7. CHAPTER FIVE ANANDAVARDHANA'S POETICS AND THE PHILOSOPHY OF LANGUAGE

CHAPTER FIVE ANANDAVARDHANA'S POETICS AND THE PHILOSOPHY OF LANGUAGE

We have seen that Anandavardhana justifies his radical

departure from the methods and concepts of previous literary

theory by appealing to bodies of thought and practice outside

the field of Alamkārasāstra. He resorts to the analysis of

existing literary works to demonstrate that dhvani, although

unrecognized by previous theorists, has played a central role

in the works of great poets from the very origins of literary

art (in Valmīki's Ramayana, the "first poem" [adi-kāvya]). In

addition, he draws on the concepts and methods of linguistic

philosophy, principally on the teleological discourse analysis

of the Mimamsakas, in formulating and defending his innovative

approach to literary analysis. Anandavardhana's application

of models drawn from linguistic philosophy in the realm of

Alamkārasāstra, while it gives him access to a sophisticated

and ready-made set of tools for text-analysis and thus greatly

expands the possibilities of literary criticism, also requires

him to account for the phenomena of literary language in

philosophical terms.

In the centuries preceding Anandavardhana's work,

Ālamkārikas had devised a detailed and sophisticated

conceptual vocabulary for the description and analysis of the

elements of literary language. This vocabulary of literary

analysis had developed more or less independently of the

system of concepts and terms deployed in the analysis of non-

literary language by Mīmamsakas, grammarians, and logicians

194

195

(the major practitioners of linguistic philosophy in

Sanskrit) . Adherents of these schools of philosophical

thought, while differing widely in their views on the nature

of word- and sentence-meaning, had, through centuries of

argument, developed in common a system of terms through which

linguistic phenomena could be analyzed and discussed. The

Ālamkārikas, taking a specialized variety of language as their

object of study, had taken few pains to integrate their

terminology and descriptive categories with those employed by

philosophers of non-literary language. In effect, two

distinct vocabularies had developed, one for the description

and analysis of language in general, and another for the

description and analysis of language crafted for aesthetic

effect.

Applications of the terms of general linguistic

philosophy in the works on Alamkāraśāstra prior to

Anandavardhana's time are rare, and do not play any major

theoretical role for the authors of these works. Where there

is a clear parallel between a particular figure of poetic

language and some recognized feature of ordinary language, the

terms of linguistic philosophy are occasionally employed by

these authors. Udbhata, for example, describes the figure

rūpaka (metaphor) as a type of gunavrtti (secondary

signification) .1 Vāmana explains that the figure ślesa

1Kāvyālamkārasārasamgraha, 1.11, Banhatti's ed., p. 10. Gunavrtti is the term employed by philosophers to describe cases in which a word, when its literal referent is

196

(double entendre) occurs when a single word or set of words

simultaneously describes two referents by a process of

tantra, 2 and defines the figure vakrokti ("oblique

expression") as "laksana [figurative signification] based on

similarity".3 In discussing these alamkäras, the use of well-

known philosophical terms is convenient for descriptive

purposes. But, for the most part, the alamkaras and other

aspects of poetic language dealt with by these authors have no

obvious parallels in the philosophers' analysis of non-poetic

language, and there is no systematic effort on the part of

these Alamkarikas to account for the poetic phenomena they

describe in philosophic terms.

Anandavardhana focusses his attention not on the typology

of poetic features, as his predecessors had for the most part

done, but on the processes by which the words of a poem convey

to its readers or hearers a particular set of meanings, and on

the hierarchical interrelations between these multiple

contextually impossible, metaphorically ascribes qualities (gunas) of its literal referent to its actual referent. example, in the expression "That boy is a lion", the word For

"lion" causes us to understand that the boy being described has the qualities of a lion (fierceness, bravery, etc.). Cf. Sabarabhasya on MImamsasutra 1.4.23 and 3.2.1, and above, Chapter Three, pp. 109-110.

2Kavyalamkarasutra, 4.3.7, Kashi Sanskrit Series #209, pp. 162-3. Tantra is an interpretive principle employed by the Mimamsakas to explain how a single act or material in a ritual, or a single word or passage in a text, simultaneously perform two or more separate functions. may Cf. Mimāmsāsutra, 11.1.1ff.

3Ibid., pp. 164-5. Cf. above, Chapter Three, pp. 109ff.

197

meanings. In doing so he applies the concepts and tools of

philosophical language analysis to the phenomena of literary

language in a more thoroughgoing and systematic fashion than

did any of the Alamkärikas who preceded him. This approach to

the analysis of poetry requires him to bridge the

aforementioned terminological and conceptual divide between

the paradigms of literary and non-literary language theory.

As we shall see, alongside the tripartite classification of

dhvani into vastu, alamkāra, and rasa (categories derived from

earlier literary analysis), Anandavardhana provides a second

typology of dhvani based on philosophically derived

categories.

In addition, Anandavardhana's contention that the realm

of suggestion he examines has gone unnoticed by earlier

theorists and must be distinguished from already recognized

modes of verbal signification places him in conflict with all

the major schools of Sanskrit linguistic philosophy, as they

admit only two modes of signification, literal (abhidha) and

metaphorical/metonymical (gunavrtti / laksanā) .

Ānandavardhana's third mode of signification (vyañjana --

suggestion) must be defended against the objections (actual or

hypothetical) of grammarians, Mimamsakas, and logicians (both

Naiyayikas and Buddhists), all of whom wish either to

incorporate it within one or both of the accepted modes of

signification, or to exclude it altogether from the realm of

verbal signification by explaining cases of "suggestion" as

198

inferences which operate after the meaning of a text has been

understood. The defense of vyañjana as an independent mode of

signification becomes a major task for Anandavardhana and his

defenders in the ensuing centuries (chiefly Abhinavagupta and

Mammata), and much of the criticism of the dhvani-theory in

this period is focussed on this point (e.g. Mukulabhatta's

Abhidhavrttamatrka and Mahimabhatta's Vyaktiviveka).

The Categorization of Dhvani in Philosophic Terms

In the Dhvanyaloka, Anandavardhana presents two separate

and unrelated typologies of dhvani. The first subdivides

cases of dhvani on the basis of the type of meaning which is

suggested. Verses or passages containing dhvani may suggest

an unexpressed vastu, alamkāra, or rasa. These three

categories of possible suggested meaning are drawn from within

the tradition of Alamkarasastra.4 The second typological

schema classifies cases of dhvani based not on their suggested

content, but rather on the process by which this suggested

content is apprehended, and on the relationship, temporal as

well as functional, between the literal and suggested contents

of a passage.

In this second schema, instances of dhvani are first

divided into two broad categories: those in which the directly

expressed meaning of a verse, being contradicted or supplanted

'Although the terms rasa and vastu ("plot" or "narrative content") are drawn originally from dramaturgy (Natyasastra) rather than Alamkarasastra, both had been introduced into the vocabulary of the Alamkarikas well before Ānandavardhana's time.

199

by its suggested meaning, is not literally intended

(avivaksita-vacya), and those in which the expressed meaning,

although subordinated to the suggested meaning, is literally

intended (vivaksitānyapara-vācya) .5

Each of these categories is further subdivided. Cases of

dhvani in which the expressed meaning is not literally

intended (avivaksita) may be divided into two subclasses:

those in which the expressed meaning is totally displaced

(atyanta-tiraskrta), and those in which it is transformed into

some other (suggested) meaning (arthantara-samkramita). The

following is given as an example of the first type:

The moon, its disk obscured by frost and its beauty transferred to the sun, does not shine, like a mirror blinded by breath.6

Here the relevant expression is "blinded by breath"

(nihsvasandha). The use of the word here suggests the

obscurity and inability to reflect images clearly of a mirror

fogged by breath. A mirror is insentient by nature, and hence

the word "blind" cannot be literally applied to it. The

expressed meaning of the word is therefore totally displaced

5All cases of rasa-dhvani fall into the latter class. In the verse "Siva, his composure somewhat disturbed ... " (quoted above, cf. Chapter Four, p. 130), the description of Siva's agitation as he gazes at Parvati's face is meant to suggest srngara rasa (which serves as the primary meaning of the verse), and is in no way falsified or contradicted by this suggested rasa.

6Dhvanyaloka, op. cit., p. 172. ravisamkrāntasaubhāgyas tusārāvrtamandalah/ nihśvasandha ivādarśaś candramā na prakāśate// (Ramāyana, 3.15.13)

200

by its suggested sense.7

The second variety is exemplified by the following verse:

The clouds, with cranes wheeling against them, smear the sky with dark, wet, splendor; the winds contain droplets of water; the joyful cries of the friends of the clouds [i.e. the peacocks] are sweet. All right, let them be so! I am Rama, whose heart is very hard; I can endure all! But how will Sita survive? Alas, my queen, be strong!8

The clouds, the moist winds, and the cries of the peacocks are

signs of the approaching rainy season, a time when travellers

typically return home and separated lovers are reunited. For

lovers who cannot be united, the onset of the rains

intensifies their longing and heightens the pain of

separation. In this verse Rama laments his separation, during

the rains, from Sita, who has been abducted by Ravana. Here,

the literal meaning of the statement "I am Rama," although

true, is not intended (vivaksita). The intention of this

expression is not to communicate the identity of the speaker

(which is already known from context), but rather emphasize

Rama's capacity to endure intense suffering by reminding the

audience of the many calamities he has endured in the past

7Ānandavardhana's arguments for the distinction of such cases of dhvani from ordinary metaphor are examined below (pp. 226-230).

8Ibid., pp. 167-8. snigdhasyāmalakāntiliptaviyato velladbalākā ghanā vātāh šīkarinah payodasuhrdām ānandakekāh kalāh/ kamam santu drdham kathorahrdayo ramo 'smi sarvam sahe vaidehī tu katham bhavisyati hahāhā devi dhira bhava// The source of the verse is not definitely known, but Raghavan suggests that it is probably from Act Two of Yasovarman's Ramabhyudaya (cf. Raghavan, Some Old Lost Rāma Plays, p. 20).

201

(his exile from his father's kingdom, the abduction of SIta,

the death of his father, and so on). Thus the literal meaning

of the word "Rama" is not totally displaced, but is

transformed into a further suggested meaning (arthantara-

samkramita).

Like avivaksita-vācya dhvani, the second major category

of dhvani, vivaksitanyapara-vacya (in which the literal

meaning is intended, but subordinated to another, suggested,

meaning) contains two subtypes. In this case, where both the

literal and suggested meanings of a verse are intended, it is

the apparent temporal relationship between these two meanings

which serves as the basis for subdivision. When a particular

rasa is suggested by a verse, it is not experienced by the

reader or hearer as something separate from, and subsequent

to, the directly expressed meaning of the verse. It may be

shown by careful analysis that this rasa is a separate meaning

occasioned by the presence, in the expressed meaning, of a

particular set of aesthetic factors (vibhavas, anubhavas, and

vyabhicaribhavas), and that the audience's apprehension of

rasa is therefore necessarily subsequent to its understanding

of the expressed meaning of the verse.9 However, for a

sensitive observer, the sequence between the apprehension of

the expressed meaning and the rasa it suggests will be so

rapid as to go unnoticed. Instances of dhvani in which rasa

is suggested are therefore termed asamlaksya-krama ("in which

9Cf. Dhvanyaloka, op. cit., pp. 403-413.

202

sequence is unnoticed").

Where the expressed meaning is intended, but

subordinated to some suggested meaning other than rasa (i.e.

a suggested vastu or alamkära), the succession between the two

meanings will necessarily be apparent. Such cases are

therefore termed samlaksya-krama ("in which sequence is

noticed"). For example,

While the divine sage [Narada] was speaking thus, ParvatI, at her father's side, counted the petals on her play-lotus.10 with face downcast,

Narada has come as an envoy to arrange Parvati's marriage with

Śiva. The expression "counted the petals on her play-lotus"

describes Pärvati's actions, and thereby suggests that she is

embarrassed by the discussion of her upcoming marriage. Here

the suggested meaning is understood from conscious reflection

on the literal meaning of the verse, and the sequential

relationship between the two meanings is thus apparent.

Because of the noticeable interval between our understanding

of these two meanings, this type of dhvani is also described

as anuranana-rupa ("having the form of a reverberation").

Samlaksya-krama dhvani is further subdivided into two

classes depending on whether the suggested meaning depends on

the suggestive power of a particular (expressed) meaning

(arthasakti-mūla) or depends on the suggestive power of a

10Ibid., p. 248. evamvādini devarsau pārśve pitur adhomukhī/ līlákamalapattrāni ganayām āsa pārvatī// (Kālidāsa, Kumārasambhava 6.84)

203

particular word or set of words (sabdasaktimula). The verse

just quoted is an example of the first type. The second type,

in which the presence of polysemic words suggests a double

entendre, is exemplified by the passage "At that time the

great season ... " (cf. above, Chapter Four, p. 170) .

Ãnandavardhana's second typology classifies dhvani on the

basis of the process by which we apprehend suggested meaning,

rather than the content of this suggested meaning. This

emphasis on the processes of linguistic cognition is largely

absent from earlier works on Alamkarasastra, and reflects an

unprecedented effort on Anandavardhana's part to give an

account of the semantic functioning of poetry in philosophic

terms. The distinction between meanings which are "meant to

be expressed" (vivaksita) and "not meant to be expressed"

(avivaksita), which serves as the primary basis of

classification, plays a major role in the Mīmamsakas' system

of textual interpretation, as we have seen.11 The close

attention devoted to the succession of multiple stages of

meaning likewise shows a Mimamsa influence.

Vyañjana as a Distinct Mode of Verbal Signification

While this classificatory system thus applies the

conceptual tools of linguistic philosophy, particularly

Mimāmsa, to the analysis of meaning in poetry, it applies them

to a variety of meaning hitherto unrecognized by philosophers

of language. Anandavardhana contends that dhvani depends on

11Cf. above, Chapter Three, pp. 98ff.

204

a mode of verbal signification distinct from those modes

already known to philosophers of language, abhidhā

(denotation, direct expression) and gunavrtti (secondary or

figurative expression -- a category comprising both metaphor

and metonymy). This third mode of signification he calls

vyañjanā (suggestion, manifestation) .12

Anandavardhana's defense of this third mode against the

attacks of the upholders of the various established schools of

linguistic philosophy is two-pronged. He attempts to show

that vyañjana, because its importance is confined largely to

poetic rather than ordinary language, lies properly outside

the scope of general linguistic philosophy.13 In addition,

he attempts to show that the theory of vyañjana does not truly

12Vyañjana is conventionally rendered as "suggestion" in secondary literature on Alamkarasastra in English. This translation, while it Anandavardhana's use of the term, fails to capture the more accords fairly well with

general meaning of the term as employed in other fields. As a technical term it appears to have its roots in Samkhya (cf. Samkhyakārika 10 and Vacaspatimisra's Sāmkhyatattvakaumudī thereon). It refers to the process by which a pre-existent but imperceptible entity becomes apparent; for example, when a lamp illuminates a pot placed in a previously darkened room. The term might be more accurately, albeit somewhat awkwardly, be translated as "revelation", or "manifestation". The conventional translation will be retained here, but these further connotations should be borne in mind. For a critique of Anandavardhana's application of the term in the realm of linguistic comprehension cf. Mahimabhatta's Vyaktiviveka (Kashi Sanskrit Series 121, pp. 80ff) and below Chapter Ten (pp. 478-481) .

13Although, as we shall see below (cf. pp. 216-221) , vyañjana plays a role in our understanding of expressions in ordinary language as well, Anandavardhana argues that the meaning conveyed thereby is separately intended (vivaksita) only in poetic contexts, and therefore tends to go unnoticed in ordinary language.

205

conflict with the theories of language advanced by the

grammarians, Mīmamsakas, and logicians, and therefore can be

accepted without any difficulty by adherents of any of these

theories. These theorists, being concerned either with

ordinary language or the language of the Veda, have made no

serious attempt to account for those linguistic processes

specific to poetry; the concept of vyañjana is introduced only

to explain these specifically poetic processes, and its

adoption thus need not conflict with theories of ordinary or

Vedic language offered previously.

The several varieties of suggestion manifested in

literature and the other arts are part of they everyday

experience of cultured persons (although they have not, before

Anandavardhana, been grouped together and described by a

single term) . The proper concern of philosophical

speculation, Anandavardhana contends, is that which lies

beyond our ordinary experience.

All the disputes of philosophers [tārkikānām] are concerned with something extraordinary, not with what is ordinary [laukika]. For they are not seen to disagree with one another about some entity which is perceptible to the whole world and which is uncontradicted [by any other perception], such as blueness, sweetness, etc. For, one who calls a blue thing "blue", when there is no conflicting perception, is not contradicted by another saying, "This isn't blue, this is yellow." In just the same way, the suggestiveness of all these -- of words, which convey an expressed meaning, of the sounds of a song, which convey no expressed meaning, and of gestures and so on, which do not consist of sounds at all -- is established by experience; who could reject this? Expressions and actions of various sorts which convey

206

some beautiful, non-verbal14 meaning, either in combination or separately, are seen in the assemblies of cultured people. What rational person, wishing to avoid ridicule for himself, would deny these things?15

The communication of unstated meanings plays an important role

not only in poetry, but in music and dance as well. The

capacity of words, sounds, and gestures to convey such

meanings is well known to all cultured people, and should thus

be accepted by adherents of all philosophical schools, just

as, whatever their metaphysical views may be, they all accept

the validity of ordinary perception.

This appeal to everyday experience, however, while it

plainly establishes the reality of the communicative processes

Anandavardhana wishes to designate by the term vyañjana, does

not conclusively demonstrate the need to postulate vyañjana as

a separate mode of signification. The poetic processes

referred to by the term vyanjana, although they have been thus

far neglected by philosophers of ordinary and Vedic language,

could possibly be accounted for as instances of one or another

of the well-known modes of signification dealt with by earlier

language theorists. Anandavardhana is forced to defend the

independence of vyañjana against the arguments of those who

would explain cases Of dhvani as instances of direct

expression (abhidha, vacakatva), of metaphorical/metonymical

usage (gunavrtti or laksana), or of inference which operates

14That is, not directly stated.

15Dhvanyaloka, pp. 446-7.

207

subsequent to the understanding of the expressed meaning of

the sentence in question.

The Differentiation of Suggestion from Direct Expression

It may seem paradoxical for an objector to assert that

dhvani can be explained as a type of direct expression

(vacakatva). The argument depends upon the view, common to

the grammarians and the Prabhakara Mīmamsakas, that the

complete sentence, rather than the individual word, is the

basic unit of linguistic expression.16

For the grammarians (chiefly Bhartrhari and his

16Anandvardhana seems to intend this argument to apply only to the Prabhakaras. At one point (pp. 443-4) he asserts that, since the concept of dhvani itself was drawn from the grammarians (who use the term to refer to speech sounds which manifest verbal meaning), there can be no question of a dispute with the grammarians over the existence of dhvani. But even Anandavardhana's staunchest defender, Abhinavagupta, realizes that this view is untenable. Like the Prābhākaras, the grammarians believe that complete complete sentence meanings by a single sentences express semantic process; Anandavardhana's theory of suggestion demands the existence of several distinct semantic processes, and so must be defended against the views of both grammarians and Prābhakaras. The arguments offered against the inclusion of suggestion within the scope of direct expression apply equally well to both -- (cf. Abhinavagupta's commentary [Locana] on Dhvanyaloka, op. cit., pp. 16 and 415-416, and below, Chapter Nine, pp. 444- 446). Abhinavagupta (Locana, p. 415) claims that this argument refutes the view of the Bhatta Mimamsakas as well, but in this he is clearly mistaken: the Bhattas would not explain Anandavardhana's suggestion as of direct expression and should be classified among those who would type

include suggestion within the scope of figurative expression (laksanā) -- cf. below, pp. 221ff) . Abhinavagupta asserts (p. 416) that the Bhattas regard the overall meaning of a sentence meaning as something directly expressed (vacya), yet we have seen (above, Chapter Three, pp. 115-117) that they believe sentence meanings (and, indeed, all meanings apart from the universal class-properties denoted by individual words) to be conveyed by a process of figurative expression.

208

followers), the sentence and its meaning are indivisible,

partless entities. The division of sentences into component

words, like the division of words themselves into

morphological components (stem, root, suffix, etc.), is merely

an artificial construction of language theorists, an

abstraction useful for purposes of grammatical analysis and

instruction, but not corresponding to any real division in the

sentence, which is conceived of as a metaphysical unity.

The Prabhakaras do not deny the metaphysical reality of

individual words, but contend that words are able to express

meanings only when construed in the form of sentences (the

doctrine of anvitabhidhana-vada -- cf. above, Chapter Three, p.

116). Human discourse, through observation of which we

acquire our understanding of language, is invariably conducted

in the form of sentences. By observing the use of words in a

variety of sentences we are able, by a process of abstraction,

to understand the semantic function of specific words, and

thereby to understand even previously unheard sentences.

Nevertheless, in practice these words are invariably employed

and understood as parts of complete sentences. It is

therefore, properly speaking, the sentence itself, rather than

the individual words, which expresses meaning (although the

individual words play a causal role in producing our awareness

of this sentence meaning) .

Thus for both the Prabhakaras and the grammarians it is

the sentence as a whole which expresses meaning. Given this

209

view, it becomes possible to explain even cases of dhvani

through an all-encompassing expressive function (vacakatva).

The opponent grants that Anandavardhana has established that

words may deviate from their ordinary meaning in certain

contexts; it is only to the characterization of such meaning

as suggested (vyangya) that he objects.17 If, as

Anandavardhana contends, the "suggested" meaning, in sentences

containing dhvani, is invariably the predominant meaning, it

is improper and indeed contradictory to describe this meaning

as anything other than expressed (vacya). The defining

characteristic of a sentence, as we have seen (cf. above,

Chapter Three, pp. 75ff), is that it is directed toward a

single meaning or purpose (artha). If the meaning conveyed by

dhvani is what the sentence as a whole is primarily meant to

convey, it is, ipso facto, what is expressed by that sentence.

The (hypothetical) proponent of this view argues as follows:

Where this ["suggested" ] is established as primary, it may meaning appropriately be termed vācya [expressed], since the sentence is primarily directed toward it. And therefore the semantic function of the sentence which reveals this meaning is only expressiveness. What need is there to postulate any other function [such as suggestion] for this sentence?

speaker's] intent, that is, primarily, what is expressed. Therefore, whatever meaning is the object of [the

Whatever intermediate understanding there may be of other expressed meanings in such a case is merely a means for the understanding of the primary meaning; just as the understanding of word meanings is merely a means for the understanding of sentence meaning.18

17Dhvanyaloka, p. 414.

18Ibid., pp. 414-15.

210

Anandavardhana himself earlier drew an analogy between

the understanding of sentence meaning from word meanings and

the understanding of suggested meaning from expressed

meaning. 19 When the meaning of the sentence as a whole is

understood, the meanings of its individual words are no longer

separately apparent. Likewise, once we understand fully the

intended meaning of a sentence, including any implied or

"suggested" sense, any intermediate stages of awareness or

"meaning" which have led to this understanding will be

displaced by our awareness of the meaning of the whole. Any

sentence must, by definition, convey a single, coherent

meaning. However complex the series of cognitions by which we

come to understand it, it is this final, complete, meaning

which is "expressed" by a sentence. Because every coherent

sentence, in poetry or in any other form of discourse, must

convey one and only one meaning, and because all discourse is

conducted in the form of sentences, we need speak only of one

all-encompassing semantic function, namely "expressiveness"

(vacakatva). There is consequently no need to postulate any

additional mode of signification, such as vyañjana, to account

for those varieties of poetic meaning designated by the term

dhvani.

This attempt to explain all semantic processes by a means

of a single, "expressive" verbal function cannot,

19Ibid., p. 100; cf. also above, Chapter Four, pp. 135- 136.

211

Anandavardhana contends, adequately account for the process of

verbal understanding involved in cases of dhvani.20 Given

that there are intermediate stages of understanding, in cases

of dhvani, between our hearing the words of a verse and our

grasping of its full meaning (including any implied or

suggested sense), one must necessarily accept that these words

perform more than one semantic function. Wherever such

multiple meaning-stages exist, some word (or group of words),

having conveyed its own meaning, further causes us to

understand some other meaning. For example, in the verse "The

moon, its disk obscured by frost and its beauty transferred to

the sun, does not shine, like a mirror blinded by breath, "

(cf. above, p. 199), the word "blinded" (andha), having

conveyed its literal sense, causes us to understand that the

mirror in question is clouded and non-reflective.

The communication of literal and non-literal meanings by

the same word or words cannot be attributed to the operation

of a single expressive function. In such cases we are aware

of two distinct, successive semantic operations. These two

operations differ both in their nature (rüpa) and in their

semantic content (visaya), and hence cannot be identified with

one another.21 Expressiveness is a property of words and

words alone, while even wordless songs and silent gestures may

be suggestive of some meaning (i.e. rasa). Given the

20Ibid., pp. 416ff.

21Ibid., p. 416.

213

itself.23 The fact that words convey two sorts of meanings

(invariable and context-dependent) cannot be explained unless

we postulate more than one semantic function. Thus the

expressive and suggestive functions of words can be

differentiated on the basis of the semantic content they

convey, even in situations where both are operative.

The relationship between the expressed meaning of a

sentence and the suggested meaning we understand from it is

not, as the objector contends, fully analogous to that between

the meanings of individual words and of the sentences they

comprise.

The analogy of word meaning and sentence meaning does not apply to expressed and suggested meanings, since some wise people [e.g. Bhartrhari and the grammarians] contend that the understanding of word meaning is unreal, and since even those who do not admit its unreality [i.e. the Prabhakaras] 24 must accept that the analogy of a pot and its material cause [i.e. clay] meanings and word meanings. applies to sentence For just as, when a pot has been formed, its material cause is not perceived as something separate from it, in the same way, when a sentence or its meaning is understood, the words and their meanings are not perceived as something separate. If they were perceived separately, the understanding of the sentence meaning itself would be displaced [given the requirement that a sentence convey a single, unitary meaningl . But this analogy does not apply to expressed and suggested meanings; for, when the suggested meaning is understood, understanding of the expressed meaning is not displaced, since the suggested meaning appears along with the expressed meaning.25

23The argument is strongly reminiscent of Sabara's defense of the notion of secondary (metaphorical) signification -- cf. Bhāsya on Mīmamsāsūtra 3.2.1.

24Abhinavagupta, as before, includes the Bhātta Mīmamsakas as well -- cf. above, p. 207.

25Ibid., pp. 419-421.

214

Unlike word and sentence meanings, the expressed and

suggested meanings of a verse may be simultaneously and

distinctly cognized. When a verse, having communicated its

literal meaning, suggests some rasa, for example, our

understanding of this rasa does not in any way obstruct or

supplant our understanding of the literal meaning. A proper

analogy for the relationship between these two meanings would

be not that of word- and sentence-meaning (or of a pot and the

clay it consists of), but that of a lamp and an object

illumined by it. When a lamp is lit in a darkened room it

becomes visible, and further makes visible any other object in

the room, without itself ceasing to be visible. Likewise, our

understanding of the expressed meaning of a verse, even when

some rasa is suggested by means of it, does not itself cease

to be apparent.

This model of the suggestive process, however, raises a

further objection; if the expressed meaning of a sentence does

not recede from our awareness once its suggested meaning has

been understood, and the two meanings are therefore

simultaneously apparent, the sentence as a whole will be

incoherent -- will, in fact, not be a sentence at all.

The sentence would be simultaneously connected with two meanings, and if this were the case the sentence could not even be a sentence, since this has unity of meaning as its defining characteristic.26

We have seen that, for the MImamsakas, the necessary and

26Ibid., p. 421.

215

defining characteristic of the sentence is that it convey a

unitary sense. Any expression which fails to do so will

suffer from the fault of vakyabheda ("splitting of the

sentence" -- cf. above, Chapter Three, pp. 85ff) .

Anandavardhana's response to this objection itself relies

on Mīmāmsā interpretive principles. The simultaneous

awareness of the expressed and suggested meanings of a single

sentence is acceptable so long as these two meanings are

integrated in a single purposive framework.27 When a

sentence causes us to understand both an expressed and a

suggested meaning, these two meanings will invariably be

related as primary and subordinate meanings of the sentence in

question.28 If the suggested meaning is predominant and the

expressed meaning is subordinated to it, the sentence in

question will be described as an instance of dhvani. If the

suggested meaning is subordinate and the expressed meaning

predominant (as in the case of alamkāras such as paryāyokta,

dīpaka, or rasavat -- cf. above, pp. 157ff), the sentence will

be described as an instance of gunībhūta-vyangya (" [poetry] in

which the suggested meaning is secondary") .29 Thus, the

27Cf. the discussion of visista-vidhāna ("qualified injunction") -- above, Chapter Three, pp. 89ff.

28Dhvanyaloka, pp. 421-422

29Anandavardhana classifies kāvya (poetry) into three broad types, based on the status of suggested meaning within it -- dhvani kavya, in which suggested meaning predominates, gunībhuta-vyangya kavya, in which it is present but subordinate, and citra kävya, in which there is no suggested meaning whatsoever. The first of these represents, for

216

simultaneous apprehension of expressed and suggested senses

does not vitiate the unitary character required for a coherent

sentence.

Anandavardhana contends that the Mimamsakas must accept

suggestion as a distinct semantic function not only to account

for the multiple non-literal meanings encountered in poetry,

but even to explain the distinction between Vedic and human

language.30 The crucial property of Vedic language is its The unfailing reliability, and the argument for the impossibility

of error in the Veda depends on its being impersonal

(apauruseya) and not the product of any author, either human

or divine. Falsity is always a possibility in personal

(pauruseya) language, due to some flaw in its author, such as

error or deceit. If a text has no author, there is no

possibility that it will be vitiated in this way.

We have seen that for the Mimamsakas the connection of a

word with its (expressed) meaning is necessarily eternal and

uncreated. If words can express only these eternally fixed

meanings, and if a sentence's meaning is purely a function of

the expressed meanings of its component words, there can be no

meaningful distinction between authored and authorless

Anandavardhana, the ideal to be striven for by contemporary poets as well as the norm in the works of the greatest poets in the past, and consequently receives the lion's share of his attention in the Dhvanyaloka. The latter two classes of poetry, because they diverge from this ideal, give rise to certain theoretical problems which will be considered in detail in Chapter Six.

30Dhvanyaloka, pp. 438ff.

217

sentences as regards their truth or falsity. If this

distinction is to be maintained, it is essential to postulate

some conditional (aupadhika), non-invariable semantic capacity

in words when they are employed by speakers or authors.

If this [conditional semantic function] is not accepted by the Mimamsaka, given the eternality of the connection

between between word and meaning, there can be no distinction impersonal and personal sentences in the communication of their meanings. [conditional function] is accepted, falseness will be However, if this

possible for personal sentences, on which another, conditional semantic function has been superimposed in conformity with [the speaker's] personal intention, even without the deviation of the words from their expressed meanings.31

When we hear personal (i.e. non-Vedic, non-authorless)

sentences, what we understand from them primarily is the

intention of the speaker who employs them. In support of this

claim Abhinavagupta (Locana, p. 440) cites a remark from

Śabara's Bhāsya on Mīmāmsāsūtra 1.1.2 (Ānandāśrama Sanskrit

Series 97 [second ed.], p. 18): "The understanding that arises

from a personal utterance is 'This person believes thus', not

'This thing is so.'" Such statements inform us not about the

true state of things, but about the state of mind of the

speaker who makes them. Any conclusion we draw about the

actual state of things based on such statements requires a

further process of deduction based on our estimate of the

reliability and competence of the speaker in question.

The speaker's expressive intent is not directly expressed

by the words of the sentence, yet it is understood when we

31Ibid., pp. 438-9.

218

hear these words and must therefore be regarded as part of

their meaning. Thus , if the Mimamsakas wish to maintain a

distinction between infallible, impersonal, Vedic language and

fallible human language, they must accept some context-

dependent, conditional semantic function for words in addition

to their expressive function. This conditional semantic

capacity is nothing other than suggestiveness.

This argument, however, gives rise to a further

difficulty.32 If this suggestiveness is a feature

necessarily belonging to all personal, non-Vedic utterances,

it can no longer be considered a special feature of poetic

language, as Anandavardhana has so far presented it. All non-

Vedic utterances would have to be regarded as instances of

suggestion.

While admitting that suggestiveness is a property of all

human language, Anandavardhana contends that in poetry this

function of language takes on a special and unique importance

which distinguishes it from the suggestive aspect of non-

poetic language. Dhvani is not coterminous with

suggestiveness. The suggested meanings conveyed by everyday

language, while they must necessarily exist, are not

ordinarily recognized separately from the expressed meanings

with which they co-occur. For example, if a speaker tells us

"There are fruits on the river-bank" (to employ a common

Mīmamsa example), we understand (by suggestion) that speaker's

32 Tbid., p. 441ff.

219

communicative intention (i.e. "This person believes that there

are fruits on the river-bank," or "This person wishes me to

believe there are fruits on the river-bank"). This kind of

suggestion is so much a part of the ordinary process of

understanding human language that it generally passes

unnoticed.

However, this is not the case in poetry. As we have

seen, the suggested meaning conveyed by a verse may actually

contradict its expressed meaning. In the verse,

Wander confidently, Monk; that dog [which frightened you] has been killed today by the lion who dwells in the creeper-bower on the banks of the Godāvari, 33

the expressed sense is a command that the monk should wander,

while we understand by suggestion that the speaker intends to

prevent him from so doing by frightening him. In such a case

the expressed and suggested senses, because they stand in

contradiction with one another, cannot but be recognized as

distinct meanings. Even where there is no such contradiction,

the suggested meanings conveyed by poetry are ordinarily

apprehended separately from the expressed meanings with which

they co-occur. Anandavardhana has already stipulated that the

term dhvani applies only to verses or works in which the

suggested meaning is predominant and the expressed meaning of

only secondary importance. This predominance in and of itself

necessitates that the suggested meaning in such cases be

apprehended as something distinct from what is directly

33Cited above, Chapter Four, p. 128.

220

expressed and thus serves to differentiate suggestion as it

functions in poetry from that type of suggestion which is

necessarily a factor in all human discourse.

Änandavardhana explains this differentiation between the

role of suggestion in everyday and in poetic discourse in

terms of the Mimamsa concept of vivaksa. 34 Sentences

employed in everyday human discourse necessarily convey a

suggested meaning, but this suggested meaning is not

ordinarily what is meant to be expressed (vivaksita) by such

sentences. To return to our earlier example, when one states

that "There are fruits on the river-bank, " what one means to

express is not that one believes or wishes one's hearers to

believe that there are fruits on the river-bank, but rather

that there are, in fact, fruits on the river-bank. One's

belief that this is so and one's intention to convey this to

one's hearers can be understood from the very nature of the

communicative act, but are incidental to the meaning one

wishes to express. Poets, by contrast, use the suggestive

capacity of language consciously and deliberately. The

suggested meanings they convey through their verses are part

of what is meant to be expressed (vivaksita) by these verses,

and in cases of dhvani, are by definition what is meant to be

expressed primarily.

This line of argument serves a dual function, defending

34Ibid., pp. 441-443. For the role of vivaksā in Mīmāmsā, cf. above, Chapter Three, pp. 98ff.

221

the autonomy of suggestion (vyañjana) against the attacks

(real or hypothetical) of those who would subsume it under an

all-encompassing expressive (vacaka) function of language by

showing it to play necessary role in all human

communication, while simultaneously maintaining that it is a

function of language which becomes uniquely significant only

in the language of poetry. Thus the existence of suggestion

as a separate semantic function must be conceded by language-

theorists in general, but the detailed study of its nature and

function remains the special concern of poetic analysis.

The Differentiation of Suggestion from Figurative Expression

Anandavardhana's attempt to formulate a clear distinction

between suggested and figurative meaning is somewhat more

problematic than his effort to differentiate suggestion from

direct expression, and becomes a focus of considerable

controversy among Ālamkārikas in the ensuing centuries.

Indian linguistic philosophers had, for many hundreds of years

before Anandavardhana's time, accounted for the non-literal

usage of words (in both Vedic and human discourse) under the

general rubric of "metaphorical" or "figurative" language

(gunavrtti, upacāra, bhakti and/or laksana) .35 Suggestion,

as conceived of by Anandavardhana, bears certain obvious

similarities to such figurative usage. In both, words are

35Typically, the terms upacara and gunavrtti specifically denote metaphorical usage, and laksana non-metaphorical figurative usage (i.e. metonymy, synecdoche, irony), but all four terms are sometimes applied to figurative language in general.

222

employed so as to convey some meaning above and beyond the

ordinary, literal meaning with which they are invariably

associated. In both, the capacity of words to convey such

non-literal meanings depends on the context in which they are

used. Thus, an obvious course of argument for those objecting

to the introduction of vyañjana (suggestion) as a new semantic

function is to account for the non-literal, context-dependent

poetic meanings termed "suggested" by Anandavardhana in terms

of those concepts already developed in philosophical circles

to explain non-literal, context-dependent meanings as they

occur in non-poetic language

Anandavardhana, anxious as always to emphasize the

novelty of his theory and sensitive, as we have seen (cf.

above, Chapter Four, pp. 122ff), to the charge that he is

merely redescribing and renaming phenomena already known,

rejects any effort to include either dhvani or, more

generally, vyañjana within the scope of figurative expression.

The distinction between suggestive and figurative usage is

discussed at length in two passages of the Dhvanyaloka. The

first (kārikas 1.14-19 with vrtti [pp. 141-164]) rejects the

identification of dhvani in particular with figurative

expression, and does so primarily on empirical grounds, citing

illustrations of figurative expression which must be excluded

from the realm of dhvani and vice-versa; the second (pp. 423-

436 -- part of the extremely long vrtti on kārikā 3.33, which

considers and rejects a long series of philosophical

223

objections to the existence of suggestion) argues for the

differentiation of suggestion in general36 from figurative

usage, and examines in greater detail the mechanisms by which

suggestive and figurative expressions convey their meanings.

These two discussions, while they overlap to a certain extent,

raise slightly different issues, and will be treated

separately.

The argument for the differentiation of dhvani in

particular from figurative expression proceeds as follows:

Dhvani cannot be identified with figurative expression, nor

can figurative expression be regarded as a defining feature

(laksana) of dhvani, as these two modes of expression are

simply not co-extensive.37 Any attempt to define dhvani by

means of figurative expression will suffer from the flaws of

over-extension (ativyapti) and under-extension (avyapti) ; that

is to say, there are cases of figurative usage which cannot be

designated as dhvani and cases of dhvani which cannot accepted

as instances of figurative usage.

To illustrate the over-extended scope of figurative usage

with respect to dhvani, Anandavardhana cites numerous verses

in which poets employ metaphors or other figurative

expressions which, because they are thoroughly conventional

and add no special beauty to the verses in question, cannot be

36Of which, as we have already seen, dhvani is only a sub- type, albeit one of paramount importance for literary language as Ānandavardhana understands it.

37Ibid., pp. 141ff.

224

called dhvani. For example:

Wilted on both ends by contact with her swelling breasts and hips; fresh in between, not having made contact with the middle of her body; disarranged by the flailing movements of her creeper-like arms -- this bed of lotus petals tells [vadatil of the affliction of this emaciated woman.38

Here, the word vadati ("tells") is employed metaphorically;

the bed of lotus petals does not actually speak, but its

wilting where it has been pressed by her body and its

dishevelled condition make known to the observer the love-

fever of the woman who has lain there, and it can thus be

said, figuratively, to "speak". This metaphorical expression,

however, is merely a conventional manner of speaking, and the

beauty of the verse in no way derives from it.

Some figurative usage is so thoroughly conventionalized

that it may not even be recognized as figurative. The word

lavanya, for example, is used very frequently in the sense of

"beauty" or "charm", but its literal, etymologically derived

meaning is "saltiness". In c classical Sanskrit poetry,

however, the word is employed almost exclusively in its

metaphorical sense of "beauty", to such an extent that

"beauty" becomes, for all practical purposes, the primary

meaning of the word. Such "dead metaphors" obviously add no

particular beauty to an expression, being, in effect, no

38Ibid., p. 143. parimlānam pīnastanajaghanasangad ubhayatas tanor madhyasyantah parimilanam aprāpya haritam/ idam vyastanyasam ślathabhujalatāksepavalanaih krsāngyāh samtapam vadati bisinīpatrasayanam// (Harsa, RatnavalI 2.11)

225

different from literal usage.39

Anandavardhana would restrict the application of the term

dhvani to those non-literal forms of expression which lend

some special beauty to a verse.

A word which, functioning suggestively, reveals some beauty which could not be conveyed by any other mode of expression will fall within the scope of the term dhvani.40

Conventional and "dead" metaphors of the sort just exemplified

are thus excluded from the realm of dhvani. The existence of

figurative usage which cannot be included within the category

of dhvani establishes the over-extensiveness (ativyapti) of

the former with respect to the latter.

The under-extensiveness (avyapti) of figurative usage

with respect to dhvani is explained as follows: It is an

essential feature of all figurative expression that the

literal meanings of words employed figuratively should be

supplanted, rather than supplemented, by the figurative

meanings they convey (cf. above, Chapter Three, pp. 109ff and

below, pp. 232-235). By this standard, the entire category of

vivaksitānyapara-vācya dhvani (cf. above, pp. 199ff) would be

necessarily excluded from the realm of figurative

expression.41 As we have already seen in the discussion of

the differentiation between suggestion and direct expression

39As Kumārila explains -- cf. Tantravārttika (on 3.1.12), op. cit., Vol. IV, p. 46 (also below, Chapter Seven, p. 330).

4°Dhvanyaloka, op. cit., p. 146.

41Dhvanyaloka, pp. 151-2.

226

(above, pp. 213-214), Ānandavardhana explains the relationship

between literal meanings and the further suggested meanings

that arise from them as analagous to that between a lamp and

the objects illumined by it -- our awareness of the former is

not displaced by our subsequent awareness of the latter;

rather, we remain aware of both simultaneously. This in

itself is enough to demonstrate that there are cases of dhvani

which cannot be explained as instances of figurative usage,

and thereby that the latter category is under-extensive with

respect to the former.

Even where the two categories appear to overlap (as in

instances of avivaksita-vacya dhvani -- cf. above, pp. 199ff),

they cannot be identified with one another. For figurative

expression of any kind to take place it is necessary that

there should be some obstacle to our understanding a word or

phrase in its literal sense, as a word will always, insofar as

this is possible, be understood literally. Words used

figuratively, because they require some such obstacle to

prevent their being understood literally, are said to be

skhaladgati ("proceeding with a stumbling gait," i.e.

conveying their meaning by an indirect path) .42

In every case of figurative usage which is not purely

42Ibid., p. 148. The term skhaladgati is applied to words conveying a figurative sense by Dharmakirti (Pramanavārttika, 2.37 [ed. R.C. Pandeya, Motilal Barnarsidass, Delhi, 1989, p. 72]). I can find no other use of the term in this sense prior to Ānandavardhana.

227

conventionalized, there must be some motive (prayojana43)

which justifies the application of words outside of their

literal meanings. While our understanding of a secondary,

non-literal meaning in such a case is arrived at through

figurative expression, our understanding of the prayojana on

which this figurative expression depends cannot be so.

[Kārikā:] When a meaning is revealed by figurative expression [gunavrttil, bypassing the primary [i.e. literall mode of expression, the word figuratively] does not operate indirectly [skhaladgati] [employed

with respect to the result [phala] figurative usage is directed. toward which the

[Vrtti:] For, in such a case, if the word operated indirectly with respect to the purpose [prayojana] to be achieved -- namely, the revelation of some meaning such that it is characterized by a high degree of beauty -- then the use objectionable.44 of the word would in fact be

43Prayojana is understood here in a sense quite different from that in which the Mimamsakas explain prayojana as a necessary condition of figurative usage (cf. above, Chapter Three, pp. 111-112) . expressed For the Mimamsakas, the figuratively must (prayojana) in the appropriate (ritual) context. It is this meaning have some useful application

criterion of usefulness that allows us to determine which of the many possible figurative senses of a term should be understood in any particular case. Anandavardhana and the Ālamkärikas who come after him understand the need for a prayojana in cases of figurative usage not as the requirement that the particular meaning expressed figuratively should apply usefully in the relevant context (this seems to be taken for granted), but rather that the resort to figurative expression itself should serve some communicative purpose (i.e. that there should be some reason for describing something figuratively rather than by a term which literally denotes it). While I have not been able to trace its history in any detail, this different understanding of prayojana is not original to Anandavardhana; Dharmakirti (c. 650) uses the term in this sense in his own commentary on Pramanavarttika 3.1 (ibid., p. 170).

44Dhvanyaloka, op. cit., pp. 149-50.

228

When metaphors or other figurative expressions are

employed in poetry for aesthetic effect (and not purely

through conventional usage, as in the examples given above),

their purpose is to impart some special beauty which could not

be achieved by literal usage.45 The indirection which

characterizes figurative expression depends on the existence

of some obstacle preventing us from understanding a word

literally, and on our circumvention of this obstacle by

concluding that the word in question is being applied to some

referent other than its literal one. Once we have understood

this figurative application, our further apprehension of the

prayojana of this usage, the beauty it imparts to the verse or

sentence in which it is employed, does not arise in a

45It is clear from the remarks just cited as well as from his later comments on the role of suggestion in poetically effective metaphors (pp. 432-436) that, for Anandavardhana, the prayojana involved in such figurative expressions is nothing more nor less than the beauty (carutva) they impart to the verses in which they are employed. Modern commentators (cf. De, History of Sanskrit Poetics, op. cit., Vol. II, p. 148; Ingalls, The Dhvanyaloka of Anandavardhana, p. 186, n.1), reading Abhinavagupta's and Mammata's understanding of prayojana back into Anandavardhana's work, generally take the prayojana referred to by Anandavardhana to be some further suggested content which the figurative usage causes us to understand. In the stock example gangayam ghosah ("A village on the Ganges"), the metonymical use of the word Ganges in the sense of "the bank of the Ganges" is taken to suggest that the village in question partakes of the properties of the river (coolness, purity, holiness, etc.), and the suggestion of these properties is taken to be the prayojana which justifies the metaphorical usage. There is nothing in Anandavardhana's remarks on prayojana to suggest that he understands it in this way; this explanation of prayojana (and this interpretation of the example gangayam ghosah) seems to originate with Mukulabhatta (cf. below, Chapter Seven, pp. 351-354 and

53.). Abhidhāvrttamātrka [Vidyābhavana Samskrta Granthamāla 165], p.

229

similarly circuitous manner -- there is no obstacle to our

apprehending it directly. If there were such an obstacle to

our apprehending the beauty of the expression, this would

constitute an aesthetic flaw (dosa) .46 The prayojana of any

figurative expression, because it is not stated directly and

cannot itself be understood through figurative expression,

must be conveyed by some additional semantic function, namely

suggestion. Thus in all cases of avivaksita-vācya dhvani

there are necessarily two semantic operations, one figurative

and one suggestive. In the example "like a mirror blinded by

breath" (cf. above, p. 199) the word andha ("blind") is

employed metaphorically in the sense of "clouded", "obscured".

The prayojana for this metaphorical usage is the beauty it

imparts to the verse, which is conveyed by suggestion and on

the basis of which the expression is regarded as an instance

of dhvani.47 Thus, even where dhvani and figurative

46As Abhinavagupta further understanding of the prayojana of a figurative expression were points out, if our

itself arrived at through figurative expression, this second figurative operation would require a prayojana of its own, and an infinite regress would result -- cf. below, Chapter Nine, pp. 438-440.

47Exactly why it is that such a metaphorical expression is more beautiful than a direct statement is something not clearly explained here (or, to my knowledge, elsewhere in the Alamkāra tradition), but it is a basic and largely unquestioned principle for many Alamkärikas that indirect or oblique modes of expression are, generally, to be preferred to direct ones. Cf. for example Bhamaha 2.84, and Dhvanyaloka on 4.5 [Vyaktiviveka, Kashi Sanskrit Series 121, p. 60]). Vakrokti (p. 533) (cited with approval by Mahimabhatta

(oblique or indirect expression) is put forward by Kuntaka as

Eight) . the basis of all beauty in poetic language (cf. below, Chapter

230

expression co-occur, they cannot be identified with one

another and must be recognized as distinct semantic

operations.

However, having argued in this manner for the distinction

between dhvani and figurative expression, Anandavardhana goes

on to assert that even if the identity of the two is accepted,

this will in no way invalidate his own enterprise, namely the

examination of dhvani with all its sub-varieties and the

consideration of its role in poetry.

[Vrtti:] And if it is said that "Dhvani is defined by figurative expression [gunavrtti] alone, " then the entire class of alamkaras -- apart from this [dhvani/gunavrtti] -- would be defined by the function of denotation [abhidhāvyapara] and the uselessness of defining the alamkāras individually would follow. And, furthermore,

[Karika:] If the definition of this has been made by others, then this is the very establishment of our view.

[Vrtti:] Or, even if the definition of dhvani has already been made by others, this is the very establishment of our view; since our view is that "Dhvani exists." And this has been established already [by those who have defined gunavrttil -- thus we have achieved our desired object effortlessly.48

The identification of dhvani with figurative expression in no

way impugns the possibility or the usefulness of

Anandavardhana's work. Anandavardhana has shown that certain

varieties of non-literal meaning (vastu, alamkara, and

especially rasa) are of central importance in the creation and

appreciation of great poetry; whether they are regarded as

being understood through an established mode of non-literal

48Dhvanyaloka, pp. 162.

231

signification, namely figurative expression, or through some

distinct, previously unrecognized, mode of signification,

namely suggestion, is ultimately of little consequence for an

appreciation of their role in poetry.

Thus, if it is contended that the non-literal meanings

understood by the process Anandavardhana calls dhvani are in

fact apprehended via the already well-known and generally

accepted process of figurative expression, this should be

regarded as bolstering,. rather than hindering, the

Anandavardhana's central contention -- that such meanings do

exist, and that they play a central role in the aesthetic

effectiveness of poetry.

The figures of speech dealt with by earlier Alamkārikas

(at least those which do not themselves rely on the processes

of suggestion and/or figurative expression) can be defined by,

and subsumed within, the primary, denotative mode of verbal

expression (abhidha). It does not follow from this that it is

useless for Ālamkārikas to define and illustrate these many

figures individually; these particular sub-varieties of

denotation, because of their special role in poetic

composition, merit individual attention. Likewise, the

several varieties of dhvani defined and illustrated in the

Dhvanyaloka, even if they are subsumed within the overarching

category of figurative expression, are worthy of separate

analysis, because of their central role in poetry.

Anandavardhana's provisional acceptance of the view he at

232

first set out to oppose shows a certain ambivalence in his

approach toward philosophically based criticism. He takes

great pains to argue for the differentiation of suggestion

from the figurative mode of expression already accepted by

most language theorists, but seems unwilling to rely entirely

on these arguments, and, despite the great effort he has made

to resist the inclusion of dhvani within the category of

figurative expression, argues that it doesn't really matter,

for his purposes, whether or not it is so included. This

divided and somewhat contradictory approach is repeated, as we

shall see, in the discussion of the distinction of suggestion

from inference (cf. below, pp. 242-243).

The discussion of figurative usage in the commentary on

kārikā 3.33 argues for the differentiation of suggestion in

general from figurative usage. Suggestion, Anandavardhana

contends, differs from figurative expression, just as it does

from direct expression (cf. above, pp. 211ff), in both its

nature (rupa) and the semantic content (visaya) it conveys.49

As we have seen, one of the distinctive features of

figurative expression is that it operates secondarily,

depending for its functioning on some impossibility or

impropriety in the primary, literal sense of the words used

figuratively. When we hear the expression "A village on the

Ganges" (gangayam ghosah -- a stock example of non-metaphorical

figurative usage [laksanā]), we realize that it is impossible

49Dhvanyaloka, pp. 423ff.

233

for a village to be literally on a river, and therefore

understand the word ganga ("Ganges") to refer, metonymically,

to the bank of the Ganges, rather than to the river itself.

All figurative usage, both metaphorical and non-metaphorical,

depends upon some such obstacle to our understanding the

literal sense of the expression in question. Because the

understanding of a figurative meaning from any word or phrase

requires the impossibility of its literal sense, figurative

meaning invariably displaces, rather than supplements, the

literal meaning of the word or phrase which conveys it.

Suggestion, unlike figurative usage, does not require any

obstacle in the literal sense i in order to function, and

therefore may convey a meaning over and above the literal

sense without displacing it. Ānandavardhana explains the

distinction as follows:

In figurative expression, when a meaning indicates [upalaksayati] another meaning, it is transformed so as to be identical with the indicated meaning, as in expressions such as "A village on the Ganges". In the case of suggestion, however, when a meaning reveals [dyotayati] another meaning, it is perceived as showing this other while still showing itself, like a lamp; as in expressions such as "ParvatI counted the petals on her play lotus" [cf. above, p. 202].50

In the example of suggestion given, Pärvati's counting of her

lotus petals suggests to us that she is embarrassed by the

discussion of her marriage to Siva, and this embarrassment

further suggests the srngära-rasa which it is the primary

object of the verse to convey. Here there is nothing at all

50Ibid., p. 424.

234

impossible in the literal meaning -- that Parvati is counting

lotus petals. The second meaning comes to mind not as a

suitable substitute for literal meaning which is

contextually impossible or incomprehensible, but rather as an

additional meaning that appears along with the first, literal,

meaning of the phrase in question without displacing it.

Likewise, this suggested meaning, Parvati's embarrassment,

without itself being in any way impossible or

incomprehensible, gives rise to a third "meaning" -- the

srīgara-rasa deriving from Parvati's love for Siva; here too

the suggested meaning appears along with the meaning which

gives rise to it, rather than displacing it.

So, the mechanism by which words, via their literal

meanings, convey figurative meanings differs markedly from

that by which they convey suggested meanings. The most

conspicuous features of figurative expression -- namely the

necessity of some obstruction to the understanding of the

literal sense and the replacement, rather than

supplementation, of the literal by the non-literal sense -- are

both absent in the case of suggestion. It is therefore not

possible to treat suggestion as identical with, or as a

subtype of, figurative expression.

If the term laksana51 is applied to cases in which one meaning indicates [laksayati] another without the awareness of itself being suppressed, then, this being the case, it would have to be accepted that figurative

51Here meaning figurative usage in general, rather than specifically non-metaphorical figurative usage.

235

expression itself is the primary function of words, since sentences, for the most part, reveal some meaning -- consisting of the speaker's intention -- apart from what is directly expressed by them. 52

The concept of figurative expression was devised to

account for extraordinary and seemingly aberrant usage. It

presupposes that words have a non-figurative sense in which

they are ordinarily used, and that the use of a word in some

figurative sense is thus the exception rather than the rule.

If figurative expression is understood so broadly as to

include what Anandavardhana would call suggestion, it becomes

a universal feature of human utterance rather than an

exceptional and aberrant mode of expression. We have already

seen that, for Anandavardhana, all human (i.e. non-Vedic)

sentences necessarily convey, in addition to their directly

expressed meaning, the communicative intent of their speakers.

If it is argued that this second, non-literal meaning is

conveyed by figurative expression rather than suggestion, then

the figurative use of words would be the norm, rather than an

exception. 53 Suggestion, then, cannot be identified with

figurative expression as it has ordinarily been understood.

52Ibid., p. 424.

53As we have seen (cf. above, Chapter Three, pp. 115-117) , the Bhatta Mimamsakas do, in fact, argue that all sentence meaning (both Vedic and non-Vedic) is conveyed figuratively. It is unclear whether or not Anandavardhana is aware of this contention. His objection seems to be based not on the logical impossibility of this view, but rather on the fact that it, counter-intuitively, deprives figurative usage of the exceptional character with which it is conventionally associated.

236

Having explained the distinction in nature (rüpa) between

suggestion and figurative expression, Anandavardhana goes on

to argue that they differ as well in the semantic content

(visaya) that they convey. Anandavardhana has identified and

illustrated three possible types of suggested meaning: vastu

(narrative element), alamkāra (figure of speech), and rasa

(aestheticized emotion). Among these, Anandavardhana contends

that no one has ever, or could ever, explain our understanding

of either rasa or alamkära as a kind of figurative meaning.54

It is true that rasa and alamkäras which are understood

without being directly stated had never before

Anandavardhana's time been explained under the rubric of

figurative expression. This resulted, however, not so much

from any insurmountable theoretical difficulty in doing so, 55

but because of the general bifurcation of literary and non-

literary language theory noted earlier. The theory of

figurative expression had been developed by MImamsakas,

grammarians, and logicians in the course of their

philosophical reflection on language. It sought to account

for the non-literal use of words in everyday or in Vedic

discourse. Rasa and alamkāra, as features of poetic language

exclusively, were unnoticed or ignored by non-Alamkārika

54Dhvanyaloka, pp. 425-6.

55As we shall see (below, Chapter Seven), Mukulabhatta, an early and influential opponent of the dhvani theory, advocates precisely the view here dismissed as impossible, explaining what Anandavardhana would call rasa- and alamkāra- dhvani as instances of figurative expression.

237

language theorists. Anandavardhana seems to take the silence

of previous theorists of figurative language on the status of

rasa and alamkāras not expressed directly as sufficient

evidence for the impossibility of explaining them as instances

of figurative meaning. He does not present any other argument

to justify their exclusion from the realm of figurative

expression. We shall see subsequently that this casual

dismissal is, at the very least, premature.

The analysis of cases of avivaksita-vacya dhvani, 56 in

which figurative expression and suggestion necessarily co-

occur, largely recapitulates the arguments given in the first

uddyota (cf. above, pp. 226ff), again stressing that it is the

beauty such expressions lend to the verses in which they occur

which distinguishes them from everyday or conventional

metaphors and which serves as the mark Of their

suggestiveness.

Avivaksita-vacya dhvani, although it depends upon the process of figurative expression [gunavrtti], is not identical with figurative expression. For figurative expression is seen even in the absence of suggestiveness. And suggestiveness cannot exist without some suggested meaning which is a cause c previously described .. of beauty in the manner But where this [figurative expression] does cause us to understand some beautiful suggested meaning, this will come about only through the introduction of suggestiveness, just as in the case of expressiveness 57

Our understanding of any suggested meaning is occasioned by

our prior understanding of some non-suggested meaning, either

56Ibid, pp. 429-436.

57Ibid., pp. 432-434.

238

directly expressed or figurative. In a case of vivaksita-

väcya dhvani, the co-occurrence of a suggested meaning with

the directly expressed meaning that gives rise to it does not

lead us to identify the two. Similarly in a case of

avivaksita-vacya dhvani, the co-occurrence of the suggested

meaning with the figurative meaning which gives rise to it

should not lead us to conclude that there is no distinction

between them.

That suggestion may operate in dependence on either

directly or figuratively expressed meanings is offered as a

further defense of its distinctness from both direct and

figurative expression.

Because this [suggestion] depends upon both of these [direct expression and figurative expression], it cannot be said to be identical with either. For, it cannot be identical with direct expression [vācakatva] because in some cases it functions in dependence on figurative expression [laksanā]; and it cannot be identical with figurative expression, established in dependence on direct expression. Nor is because elsewhere it is

it only because of its being a property of both [ubhayadharmatvena] that it cannot be identified with either, since it also exists as a property of sounds devoid of both direct and figurative expression. That is to say, suggestiveness having rasa as its content exists even in the sounds of a song; and neither direct nor figurative expressiveness is seen to belong to these in any way ... And thus in verbal usage there are three modes: direct expression, figurative expression, and suggestiveness. 59

Ānandavardhana thus claims to have established beyond doubt

that the poetic phenomenon he terms dhvani cannot be reduced

to or incorporated within either of the modes of semantic

58Ibid., pp. 427-429.

239

operation generally accepted by philosophers of language.

Since the process by which we understand, in cases of dhvani,

meanings not literally stated cannot be adequately explained

in terms of either of these well-known modes of signification,

it is essential that we postulate a third, previously

unrecognized, mode to account for such instances, namely

vyañjana

The Differentiation of Suggestion from Inference

There is, however, a further possible objection to the

introduction of vyañjanā as a third mode of verbal

signification. One might attempt to explain our understanding

of various unstated meanings in cases of dhvani as the outcome

of a process of inference (anumana), and thus avoid the

necessity of positing any such additional semantic mode. Our

understanding of such additional "meanings" then would not,

properly speaking, result from any mode of verbal

signification whatsoever. Rather, once we have understood the

meaning (whether directly expressed or figurative59) of a

particular verse, we arrive, by a further deductive process,

at an awareness of some fact (or figure of speech [alamkāra]

or aestheticized emotion [rasa]) which is not expressed by the

words of the verse per se.

59One might go so far as to explain even our understanding of figurative and directly expressed meanings as the result of an inferential process; the Buddhist logicians do so, and Mahimabhatta follows them in this respect (as in many others) - - cf. Vyaktiviveka, op. cit., pp. 80-81, 114ff; also below, Chapter Ten, pp. 476ff, 485-486.

240

Anandavardhana identifies this as the view of "those who

understand reasoning" [yuktivid] and "who believe that the

connection between words and their meanings is produced60"

(krtrimasabdarthasambandhavadin) .61 This seems to be meant

(as Abhinavagupta takes it) to refer to the views of both the

Buddhist logicians and the Naiyayikas; both believe the

connections between words and their literal meanings are

conventional; the Buddhists believe that these conventional

associations were established by human beings, the Naiyayikas

that they were created by God) .

What Anandavardhana calls "suggestiveness" is the

capacity of words to create in a hearer the awareness of

something above and beyond the meaning they explicitly convey.

This, the opponent contends, is nothing more nor less than

their ability to lead us to draw a certain inferential

conclusion.

The suggestiveness [vyañjakatva] of words is their ability to cause us to understand something [gamakatva], and this is the property of serving as an inferential mark [lingatva]; and therefore the understanding of a suggested meaning [vyangya] is only the understanding of an inferential conclusion [lingin]. Thus the relation between suggestive words and their suggested meanings is simply the relation between inferential marks and conclusions, nothing more. And this must certainly be accepted, since you yourself [that is, Anandavardhana]

6°Rather than eternal and uncreated, as the Mimāmsakas believe.

61Dhvanyaloka, op. cit., p. 444.

241

have just now62 explained that suggestiveness operates with respect to a speaker's intention [vaktrabhiprāya], and since the intention of a speaker is, in fact, inferred.63

Here, Anandavardhana's previous effort to show that

suggestion plays an essential, though undetected, role even in

non-literary language is turned against him. For the most

part, the sort of meanings Anandavardhana has resorted to the

theory of suggestion to account for are those which are unique

to poetry (rasas, implied figures of speech, and so on). As

previously noted, non-literary language theorists before

Anandavardhana's time, being concerned with the analysis of

either everyday or Vedic language, had generally ignored such

specifically poetic features of language. The question of

speaker's intention (vaktrabhipraya), however, had not been so

ignored; conventional, non-Alamkarika, language analysis would

explain our understanding of this communicative intent as the

result of inference64. Anandavardhana attributes a hearer's

apprehension of a speaker's communicative intent to the

process of suggestion and, in so doing, invites the objection

that this so called "suggestion" is nothing other than

62In the argument for the differentiation of suggestion from direct expression -- cf. Dhvanyaloka pp. 438ff, and above pp. 216ff.

63Ibid., p. 448.

64Or possibly, for the Bhātta Mīmāmsakas, arthapatti. In any case, the Buddhist logicians and the Naiyayikas, the theorists to whom these objections are attributed, would reject arthapatti as a distinct source of valid knowledge (pramana) and include it entirely within the domain of inference.

242

inference.

Anandavardhana's first response to this line of argument

is to provisionally and hypothetically accept the inclusion of

suggestion within the domain of inference, and to argue that

this would in no way vitiate the central contentions of the

Dhvanyaloka (just as he earlier contended that the inclusion

of suggestion within the domain of figurative expression would

do no real harm to his theory).

But, even if this were the case [that suggestion is nothing more than inference], what harm would this do us? We have argued that there is a mode of verbal signification, distinct from direct and figurative expression, called suggestiveness. Even if it were as you say, there would be no damage to this thesis. Let this suggestiveness be the property of serving as an inferential mark [lingatva], or something different. In any case it will be distinct from the well-known verbal modes and will be t the object of the operation of words65 -- thus there is really no disagreement between us.66

However, just as in Anandavardhana's discussion of the

inclusion of suggestion within figurative expression, the

argument that acceptance of the opponent's view would do no

real harm is coupled with a determined effort to refute this

view. Immediately following the passage just quoted

Ānandavardhana says, "But it is not really the case that

suggestiveness is only the property of serving as an

inferential mark and that the understanding of a suggested

65That is to say, our understanding of the inferred meaning in such a case is a result, albeit not an immediate one, of our hearing a particular series of words.

66Ibid., pp. 448-449.

243

meaning is only the understanding of an inferential

conclusion."

Anandavardhana's defense of a domain for suggestion

independent from that of inference begins with a very general

analysis of the content conveyed by language. 67 That which

we understand when we hear a speaker employ language is

broadly divisible into two categories: that which is inferable

(anumeya) and that which is to be communicated (pratipādya) .

The inferable content is the speaker's expressive intent

(vivaksā) 68; which may be further divided into two

components: the intention to produce sounds (which may be

inferred simply from the fact that the speaker is an animate

creature, whose observed behavior may be presumed to conform

to some intention), and the intention to convey some meaning

by means of these sounds (which may be inferred from the

speaker's utterance of a recognizable, meaning-bearing series

of sounds, i.e. particular words).

The communicated content of a speaker's utterance is the

specific meaning he or she intends to convey, and may likewise

be divided into two components: that which is directly

67Ibid., pp. 449ff.

68In admitting that the speaker's intention is inferable, Anandavardhana has of course already partly conceded the opponent's objection; his argument for suggestion as distinct from inference depends on the contention that not all suggested meanings are inferable.

244

expressed (vacya) and that which is suggested (vyangya) .69

This intended meaning is not understood through a process of

inference.

This communicated content of words, of both varieties, does not by itself [svarupena] become apparent as an inferential conclusion, but rather [appears] by means of some other connection, whether artificial or non- artificial.70 The fact that this meaning is the object of the speaker's intent is understood, by means of his words, as an inferential conclusion, but the form of this meaning is not.

That a speaker intends to convey some meaning can be inferred,

but what he intends to convey cannot.

If the meaning of a sentence were inferred from its

words, Anandavardhana contends, no disagreement would ever

arise regarding the truth or falsity of verbal statements. If

a valid inferential mark (linga) is clearly perceived, there

can be no doubt regarding the reality of that which we infer

from it; to take the stock example of inference in Sanskrit

philosophical literature, if we see smoke rising from a hill,

and infer from it the presence of fire there, no one could

69Here the directly expressed content must be understood to include figurative expression as well! Ānandavardhana is somewhat inconsistent in his treatment of figurative expression, sometimes, as here, seeming to regard it as merely a specialized subvariety of direct expression (cf. also Dhvanyaloka 1.18 [p. 151]), at other times treating it as a third mode of semantic operation, distinct from both direct expression and suggestion (ibid., pp. 428-429).

70Artificial: i.e. created by either human or divine convention, as the Buddhists and the Naiyayikas (respectively) contend. Non-artificial: i.e. natural and eternal, as the Mīmamsakas contend.

71Ibid., p. 151.

245

reasonably dispute the validity of this conclusion. If words

served as inferential marks for the meanings they convey,

these meanings, as valid inferential conclusions, would be

similarly beyond dispute. 72

Thus, the connection between words and their communicated

content must be recognized as distinct from that between

inferential marks and the conclusions drawn from them. And

this applies whether the communicated content is directly

expressed or conveyed by suggestion.

The suggested meaning, being implied by the capacity of the directly expressed meaning [vācyasāmarthyaksiptatayal, is connected with words in the same manner as the directly expressed meaning. For the mediacy or immediacy of this connection is of no importance73 Thus, words operate as inferential marks with respect to a suggested meaning only when it takes the form of the speaker's intent; with respect to the meaning which is the object of this intent they

72The argument does not really hold water. It is widely accepted in philosophical circles that what sentences (or, at least, non-Vedic sentences) convey is not information about the actual state expressive intentions of the speakers who utter them; only by of things, but about the beliefs or

a further process of inference can we draw conclusions about the actual state of the world on the basis of linguistic reports (i.e., if the speaker is judged to be reliable, we may conclude that what he tells us is accurate) -- cf., for example, Pramanavarttika 1.3-4 Sabarabhasya on 1.1.2 (op. cit., Vol. I, p. 18 -- cited above, (op. cit., pp. 2-3) ,

p. 217), and Salikanāthamiśra's Prakaranapañcikā (Benaras Hindu University Darsana Series 4, 1961), p. 264. that either directly expressed or The view

inferential mark which allows us to validly deduce the meaning inferred suggested meanings are demands only that the sentence serve as an

or meanings the speaker wishes to convey -- it need not be sufficient to allow us to validly deduce the truth of this intended meaning.

73That is, of no importance for purposes of distinguishing it from the connection between an inferential mark and the conclusion drawn from it.

246

operate as toward [pratipādya] .74 something to be communicated

Anandavardhana has conceded the opponent's argument in part:

upon hearing any human utterance, our understanding of the

speaker's intention is arrived at by a process of inference

(although he still describes it as vyangya ["suggested"] ) .

Thus the domain of suggestion as Anandavardhana has described

it is partly reducible to inference -- but only partly. Apart

from the single exception of the speaker's communicative

intent, suggested meanings, being part of the communicated

content of language, cannot be said to be inferred. And it is

only the latter, non-inferable, type of suggested meaning that

is important for Anandavardhana's analysis of poetry.

Any human utterance will invariably lead its hearers to

understand that its speaker wishes to communicate something,

but this awareness is incidental to the communicated content

of this utterance; no instance of such an inferable suggested

meaning would ever merit the designation dhvani. Discussion

of the role of suggestion in our understanding of the

communicative intent of speakers in everyday, non-poetic

discourse was, Anandavardhana explains, introduced only to

convince the Mimamsakas that words must have some suggestive

capacity in addition to, and distinct from, their ability to

convey the literal meanings with which they are invariably and

eternally associated. So even if it is granted that suggested

74Ibid., pp. 452-453.

247

meaning, when it takes the form of a speaker's intent, is

inferable, this will have no consequences for the analysis of

suggestion as it is deliberately employed for aesthetic effect

in poetry. It is dhvani that is of primary concern to

Anandavardhana, and in cases of dhvani the process of

suggestion, for the reasons given above, will not be reducible

to inference.75

Anandavardhana's Relation to Linguistic Philosophy

Anandavardhana's overall approach to philosophically

based criticism of the theory of suggestion is unusual in

several respects. We have already seen that in his encounter

with general linguistic theory Anandavardhana often proceeds

somewhat tentatively. In his discussions of figurative

expression and inference in their relation to suggestion he is

unwilling to rely totally on the arguments he has offered in

defense of the existence of suggestion as an autonomous

semantic function. In both cases, he couples his defense of

the autonomy of suggestion with the claim that this autonomy

is not, in any case, necessary; even if one accepts that

suggestion is merely a type of figurative expression or

inference, the existence of dhvani and its central importance

in poetry will not thereby be called into question. At the

75The concession Anandavardhana here makes that the speaker's intent is inferable would seem to totally undercut the argument earlier made that the Mimamsakas must accept vyañjana in order to account for our understanding of such intent in non-Vedic language, but no notice is taken of this difficulty by either Abhinavagupta. Anandavardhana or his commentator

248

conclusion of his detailed discussion of philosophical

objections to the existence of suggestion he returns once

again to this concessive approach:

Even if it is stubbornly asserted that this [suggestion] is included within these [already well-known modes of signification -- direct expression, figurative expression, etc.], nevertheless, if an examination is made of dhvani as a particular sub-type of one of these, in order to remove disagreement and instruct people of taste [sahrdayas], this should not be criticized. For merely by the definition of something in general one does not preclude the definition of specifics, when this is useful. If this were so, once existence itself was defined, the definitions of all existing things would become redundant. 76

Anandavardhana takes great pains to offer arguments which he

feels adequately establish the autonomous existence of

suggestion, but recognizes that these arguments are unlikely

to prove entirely convincing to the adherents of the various

existing language theories, and attempts to justify his

approach to the analysis of poetic language even in the eyes

of those who refuse to accept suggestion as a separate

semantic function.

Anandavardhana's defense of suggestion as a mode of

expression unique to poetic language (or, at the very least,

one which plays a significant role and demands analysis only

in its poetic applications) requires that he consider the

objections offered by adherents of various general, non-

poetic, language theories, but he manages to do so without

himself making a specific commitment to any general language

76Ibid., p. 457.

249

theory. He neither clearly adheres to, either as is or in

some modified form, any of the well-known language theories

offered by the grammarians, Mīmāmsakas (Bhatta and

Prabhakara), and logicians (Buddhist and Nyaya), nor offers

his own theory of everyday, non-poetic language. He takes no

stand on the major linguistic controversies of his day (e.g.

on whether the connection between words and their literal

meanings is natural or conventional, on whether words convey

their individual meanings before or after they are construed

as sentences [abhihitanvayavada vs. anvitābhidhanavāda],

etc.). Rather than attempt to articulate and defend a

particular theory of language, Anandavardhana endeavors to

show that adherents of any of the major current language

theories, if they wish to account for the poetic phenomena

Anandavardhana describes under the rubric of dhvani, must

accept suggestion as a supplement to whatever semantic

functions their theories already allow for. Thus he manages

to argue for the general acceptance of suggestion by all

language theorists while remaining aloof on those questions

which divide these language theorists among themselves.

It is also important to note that while Anandavardhana

contends that adherents of all language theories must accept,

in addition to the well known semantic modes of direct and

figurative expression, a third mode, namely suggestion, he

does not provide a detailed and specific account of the

mechanism by which words convey their suggested meanings.

250

The specific processes by which words convey their

directly expressed and figurative meanings had been analyzed

in great detail by earlier linguistic philosophers. In the

case of direct expression, philosophers had for many centuries

debated regarding the ontological status of the connections

between words and their literal, directly expressed meanings,

the origin (if any) of these connections, and the processes by

which new members of the linguistic community acquire an

understanding of them. While the various schools of language

philosophy differ widely in their views on these matters, each

provides some specific account of the nature of these

connections and the manner in which they come to be understood

and employed by members of the linguistic community.

The exact processes by which words convey figurative

meanings had been subjected to similarly intense scrutiny.

The Mimamsakas (whose account of the mechanism of figurative

expression seems to have gained widespread acceptance)

determined the necessary conditions which must apply for a

word to express a particular figurative meaning: there must be

some obstacle (badha) to our understanding the word literally

in a given context, the figurative meaning must be connected

with the literal one in some way, and the figurative sense

must have some useful application (prayojana) in the given

context. They attempt to specify the types of connection

which may obtain between the figurative and literal referents

of a word (similarity, physical proximity, part/whole

251

relation, etc.) .

Anandavardhana does not attempt to give such a specific

account of the necessary preconditions required for a word or

group of words to convey a particular suggested sense in a

particular context. His argument in defense of suggestion

proceeds, , rather, by process of eliminantion. If it is

granted that a particular verse or expression conveys some

sense above and beyond its literal one, and if our

understanding of this additional sense cannot be explained by

either of the well-known semantic modes (or by some extra-

linguistic mental process such as inference), then some

additional semantic mode must be operative in such cases.

Thus, by demonstrating the impossibility of accounting for

dhvani in terms of any of the recognized semantic modes,

Anandavardhana is. able to assert the necessity of suggestion

as an additional semantic function without explaining

precisely how suggestion works.

All these aspects of the treatment of philosophical

issues in the Dhvanyaloka serve to reinforce the impression

that these issues are in some way peripheral to

Anandavardhana's primary concerns, and are dealt with more

from necessity than from choice. Anandavardhana's insistence

that dhvani represented a previously unrecognized phenomenon

of poetic language (rather than merely a re-description of

phenomena already dealt with by earlier Ālamkārikas) led him

to argue for the existence of a previously un-theorized realm

252

of meaning, the "suggested". This posed a challenge not only

to orthodox Alamkārasāstra, but to linguistic philosophy as

well, as all existing philosophies of language claimed to

account fully and adequately for the phenomena of language and

none (currently, at least) allowed for the existence of

suggestion.

Because up to this time the Ālamkārikas' analysis of

poetic language had proceeded for the most part independently

of the analyses of non-literary language offered by the

philosophers, the agreement or disparity between their models

of language and those offered by general language theorists

was r not seriously examined. Anandavardhana, because he

introduces concepts and analytic techiniques drawn from

philosophical language analysis into the analysis of poetry

and thereby to a great extent bridges the conceptual and

terminological divide which had so far existed between

literary and non-literary language theory, is forced to

consider more carefully those respects in which his own model

of language diverges from those of various theorists of non-

literary language, and to defend this divergence against the

objections offered or likely to be offered by these theorists.

In his treatment of these objections, however,

Ānandavardhana tries to maintain a certain distance between

Ālamkarasastra and the philosophy of language. He employs the

techniques of philosophical language analysis in his own

examination of poetry, but does so without himself attempting

253

to become a philosopher of language. By defending the

existence of a mode of signification whose significance is

limited to poetic language, and, conversely, by refusing to

take any definite stand on the controversies that divide

language philosophers among themselves, he keeps the domain of

poetic language analysis separate from that of the analysis of

everyday and Vedic language. Anandavardhana's arguments

against the theorists of non-poetic language are not designed

to challenge or replace their theories of language, but to

ensure that, alongside these theories, a place is secured for

his own special theory of poetic language, and for suggestion,

the specially poetic mode of signification which it takes as

its object.

Prospectus

Ānandavardhana's arguments for suggestion as a new and

distinct function of language are, for the most part, totally

ignored by later philosophers of language. The one notable

exception is the great Kashmiri logician Jayantabhatta (c. 900

AD) . In his Nyayamañjari, after refuting the Mimamsakas'

arguments for the existence of arthapatti as a means of valid

knowledge distinct from inference, Jayantabhatta briefly

considers, and summarily dismisses, Anandavardhana's dhvani:

And by virtue of this capacity [samarthya] of words [to lead us to infer some unspoken meaning"], we may also

77On account of which, as Jayantabhatta has just explained, the supposed examples of arthapatti given by the Mīmamsakas can be explained as inferences (cf. Nyayamañjarī, ed. Gaurinath Sastri, Sivakumārasāstri Granthamālā, 1982, Vol.

254

exclude that "dhvani" which another person, himself wise, has accepted. thinking The understanding of a prohibition arises from a command, and awareness of a command from a prohibition: "Wander confidently, Monk ... 11 78 "Traveller, do not enter my house .. #f 79 In each case, it is the capacity of words -- which indicate the form of things which are qualified by some other means of knowledge [i.e. inference] -- which functions in this way.80 In any case, this discussion with poets is not appropriate; even learned people become confused in this difficult path of sentence meaning.81

For Jayantabhatta, Anandavardhana's dhvani is little more

than a minor variation on arthāpatti (which, for the

Mīmāmsakas, provides the basis for all types of figurative

usage), and requires nothing new by way of refutation. The

theory of dhvani draws his scorn, in large part, because of

its source. The analysis of sentence meaning and sentence

cognition poses serious problems even for dedicated

philosophers, and Jayantabhatta seems to regard it as

presumptuous for a mere poet to deal with such questions. The

1, pp. 67-68).

78Anandavardhana's first example of dhvani -- cf. above p. 128.

"I cannot definitely trace the source of this example; possibly a condensed paraphrase of Rudrata 7.41. In any case, it is meant to exemplify a common motif in single-verse compositons, in which a housewife whose husband is absent overtly sends away a traveller seeking lodging, but in such a way as to imply, covertly, that she wishes him to return in secret so they can make love.

80That is to say: words convey their own meanings, and under appropriate conditions these meanings allow us to infer other, unstated, meanings. Thus the words have the capacity, via an intervening process of inference, to convey meanings other than those with which they are directly associated.

81Ibid., pp. 76-78.

255

general silence of other philosophers on the question of

dhvani, even long after it comes to be universally accepted by

the Alamkarikas, would seem to indicate that they share this

dismissive and contemptuous attitude.

Within the field of Alamkarasastra, however, the

philosophical issues dealt with by Anandavardhana assume

increasing prominence in the ensuing centuries. The question

of the distinctness of suggestion from other modes of

signification becomes the major point of controversy in

Kashmiri Alamkārasāstra from Ānandavardhana's time to that of

Mammata (c. 1100 AD). Anandavardhana's more fundamental

innovation in the realm of Alamkarasastra -- the assumption

that in any aesthetically coherent poem all constituent

elements must be directed toward a single, overriding goal,

typically the effective communication of a single, predominant

rasa -- is, for the most part, accepted even by his harshest

critics. It is the more specific question of whether dhvani

is or is not explicable without positing suggestion as a newly

discovered mode of signification that increasingly absorbs the

attention of Anandavardhana's supporters and detractors during

this period.

Subsequent opposition to the introduction of suggestion

proceeds largely along the lines laid down by Anandavardhana

himself. Of the three major objections to the theory of

suggestion set forth in the Dhvanyaloka and briefly outlined

above, two find serious adherents among the anti-dhvani

256

Ālamkarikas. The inclusion of "suggestion" within the domain

of figurative expression is asserted by Mukulabhatta in his

Abhidhavrttamatrka, and Mahimabhatta, in his Vyaktiviveka

argues that all supposed cases of suggestion may be explained The as inferences. The argument for the inclusion of suggestion

within a single, all-encompassing expressive function

(vacakatva) has no adherents in the extant Alamkarasastra

literature, but is regularly re-examined and rejected by later

Ālamkārikas. 82

This increasing absorption in the philosophical issues

and arguments raised by the theory of suggestion shifts the

focus of attention in Alamkarasastra from the composers of

poetry to its audience. Both critics and adherents of

Anandavardhana are largely in agreement about the content

conveyed by poetry (both generally and in particular example

verses); it is the specific cognitive process by which one

understands this content that is disputed. Thus the efforts

of both supporters and detractors of the theory of suggestion

are directed primarily toward a careful analysis of the

psychology, and even the epistemology, of poetic

comprehension.

Although it is Anandavardhana's own importation of

philosophically derived models and techniques of language

82Both adherents and detractors of the theory of suggestion. Cf. Locana on Dhvanyaloka, op. cit., pp. 64-67, 413-422, 438-443; Vyaktiviveka, Kashi Sanskrit Series 121, pp. 137-141; Kāvyaprakāsa, Sārasvatī Bhavana Granthamālā 46, pp. 154-158.

257

analysis into the field of Alamkārasāstra that precipitates

this shift in focus from the process of poetic composition to

that of poetic comprehension, this transformation is somewhat

at odds with the approach to poetic analysis he seems to

strive for. We have seen that the greatest strength of

Ānandavardhana's teleological model of literary aesthetics is

that it provides, for the first time in the history of

Alamkārasāstra, a general criterion in terms of which the

appropriateness of particular elements of a poem (ranging from

individual phonemes to broad features of plot organization) to

the whole composition of which they form a part may be judged.

This creates important new possibilities in the area of

literary criticism and, as we shall see in the next chapter,

Anandavardhana himself expoits these possibilities by engaging

in critical analyses of existing works of Sanskrit literature

in a wholly unprecedented manner. This kind of literary

criticism is, for the most part, marginalized in later

Alamkāra literature, in large part because of the transference

of attention from processes of composition to processes of

comprehension.