Philosophy of Language in India
PHILOSOPHY OF LANGUAGE IN INDIA
The earliest Indian thinking about language, found in Vedas (Arapura and Raja 1990), is speculative, but later discussions involve sophisticated arguments among various schools of thought. These discussions, which concern speech units (Sanskrit śabda, "sound, speech element, word") and associated meanings (artha ), share certain themes. One is epistemological. Sounds are evanescent; an instant after they are pronounced they disappear. Consequently, the question arises: How can one rightly speak of complex units like words (pada ) and sentences (vākya ) as perceptible entities? Similarly, though one speaks of actions and things involved in them, it is also arguable that acts and things which are thought to be perceived as wholes actually are not so; there is a stream of instants, none of which lasts long enough to enable a qualified cognition of complex external entities. How, then, can one maintain that speech units signify actual actions and things? The second point concerns theory and procedure. Indian scholars operate with constructs in order to account for facts and behavior. This approach was evident already at an early period (ca. 7th c. BCE), when Vedic scholars posited constructed analyzed texts (padapāṭha ) from which the Vedic texts as continuously recited (saṁhitāṭha ) were derived by rules.
Indian thinkers accept certain means of acquiring knowledge, referred to as pramāṇa (a derivate of pra-mā [3rd sg. pres. pramimīte ], apprehend"). At least two pramāṇa s are generally accepted: direct perception (pratyakṣ ) and inference (anumāna ). A third, verbal transmission (śabda āgana ), is accepted by others, including Patañjali's yoga system. A means of knowing through similarity of one thing to another (upamāna ) makes up a set of four pramāṇa s adopted by a major school of logicians, Nyāya. Not all thinkers, however, accept śabda/āgama as a separate pramāṇa ; some account for knowledge acquired verbally through inference.
Meaningful Units and Symbols
Systematic speech sounds—vowels (a, ā, i, ī, etc.) and consonants (k, kh, g, gh, ṅ, etc.) are distinguished from mere sounds (dhvani ) such as the noise made by a drum. Classes of larger units are also recognized, the major ones being nominal forms (nāman ), verbs (ākhyāta ), preverbs (upasarga ), and particles (nipāta ); for example, gauḥ (nom. sg.), "cow, ox," gacchati (3rd sing. pres.), "goes, is going," upa in upa gacchati, "approaches," and vā, "or," respectively.
At an early stage, represented in pre-Pāṇinian texts and alluded to in later works like Patañjali's great mid-second-century BCE commentary (Mahābhāṣya ) on Pāṇini's c. fifth-century BCE grammar, verbs and nouns were defined semantically. In one view, verbs signify varieties of being (bhāva ): something comes into being (jāyate, "is born"), continues to be (asti, "is"), undergoes change while remaining the same entity (viparaṇamate, "changes"), increases (vardhate, "grows"), decreases (apakṣīyate, "diminishes"), and ceases to be (vinaśyati, "perishes"). Some scholars reduce these to three stages, with the second encompassing the third, fourth, and fifth. Alternatively, verbs are considered to signify particular actions (kriyā, karman ), the most general action being signified by kṛ, "do." This definition is supported by usage: (1) devadattaḥ pacati, "Devadatta is cooking," is an appropriate answer to (2) kiṁ karoti devadattaḥ, "What is Devadatta doing?" These two views are superseded by considering that whatever a verbal base (dhātu ) signifies—now spoken of as kriyā or bhāva —is conceived of as involving continuity in time, always associated with some time. As a consequence, not only terms such as kṛ (karoti ), "do," pac (pacati ), "cook," and vraj (vrajati ), "go," but also ones like as (asti ), "be," ās (āste ), "be seated," and sthā (tiṣṭhati ), "come to a stand, be in place," are now part of a single class of units signifying kriyā/bhāva. The canonical statement of this position, which can be seen already in the Mahābhāṣya, appears in Bhartṛhari's mid-fifth-century Vākyapadīya (3.8.1): whatever is always spoken of as something to be brought to accomplishment, whether it is already accomplished or not, is referred to as kriyā ("action") by virtue of its taking on a sequential status.
Contrasting with such semantic definitions, there is a formal approach, epitomized by the grammarian Pāṇini, who assigns to his class of units called dhātu verb bases listed in an appendix to the main corpus of rules and to items derived from both verbs and nominal forms (Cardona 1997).
There is a conception of units under which words are groups of sounds and larger units are groups of words. This view is represented in the section of Kauṭilya's Arthaśāstra (disputed date but not later than the third century CE) that deals with writing edicts (2.1.13–14) and in other works. It is already reflected in an argument Kātyāyana (third century BCE) mentions when he speaks of a group of sounds (varṇasaṅghāta ) as being meaningful (arthavat ). The same view is presupposed in the Ṛgvedaprātiśākhya (2.2), which describes the continuous recitation of a text (saṁhitā ) as consisting in one's continuously putting together the last sounds of words (padāntān ) with the initial sounds (padādibhiḥ ) of following words, without any temporal separation (kālāvyavadhānena ). Pāṇini himself (1.4.109: paraḥ sannikarṣaḥ saṁhitā ) states that the maximum drawing together (paraḥ sannikarśaḥ ) of sounds is called saṁhitā. One may consider that such a procedure accepts that sounds do actually come together to form larger units, but it is also possible that this is an artifact necessary for the proper description by rules of what is found in a language.
From an argument presented by Kātyāyana and Patañjali, it is clear they were aware that one cannot speak of physical sounds truly co-occurring in immediate succession, for example, gauḥ, "cow" = g -au -ḥ. For speech does not produce two sounds simultaneously, since sounds (varṇānām [gen. pl.]) have the property of disappearing immediately upon being pronounced (uccaritapradhvaṁsitvāt [abl. sg.], Kātyāyana's vārttika 10 on 1.4.109: … uccaritapradhvaṁsitvāca varṇānām ). As Patañjali explains: when g is pronounced, there is no au or ḥ ; when au is pronounced, there is no g or ḥ ; and when ḥ is pronounced, there is g or au.
If there is no physical composite unit such as gauḥ, the question arises: What is it that it is understood to signify? Two approaches were taken on this issue. One involves memory. It is assumed that when a sound is perceived, this experience leaves in one a lasting trace (saṁskāra, vāsanā ); the last sound uttered in a given stretch produces a cognition accompanied by the traces left from preceding cognitions of sounds, and this final cognition is what produces an understanding of meanings of words and the sentences they make up. Alternatively, sounds are considered merely to manifest (vyañj ) actual meaning bearers. These signifiers are posited elements called sphoṭa, distinct from physical sounds but manifested (vyaṅgya, "to be manifested") by them. Three major sphoṭa types are assumed: sentence (vākyasphoṭa ), word (padasphoṭa ), and subword meaningful elements (varṇasphoṭa ) such as bases and affixes.
The first of these views was proposed at least by the time of the Mīmāṃsaka commentator Śabara (second century) and was accepted by adherents of different schools. The sphoṭa theory was first expounded fully by the grammarian-philosopher Bhartṛhari and remained basically the position of grammarians. Each of these positions was subjected to criticisms. Arguments against the first revolve about the nature of memory, what is recalled, and in what manner; the main argument against the sphoṭa position is that it requires positing units which one can do without.
Speakers and hearers communicate and understand messages by means of words and sentences of a language they share. It is therefore universally accepted that a relation (sambandha ) holds between words (śabda, pada ) and meanings (artha ) and that this relation can be a direct or indirect one, respectively called śakti ("capacity") or abhidhā ("signifying") and lakṣaṇā ("secondary meaning relation, metaphor"). A term that directly signifies (vācaka ) a meaning is qualified as śakta ("capable") and its meaning as śakya, the object of this capacity. For example, gaṅgā directly refers to a flow of water, the river Gaṅgā. By lakṣaṇā, the same term can refer to the banks of the river. Thinkers of different schools engaged in arguments concerning both the nature of what is signified and the relations that link words and their meanings.
Concerning what words signify, at one extreme there is the view that terms like ghaṭa, "clay pot," aśva, "horse," pac, "cook, bake," refer to actual external entities, including actions one can witness. Other positions start from the observation that what one can actually perceive is not such an external thing (vastu ) or action (kriyā ): The latter is a stream of moments (kṣaṇa ) that are beyond direct perception, and the former also can be broken down into such moments. The putative wholes treated as having identity are mental constructs (vikalpa ).
One view consequent on this observation, adopted by certain Buddhistic thinkers, is that signification applies negatively, being a removal or differentiation (apoha ) of all that is not the momentary entity in question, which is thus differentiated (apoḍha, "removed") from all others. The relation between a word—itself a construct—and its significand is then one of cause (kāraṇa ) and effect (kārya ): Words have mental constructs as their sources and bring about a comprehension of mental constructs. Although they accept that words and their meanings are related as signifier and significand (vācyavācakabhāvasambandha ), Pāṇinian grammarians such as Bhartṛhari—with earlier precedents—also consider the cause and effect relation acceptable and conceive of the significands as word-meanings (śabdārtha ) that are mental (bauddha ) and not necessarily external objects (vastu ).
In this connection, grammarians speak of a vivakṣā, a desire to speak about things in a particular manner. For example, it is in the nature of things that a sword (asi ) serves as a means of cutting; one says, for example, (3) devadattḥ asinā chinatti, "Devadatta is cutting with a sword," using the instrumental asinā to refer to a sword as a means. If a sword is quite sharp, one may also appropriately say (4) asiḥ sādhu chinatti," the sword (asiḥ, nom. sg.) cuts well (sādhu )," speaking of a sword as an agent of cutting in the same way that (3) refers to Devadatta as an agent. In order to account for such usages, Pāṇini orders a group of rules that assign direct participants in the accomplishment of actions (kāraka ) to particular categories in such a manner that the participant in question is eligible to be assigned both to the category of participants called karaṇa, "instrument," by virtue of being the means (sādhakatama, "which is means more than any other") of accomplishing an act and, by a later rule, to the category of agents (kartṛ ) by virtue of being an independent (svatantra ) participant. Since a sword cannot be spoken of as an independent participant in the act of cutting without one's simultaneously considering it a means used by someone, this involves a conflict (vipratiṣedha ), and Pāṇini provides explicitly that in case of such conflicts, what is provided for by a subsequently stated rule takes precedence over the provision of a preceding rule (see Cardona 1974). In connection with such situations, Patañjali notes that this involves what he calls laukikī vivakṣā, "communal desire to speak"; that is, it is not a matter of individual preference but of the way a community of speakers (loka, "world") expresses itself.
There is also the point of view that words have a natural relation of fitness (yogyatā ) with their meanings, comparable to the fitness of different sense faculties with respect to what is perceived. Moreover, words and their meanings are commonly identified with each other.
However one conceives of the relation, each generation acquires a knowledge of words related to their meanings by observing how people interact. For example, a child witnesses an interaction between his father (F) and his grandfather (G): G says (5) gām ānaya, "bring the cow," to F, who then brings a cow, but F brings a horse when G says to him (6) aśvam ānaya. The child learns therefrom that gām and aśvam respectively designate a cow and a horse. This is an instance of reasoning from concurrent presence (anvaya ) and absence (vyatireka ): (a) x → y, (b) x̄ → ȳ. If both hold, then x which precedes y is its cause. Thus, if a given meaning is understood when a given term or member of a set of terms is used and not understood when this is not used, then the comprehension of the meaning in question is said to be caused by the term, to which this meaning is attributed.
Assuming that words designate positive significands, in ordinary usage one thinks of the term go, "cow," as referring to something that one can see and speak of repeatedly, using the same term. Moreover, in order to account for the repeated cognition of a cow each time one is seen, which can be verbalized saying (7) iyaṁ gauḥ, "this is a cow," it is assumed that each cow belongs to a class characterized by a class property (jāti, "generic property") that inheres in every member: gotva ("the property of being a cow").
If one assumes that a word-meaning relation is learned between an instance of the term go and a particular cow and also assumes that when another instance of go is used it too can refer to this particular cow, then the reasoning procedure shown above is violated, since one now has y in the absence of x. To assume that a separate relation is grasped between each instance of go and each individual (vyakti ) cow has the consequence that no speaker can acquire the knowledge of such an infinite number of relations. Various solutions are proposed to remedy the situation (see Deshpande 1992 and Scharf 1996). One view, espoused by Mīmāṁsakas, is that the primary word-meaning relation is between a term and the class property (jāti ). A sentence such as (5) is used, however, with the intention (tātparya ) that someone bring a cow, not a class property. This is accounted for by assuming that in such an utterance gām signifies not only a class property, through a primary relation (śakti ), but also a particular cow, through the secondary relation called lakṣaṇā. An alternative to this position is adopted by grammarians and logicians of the Nyāya school: A term like go signifies an individual (vyakti ) qualified (viśiṣṭa ) by its class property.
There are other instances where lakṣaṇā is said to operate. Consider (8) kuntān praveśaya, "have the javelins (kuntān [acc. pl.]) come in (praveśaya )." Praveśaya is a form (2nd sg. imper.) of a causative verb whose non-causal is praviś (3rd sg. pres. praviśati ) "enter." Javelins cannot enter a room of themselves, so they cannot be caused to perform this act in the same way that one can cause people to enter a room. In order to make sense of the intent (tātparya ) of a speaker who uses (8), it is accepted that kunta here bears a secondary relation with the men who bear javelins. In the same vein, consider (9) gaṅgāyāṁ matsyāḥ, "there are fish in the Gaṅga," and (10) gaṅgāyāṁ ghoṣaḥ, "there is a dairy colony on the Gaṅga." Assuming that gaṅgāyām (loc. sg. fem.) is used to refer to a locus in or on which something is located, (9) makes immediate sense, but (10) is hard to understand: fish can live in a river but a village of dairymen cannot be located physically in or on a body of flowing water. It is assumed, then, that in (10) gaṅā, which bears a primary word-meaning relation with a river, now bears a secondary relation with its bank (tīra ).
(10) involves an assumed semantic incompatibility such that it is not possible for the primary meanings of gaṅgāyām and ghoṣaḥ to be related. However, it is not sufficient to say that what prompts one to understand a secondary meaning here is solely the impossibility of the referents being connected (anvayānupapatti ). For this could be resolved also under the assumption that ghoṣa has a secondary relation with fish, so that (10) is understood to say what (9) says. Yet this is not the case: A person who hears (10) understands it to say that a dairy colony is located on the edge of the Gaṅga. Accordingly, the major reason prompting a secondary word-meaning relation is considered to be the impossibility of reconciling the primary meaning of a term with the intention (tātparya ) of a speaker.
Understanding (8) and (10) in the way shown involves setting aside the primary meanings of kunta and gaṅgā. Consider now (11) arko' staṁ gataḥ, "the sun (arkaḥ ) has set (gataḥ, "gone," astam, "home")." This can have its literal meaning. Without rejecting this meaning, moreover, there are several possible meanings that can be suggested (vyaṅgya, "to be made manifest"), depending on contexts and the persons uttering (11). For example, a go-between saying this to a woman who is to meet a lover suggests it is time to set out, but a servant saying this to a Brahmin means to imply that it is time for his master to perform the evening prayer. Another function of words is therefore considered, called vyañjanā (usually translated "suggestion"). This is principally accepted by theoreticians of poetics, though later Pāṇinian grammarians accept it, mainly because under the theory that a meaning bearer is a sphoṭa, which is manifested (vyaṅgya ) by physical sounds.
Sentence and Sentence Meaning
Adherents of various schools of thought in ancient and medieval India adopted different views concerning sentences and their meanings. One position—most systematically elaborated and defended first by Bhartṛhari—is that the true unit of communication is an atomic (akhaṇḍa ) sentence (vākya ), associated with an equally atomic sentential meaning, considered to be the object of a single flash of knowledge, hence referred to as pratibhā ("flash"). This thesis can be justified in so far as actual communication involves whole utterances, but it encounters the problems mentioned earlier in connection with words and their meanings: it is not possible for one to acquire a knowledge of all relations between all possible atomic sentences and their meanings. Moreover, a grammarian's aim is to give a generalized description of all possible sentences in terms of their structures, both formal and semantic, which is impossible if this thesis is taken strictly. Hence, Pāṇinians agree that at least one lower level—of words—must be accepted in terms of both language learning and description. They maintain, however, that words and their constituent bases, affixes, and so forth are constructs posited in order to account for whole utterances.
Under another view, held by some Mīmāṁsakas, there are no sentences qua distinct meaningful units. The sentential meaning of any stretch one calls a sentence is now accounted for indirectly, through the meanings of individual words. A parallel is drawn with the effect produced by utterances such as (12) putras te jātaḥ, "You've had a son" (putraḥ [nom. sg.], "son," jātaḥ [pptcple. nom. sg. m.], "born," te [dat. sg.], "to you") or (13) garbhiṇī te duhitā, "Your (te [gen. sg.]) daughter (duhitā ) is pregnant (garbhiṇī )." Each of the words of these sentences signifies its particular meaning. These word meanings are then related to each other in accordance with the speaker's intention (tātparya ) to convey a message and the hearer's semantic expectation (ākāṅkṣā ) that each meaning has to be linked to other meanings of words in the utterances. The effects are happiness on the part of a man who learns he has had a son and sadness on the part of a man who learns his unmarried daughter is pregnant. Similarly, the words of all utterances denote only their individual meanings, which are then related to each other. An intermediate position is taken by logicians of the Nyāya system, who consider that the meaning of an utterance is apprehended through the intermediary of related words: The first word is first cognized as shown earlier, with the consequent memory of the word-meaning relation and a memory trace of the word and its meaning, then this process is repeated until, with the perception of the last word, a cumulative memory trace results of all the words and their meanings related to each other.
Whatever position one takes, two requirements apply to sentences. First, constituents must be in proximity (āsatti ): each word following the first word of a sentence is uttered immediately after the preceding word, without the intervention of any term that is not syntactically related to the others. Secondly, there must be semantic expectancy (ākāṅkṣā ), so that a hearer expects that the meaning signified by a word such as gām in (5) is connected with an action, since it contains an object-signifying suffix, and ānaya requires an object. As shown, the intention of a speaker (tātparya ) also comes into play. Another requirement must be met if one is mainly interested in an utterance's serving as a means of conveying true knowledge: semantic compatibility (yogyatā, "the property of being connectible"). For example, each word of (14) agninā puṣpāṇi siñcati, "… is irrigating (siñcati ) flowers (puṣpāṇi [acc. pl.]) with fire (agninā [instr. sg.])," conveys a meaning that is immediately understood. (14) cannot, however, convey a meaning acceptable in our world, where the act of irrigating requires a liquid (dravadravya ) as a means. Accordingly, Naiyāyikas would deny that (14) has the status of pramāṇa. One might go so far as to deny that (14) produces a verbal cognition (śābdabodha ). Against this, the following is pointed out. Upon hearing (14), a person would respond by asking how one can speak of irrigating with something that is not a liquid? The hearer has indeed related the meanings of the words in the well-formed utterance (14) according to their syntax, but the resulting sentence meaning is not acceptable in the world as we experience it.
Adherents of different schools differ also concerning the ways in which verbal cognitions (śābdabodha ) are portrayed. Pāṇinian grammarians, logicians of the Nyāya school, and Mīmāṁsakas of the Bhāṭṭa school can agree that (15) devadattaḥ kaṭaṁ karoti, "Devadatta is making (karoti [3rd sg. pres.]), a mat (kaṭam [acc. sg.])," speaks of a given man making a mat. On the other hand, they give different paraphrases reflecting what they consider to be the śābdabodha prompted by this sentence, reflecting the preoccupations and theoretical premises of different schools of thought (see Cardona 1975 and Matilal 1985). Pāṇini accounts for the structure of Sanskrit through a set of derivational rules starting from semantics, and this is most efficently done under the assumption that the principal meaning of (15) is the action. Naiyāyikas are principally interested in the values of utterances as conveyors of valid knowledge, and within this system they operate with subjects and predications, so that the main qualificand in (15) is the person referred to by devadattaḥ. Mīmāṁsakas deal chiefly with the exegesis of Vedic utterances related to ritual performance, and in this context the principal meaning of an utterance is the act of bringing about a result.
These different interests and the fact that adherents of these systems and others either accepted the authority of Pāṇinian grammar or reacted to it led to ongoing arguments and counterarguments, with successive refinements over millennia, making India a center for the intense study of language and the philosophy of language.
See also Brahman; Knowledge in Indian Philosophy; Liberation in Indian Philosophy; Logic, History of: Logic and Inference in Indian Philosophy; Mind and Mental States in Indian Philosophy; Truth and Falsity in Indian Philosophy; Universal Properties in Indian Philosophy.
Aklujkar, Ashok. 1990. "Bhartṛhari: Trikāṇḍī or Vākyapadīya, with Vṛtti on Books 1 and 2." In The Philosophy of the Grammarians, edited by Harold G. Coward and K. Kunjunni Raja, 172–174. Princeton, NJ: Princeton University Press, 1990.
Arapura, John G., and K. Kunjunni Raja. "Philosophical Elements in Vedic Literature." In The Philosophy of the Grammarians, edited by Harold G. Coward and K. Kunjunni Raja, 101–106. Princeton, NJ: Princeton University Press, 1990.
Biardeau, Madeleine. Théorie de la connaisance et philosophie de la parole dans le brahmanisme classique. Paris: Mouton, 1964.
Cardona, George. "Pāṇini's kārākas : Agency, Animation and Identity." Journal of Indian Philosophy 2 (1974): 231–306.
Cardona, George. Pāṇini: His Work and Its Traditions. Vol. 1, Background and Introduction. 2nd ed. Delhi: Motilal Banarsidass, 1997.
Cardona, George. "Paraphrase and Sentence Analysis: Some Indian Views." Journal of Indian Philosophy 3 (1975): 259–281.
Coward, Harold G., and K. Kunjunni Raja, eds. The Philosophy of the Grammarians. Vol. 5 of Encyclopedia of Indian Philosophies. Princeton, NJ: Princeton University Press, 1990.
Coward, Harold G., and K. Kunjunni Raja. "Introduction to the Philosophy of the Grammarians," In The Philosophy of the Grammarians, edited by Harold G. Coward and K. Kunjunni Raja, 1–97. Princeton, NJ: Princeton University Press, 1990.
Deshpande, Madhav M., trans. The Meaning of Nouns: Semantic Theory in Classical and Medieval India = Nāmārthanirṇaya of Kauṇḍabhaṭa. Translated and annotated. Dordrecht and Boston: Kluwer, 1992.
Ganguli, Hemanta Kumar. Philosophy of Logical Construction: An Examination of Logical Atomism and Logical Positivism in the Light of the Philosophies of Bhartṛhari, Dharmakīrti and Prajñākaragupta. Calcutta: Sanskrit Pustak Bhandar, 1963.
Matilal, Bimal K. Logic, Language and Reality: An Introduction to Indian Philosophical Studies. Delhi: Motilal Banarsidass, 1985.
Raja, K. Kunjunni. Indian Theories of Meaning. 2nd ed. Adyar, Madras: Adyar Library and Research Centre, 1969.
Ruegg, David Seyfort. Contributions à l'histoire de la philosophie linguistique indienne. Paris: Boccard, 1959.
Scharf, Peter M. The Denotation of Generic Terms in Ancient Indian Philosophy: Grammar, Nyāya and Mīmāṁ. Philadelphia: American Philosophical Society, 1996.
George Cardona (2005)