).

509

a specific critique of Ānandavardhana's sabdasaktimūla dhvani.

Punning and the Expressive Power of Words

A substantial portion of Mahimabhatta's discussion of

vacya-avacana is devoted to an examination of the poetic flaws

produced by improperly constructed puns.82 The mere presence

of a series of sounds capable of bearing more than one

possible meaning is not sufficient to create a viable pun

(slesa); there are further conditions which must be met if a

single set of sounds is to effectively convey multiple

meanings within a single sentence. Again, the argument rests

fundamentally on the assumption that poetic language must play

by the same rules as non-poetic language; poetry, even

paronomastic poetry, is not exempt from the ordinary

requirements sentential coherence.

As an example of an attempted pun which fails to achieve

such coherence, Mahimabhatta offers the following:

Even a little bit of misfortune [vipal-lavo 'pi], ceaselessly spinkled with the water of tears/sprinkled with water ceaselessly brought, grows a thousandfold, like a tree even without a creeper [vi-pallavo 'pi] .83

Here the poet has attempted to use a single set of sounds

82Tbid., pp. 399-425. This passage is not, as Chakravorty (op. cit., pp. 132-133) mistakenly asserts, a philosophically motivated attack on the use of puns in general. rather, an examination of the conditions which must apply if It is,

the multiple meanings expressed by a set of words used punningly are to be integrated into a single, coherent sentence-meaning.

83Ibid., p. 406. anavaratanayanasalilasicyamanas tarur iva vipallavo 'pi sahasradā prarohati.

511

radiant lap."86

In this example the two major terms of the comparison, the

mountain slope and Sita, are each expressed by a term which is

not itself used punningly. The word "like" (iva) is the

nibandhana, the verbal cue which leads to the understanding of

the double meaning. It leads us to construe the compound

bhasvadankollapallavataya as a qualifier of both the mountain-

slope and Sita, and this construal will be intelligible only

if we interpret the compound differently in connection with

each qualificand: the slope has "ankolla-shoots" (añkolla-

pallava) which are "shining" (bhasvad), while Sita has in her

"shining lap" (bhasvad-anka) her infant son the "crying"

(ullapat) Lava (lava) .87 Thus the proper understanding of

the simile, expressed by the word iva, requires that

bhasvadankolla-pallavataya be interpreted paronomastically.

Syntactically, however, this pun-based simile is constructed

like any other simile: two elements are shown to resemble one

another through their association with a single qualifier.

Thus the verse, despite its use of paronomasia, forms a

single, coherent proposition. Furthermore, there is no

question of "suggestion" operating in such a verse; a single

86Tbid., p. 407. sänusthitir janakarājasūteva bhāsvadankollapallavatayā śriyam eti yasya// (Ratnākara, Haravijaya 4.23cd)

87The rules of Sanskrit euphony (samdhi) account for the sound changes which occur when these words are pronounced in sequence -- anka + ullapat + lava becomes ankollapal-lava.

512

qualifier, connected with two different qualificands, directly

expresses two different meanings.

The need to maintain sentential unity imposes two major

requirements on all puns: the word or words which are to

convey a double meaning must be "anchored" by syntactic

connection with one or more terms which are not themselves

used punningly, and there must be some nibandhana, some

explicit verbal cue which makes a punning interpretation

necessary. Where these conditions are not met, double

entendre will not be possible. It is Mahimabhatta's

insistence on these two features, particularly the latter, as

necessary preconditions for all double entendre which leads

him to deny absolutely the possibility of what Anandavardhana

calls śabdaśaktimūla dhvani.

Anandavardhana and Mahimabhatta are largely in agreement

regarding the nature of the figure slesa -- both argue that

wherever this figure occurs there will be some verbal cue

which prompts the search for a double meaning, and that both

meanings conveyed by such an expression will be directly

expressed. It is regarding the possibility of double entendre

in the absence of such a verbal cue that they differ. For

Anandvardhana, a set of sounds capable of bearing more than

one meaning, used without such an explicit verbal cue, once

they have expressed a single, contextually relevant meaning,

will suggest, rather than directly express, a second meaning --

513

a case of sabdasaktimula dhvani;88 for Mahimabhatta, there

will be only a failed attempt to compose the figure ślesa.

Mahimabhatta attempts to show that, by the standards he

has articulated in his analysis of slesa, examples of what

Anandavardhana calls sabdasaktimula dhvani cannot effectively

and coherently convey a double meaning at all. He considers

the following verse, one of Anandavardhana's examples of this

type of dhvani:

Giving joy to creatures/to their offspring by their water/milk released without difficulty at appropriate times, spreading out in all directions at the beginning of the day and withdrawing/coming together when the day ends -- may the rays/cows of the bright-rayed sun, [which arel ships for crossing the ocean of fear of existence arising from long suffering, produce limitless joy for you pure ones. 89

The verse is understood to refer explicitly to the rays of the

sun (which are thought to extract water from the earth through

evaporation, later to return it in the form of rain). But the

word for "rays" used here (gavah) also means "cows", and the

terms used to describe the rays, differently interpreted,

could also be applied to cows (who give milk to their

offspring, and who disperse in the morning and return home at

night). According to Anandavardhana, even though there is

88Cf. above, Chapter Four, pp. 170-172.

89Ibid., pp. 425, 478ff. Cf. Dhvanyaloka, op. cit., p. 244. dattānandāh prajānām samucitasamayāklistasrstaih payobhih pūrvāhne viprakīrnā disi disi viramaty ahni samhārabhājah/ diptamsor dīrghaduhkhaprabhavabhavabhayodanvaduttāranāvo gavo vah pāvananām param aparimitām prītim utpādayantu// (Mayūra, Sūryaśataka 9)

514

nothing in the explicit meaning of the verse which requires a

punning interpretation, the use of the word gavah and and of

descriptive terms which may apply to cows as well as rays

creates a suggested simile between the rays of the sun and a

group of cows

Mahimabhatta argues, however, that the verse, as the poet

has constructed it, cannot convey this intended double

entendre. It violates both of the requirements already

discussed: the two terms of the punning comparison (the rays

and the cows) are themselves meant to be expressed by a single

term (gavah), itself used punningly, and there is no explicit

indicator of comparison (such as the word "like") which would

cause the hearer to construe the stated qualifiers with both

referents and thereby require a punning interpretation. The

explicit meaning of the verse (referring to the sun's rays)

makes sense in itself, and there is nothing which calls for a

second, unstated meaning (relating to cows), and, in any case,

nothing to connect this supposed second meaning with the

first, explicit, one. This is not to say that one cannot see

that the poet meant to convey a double meaning in this verse --

Mahimabhatta's critique, here as elsewhere in his discussion

of poetic flaws, presupposes that one can discern the poet's

intention even where it has not been effectively carried out.

One can certainly see that the poet has carefully selected a

set of words which could refer to both rays and cows, but, as

the verse has been constructed, there is no way both meanings

515

of this set of words can be combined in a single, coherent

propositional structure -- and hence, no possibility of

ekavākyata (cf. above, Chapter Three, p. 76). The verse as a

whole could convey either of two possible meanings, but cannot

convey both simultaneously; it must be seen either as a

statement about rays or as a statement about cows, but not as

a simile comparing one with the other. By failing to

designate the two terms of the intended comparison by separate

words, and by failing to supply an explicit indicator of the

comparison, the poet has made it impossible for his verse to

effectively convey its intended double meaning.90

Other cases of Anandavardhana's sabdasaktimula dhvani are

attacked on similar grounds.91 In a few of Anandavardhana's

examples of sabdasaktimula dhvani Mahimabhatta finds that

there is, in fact, an explicit verbal cue, a nibandhana, which

allows for a double entendre. These can be accepteded as true

cases of the figure slesa (not of dhvani) .92 But, where no

such nibandhana can be found, no double entendre is possible;

if the conditions necessary for creating the figure slesa are

not met, there can be no double meaning at all. Thus, there

is no scope at all for Ānandavardhana's sabdasaktimūla dhvani,

a mode of double entendre which supposedly operates only in

90Ibid., p. 425.

91Ibid., pp. 478ff.

92Tbid., p. 401 (cf. Dhvanyaloka, op. cit., p. 241) and p. 420 (cf. Dhvanyaloka, p. 245) .

516

the absence of such conditions.

This is one of the comparatively rare instances in which

the theoretical dispute between Anandavardhana and his critics

results in difference of opinion rega a the

interpretation of actual verses. The arguments between pro-

and anti-dhvani theorists are almost wholly concerned with the

process, rather than the outcome, of verbal cognition. There

is, in general, widespread agreement regarding the unstated

meanings ultimately conveyed by particular verses; the

question around which controversy revolves is, How are these

meanings conveyed (by suggestion, by figurative expression, by

inference, etc.)? Here, however, Mahimabhatta is challenging

not merely Anandavardhana's explanation of the process by

which the verses given as examples of sabdasaktimūla dhvani

convey their meanings, but his account of what these meanings

are. For Mahimabhatta, these verses simply do not convey the

double meanings Anandavardhana believes them to convey

(although the poets who composed them evidently meant that

they should do so) .

Mahimabhatta's attack on sabdasaktimūla dhvani ties into

a broader argument he makes regarding the nature of verbal

expression. In opposition to Anandavardhana and others who

would argue that implied or unstated meanings are conveyed by

one or more "powers" (saktis) of words apart from their

capacity of direct expression, Mahimabhatta contends that

words have one and only one expressive power -- that of

517

denotation or direct expression (abhidha). If words had the

capacity to convey figurative and "suggested" meanings, in

addition to their own denoted meanings, there would be no way

to explain the sequence in which these meanings are invariably

arrived at; one always comes to understand the implied or

"suggested" meaning of an expression only after one has

understood its directly expressed meaning. If direct

expression, figurative expression and suggestion were all

functions of words, then all would operate simultaneously as

soon as the words themselves were heard -- there would be

nothing to account for the delayed operation of the latter

two.93 Because the understanding of unstated meanings always

follows upon the understanding of directly expressed meanings,

it must be these directly expressed meanings, rather than the

words which convey them, which produce an awareness of these

additional, unstated meanings. And, for reasons already

discussed (above, pp. 476-478), directly expressed meanings

may lead to the understanding of further, unstated meanings

only through a process of inference. So, word and meaning

each have one and only one expressive capacity: "Word has one

expressive capacity -- denotation [abhidha]; meaning too has

only one -- its acting as an inferential mark [linga] . " 94

Words can express only those individual, denoted meanings with

93Just as fire, which has the capacity both to burn and to shed light, does both simultaneously -- cf. Vyaktiviveka, op. cit., pp. 113-114.

94Ibid., p. 111.

518

which they are conventionally associated. Thus all talk of

multiple "verbal functions" (sabdavyaparas) is fundamentally

misguided.95

Conclusion

Despite his general adherence to the views of the

Buddhist logician DharmakIrti, Mahimabhatta's understanding of

the process of linguistic comprehension and of the

requirements of textual coherence closely mirrors that of the

Mīmamsakas in many respects. For him, as for them, words can,

by themselves, convey only the individual meanings invariably

associated with them; it is these individual word meanings,

considered in relation to one another and to the context in

which they are employed, which, by a subsequent mental

process, produce an awareness of any further meanings to be

understood from a text. Textual coherence consists in the

direct or indirect subordination of all elements of a text to

a single ultimate goal -- in the case of poetry, this goal must

always be the effective communication of rasa. The elements

of such a functionally unified text are connected to one

another through subject/predicate relations (vidhi/anuvāda

bhava), and authors must therefore strive to convey these

relations with maximal clarity. The propositional unity of

sentences (ekavākyatā) is condition of coherence in poetic

texts, as in non-poetic ones.

95Mahimabhatta reiterates the point tirelessly -- cf. ibid., pp. 16, 26, 140-141, 178, 478.

519

So, while Mahimabhatta challenges one central element of

Anandavardhana's poetics -- his insistence that dhvani cannot

be accounted for by existing semantic theory and requires the

acceptance of "suggestion" as a previously unrecognized,

specifically poetic, mode of verbal expression -- in other

respects, he consolidates and further extends Anandavardhana's

theoretical and methodological innovations. The application

of Mīmāmsā interpretive concepts and tools to literary

analysis, begun by Anandavardhana and carried further in the

works of Mukulabhatta and PratIharenduraja, reaches new

heights in Mahimabhatta's Vyaktiviveka. Anandavardhana's

preference for functionally unified, rasa-directed poetry

becomes, for Mahimabhatta, an absolute requirement -- all

literary works must have the effective communication of rasa

as their ultimate, overriding goal.

While Mahimabhatta's argument for the inclusion of dhvani

within the scope of inference finds no followers among

subsequent Ālamkārikas, other elements of his work,

particularly the criteria for propositional clarity and

coherence developed in the course of his analysis of poetic

flaws, are enthusiastically adopted by the later dhvani-

theorists.96 Mahimabhatta's rigorous effort to apply to

96Mammata incorporates Mahimabhatta's vidheyāvimarsa, prakrama-bheda, krama-bheda and paunaruktya in his own treatment of poetic flaws, explaining them in very similar terms and often adopting the very same examples and suggested emendations employed in the Vyaktiviveka -- cf. Kavyaprakāśa (ed. N.S. Venkatanathacarya, Oriental Research Institute Series 122, University of Mysore, 1977) Vol. II, pp. 50-61,

520

poetic language the same standards of sentential and textual

coherence developed for the analysis of non-poetic language

(chiefly in the Mīmamsakas' interpretation of the Veda)

continues to exert a powerful influence on later authors, even

though his denial of a specifically poetic mode of

signification is rejected.

105-117, and 123-4. Hemacandra's Kavyanusasana, Vidyanatha's Prataparudrīya, and Visvanatha's Sahityadarpana follow the Kāvyaprakasa in this respect. The entire discussion of poetic flaws in Vidyadhara's Ekavali is modelled on Mahimabhatta's treatment, adopting the same categories of flaw found in the Vyaktiviveka.