The unity of a language is represented always in relation to another unity.
—NAOKI SAKAI, “HOW DO WE COUNT A LANGUAGE?”1
Equivalence in difference is the cardinal problem of language...
—ROMAN JAKOBSON, “ON LINGUISTIC ASPECTS OF TRANSLATION”2
INTRODUCTION
The most straightforward way to determine what Prakrit was is to look at how it was represented, that is, how it appeared from within the literary and intellectual culture of premodern India. Chapters 2 and 3 offered a largely diachronic account of Prakrit’s invention as a literary language. This chapter provides an analysis of Prakrit’s synchronic position within the order of literary languages. It follows an ongoing attempt to “figure out,” by representing it in figures, Prakrit’s relation to other languages. What is remarkable is that no one seems ever to have thought that such an analysis was even necessary: scholars have focused their explanations, as reductive as they tend to be, on why certain kinds of people used Prakrit, or were represented as using Prakrit, rather than why Prakrit was available for such uses in the first place.
In what follows, I adopt Naoki Sakai’s idea of a “schema”—itself adopted from Kant—to characterize the language order of premodern India. My idea of a schema is historicist and constructivist, like Sakai’s but completely unlike Kant’s. The problem Sakai addressed with this idea is the “unity” of a language.3
On the one hand, it is second nature for us to count languages, that is, to represent them as unified objects that can be enumerated in a series. Sanskrit and Prakrit do not differ in this respect from English, Japanese, Russian, and French. Language’s discrete character is essential to almost everything that we can think to do with it. “Narrating, reciting, listening, reading, writing and translating” are all performed in a way that presupposes and reproduces the differences between languages.4 For any given language, the unity of that language, and thus its ability to be counted alongside other languages, is given as well. On the other hand, it is still second nature. We would like to believe that our representations of language “cut nature at the joints,” but the closer we look, the further we get from finding any.5 We find, instead, that what holds a language together, and what categorically separates it from others, is not any intrinsic property, but effective fictions, of which we are collectively the authors.
A schema is, in Kant’s words, “the representation of a general procedure by which the imagination supplies its image to the concept” of which it is the schema.6 It is a “mediating representation” (vermittelnde Vorstellung) that allows us to bring the messy and gradient language practices as we encounter them in “the real world” under discrete and ordered categories.7 Schemas perform the work of figuration, classification, and categorization that enable us to think of languages as objects. It is through the representational work of schemas that Prakrit became a language: an internally homogeneous and discrete object, differentiated from other such objects—and above all from Sanskrit—as a species of a genus. But the effects that schemas have thus go far beyond the representational work that they do. They provide us with concepts with which we can reflect upon, evaluate, and regulate our own uses of language, as well as the range of social practices that intersect with language use. This results in a feedback loop: concepts are based on practices, practices are based on concepts, and thus the objects and relations that a schema posits come to form part of the world that the schema is meant to represent. A schema can thus be seen as a blueprint for, rather than merely a picture of, a language order.
Schematism, the capacity or even requirement to produce schemas, may be “an art hidden in the depths of the human soul,” but a schema itself is a historical artifact.8 It belongs to those deeply embedded patterns of reasoning and representation so deeply in a culture that we generally call “common sense,” and hence it underlies the particular ways of speaking about and using language that are prevalent within that culture. The closest Sanskrit equivalent of the kind of schema I have in mind is vyavasthā, something that sets a number of other things in their place relative to one another, a single figure that encompasses and imposes order on an enormous diversity of practices.
The approach adopted in this chapter differs radically from the method by which Indological scholarship has traditionally attempted to understand “language talk” in premodern India, namely, by invoking the paradigm of sociolinguistics and reading the sources as proxies for attitudes toward and beliefs about language in the various segments of premodern Indian society. Among the many methodological and epistemological liabilities in this approach is the tendency to view language as a “dependent variable” and social distinctions as the “independent variable.” On this view, language is a reflection of more fundamental patterns in social organization. Given that religion is still thought of as the most important source of social distinctions in premodern India, this view often has the effect of reducing language to religious identity, and thus of producing facile equations between Brahmans and Sanskrit, or between Jains and Prakrit. The tendency to treat Sanskrit and Prakrit as transhistorical categories is another liability that makes it difficult to see when and how people began thinking of and representing language in these terms.9 This tendency is explained in part by Hermann Jacobi’s intentional conflation of the emic terms “Sanskrit” and “Prakrit” with the etic terms “Old Indic” and “Middle Indic,” discussed in chapter 1.
My approach differs less radically from the one developed by Sheldon Pollock, and shares with it the goal of denaturalizing such familiar concepts as Sanskrit and Prakrit by tracing out their history.10 But where Pollock minimizes the differences between Sanskrit and the other members of the “closed set” of literary languages, I am interested in the logic of internal differentiation within this set. And where Pollock assigns a nomothetic function to many of the representations discussed here, I assign them a schematic function. Precisely what this function is will become clear over the course of this chapter, but to begin with, I mean that representations of language do not simply list languages that already exist—they do not gather together languages that meet a certain criterion, such as “their availability across region, ethnie, sect, and time” as Pollock suggests—but stake out discursive positions that languages occupy vis-à-vis each other. They are as much ways of making sense of language practices, of “figuring them out,” as they are rules regarding their use.
This chapter departs from earlier scholarship in one other significant respect. Just as the preceding chapters enabled us to challenge the historical priority of Sanskrit by considering alternative points of origin for the “poetry of polity” (praśasti) and high literature in general (kāvya), this chapter enables us to challenge the conceptual priority of Sanskrit by focusing on the relational figures through which languages were represented. According to the schemas reconstructed here, Sanskrit and Prakrit defined each other, contrasted with each other, and complemented each other. This approach ties in with the slightly revisionist history of Prakrit, as well as Sanskrit, offered in this book: rather than naming timeless categories of speech, Sanskrit and Prakrit came into use as names of languages around the first century CE, when the language order they jointly constituted came into being.
THE ARCHETYPAL SCHEMA
The archetypal schema here is the underlying framework of the language practices of “classical India”—the literary and intellectual culture of India from the first to the twelfth centuries CE, in which Sanskrit and Prakrit jointly served as the parameters of textual production. This characterization closely resembles Pollock’s characterization of the “Sanskrit cosmopolis.” One reason I have adhered to the older term is simply to avoid confusion: the “Sanskrit cosmopolis” is really a metonym, based on the importance of Sanskrit to the entire cultural order, but in this chapter I am interested precisely in Sanskrit’s others.
The representations that the archetypal schema provides procedures for constructing are the statements in which participants in literary and intellectual culture articulated an understanding of their own language practices. Many of these texts are “classical” in the further sense that they are foundational within their respective discourses. They reflect an understanding of language that has a long history of effects. This is why I call the schema presented here archetypal: other ways of understanding language in India, up to the present day, presuppose it as a template.
The most common formulation of this schema is the bhāṣātraya, “the three languages”: Sanskrit, Prakrit, and Apabhramsha. This is the figure that Bhāmaha and Daṇḍin present in the two foundational works in the discourse of poetics, the Ornament of Literature and the Mirror of Literature. This is just one form of the schema—not everyone who has attempted to make sense of the language practices of this literary and intellectual culture enumerates precisely three languages—but I take it to be representative of a broad consensus regarding the number of languages, their identity, and their relationship to one another. Its archetypal status is easily illustrated by the fact that the fourfold and sixfold schemas that begin to emerge in the ninth century incorporate and expand upon the threefold schema.
Four important features characterize this archetypal schema: the opposition between Sanskrit and Prakrit; the identity of Sanskrit and Prakrit; the totality of the practices the schema represents; and the iterability of its distinctions. Together these give the language order of classical India its unique shape: the central dichotomy of Sanskrit and Prakrit, the asymmetrical relation between the two, and the peripheral position of Apabhramsha. The role and status of a language within a language order are the result of a complex configuration of factors on the level of schematic representation. “Cosmopolitan” and “vernacular” are two of the roles that may be available, but they do not exhaust all of the possibilities—Prakrit does not easily fit into either category—and it would be a mistake to understand them as universal categories that classical Indian culture just happens to instantiate.
OPPOSITION
At the core of the basic schema lies a binary opposition between Sanskrit and Prakrit. Generally, one can speak of opposing two things that already exist, or of an opposition that creates two things that did not exist before. It is the latter sense that I intend here. Sanskrit and Prakrit exist in a “schema of co-figuration,” where the representation of one determines the representation of the other.11 There are two aspects of the schema of co-figuration that I would like to emphasize at the outset, because they lead to an understanding of the relationship between Sanskrit and Prakrit that differs from what one commonly encounters in scholarship.
One aspect is the prior indeterminacy of the objects under co-figuration. The schema does not simply apply contrasting attributes to each member of the pair—although this is one of its important functions—but rather defines what each member of the pair is. Although we tend to see the opposition between Sanskrit and Prakrit as an opposition between two languages, it is only as a result of a schematic representation that we can oppose Sanskrit and Prakrit as languages in the first place. This claim opens up the possibility that Sanskrit and Prakrit were not always what they currently seem to be. For example, Sanskrit and Prakrit are figured in the Treatise on Theater not as languages, for which other terms are used, but as two distinct types of actors’ lines.
The second aspect is the lack of a prior independent existence for each of the objects under co-figuration. Co-figuration implies that the emergence of Sanskrit and Prakrit as objects of representation was more or less simultaneous. Of course there is a sense in which Sanskrit existed prior to the Sanskrit–Prakrit dichotomy. But this type of Sanskrit, the language of Vedic texts, was quite different from that which we commonly call “classical”—the language that the archetypal schema delineates—and in fact there is no evidence that it was even called “Sanskrit” much before the first and second centuries CE. Exactly the same can be said of Prakrit. Co-figuration replaces the question of whether Sanskrit or Prakrit came first—the answer to which depends entirely on one’s chosen definitions—with an answerable question about what phenomena the words “Sanskrit” and “Prakrit” were applied to.
One kind of opposition is built into the words Sanskrit and Prakrit themselves. The words form, as George Grierson noted, a “naturally correlated pair.”12 The word saṃskṛta, from the verb √saṃ-s-kṛ, means in the broadest terms “what has been elaborated.”13 The word prākṛta means what exists in, or has come from, the source (prakṛti).14 In contrast to Sanskrit, it refers to the original state of something prior to elaboration. Hence Grierson contrasted them as “artificial” and “unartificial.”
The words saṃskṛta and prākṛta did not start out as designations for languages. It seems likely that they were employed for this purpose in order to represent the practices they designated as opposites. This interpretation is consistent with the ritual connotations of saṃskṛta, according to which Sanskrit is speech that has been “purified” for ritual use. This term, as Sheldon Pollock argues, forges an association between Sanskrit and the sphere of Vedic ritual, where the language was used both in actual ritual practice, in the form of hymns and prayers, as well as to talk about those rituals and the forms of knowledge that they presupposed. But it is important to note that “Sanskrit,” as a designation for a language, is used only after the “prestige economy” of this language had expanded far beyond the sphere of ritual alone.15 One of the earliest known uses of the word saṃskṛta to refer to a language occurs in the Rāmāyaṇa. In the Sundarakāṇḍa, Hanumān considers how he should address Sītā, and says: “If I present a saṃskṛtā speech, like a twice-born, she will mistake me for Rāvaṇa and get scared. I must address her with a human [māṇuṣaṃ] speech, full of meaning.”16 This passage contrasts Sanskrit as the language of twice-born Brahmans, such as Rāvaṇa, with the language of humankind as a whole. We can view this passage, as Pollock does, as a reflection of the social and discursive limitations that applied to the use of Sanskrit in the centuries preceding the Rāmāyaṇa’s composition. But we can also view it as a reflection of a set of circumstances that did not exist long before this passage itself was composed. The first circumstance is an increased distance between languages, in Heinz Kloss’s sense of Abstand, or at least an increased awareness of this distance, relative to Patañjali’s time. As is well known, Patañjali represented incorrect words as local deviations from the corresponding correct words rather than systemic deviations that might possess a logic and structure of their own.17 This distance allowed people to think of languages as distinct systems, rather than as a single system that included arbitrary variation within it. The second circumstance, closely linked to the first, is choice. The necessity of choosing a language, and the awareness of doing so, is a special feature of literature, and radiates from literature into other discourses. Pollock is right to connect the Rāmāyaṇa’s consciousness of its own language with its self-declared status as the first work in an entirely new type of expressive literature.18 Hanumān’s dilemma of what language to frame his speech in is the same as that of Vālmīki, the author of the Rāmāyaṇa. Whenever language is an object of choice, we require a schema to tell us what the choices actually are.
We don’t know when the Rāmāyaṇa was composed, but it was likely in the first century BCE. Around this time, and continuing into the early centuries of the common era, Jain monks were collecting, revising, and expanding a body of canonical literature. In a long discussion of music that several canonical texts share, it is observed that the language of song can be either Sanskrit or Prakrit.19 This rather accidental passage reveals to us both the circumstances in which language is an object of choice, and what the choices were in such circumstances. Just as the Vedic scriptures never proclaim that they are composed in Sanskrit, the Jain scriptures never proclaim that they are composed in Prakrit, and only mention Sanskrit and Prakrit in a passage that clearly concerns the practices of a different cultural realm: that of literature and music.
The most compelling illustration of co-figuration occurs in a passage from Kālidāsa’s Birth of Kumāra (early fifth century CE). During the celebration of Śiva and Pārvatī’s wedding, Sarasvatī congratulates the couple:
Sarasvatī praised the couple with a speech
that she delivered in two ways:
one purified by saṃskāra to the excellent groom,
and one that could easily be understood to the bride.20
Kālidāsa here imagines the speech of Sarasvatī, the goddess of language and literature, in accordance with the same schema that distinguished Sanskrit and Prakrit as literary languages. In the literary culture that Kālidāsa inhabited, Sarasvatī did in fact speak two languages. Kālidāsa composed the Birth of Kumāra in Sanskrit within generations of Sarvasena composing another court epic, Hari’s Victory, in Prakrit. The earliest available commentary on this passage of the Birth of Kumāra, Vallabhadeva’s, explicitly identifies Sarasvatī’s “speech delivered in two ways” with Sanskrit and Prakrit.21 This passage is therefore a self-conscious reflection, from one of the foundational figures of kāvya, on the language practices of kāvya itself. Its wording even anticipates the wording of later works of poetics that sought to divide up the sphere of “textuality” (vāṅmaya) on the basis of language.
Kālidāsa’s image shows us not just the dichotomization of literary language into Sanskrit and Prakrit, but some of the specific contrasts that create this dichotomy. One contrast etymologically defines Sanskrit as the language that is “purified by saṃskāra”; Prakrit’s lack of saṃskāra is implicit here, but is explicitly stated in other texts.22 It has proven difficult to say what saṃskāra means here because the word originally referred to the consecration of ritual objects and only by extension to language. There were many ways in which a language might be thought to possess saṃskāra: it could be consecrated for ritual use; it could be endowed with a certain kind of power or prestige; it could be validated by the teachings of grammarians; it could be produced by people who have been instructed in these teachings; it could be produced with care and attention; or it could be all of these things. In this context, saṃskāra likely refers in the first place to the rules enunciated by Pāṇini, around the fourth century BCE, that defined Sanskrit as a discrete, unitary language—without, however, using the name “Sanskrit” in reference to it. Co-figuration implies that Prakrit is projected as the opposite of Sanskrit across all of these senses.
A verse from Vākpatirāja’s Gauḍa’s Demise (early eighth century) provides a further example of these contrasts: “The loveliness of Sanskrit words unfolds through the beauty of Prakrit, and the splendor of Prakrit through the excellence of Sanskrit’s saṃskāra.”23 What Prakrit uniquely contributes to a work is “beauty,” whereas Sanskrit’s unique contribution is saṃskāra, which in this context might mean grammatical perspicuity—the quality that enables Vākpati’s work to be appreciated in a court where the preferred medium is Sanskrit. For Vākpatirāja, Prakrit can possess saṃskāra, but only by borrowing it from Sanskrit.
Another contrast that emerges from Kālidāsa’s verse is that Prakrit is simple and Sanskrit is difficult. A Sanskrit sentence is conceived as an elaborate complex of discrete grammatical elements; it was defined by this complexity, a literal “putting-together” or saṃskāra. Thus a topos in Prakrit literature is that Prakrit is easier than Sanskrit because it does not require the in-depth grammatical knowledge that Sanskrit does.24 Earlier we encountered a similar representation of Prakrit among Jain writers. They wished to depict their scriptures, which they claimed were composed in Prakrit, as inherently more accessible to the unlettered masses than the scriptures of other religious traditions. “Those who know the truth,” Haribhadra wrote around the seventh century, “have produced scriptures in Prakrit for the benefit of children, women, the slow-witted and the uneducated, and for men who strive after good conduct.” I argued in chapter 3 that such representations depend on and reinforce a myth of continuity between Prakrit and demotic language practices. It will be clear from the following chapter that for nearly the entire period with which we are concerned here, Prakrit was no less of a learned language than Sanskrit was, and Prakrit had grammars and lexicons just as Sanskrit did. And difficulty and complexity are, of course, relative concepts: there were no doubt people for whom Sanskrit was more easily intelligible than Prakrit and vice versa. The important point here, however, is that Prakrit was consistently represented as essentially different from Sanskrit in this respect, from its first literary monuments onward.
The ways in which the earliest Prakrit literature explicitly positioned itself against Sanskrit—representing itself as a discourse that was about, if not exactly for and by, common people (prākṛta-jana), rather than scholars and ritual specialists—are discussed in chapter 3, citing the following programmatic verse from Seven Centuries:
Prakrit poetry is nectar.
Those who don’t know how to recite it or listen to it
make love into a science.
How are they not ashamed?25
This passage is among the earliest examples of the word prākṛta (pāua) used in connection with a language, and hence complements the earliest use of the word saṃskṛta in the passage from the Rāmāyaṇa discussed above. This verse turns on a contrast that illuminates what “Prakrit poetry” is. On the one side stand those who exercise themselves in scholarly disputes. On the other side stand those who compose and appreciate “Prakrit poetry,” a phrase that could imply the poetry of common people in contrast to scholars, or common poetry in contrast to sophisticated scholarly discourse, besides poetry in the Prakrit language.26 Prakrit and its other, Sanskrit, thus align onto the discourses of kāvya and śāstra and the personas stereotypically associated with them: sensitive litterateurs and fastidious, fault-finding scholars. This verse hints at the possibility that these two languages can complement each other and inhabit the same social space.
The most extensive early discussion of this shared social space, jointly inhabited by Sanskrit and its others, is Vātsyāyana’s Kāma Sūtra (late third to early fourth century). In the course of describing the day-to-day activities of the urbane man (nāgaraka), Vātsyāyana has him attend a goṣṭhī, which is “when men of equal knowledge, intelligence, character, wealth and age, accompanied by courtesans, sit down together to discuss suitable matters, either in a courtesan’s house, the court, or one of their own houses.” What takes place there is “critical discussion of literature and fine arts,” followed by the appreciation of beautiful women.27 Later on, Vātsyāyana cites a few verses concerning goṣṭhīs from an older source. One of them claims that “one who participates in discussions in goṣṭhīs, neither exclusively in Sanskrit (saṃskṛtena) nor exclusively in the regional language (deśabhāṣayā), will become highly esteemed in the world.”28 This verse is another early use of the word saṃskṛta in reference to a language. The opposition is between the “regional language” (deśabhāṣā) and Sanskrit, which is figured as transregional in contrast. Prakrit is not explicitly mentioned here, although I consider it likely that the term “regional language” here refers to Prakrit, which is the only Indian language besides Sanskrit and probably Tamil for which we have evidence of literary production in the early first millennium.
This verse commends a “middle way” between the exclusive use of Sanskrit and the exclusive use of the regional language. This might mean that Sanskrit should be used in some contexts and that the regional language should be used in others, or it might mean that both Sanskrit and the regional language should be employed in similar contexts.29 In either case, this verse locates both of them in the same social space, namely, the goṣṭhī, and in the same social actor, namely, the nāgaraka. The fact that Sanskrit and Prakrit were figured as opposites does not mean that they were relegated to entirely different social and discursive spheres.
The literary culture that Prakrit partially constituted was overwhelmingly dominated by men, as Vātsyāyana’s descriptions of goṣṭhīs show. But Prakrit was represented as being more understandable to women and more open to women’s participation than Sanskrit, and for these reasons preferred by women to Sanskrit, as we see in the verse from the Birth of Kumāra. Sanskrit and Prakrit conform to a pattern in which the social exclusivity of high culture generates parallel traditions purporting to offer the same kind of content but with fewer restrictions. Sanskrit was “high,” and accessible only to people of a certain social status, while Prakrit was “not quite so high” and in principle open to everyone.
The comparative accessibility of Prakrit is a commonplace in Prakrit literature. A verse from the Vajjālagga, a collection of Prakrit poetry compiled near the end of the first millennium, says: “Prakrit poetry is playful and has sweet syllables; it is adored by young women and is erotic. So who is going to recite Sanskrit?”30 The effect, as in the other programmatic passages we have seen so far, is to claim the territory of poetry for Prakrit, and especially poetry that has love as its central theme. Prakrit poetry is a discourse that notionally includes men and women; it is a poetry that not only speaks about women, but a poetry in which women speak and are spoken to.
Prakrit was not just favored by young women, according to these representations, but figured as a young woman. Some manuscripts of Rājaśekhara’s Karpūramañjarī, a stage play of the early tenth century, read a verse in the prologue that claims that “Sanskrit compositions are harsh, but a Prakrit composition is soft; the difference between these two is as great as between a man and a woman.”31 A verse from Jayasiṃha Sūri’s Explanation of the Garland of Advice (860 CE) uses an impressive triple entendre to imagine the Prakrit language—here called “the language of Maharashtra,” marahaṭṭhayabhāsā—as a beautiful woman:
Teeming with charming words,
manifesting the theme of love,
and bejeweled with lovely sounds,
the language of Mahārāṣṭra is like a woman—
walking attractively,
revealing her intentions,
and decked with gold and jewels,
and like a forest—
laced with lovely paths,
where you can see mynah birds,
and clothed in beautiful leaves.32
Prakrit is here, as in the verse just quoted from the Vajjālagga, figured as “soft,” referring to its characteristic lenition (“softening”) of intervocalic consonants (see chapter 4). But the comparanda that Jayasiṃha Sūri chooses are motivated by the content of Prakrit poetry just as much by its form: Seven Centuries is full of women arranging meetings with their lovers in the forest.
It is the nature of “not quite so high” culture that there is something higher than it. What Prakrit gained in being represented as more broadly accessible than Sanskrit (whether or not it actually was more accessible), it lost in exclusivity and thus prestige. Prakrit authors attempt to close the prestige gap by presenting the differences between Sanskrit and Prakrit as superficial and irrelevant to the meaning that the text itself conveys. One verse from the Vajjālagga figures Sanskrit and Prakrit as two equivalent options for expressing a given sense: “Sanskrit or other than Sanskrit, depending on who has come to listen, it is the meaning that produces a special kind of rasa, never before experienced. Isn’t it amazing?”33 The form of the binary here, Sanskrit and non-Sanskrit, has two implications. One is that the Sanskrit–Prakrit binary becomes a merism for all language: there is nothing not encompassed by either “Sanskrit” or “non-Sanskrit.” The second is that Sanskrit is the unmarked member of the Sanskrit–Prakrit pair. This asymmetry comes out of an older view, represented, for example, by the grammarian Patañjali, that makes the language that Pāṇini described language as such without any further specification. For the entire classical period, composing a text in Sanskrit required no apology or explanation, whereas composing a text in Prakrit often did. This is one symptom of Sanskrit’s discursive dominance, and of its superposition within the language order that Pollock has referred to as “hyperglossia.”34
One Sanskrit work that does comment on its own choice of language is Govardhana’s Seven Centuries of Āryās, a collection of lyrics in Sanskrit produced in eastern India around 1200 CE. But this is because Govardhana conceived his work as a Sanskrit response to Hāla’s Seven Centuries:
It took force
to turn this poetry, whose rasa is most suited to Prakrit,
toward Sanskrit,
just as it took Balarāma
to turn the Yamunā, whose water naturally flows down,
toward heaven.35
This comparison may carry a suggestion that Sanskrit represents a diversion from the “natural” course of language represented by Prakrit, or it may simply have served to situate Sanskrit, the “language of the gods,” in its rightful heavenly place. The purpose of the comparison, however, is to emphasize the difficulty in transforming the kind of “speech” (vāṇī) for which Prakrit had long been thought appropriate or even obligatory—namely, stand-alone verses of a predominantly erotic character in the gāthā meter—into Sanskrit.
Most of the above passages that help us recover the representations of Prakrit current in the language order of classical India come from literary texts. But the opposition of Sanskrit and Prakrit is not limited to these sources. When I describe the schema as “archetypal,” part of what I mean is that it supplies a general framework for thinking about and talking about language within all of the domains of culture. One particularly important domain, besides the literary, is systematic thought about language. The discussion that I highlight here comes from Bhartṛhati’s On Sentence and Word, a seminal work on the philosophy of language from around the fifth century CE.
Bhartṛhari implicitly juxtaposes Sanskrit and Prakrit by presenting two opposing views about what is correct and what is incorrect in language use:
“The language of the gods was brought into confusion by incompetent speakers.”—but on this point, people who hold it to be non-eternal have the opposite opinion.36
The prose commentary on this slightly obscure verse seems to get Bhartṛhari’s intention right. The first half represents a view according to which Sanskrit, the “divine language,” was once pure, but over time became corrupted by the accumulated mistakes of careless speakers. This view places Sanskrit at the root of all current language practices, and also accounts for the deviation (apabhraṃśa) of those language practices from each other and, of course, from Sanskrit. The “opposite” view referred to in the second half sees Sanskrit, not as the root of all language practices, but as a secondary elaboration and codification of preexisting language practices. Proponents of this view call these originary practices “Prakrit,” which can be analyzed as meaning “existing in the original.” Bhartṛhari also alludes to this position in his Light on the Great Commentary, an incomplete gloss on Patañjali’s treatise.37 In this view, words are correct, not because their use leads to merit (dharma), as Patanjali had argued when trying to establish the purposes of grammar, but only because they accord with conventions. Accordingly, it is the “original” Prakrit words that are correct, while Sanskrit words represent an unsuccessful attempt to “dress up” language.38 It is nearly certain that the “others” to whom Bhartṛhari refers are Jains who employed Prakrit for literary, religious, and philosophical texts and who defended their language practices with arguments similar to those summarized in the prose commentary to On Sentence and Word.39 It is because Prakrit had become an important counterweight to Sanskrit in Jain intellectual circles, as well as in literary circles beyond Jainism, that Bhartṛhari can represent an argument for its originary status. Bhartṛhari’s Prakrit, in other words, is not just any language that deviates from Sanskrit, but the specific language or languages that Jains defended as legitimate for religious and philosophical use.40
The co-figuration of Sanskrit and Prakrit is one of the key features of the archetypal schema of language in classical India. Sanskrit and Prakrit are two discrete objects, and objects of broadly the same type, but they contrast across multiple dimensions. The dimensions highlighted in this brief survey include the social (the comparative accessibility, however notional, of Sanskrit and Prakrit to women), the aesthetic (the harshness of Sanskrit and the softness of Prakrit), the discursive (the affinity of Prakrit for kāvya and of Sanskrit for śāstra), the grammatical (the presence of absence of saṃskāra). Sanskrit was figured as “the language of the gods,” and at this stage, Prakrit was contrastively figured as “the language of men.” These differences render them complementary rather than incomparable; they constitute the twin parameters of discourse.
IDENTITY
The archetypal schema also represents Sanskrit and Prakrit in a particular and at first glance paradoxical relationship that I call “identity-in-difference.” All schemas represent languages as identical in the minimal sense in that they are species of a genus. But a more substantive kind of identity obtains between Sanskrit and Prakrit, which are considered to be made out of the same linguistic stuff.
The strongest case for the identity of Sanskrit and Prakrit was made by the tenth-century poet Rājaśekhara in the prologue to his Prakrit play, Karpūramañjarī:
The particular meanings are the same,
and the words are the same—
even if they undergo some change.
A literary work is a special kind of composition,
whatever language it happens to be in.41
The conclusion of this verse might lead us to think that the poet can choose whatever language he wishes, since every language has words and meanings that can be combined to make literature. But that is not the argument that Rājaśekhara makes, nor is it an argument that he would make. For Rājaśekhara makes very clear in his other works his opinion that literature could only be composed in four languages—Sanskrit, Prakrit, Apabhramsha, and Paishachi (see below)—and this verse is a defense, in Prakrit, of writing a play in Prakrit.42 The argument is rather that if the definition of literature applies to a work in Sanskrit, then it should apply equally to a work in Prakrit. It is not simply that Prakrit is capable of conveying the same meanings as Sanskrit, or that Prakrit words differ only superficially from the corresponding Sanskrit words, but that Prakrit shares with Sanskrit the particular (visesa) words and meanings in which their literariness consists. Their underlying identity ensures that Sanskrit can be “transformed” (pariṇamantā) into Prakrit, in the way that milk, and only milk, can be transformed into curd.
Transforming Sanskrit into Prakrit is precisely what the discourse of Prakrit grammar accomplishes: it explicitly figures Sanskrit as an archetype (prakṛti) that can be systematically modified to produce Prakrit as an ectype (vikṛti), although the domain of such relations included only a part of the Prakrit language. I will limit my discussion here to one text which includes the earliest available Prakrit grammar, the Treatise on Theater ascribed to Bharata; chapter 6 will discuss other texts in this tradition.
The Treatise on Theater is a compilation of knowledge related to theater probably produced between the third and fourth century CE. It offers one of the earliest systematic accounts of literary language in India. Language was a primary concern to the compilers because “verbal representation” (vācikābhinaya) was essential to all ten major forms of theatrical performance, and was thus considered to be “the body of theater.”43 The Treatise on Theater is the earliest text to clearly and systematically distinguish between Sanskrit and Prakrit, and it is the text that most clearly presents the relationship of “identity-in-difference” of Sanskrit and Prakrit.44
The discussion of language occupies the first sixty-two verses of the Treatise’s seventeenth chapter. In this section, “Sanskrit” and “Prakrit” are terms used as modifiers, not of language (bhāṣā), but of pāṭhya, the actors’ lines. Abhinavagupta’s detailed eleventh-century commentary makes it clear that pāṭhya is not just the text of a play, something the Treatise on Theater generally calls kāvya, but the precise way in which the text is realized on the stage.45
There are exactly two kinds of lines, Sanskrit and Prakrit.46 The Treatise defines Prakrit as follows:
A Prakrit line is exactly the same as Sanskrit, but reversed:
it is devoid of the quality of saṃskāra.
It consists of various intermediate grades.47
Prakrit is, paradoxically, both “the same as” and the “reverse of” Sanskrit. What distinguishes them, as we saw above, is the presence or absence of saṃskāra, which Abhinavagupta plausibly understands in this context to be the “care” that results in the “maintenance” of the language in an identical state. Abhinavagupta explains that Sanskrit and Prakrit have an identical linguistic substratum (prakṛti), but Prakrit “comes from” that substratum “in the form that it takes without saṃskāra”—invoking the standard analysis of prākṛta as “what has come from the prakṛti.”48
The Treatise on Theater’s definition of Prakrit involves a further paradox. If Prakrit lacks the very quality of saṃskāra that provides language with stability, it must be a “deviation” (apabhraṃśa), a practice that is characterized by the absence of those regularities (niyama) by which a language is constituted as a unity. And if this is the case, then any attempt to explicitly formulate the regularities of this practice—as the Treatise on Theater set out to do—is doomed to fail. Abhinavagupta poses the problem succinctly: “what regularity can a ‘deviation’ possibly have?” He answers with a creative interpretation of the last quarter of the verse. Prakrit owes its regularity to its conventional acceptance (prasiddhi) within specific regions (deśaviśeṣa), in contrast to Sanskrit, whose regularity is prior to its conventional acceptance in any particular place.49
The Treatise on Theater’s definition of Prakrit raises the question of how can we think about regularity outside of the paradigmatic regularity of Sanskrit. There was, however, no need for its compilers to reinvent the wheel. To answer this question, they availed themselves of existing literature about the definition and analysis of Prakrit. First, the Treatise on Theater presents the standard threefold classification of Prakrit words that was also presented in early grammars of the language that are now lost (see the discussion in chapter 6): Sanskrit-identical (samānaśabdaṃ), Sanskrit-derived (vibhraṣṭaṃ), and regional (deśīgatam).50 Then it quotes from and adapts some of these lost grammars to produce a “mini-grammar” of Prakrit in two complementary sections.51
In connection with Treatise on Theater, it is worth noting one other important passage in which Prakrit furnished an example, or rather the example, for thinking about regularity outside of Sanskrit. That is Kumārila Bhaṭṭa’s discussion of the language of Buddhist scriptures in his Explanation of the System (ca. seventh century). He claims that the authority of the Buddhist scriptures must be rejected because they fall under the Mīmāṃsā Sūtras’ category of “illegitimate compositions.” They are illegitimate, he claims, because they are “not even Prakrit.” “Those texts are composed in mostly incorrect words from the Māgadha and Dākṣiṇātya languages and their degraded forms,” he says, and after quoting a verse in a Middle Indic language, he complains that it is “more degraded than the degraded regional languages with which we are familiar.” The examples that he gives show his familiarity with literary Prakrit and Apabhramsha. One of these examples is the word saṃskṛta-, which appears in the degraded language of the Buddhists as saṃkaḍa-. He says that the “correct incorrect” form, as familiar from Prakrit and Apabhramsha, should be sakkaa-.52 Prakrit provided Kumārila with a model of how words could be correct in the sense of conforming to some standard while at the same time being incorrect in the sense of deviating from Sanskrit.
To return to the Treatise on Theater, we have almost no evidence as to what languages were in fact used on stage before this text was compiled. A few fragments of Aśvaghoṣa’s otherwise-lost plays from the early second century seem to use a more archaic version of the languages we find in later plays.53 The Treatise on Theater itself provides many examples of dhruvā songs in the thirty-second chapter that are composed in what also appears to be a rather archaic language.54 It is difficult to speak with confidence about these texts—one on account of its fragmentariness, the other on account of its corruption—but it certainly appears that their language does not agree in all of its particulars with the language that the Treatise describes in the seventeenth chapter, as Luigia Nitti-Dolci was among the first to note.55 I do not think that this difference can support detailed claims about the historical development of the Prakrit language, or languages, such as Manomohan Ghosh’s argument that Śaurasenī is merely an older form of Prakrit than Mahārāṣṭrī, the standard literary language. Rather, it appears that the compilers of the Treatise on Theater had defined one kind of “Prakrit” by reference to another. Their goal must have been to categorize and describe the languages that were used in stage plays, including Sanskrit and its others. But the world in which the Treatise on Theater took shape was one in which Prakrit was already a literary language of some standing. Its compilers appeared to borrow the name, as well as the basics of a grammatical description, from the discourses of “literature heard” (śravyakāvya)—Prakrit lyric and courtly epic—in order to characterize the language practices of the stage play, or “literature seen” (dṛśyakāvya). As Abhinavagupta tells us, Bharata’s purpose is not to describe the languages of the stage in very precise detail, but simply to give a general indication of how they sounded. For this purpose, the rules formulated by other texts and integrated into the seventeenth chapter served that purpose adequately.56 The use of “real” Prakrit—that is to say, the language of Seven Centuries and Rāvaṇa’s Demise—in plays is commonly thought to be a later innovation, found in Kālidāsa and later playwrights. The evidence for earlier practices, however, is very slight, essentially limited to the fragments of Aśvaghoṣa’s plays and the difficult-to-date Little Clay Cart by Śūdraka.
The next sections map the distinction between Sanskrit and Prakrit onto the plurality of language practices of the theater. Scholars usually take for granted a model that organizes these language practices into two sets: Sanskrit, which contains only itself, and “the Prakrits,” which contains all of the languages besides Sanskrit, such as Śaurasenī, Māgadhī, and so on.57 This model has come to dominate modern scholarship in part because it came to dominate premodern thinking about language. For this reason it is important to note that it is completely absent from the Treatise on Theater itself. The work instead offers two alternative models, one for relating the specific language economy of the theater to the dichotomy of Sanskrit and Prakrit in the literary-cultural sphere, and one for relating it to the messy world of regional languages beyond it.58
The first model involves a fourfold classification of language (bhāṣā) which supervenes upon, rather than replaces, the twofold classification of lines into Sanskrit and Prakrit.59 This relates to a distinctive feature of theater vis-à-vis other kinds of literature: it alone has “speakers” (vaktṛ) who pronounce its “text.”60 The four types are “superlanguage” (atibhāṣā), “noble language” (āryabhāṣā), “birth language” (jātibhāṣā), and “other-origin” (yonyantarī). The first two types are identified with Sanskrit.61 The last type is spoken by animals; all that is said about it is that it “rests upon theatrical convention” (nāṭyadharmīpratiṣṭhitā). The third type, “birth language,” is spoken by human beings, and it is said to be “twofold,” involving both Sanskrit and Prakrit. The following verses specify the “birth language” by assigning either Sanskrit or Prakrit to human speakers. These assignments are well-known and do not need to be reviewed here.62
The Treatise then presents a second model that does not involve the categories of Sanskrit and Prakrit at all: “Alternatively, if they so choose, producers may employ the regional languages, for the text [kāvyam] of a play arises in various regions.”63 The category of “regional languages” includes seven “languages” (bhāṣā: Māgadhī, Āvantī, Prācyā, Śaurasenī, Ardhamāgadhī, Bāhlikā, and Dākṣiṇātyā) and seven “sublanguages” (vibhāṣā: Śakārī, Ābhīrī, Cāṇḍālī, Śābarī, Drāmiḍī, Āndhrī, and Vānaukasī). The names of the languages refer to regions, but it is important to keep in mind that “regions” in this sense are constituted by people rather than places: Māgadhī is the language of the Magadhas, not of Magadha. The names of the sublanguages refer to groups of people who are either not associated with a particular region, or associated with regions outside of a core cultural area. This model has its own rules of language assignment, but they refer to theatrical rather than social roles: leading men, leading ladies, rogues, jesters, and so on. The default language of this model appears to be Śaurasenī.64
These two models might represent different traditions of theatrical practice. But whatever their origins, it is only by combining them into one that we can produce the familiar model in which a unitary Sanskrit is set over a plurality of Prakrits. Dhanañjaya, a scholar of dramaturgy of the tenth century, is perhaps the first to make this combined model explicit. He understands “Prakrit” and “regional language” as synonyms—making Śaurasenī and Māgadhī varieties of Prakrit—and says that “Prakrit, particularly Śaurasenī, is used by women and low-status men,” in contrast to high-status men, who use Sanskrit. As one moves from the top to the bottom of the social hierarchy, the language practices become less unified and more regionalized: “low characters speak the language of the region to which they belong.”65
The Treatise on Theater’s discussions of language raise important questions about representation: how a schematic model can represent the language practices of a literary form, and how these language practices themselves represent the world outside. This section ends with a recommendation to “take from the world whatever is not spoken of here,” and most scholars have assumed that the languages the Treatise describes are “literary versions of the actual languages.”66 But imitating is not the only way of representing, and it seems impossible to regard the literary languages as “versions of” the spoken vernaculars for which they are named in any significant sense.67 The Treatise on Theater gives us to know that certain characters are entitled to use a transregional language, as Sanskrit is unambiguously characterized by its contrast with the regional languages. At the same time, they give us to know that other characters are not entitled to use this language; we must therefore imagine them as speaking the language of the region to which they belong. But it does not follow that these characters must actually speak some form of the language of the region to which they belong. A commitment to linguistic realism of this kind would entail enormous practical problems: everyone, from the author of the play to the actors to the audience, would be required to master an impossibly broad variety of language practices. Abhinavagupta gestures towards this explanation when he remarks that the limitation of “languages” and “sublanguages” to seven each serves to exclude the infinite variety of spoken dialects.68
In my view, the models presented by the Treatise on Theater offer a compromise solution to this problem. Sanskrit and Prakrit would become the principal languages employed in the theater. This maneuver brought the language practices of the theater into conformity with those of the wider literary culture to which the theater now belonged, where Sanskrit and Prakrit had long since been established as the primary languages of expressive textuality. As noted above, the languages considered to be “Prakrit” in the theater were not exactly the same as literary Prakrit. These languages were named for regions and represented the speech of those regions according to theatrical conventions. The differences between them, however, as well as the differences between them and the literary Prakrit that served their archetype, were carefully constrained so as not to transgress the limits of intelligibility. The language practices of the theater were thus limited by the principle of identity-in-difference: the different languages were minor modifications of the same linguistic substratum. Nowhere is this clearer than in Bhavabhūti’s Mālatī and Mādhava, where the Sanskrit-speaking hero Mādhava, impersonating Mālatī’s Prakrit-speaking friend Lavaṅgikā, pronounces a verse that can be understood in both languages simultaneously.69
The last section of the discussion of language in the Treatise on Theater is concerned to reintroduce regional characteristics that otherwise would not find expression in a theater, which primarily employed the standardized and increasingly transregional languages of Sanskrit and Prakrit. This section begins with a proscription on the representation of the languages of certain groups (jātis): “in theatrical productions, the text should not be made to reflect the language in the case of groups such as Barbaras, Kirātas, Āndhras, and Dramilas.”70 What these groups may have in common is their outsider status, at least in the social imaginary of Sanskrit drama. But it is naïve to read this statement as evidence of a sociolinguistic attitude according to which the language practices of these despised groups were denigrated and avoided. It simply states that the languages of these groups—including at least a few Dravidian languages—are too distant from Sanskrit and Prakrit to share a stage with them: it enforces the principle of identity-in-difference. Regional languages that differed less radically from Sanskrit and Prakrit could be represented, but only according to certain conventions that simplified their bewildering diversity and multiplicity into a small number of diagnostic differences. These conventions would allow a listener to recognize, for example, the word māṇavaü as “northern,” māṇavao as “western,” and māṇavae as “eastern,” like similar shibboleths in English (“y’all” indicating the American south, “yous guys” Philadelphia, “yinz” Pittsburg, and so on).
The Treatise on Theater gives an exhaustive account of what it means for Sanskrit and Prakrit to be “the same” and yet “opposite” each other. Its redactors used Sanskrit and Prakrit to anchor a continuum of literary language practices. Given that verbal representation was the “body of theater,” the continuity of language practices was essential to maintaining theater’s bodily integrity. This continuity can be seen as a space of translation, in the etymological sense of moving back and forth, across the divisions instituted by the schema. This kind of translation, however, forecloses the possibility of translation in the sense familiar to us: precisely because Sanskrit and Prakrit are figured as an underlying unity under different kinds of transformation, there was no need to actually translate a Prakrit text into Sanskrit or vice versa. And in fact the earliest translations from Prakrit into Sanskrit—never the reverse—known to me date from the eleventh century, when the language order begun to shift in such a way as to marginalize Prakrit.71
TOTALITY
Another basic feature of the schema under consideration here is the totality of the practices it schematizes. The space constituted by Sanskrit and Prakrit expands to fill the entirety of literary language; any languages that are not encompassed within this space are not literary. Or, as a verse anthologized in the Verses of the Chappaṇṇayas states: “The person who knows how to speak neither Sanskrit nor the purest kind of Prakrit has one refuge: silence.”72
There are different ways of representing this totality, for example, the merism “Sanskrit and non-Sanskrit.”73 By far the most important representation is what I call the “enumerative totality,” which expands the binary structure of Sanskrit and Prakrit into an n-ary structure. The earliest and most influential example of such an enumerative totality is the “three languages”—Sanskrit, Prakrit, and Apabhramsha—espoused, if not formulated, by the founding fathers of the discourse of poetics, Bhāmaha and Daṇḍin, before the beginning of the eighth century.74
Bhāmaha was perhaps the first to claim that literature as a whole (kāvya) can be exhaustively divided up into Sanskrit, Prakrit, and Apabhramsha.75 Daṇḍin invoked a metaphor to make the status of this division clear: it is the “body of literature” (śarīraṃ kāvyānāṃ) that can be analyzed in terms of language, in contrast to “ornaments” (alaṅkāraḥ), the term under which the tradition had gathered figures of sound and sense and which supplied the title of Bhāmaha’s work.76 The body of literature was textuality itself, “what was made of language” (vāṅmayam), which in Daṇḍin’s schema was “Sanskrit, Prakrit, or Apabhramsha, or mixed.”77
The “body of literature” was a metaphor of substance as opposed to accident: a text without figuration was plain, and perhaps not even literature, but a text without language was impossible. It was also a metaphor of unity. So long as “the whole of literature” is conceived of as an “organic unity of the highest order”—a unity that the discourse of poetics presupposed and sought to theorize—then the languages in which literature subsists can be thought to constitute an “organic unity” as well.78 Rājaśekhara’s famous image of “literature man” (kāvyapuruṣa) is a reinterpretation of Daṇḍin’s metaphor that makes the “four languages” (Daṇḍin’s three with the addition of Paishachi) into actual body parts: Sanskrit is the face, Prakrit the arms, Apabhramsha the groin, and Paishachi the feet.79
The “three languages” served as a top-level classification of literature. The word bhūyaḥ in Daṇḍin’s formulation does not mean that literary works may rarely be composed in other languages (“primarily”), but, as the commentator Ratnaśrījñāna notes, simply serves to introduce a new classification (“moreover”). Alternatively, we could take it as referring to the fact that every single literary work is either predominantly composed in one of the three languages—which Pollock has therefore called “primary languages”—or, in the case of stage plays, involves a tightly constrained “mixture” of languages.80 Bhāmaha implicitly and Daṇḍin explicitly map these languages onto literary genres.81
To enumerate is to exclude, as any speaker of Sanskrit would recognize.82 Sanskrit, Prakrit, and Apabhramsha never fully comprehended the domain of language practices, even textual language practices, at any point in Indian history. We can make sense of this apparent disconnect between theory and practice by highlighting two related features of enumerative totalities in general.
First, they are totalizing representations rather than representations of a totality. Take, for example, the story of Guṇāḍhya’s renunciation of the “three languages” related in the Ocean of the Rivers of Story, a twelfth-century collection of tales in the tradition of the Great Story attributed to Guṇāḍhya. In the Ocean, Guṇāḍhya is said to lose a bet with his colleague Śarvavarman about how long it will take to teach Sanskrit grammar to King Sātavāhana, and in consequence he gives up “Sanskrit, Prakrit, and the regional language, the three languages that are possible for human beings.”83 This leads him to learn “the fourth language,” that of inhuman ghouls called Piśācas, while living with them in the forest (see below).84 This story uses the rhetoric of n-ary structures to make the “three languages” representative of human culture as a whole, in contrast to the “fourth” language, which represents its very opposite.85 Despite the claim that they represent all of human culture, the figure of the “three languages” foregrounds Sanskrit and Prakrit and thus represents human culture from a privileged, educated, and courtly perspective. His story transforms the languages of the Sātavāhana court into the languages of literary culture and then into the languages of human civilization.
Rājaśekhara makes the same point even more clearly:
The language of the gods is worth hearing,
and the Prakrit languages are naturally sweet.
Apabhramsha is very pleasant,
and there are choice works in the language of the ghouls.
There are different paths,
but these are the ones that are preferred.
The one who writes in all of these is indeed a master poet.86
There are more languages than those enumerated in the schema, but these four are the only ones that matter. Nor do all four matter equally. Rājaśekhara called himself “skilled in all languages,” but he did not write any significant works in Apabhramsha or Paishachi.87 He advanced his claim to total expertise on the basis of his Prakrit compositions: for many poets could write in Sanskrit, but few—perhaps even none—had attempted to write an entire play in Prakrit, as Rājaśekhara did. Sanskrit and Prakrit metonymically represented the totality of literary languages, and even if Sanskrit remained Rājaśekhara’s preferred medium, Prakrit represented for him the seldom-gained summit of literary expertise.
Second, the enumerative totality is an integrated unity. Daṇḍin was more concerned than Bhāmaha to demonstrate that the languages of the schema were internally related. Perhaps this is because, as a resident of Kāñcīpuram in the Tamil country around 700 CE, he was exposed to different literary cultures that each had their own linguistic parameters. In contrast to Bhāmaha, Daṇḍin offers the standard threefold classification that systematically relates Prakrit to Sanskrit. He also proposed a solution to the slight disjuncture between what Prakrit meant in the context of “literature heard” and what it meant in the context of “literature seen.” He noted that it is the former that Prakrit was primarily associated with: this kind of Prakrit was, after all, the language in which “were composed works such as the Building of the Bridge, an ocean full of jewels of beautiful sayings.” But he added something to this characterization of the language, namely, that it was based in the region of Maharashtra.88 As we will see in chapter 6, this is also a relatively conventional description of the literary language (see the verse of Jayasiṃha above), and true to its historical origins in the western Deccan. But in the context of Daṇḍin’s discussion, this remark gave Prakrit a “regional” character that distinguished it from Sanskrit and brought it closer to another set of languages: namely, the theatrical languages notionally derived from Prakrit and given names that associate them, just as notionally, with particular regions. Daṇḍin says that Śaurasenī, Gauḍī, and Lāṭī—respectively associated with the northern midlands, the Ganges plain in the east, and present-day Gujarat in the west—can also be considered Prakrit in the context of representing conversations (vyavahāreṣu) in stage plays.89 He includes “other languages similar to them” (tādṛśī) in this set, reinforcing the Treatise on Theater’s constraint that the languages employed on the stage need to be more or less mutually intelligible. Daṇḍin’s discussion, especially compared to Bhāmaha’s relatively brief remarks, significantly expands the rubric of “Prakrit” and the languages it encompasses, but at the same time insists on the internal relationships between the languages that belong to this category: firstly, in terms of the preeminent position of the literary Prakrit, now increasingly regionalized as “Mahārāṣṭrī,” and, secondly, in terms of the criterion of similarity that applies to the languages of stage plays.
Within the literary culture whose practices it schematizes, the figure of the “three languages” was widely understood to be total in these senses. Uddyotana’s Prakrit romance Kuvalayamālā (778 CE) furnishes an important example in which Sanskrit, Prakrit, and Apabhramsha represent all of the languages that are “possible among human beings.” Dhanadeva is a merchant who has been shipwrecked in a distant land, and finally finds a quiet place in the forest to rest, after escaping cannibals and man-eating birds. He falls asleep under a tree, but immediately wakes up to the chattering of the ghouls (piśācas) who inhabit the forest. It takes him some time to identify the language that he hears, because he needs to compare it to Sanskrit, Prakrit, and Apabhramsha before finally deciding that it must be the “the fourth one, the language of the ghouls” (caütthā bhāsā pesāyā):
He listened and thought: “Wait a minute. What is this language that I hear being spoken? Hmm. Well, it can’t be Sanskrit, because that is harsh like the heart of a wicked person, difficult to understand with its hundreds of horrible options for forming all of the different words, compounds, indeclinables, prefixes, case endings, and genders. And this isn’t like that. So could it be Prakrit? Hmm, that’s not it, either, because that is pleasant like the words of good people, made up of the nectar that streams forth when great men churn the ocean of life that constantly surges with the waves of all learning, with compositions of various types that perfectly join their sounds and words together. And this certainly isn’t like that. So might it be Apabhramsha, then? Hmm, it’s not that either, because that is a mountain stream that gushes with floodwaters from the downpours of the first springtime clouds, rolling and swelling with the steady and unsteady waves that are the words of Sanskrit and Prakrit both pure and combined, alluringly harsh and gentle like the words of a lover in playful anger. And this isn’t like that at all...”90
The basic principle of this representation is the opposition between Sanskrit and Prakrit. Sanskrit is the sum of its grammatical parts much in the way that Latin was an assemblage of third-person passives and ablative plurals to generations of British schoolchildren, and associated with the tedium and terror of learning those distinctions. Prakrit, the language in which Uddyotana composed the Kuvalayamālā, is not necessarily natural and spontaneous, but it is figured as more closely aligned with lived experience, and thus more pleasant and more appropriate to literary compositions. There is an ethical difference, too: Sanskrit is aligned with wicked people—perhaps the sanctimonious and hypocritical Brahmans that Uddyotana’s teacher, Haribhadra Sūri, lampooned in his Rogue Stories—while Prakrit is cultivated by good people, preeminent among whom are Jain monks like Uddyotana himself.91 Apabhramsha is not represented as an entirely distinct third language but as a recombination of Sanskrit and Prakrit.
Uddyotana is well aware that other kinds of languages exist; he even represents a number of “regional languages” in a market scene later on in the novel.92 But the “three languages” are the languages of the court—as the description of the court of Dṛḍhavarman shows—and the languages of the literary culture that Uddyotana himself, and the protagonists of his novel, participated in.93
Svayambhū offers another compelling metaphor of totality in the introduction to his Deeds of Padma (ninth century). There, he compares the Rāma story to a great river that has flowed throughout the generations, and he compares the two banks of the river to Sanskrit and Prakrit. This is likely a reference to his predecessors, Vimala’s Deeds of Padma in Prakrit and Raviṣeṇa’s Legend of Padma in Sanskrit: the literary tradition prior to Svayambhū is divided into just two languages in the same way that a river has just two banks.94
A final example of what the enumerative totality represents can be drawn from a passage in Bilhaṇa’s Deeds of King Vikramāṅka, composed in eleventh-century Karnataka, but looking back in the following excerpt on the poet’s home town in Kashmir:
What can I say about Pravarapura?
It’s a source of wonder,
filling the ears with the nectar of so many marvelous stories,
where the Sanskrit and Prakrit languages
resound in every single house
as if they were the mother languages
even of women, to say nothing else.95
Here Sanskrit and Prakrit form a binary structure that contrasts with the janmabhāṣās, literally, “birth languages,” that one might have expected housewives to speak. This binary represents “culture” with all of the tensions and aspirations of the English word: the “works and practices” in general that define us as members of a group, and those of intellectual and artistic creativity in particular.96
ITERABILITY
The distinctions that operate over a schema as a whole can be reinscribed onto its constituent parts. This process of iteration results in fractal representations, rather than the n-ary representations we have surveyed in the preceding section. In contrast to the diachronic expansion of a schema through the introduction of new distinctions, the iteration of existing distinctions is synchronic. The representations produced by iteration run parallel to each other, while those produced by expansion follow upon each other in history.
Apabhramsha furnishes the major example of iteration within the language order of classical India. The term “Apabhramsha” itself, meaning “deviation,” has a longer history than either “Sanskrit” or “Prakrit” in Indian discourses on language. Patañjali used it as a synonym for incorrect words, and his usage was recognized by Daṇḍin: “with reference to scientific works, anything other than Sanskrit is called Apabhramsha.”97
The qualification is necessary because, by Daṇḍin’s time, Apabhramsha had acquired a more specific meaning. It referred to a literary language besides Sanskrit and Prakrit, and thus Daṇḍin defines Apabhramsha, with reference to literary works, as “the language of people such as the Ābhīras.” The Ābhīras were a group who came to political prominence in the Deccan in the twilight of the Sātavāhana empire, around the middle of the third century, but Daṇḍin’s statement provides nearly all we know about their association with Apabhramsha as a literary language.98 It is significant that this newcomer to the field of literary languages was given the very name that was formerly used to denominate all non-Sanskrit language practices. Prakrit was Apabhramsha, in this basic sense of a “deviation,” before Apabhramsha was Apabhramsha. In other words, Apabhramsha slid into the position in the language order occupied by Prakrit. Not only that, but it was imagined and represented in very much the same way as Prakrit was. Daṇḍin’s tenth-century commentator Ratnaśrījñāna mentions a tradition that analyzed Apabhramsha into exactly the same four categories into which earlier teachers had divided Prakrit.99
Apabhramsha is thus seen as the result of a kind of mitosis of Prakrit. This representation aligns with the relationship between Prakrit and Apabhramsha in practice, for these languages often occupy the same discursive space: works in Apabhramsha include prologues in Prakrit (such as the Message Poem of ʿAbd ur-Raḥmān); Prakrit anthologies include verses in Apabhramsha (such as the Verses of the Chappaṇṇayas); Apabhramsha verse forms were used occasionally in Prakrit, Prakrit verse forms were used abundantly in Apabhramsha; the same authors composed works in both languages. ʿAbd ur-Raḥmān expressly represents himself as a Prakrit poet, and for good reason: not only does the Message Poem include several Prakrit gāthās, but it engages with Prakrit intertexts at nearly every turn.100 It is with some justice, then, that Herman Tieken has sought to see Apabhramsha as “a Prakrit,” by which he means that Apabhramsha literature is essentially Prakrit literature written in a different language.101
Another clear example of iteration comes from the way that Abhinavagupta understood the categories of language laid out in the Treatise on Theater. What Bharata calls a “language” (bhāṣā) is a deviation (apabhraṃśaḥ) from Sanskrit, and what Bharata calls a “sublanguage” (vibhāṣā) is a deviation (apabhraṃśaḥ) from a language.102 Another example might be drawn from the use of the concept in Prakrit grammar. In this discourse, Sanskrit figured as the archetype (prakṛtiḥ) and Prakrit as the ectype (vikṛtiḥ): Prakrit words were derived from Sanskrit words by a set of transformational rules. When Prakrit grammar grew to encompass the languages of the theater, Śaurasenī and Māgadhī occupied the position of ectypes in relation to Prakrit, which was repositioned as an archetype. Just as in the Treatise on Theater’s typology, a procedure of derivation connects Sanskrit to Prakrit, and the same procedure connects Prakrit to Śaurasenī and Māgadhī. In the influential grammar composed by Hemacandra in the middle of the twelfth century, the Siddhahemacandra, the final stop on this itinerary is Apabhramsha. Iteration within this schema comes to an end with Apabhramsha, perhaps because Apabhramsha—whatever specific practices this term referred to—is always axiomatically configured as the furthest stop away from the starting point that is Sanskrit. The same logic operates in the eastern Prakrit grammars, for example in Mārkaṇḍeya’s Sum-Total of Prakrit, although here it is the paiśācika languages that are the last stop, after bhāṣās, vibhāṣās, and apabhraṃśas.
The scope of Bhoja’s discussion of language in his Illumination of the Erotic, like the Siddhahemacandra modeled on it, is the totality of literary culture.103 But whereas Hemacandra represents each successive language as a transformation of the preceding, Bhoja proceeds by iterative divisions. The “three languages,” each of which has three further subdivisions, and each of those has two varieties, are his starting point. Regarding Apabhramsha, Bhoja arranges six notionally regional varieties under the three subdivisions of “high,” “middle,” and “low.” Regarding Prakrit, Bhoja synthesizes two existing classifications, one that recognized a number of “regional” varieties of Prakrit (Śaurasenī, Māgadhī, etc.), and one that classified Prakrit words on the basis of their derivational distance from Sanskrit (tatsama, tadbhava, deśya; see the following chapter). Bhoja’s “Prakrit” is divided into “natural” (sahajam), “derived” (lakṣitam), and “distorted” (śliṣṭam). The first category alludes to a kind of language that is independent of grammar, either because it is identical to Sanskrit (saṃskṛta-samam) or because it has no relationship to Sanskrit at all (deśyam); the second includes the main varieties of Prakrit that are grammatically derived from Sanskrit, mahārāṣṭram and śaurasenam; the third includes languages that are more distant from Sanskrit (such as māgadham) or at least more obscure to the grammarian (such as paiśācam); the latter are similar in status to the Treatise on Theater’s “sublanguages,” in that they are second-order deviations.
The principle of iteration explains why the representations of language we encounter in Indian texts, although they do differ from each other, differ in systematic and tightly constrained ways. We can formulate for them a set of “implicational universals,” a term that linguists use to describe the necessary occurrence of one feature given another feature. If a representation distinguishes two languages, then one of them must be Sanskrit. If it distinguishes three, then Sanskrit and Prakrit must be two of the three. And if it distinguishes more than three, then it must include Sanskrit, Prakrit, and Apabhramsha. These implications build in some latitude, since there is always at least one indeterminate slot, but the other slots are determined by the schema under analysis here.
THE HALF-LANGUAGE
To say that the schema described above is archetypal is, in the first place, to recognize its primacy in ordering language practices over a vast domain of textual production. In fact, the large-scale formation that has been described as “classical India,” and more recently as the “Sanskrit cosmopolis,” can be reframed in terms of these ordered language practices: it is the world in which textuality is governed by the schema of co-figuration of Sanskrit and Prakrit. It is not simply the world in which these specific languages are employed, but the world in which the use of these languages is essentially linked to the exercise and maintenance of culture-power. As Sheldon Pollock has argued at length, this was not only, and perhaps not even primarily, due to military conquest, colonization, trade, or the spread of religious ideas.104 Absolutely essential to the determination of Sanskrit and Prakrit as languages of culture power were schematic representations such as those we have seen in this chapter.
Prakrit has generally been omitted from this story, as the very phrase “Sanskrit cosmopolis” suggests. But once we recognize that languages are constituted as what they are only within larger structures that I call language orders, we must recognize also that Sanskrit depends on Prakrit and vice versa, both historically and conceptually. As I have tried to show, the names “Sanskrit” and “Prakrit” only come to be used to designate language practices in around the first century CE, and are used to designate them contrastively within a new sphere of textuality whose limits they jointly define. Apabhramsha appears somewhat later, but when it does, it appears within the framework already established by the opposition, identity, and totality of Sanskrit and Prakrit. Textuality in the Sanskrit cosmopolis was never simply Sanskrit textuality, but it was configured by the identity-in-difference of Sanskrit, Prakrit, and Apabhramsha.
This schema is archetypal in the further sense that it admits of modifications. Arguably, the language order it describes was only uprooted and replaced by European colonialism. This leaves more than fifteen hundred years of language practices that were subsumed under a wide variety of schemas that can generally be seen as ectypal modifications of the archetypal schema presented above, as well as language practices that remained more or less outside of the unified language order or constituted a kind of counterpart to it. Śrīnātha, the fourteenth- and fifteenth-century Telugu poet, can serve as a good example of both. In composing literature in Telugu at all, he was certainly breaking away from the model of the “three languages.” He was, however, not rejecting it but extending it. He styled himself a “lord among poets in the eight languages.” The following sections will explain how the schema was extended from three to eight, but for the moment it will suffice to note that Śrīnātha includes Sanskrit, Prakrit, and Apabhramsha among these languages. Despite this expansion, a number of important language practices remained unintegrated in his schema, above all those introduced by the Bahmani sultans just to the west: Arabic, Persian, and Turkish. Śrīnātha is well aware of these languages, and praises one of his patrons for his mastery of them, but does not—and perhaps cannot—integrate them into a single representational schema with the “eight languages.”105
These concluding sections will examine just two modifications of the archetypal schema: the addition of Paishachi as a “half-language,” and later as a full language, alongside Sanskrit, Prakrit, and Apabhramsha; and the expansion of this schema of three and a half or four languages into the enduring schema of six languages. I focus on these modifications in particular because the first illustrates the power of the schema to conjure an entire language from nothing, as it were, and the second represents a major redetermination of Prakrit as a concept and as a category.106
An inscription in far-off Cambodia around 900 CE described King Yaśovarman I as “a Guṇāḍhya who hates Prakrit” (guṇāḍhyaḥ prākṛtāpriyaḥ), an apparent contradiction, which resolves to “rich in virtues and no lover of what is base.”107 Guṇāḍhya was the author of the Great Story, which has been called one of the three streams of Sarasvatī alongside the Mahābhārata and the Rāmāyaṇa.108 The Great Story itself, however, is lost: all we have are retellings in Sanskrit, Prakrit, and Tamil.109 It seems to be always already translated, for the earliest mention of it in the sources available to us is an inscription in which the Gaṅga king Durvinīta claims to have rendered it into Sanskrit.110 Yaśovarman’s reference to Guṇāḍhya might lead us to think that the Great Story was composed in Prakrit. But Daṇḍin seems to have considered it an exception to the rules of textuality he himself enunciated. Stories (kathā), he tells us in the Mirror of Literature, are composed in all languages, but most commonly in Sanskrit. The exception is “the wondrous Great Story, which is composed in bhūtabhāṣā.”111
There has been an enormous amount of discussion about what this bhūtabhāṣā was and what its characteristics were. Scholars have attempted to identify this language with the spoken vernacular of one or another group. The crucial maneuver has been the identification of Daṇḍin’s bhūtabhāṣā with the language that ghouls (piśācas) are imagined to speak and are, on a few occasions, represented as speaking. The identification with bhūtabhāṣā with Paishachi, as this imaginary language was so called, rests on the interpretation of the compound as a “language of the dead.” But I believe that Daṇḍin meant to describe the language of the Great Story as a “dead language”: a language of the literary past. This bhūtabhāṣā was neither Sanskrit nor Prakrit nor Apabhramsha. It was incompatible, for reasons that are lost to us, with the principles of textuality that governed the classical language order, and that is why the only text ever known to have been composed in this language, the Great Story, seems to have always been known through translations.
The earliest surviving Kannada text, the Way of the Poet-King (ninth century), faithfully represents the circumstance of co-figuration described earlier in this chapter: besides Kannada, which the text endeavors to theorize, the only languages mentioned are Sanskrit and Prakrit, which are represented as the only languages in which high literature may be composed.112 But in the tenth century, a number of authors started to speak of “three and a half languages,” where the half was Paishachi.113 It is “half” a language precisely in the sense that Daṇḍin suggests: important literature has been composed in it, but unlike the “three languages,” no new literature could be composed in it. But does their use of the name “Paishachi” suggest that it was really thought of as the language of ghouls?
I have argued that the appearance of Paishachi within schemas of language after Daṇḍin’s time was the result of a literary joke gone wrong—or perhaps gone right. Uddyotana tells us that he included some passages in languages other than Prakrit in the Kuvalayamālā “for fun” (koūhaleṇa).114 In a scene I’ve already mentioned, the merchant Dhanadeva finds himself surrounded on a desert island by a horde of ghouls (pisāyas) who speak ghoulish (pesāyā). The language of this scene might plausibly be modeled on that of the Great Story, as a dead language that Uddyotana cleverly repurposed as the language of the undead. The Kashmiri retellings of the Great Story in the eleventh century say that Guṇāḍhya composed the work in ghoulish, precisely because he took a vow that prevented him from using the three languages current among men, but significantly this detail is absent in all of the earlier retellings of the story. In my view, this detail reflects a retrospective identification of the dead language in which the work was composed as the language that Uddyotana calls Paishachi. Whatever the truth is, Paishachi went from being a non-language in the enumerative schemas of the seventh and eighth centuries to being a half-language, and later on a full language, in subsequent representations. It is not that new literature was written in this language. On the contrary, fewer and fewer people seemed to have had direct access to the Great Story as time went on. What was new was simply that it had been included in the schemas from which it had earlier been excluded. This made it available, in principle, for literary composition, although the lack of literary models made composition in the language difficult in practice. In fact, apart from fragments of the Great Story, nearly the only writing in Paishachi we have are literary experiments like Uddyotana’s. A very similar scene to the one in the Kuvalayamālā would be included by Jineśvara in his Story of Nirvāṇa and Līlāvatī (1036), and Hemacandra in the twelfth century would write a short section in Paishachi to illustrate the grammatical rules that he collected in the Siddhahemacandra and probably culled from experiments like Uddyotana’s.
THE SIX LANGUAGES
The transformation of Paishachi from non-language to language is just one part of an important refiguring of language practices that took place shortly before the ninth century: the threefold schema of Sanskrit, Prakrit, and Apabhramsha was replaced by a sixfold schema that added Śaurasenī, Māgadhī, and Paishachi. The earliest text to exhibit this refiguration is Rudraṭa’s Ornament of Literature, composed in Kashmir in the early ninth century.115 Śaurasenī and Māgadhī, as we saw above, were used exclusively in the theater, which had in the generations before Rudraṭa become the analytical focus of Kashmiri theorists of Sanskrit literature. As is well known, during the reign of Jayāpīḍa (779–813), Bhaṭṭa Udbhaṭa began a tradition of studying and commenting upon the Treatise on Theater in Kashmir. The shift in focus to “literature seen” (dṛśyakāvya), as opposed to “literature heard” (śravyakāvya), entailed a shift of focus from monoglossic to polyglossic genres. In the theater, language was not predetermined by genre, but could be an object of choice and purposeful manipulation.
One of the techniques of language manipulation is bhāṣāśleṣa, in which a verse is spoken in two or more languages at the same time, either with the same meaning or with different meanings.116 This provides a way of manipulating the language assignments in a play—for instance, a character who is “supposed” to speak Sanskrit may speak Prakrit and vice versa—as well as a clever way of saying two different things to two notionally different groups of people.117 But it also provides a way of surreptitiously modifying the language of a composition in “literature heard,” which otherwise does not admit of such changes. Hence we find bhāṣāśleṣa sections in works such as Bhaṭṭi’s Poem and Śivasvāmi’s Rise of Kapphiṇa. Bhoja’s discussion of the “type” of language (jāti) in his Necklace of Sarasvatī reflects this new theoretical orientation according to which language is an object of choice, and therefore something about which judgments of propriety (aucitya) can be rendered. This represents a major departure from Bhāmaha and Daṇḍin. For Rudraṭa and Bhoja, language does not just constitute the “body” of literature but could itself become an “adornment.”
Rudraṭa’s “six languages” provided the basis for a new kind of linguistic knowledge that was textualized in the form of the multilingual grammar. The earliest datable text that might be called a multilingual grammar is in fact Namisādhu’s commentary on the Ornament, completed in 1069. While commenting on Rudraṭa’s exposition of the “six languages,” Namisādhu provides a short description of each of them, referring to rules that he has either taken from earlier grammars (perhaps Harivṛddha’s lost grammar, which he quotes elsewhere) or inferred from actual texts (such as Uddyotana’s Kuvalayamālā in the case of Paishachi). Other multilingual grammars from around this time include the “expanded” version of the Light on Prakrit, with chapters on Śaurasenī, Māgadhī, and Paishachi (see chapter 6), and Kramadīśvara’s Distilled Essence. The most complete and most influential grammar of this type was Hemacandra’s Siddhahemacandra, which adopts Rudraṭa’s “six languages” as its organizing principle and defines Sanskrit, Prakrit, Śaurasenī, Māgadhī, Paishachi, and Apabhramsha in turn. For most authors after Hemacandra, that there were six languages was common knowledge.118
CONCLUSIONS
The schema that I have presented in this chapter underlies the representation of language in classical India. It supplies the basic categories—including the languages themselves—and calibrates a complex set of relations, constituting a framework within which language can be thought. The overall picture that emerges from this schema should now be clear. Sanskrit and Prakrit are mutually constitutive languages, closely related to each other but contrasted across a number of dimensions. Even further from Sanskrit in the direction of Prakrit is Apabhramsha. These three languages form a coherent unity. They are the only languages in which literature can be composed, and they thus represent the linguistic parameters of a literary culture.
This picture closely matches the actual practices of literature from the second to the ninth century, from Kashmir to the Kaveri river. This picture has two particularities, in comparison with later imaginations of language in South Asia, that I will simply note here; many other particularities could be discerned if the comparative lens were turned to literary cultures outside of South Asia. The first is that language is imagined as monocentric. It does not matter whether Sanskrit or Prakrit is taken to be the center, since they are imagined to be identical at a deeper level in any case. The name “Prakrit” itself suggests a relationship to a single “source” (prakṛti). On this model, all languages are related to each other through the central source. There is no possibility of a polycentric language order of the kind that the Pāṇṭiya rulers of the area around Maturai in Tamil Nadu fashioned in the ninth century, in which Sanskrit and Tamil were accorded something approaching equal status and authority.119 The second particularity is that vernacular textuality is not just absent but unthinkable within this schema. There is plenty of evidence that Prakrit and especially Apabhramsha were thought of as regional languages (deśabhāṣās). This does necessarily imply that regional languages as we understand them were in turn thought of as Prakrit or Apabhramsha: as the following two chapters show, regional languages were indeed represented as Prakrit and Apabhramsha, but this was part of the process of vernacular literarization that took place centuries after the foundations of the Sanskrit cosmopolis, including the archetypal schema of its language order, had been laid.120 For much of the first millennium, the regional was not conceived as a source of authority or legitimacy in itself, but was rather defined negatively, as a site of difference from transregional Sanskrit.
The classical schema made Prakrit an object of imagination, representation, and knowledge. The following chapter will examine in detail the systems of knowledge that Prakrit was the object of, grammar and lexicography, and the concepts and strategies that were developed in these systems. One of these concepts is “the regional” (deśya), which links the classical language order to the vernacular language orders that followed it.