PRAKRIT KNOWLEDGE
The history of Prakrit is closely bound up with the history of knowledge about Prakrit. In this chapter I examine the discourses in which this knowledge was systematically articulated. To see precisely how these discourses constituted Prakrit as a stable and coherent object of knowledge, we need to look at them at two different resolutions. At a lower resolution, what we see are texts that are situated in traditions, and the important question is how the traditions of Prakrit grammar, metrics, and lexicography develop in tandem with Prakrit literary traditions. At a higher resolution, what we see are conceptual strands that run throughout these texts, structuring them and tying them into larger discursive configurations. The extension of concepts formulated in order to account for Prakrit into new domains of textuality was crucial to the process of vernacularization, although modern scholarship has ignored or minimized the provenance of these concepts.
Just what was systematic knowledge of Prakrit? In the middle of the twelfth century, the Jain monk Hemacandra composed a number of works in which he sought to synthesize the knowledge that was necessary to participate fully in literary culture.1 This knowledge was organized into the four domains of grammar, lexicography, metrics, and poetics, each the subject of separate works by Hemacandra himself. There is much that is new in this configuration, but it exhibits two features that characterize systematic knowledge of Prakrit over its long history: first, it is dispersed over interlocking domains; second, it is a literary-cultural knowledge, which is clear enough in the case of metrics and poetics, but must be emphasized in the case of grammar and lexicography. The “contexts of use” (prayoga) with which grammarians and lexicographers were concerned were always literary contexts. To illustrate his own rules, Hemacandra very often quotes verses from literary works such as Seven Centuries and Rāvaṇa’s Demise, and very rarely from the Jain scriptures, and he never quotes examples from the language of everyday life.
Prakrit knowledge was thus philological. For this characterization I invoke a heuristic distinction between philology, which is oriented toward texts, and linguistics, which is oriented toward language—“heuristic,” of course, because texts are made out of language, and language, for most of human history, can only be accessed through texts.2 Although the primary object of Prakrit knowledge was language, it was never language per se, but language that either was, or could be, deployed in literary texts. Prakrit knowledge was not a “model of” a linguistic reality with an independent existence, but a “model for” the continuous recreation—through reading, commenting, anthologizing, recombining, and composing anew—of literary traditions. We risk misconstruing the enterprise entirely if we conceive of it on the model of linguistics, either in its Pāṇinian or modern incarnations.3
The central component of this configuration was grammar. The “centripetalizing” force of grammatical discourse in the modern world—its ability to determine or redetermine language as a single object with a single source of authority—has long been recognized. It has been particularly important in shaping the national languages which modern subjects have identified with and cathected upon.4
But grammar is not an invention of modernity. In this chapter I adopt a two-pronged strategy for recovering what Prakrit grammar was, and, more important, what it did, in premodern India.
On the one hand, I argue that Prakrit grammar was just like any other grammatical discourse. These discourses do not simply list, or provide the rules for generating, forms of a given language. They teach people to think of the language under description, of language in general, and of culture more broadly, through a certain set of models, concepts, and relations.5 Since Prakrit grammar is seen as a tiny, obscure subject, lacking both the sophistication and dynamism of Sanskrit grammar, and hence hardly studied at all, I want to emphasize this point: anyone in premodern India who thought in any depth about the relationships between different languages, or between cultural practices delimited by language—in a word, about polyglossia—used concepts that originated in Prakrit grammar.
On the other hand, I argue that Prakrit grammar was different. We can think about these differences using the terms that grammatical discourse in India itself provides. It consists of a set of rules, called a lakṣaṇa (“that which defines”), which serves to characterize a set of linguistic phenomena, called a lakṣya (“that which is defined”). With regard to the former, Prakrit grammar is very closely related to Sanskrit grammar, but because it needs to define one language in terms of another—because it is interlingual rather than intralingual—it has certain concepts, strategies, and techniques of its own.6 With regard to the latter, Prakrit grammar describes a very different kind of language from Sanskrit or the regional vernaculars, not to speak of modern national languages. There were never, to our knowledge, any communities that defined themselves by their use of Prakrit, no “Prakritikas” comparable to Kannadigas or Tamilians, nor did Prakrit ever approach Sanskrit’s broad acceptance as a language of learning that cut across such communities. It was, for most of its history, an exclusively literary language, and the enterprise of Prakrit grammar could not but reflect the fact that the language belonged to an elective subculture of experts and connoisseurs, if it belonged to anyone.
This approach requires going behind the descriptive–prescriptive dichotomy, and by that I mean examining the complex relationships between lakṣya and lakṣaṇa, and between grammar and its uses and effects, that are preprocessed and flattened out by the terms “descriptive” and “prescriptive.” The descriptive–prescriptive distinction was never explicitly made in Indian grammatical traditions, and it dissolves upon closer analysis even in the twentieth-century projects that explicitly identify with one or the other modality.7 Yet it retains a heuristic value. Conceiving of Prakrit grammar as a “descriptive” enterprise would require us to identify the specific forms of language that it sought to describe at various points in its history; conceiving of it as “prescriptive” would require us to identify its specific practical applications. But because these conceptions are only heuristic, we should not expect to find, in the first case, a stable object language represented by a fixed corpus of texts, and in the second, a coherent regulative agenda. Ultimately these tasks will take us back to the ontology of the languages for which Prakrit grammar serves as an epistemology: where, when, for whom, in what contexts, and given what preconditions did they exist?
AN ARCHAEOLOGY OF PRAKRIT KNOWLEDGE
Our history of Prakrit knowledge starts in the middle of its history. The earliest contributions to Prakrit grammar and lexicography that we can reliably locate in time were composed in the tenth and eleventh centuries, long after these discourses first took shape. These include the Prakrit Lakṣmī of Dhanapāla (972) and Namisādhu’s commentary (1069) on Rudraṭa’s Ornament of Literature. Earlier texts survive in the discourse of Prakrit metrics, but these too carry indications of a longer prehistory that is lost to us. The scarcity of surviving works is probably due to the “Hemacandra bottleneck.” Hemacandra’s writings became the primary reference point for the systematic knowledge of Prakrit almost as soon as the ink was dry, and consequently earlier works were no longer studied and transmitted. Much has been lost, and much that survives cannot be dated with certainty. An example of the latter is Caṇḍa’s grammar, which has circulated in various forms and under various names, and has been assigned to the last centuries BCE (by Hoernle) and the early second millennium CE (by Bloch) and various times in between.8
What I offer in the following pages is an archaeology of Prakrit knowledge, although more in the spirit of Cuvier than of Foucault. It is an attempt to construct a historical narrative on the basis of texts that resist it: lost texts, fragmentary texts, poorly preserved texts, corrupt texts, authorless texts, imaginary texts, mythical texts. The fact that we cannot always link these texts to names, places, and dates does not mean that they lie outside of history. Nor is the history of Prakrit knowledge as a discourse identical with the chronology of the individual texts that constitute it. My archaeology attempts to recover the overarching goals of these texts, their scope and analytical techniques, their principal intertexts, and the changes that the discourse underwent.
The materials that do survive suggest that Prakrit knowledge began at the court of the Sātavāhana kings in the early centuries of the first millennium CE. This should come as no surprise after seeing in chapter 3 the leading role that Sātavāhanas played in inventing and patronizing Prakrit literature. It also appears that the earliest works of Prakrit literature presuppose a body of systematic literary knowledge. Seven Centuries, for example, is strikingly unified in metrical form and language. There are scattered indications that the very people responsible for giving Seven Centuries its final shape—above all the author-editor known to tradition as Sātavāhana—were also responsible for theorizing the grammatical, lexical, and metrical forms of which Prakrit literature consisted.9
On seven occasions in his Prakrit lexicon, Hemacandra refers to Sātavāhana’s Sanskrit definitions of Prakrit words. The words cannot be traced in Seven Centuries, so Hemacandra must be either paraphrasing or quoting another work. The latter seems more likely, given that most of the references can be read as parts of an anuṣṭubh verse, although Hemacandra may be using an intermediate source.10 Virahāṅka and Svayambhū, writing around the eighth and ninth centuries respectively, also refer to Sātavāhana in the context of Prakrit metrical forms, and notably forms that do not occur in Seven Centuries.11 Ghanaśyāma, an author of the eighteenth century, refers to “Śālivāhana” as a lexical and grammatical authority who wrote a work called Moonlight of Prakrit (Prākṛtacandrikā). Some, but not all, of these references involve a Prakrit word being defined with a Sanskrit synonym in an anuṣṭubh verse (or a reference that can plausibly be reconstructed as such), and it is possible—although by no means certain—that Ghanaśyāma was quoting from the same work as Hemacandra.12 This work seems to have been a practical handbook to Prakrit composition, covering the basic points of grammar as well as points of usage and vocabulary.13
Another author only known to us from fragments is Harivṛddha. He is often mentioned in the same breath as Sātavāhana, and it seems likely that he was his contemporary. A few of his verses are quoted by Ratnaśrījñāna (tenth century) and Namisādhu (eleventh century). What is notable about these verses is that they are written in Prakrit, using the gāthā verse form typical of Prakrit literature. Similar verses are quoted without attribution in other works, including the Dhavalā and Jayadhavalā of Vīrasena and Jinasena (ninth-century Karnataka), the Treatise on Theater, Nanditāḍhya’s Definition of the Gāthā, and Caṇḍa’s Definition of Prakrit. Together they show that knowledge about Prakrit was articulated, and probably was first articulated, in Prakrit. The grammatical fragments provide a broad characterization of Prakrit phonology and morphology rather than concise transformational rules in the style of either Pāṇini’s grammar of Sanskrit or later grammars of Prakrit.14
The most important, and to all appearances the most influential, idea in Harivṛddha’s fragments is the “metagrammatical” classification of Prakrit itself, which I discuss later. These verses also show, however, that knowledge of Prakrit was never limited to knowledge of the forms of the Prakrit language, but was always oriented toward literary practice. One verse of Harivṛddha enumerates eight varieties of speech (bhaṇitis), which largely coincide with what later authors would call alliterative styles (anuprāsavṛttis).
Luigia Nitti-Dolci saw in the grammatical fragments an abortive attempt, on the part of Jain scholars, to describe the language in which the texts of their tradition were composed, in contrast to the language of secular and courtly texts. She saw Caṇḍa’s Definition of Prakrit as a synthesis of this material, which was “neither abundant nor properly classified.”15 As I argued in chapter 3, however, separating Jain and non-Jain varieties of Prakrit—what scholars now call Jain Māhārāṣṭrī and Māhārāṣṭrī—would have made little sense to the people who actually wrote in these languages. Nor it is clear that the authors of these Prakrit verses were themselves Jains. What will become clear, however, is that Harivṛddha saw himself as defining a field of Prakrit literature rather than a field of Jain literature that happened to be written in Prakrit.
At least one text, Mirror of Figures (Alaṃkāradappaṇa), testifies to the existence of a discourse on poetics in Prakrit. Although it tells us little that we didn’t know from Sanskrit sources, it may well be earlier than most of those Sanskrit sources. I believe that this text represents the discourse on poetics prior to Bhāmaha (prior to 700 CE), a period concerning which we otherwise have only fragmentary evidence.16 For the moment, however, the position in the history of poetics of Mirror of Figures—and works of systematic knowledge in Prakrit more generally—must remain an open question.
We are on more solid ground when it comes to metrics. We have two major treatises on metrics written in Prakrit, Virahāṅka’s Collection of Mora- and Syllable-Counting Meters and Svayambhū’s Meters, and both refer to a handful of earlier authors. Svayambhū lived in the later ninth century; he wrote Apabhramsha epics about Rāma (Deeds of Padma) and Ariṣṭanemi (Deeds of Ariṣṭanemi). The identity of Virahāṅka remains a mystery. Velankar located him between the sixth and eighth centuries.17 Although I cannot prove it, I suspect that Virahāṅka’s Collection is an early work of the brilliant eighth-century poet, doxographer, and philosopher Haribhadra before his conversion to Jainism. The name Virahāṅka refers to his use of the word viraha as a “signature” (aṅka, cihna, or lāñchana) that poets worked into the concluding verses of their works. The only author I know to have used this signature is Haribhadra, but the signature viraha (“separation,” usually of two lovers) is slightly odd for a Jain monk, and explanations of it in Jain sources seem forced. Haribhadra might thus have used the signature viraha, “separation,” when he was young, and after his conversion to Jainism, reinterpreted it as bhavaviraha, “separation from worldly existence.”18 A possible corroborating instance is the Prakrit Lakṣmī, written in 972 CE by Dhanapāla, who would later convert to Jainism and write Tilakamañjarī and Fifty Verses for Ṛṣabha.19
Prakrit metrics is not just Sanskrit metrics in Prakrit. Although it defines and exemplifies all of the syllable-counting meters used in Sanskrit literature, called vṛttas, its real focus is on the mora-counting meters that distinctively characterize Prakrit literature, called jātis; this dual aspect is referenced in Virahāṅka’s title. Prakrit metrics defines many more of these jātis than Sanskrit metrics does, and in fact many more than are actually attested in the surviving literature. Svayambhū in particular gives us some insight into the richness of Prakrit literature at his time, quoting from authors such as Jīvadeva and Śuddhasvabhāva whose works are otherwise completely lost.
A number of other early authors are merely mentioned, or briefly quoted, in later works. Unsurprisingly, many of those who made contributions to lexicography and metrics were themselves poets, as we know from the fact that other authors have quoted their verses or from the fact that they are identified by literary noms de plume. One author whom Svayambhū quotes is Abhimānacihna (“the poet who used the signature ‘pride’ ”), the author of a lexicon in Prakrit cited frequently by Hemacandra. These quotations confirm the impression that the systematic knowledge of Prakrit developed alongside Prakrit literary practice throughout the first millennium CE.
As the distance from its original circumstances of composition grew, and as it was rearranged, integrated into other texts, and lost, this earlier material was imagined to belong to “time out of mind,” and was accordingly reattributed to sages of the mythical past.20 Sometimes such reattribution occurred even in the absence of temporal distance, for reasons that are still difficult to determine. The best-known case is that of the Vālmīki Sūtras, a grammar of Prakrit that was, as the name implies, thought to have been composed by the semi-mythical author of the Rāmāyaṇa. A. N. Upadhye has argued convincingly that these Vālmīki Sūtras are none other than the sūtras composed by the Jain monk Trivikramadeva in the thirteenth century, which were reattributed to Vālmīki by later Hindu authors.21 Another example is Pāṇini. Starting, it seems, with Bhoja in the eleventh century, a number of authors believed that the most influential Sanskrit grammarian had also written a grammar of Prakrit. The few quotations from this alleged grammar make it hard to believe that its author was Pāṇini, who in any case lived several centuries before people began thinking about Prakrit as a language.22
The attributions to Pāṇini and Vālmīki locate the origins of Prakrit knowledge in the founding figures of the Sanskrit grammatical and literary traditions respectively, and thus affirm the prevalent understanding of Sanskrit and Prakrit by making them literally cognate traditions. The “eastern grammarians” (Puruṣottamadeva, Laṅkeśvara, Rāmaśarman, Mārkaṇḍeya) likewise refer to several mythical sages—Śākalya, Bharata, Kohala, and Kapila—under whose names various systems of knowledge circulated, of which only the Treatise on Theater ascribed to Bharata survives.23
It might be argued that the ascription of works of Prakrit lexicography and metrics to Sātavāhana is parallel to the ascription of Prakrit grammars to Vālmīki and Pāṇini, in that the author’s celebrity precedes and occasions the ascription. The reason I credit the former and not the latter is that Prakrit literature was the basis for Sātavāhana’s celebrity, whereas the others were known first and foremost for their contributions to Sanskrit literature and its forms of knowledge and were only associated with Prakrit much later. Further, there are deep connections between the literary productions of the Sātavāhana court and Prakrit forms of knowledge that either did not exist, or can easily be explained otherwise, in the other cases.
The earliest Prakrit grammar that survives in its entirety—or, as we will see, in more than its entirety—is Light on Prakrit, ascribed to the legendary figure Vararuci. The earliest and most widespread traditions about Vararuci make him one of the ministers of King Nanda, who ruled the Gangetic plain just prior to Alexander the Great’s forays into India. He is, however, also counted among the “nine jewels” of the court of Candragupta II Vikramāditya. Several texts besides Light circulate under his name, most notably a one-act play called Both Go to Meet and a collection of one hundred gnomic verses. A verse commentary on Light, called A Cluster of Blossoms of Prakrit, gives Vararuci the family name Kātyāyana, which evokes—if it does not identify him with—the famous author of a set of critical notes (vārttikas) on Pāṇini’s Aṣṭādhyāyī. Cluster is hardly the first text to identify Vararuci with Kātyāyana.24
From one perspective, then, the authorship of the earliest and most important grammar of Prakrit is thus beset with philological difficulties. The fragile originary connection between a man and his work, moving forward through time, collides against the will to remember otherwise—to reach back into the past and overwrite it, to reassign identities, to constantly reauthorize the text. From another perspective, the solution to this problem is ultimately not a judgment about the historicity, or lack thereof, of these crisscrossed traditions, but an understanding of the motivations, logics, and mechanisms of attribution. For these we have a parallel in the oldest extant grammar of Pali, which is likewise attributed to Kātyāyana (Kaccāyana in Pali). Centuries after the historical Kātyāyana composed his vārttikas on the Aṣṭādhyāyī, his name—and that of Vararuci, with whom he was identified—was attached to projects that sought to apply the principles and techniques of Sanskrit grammar to Middle Indic languages.
These projects can be seen as part of a broader movement to “liberate” these techniques, so to speak, from the tradition of the Aṣṭādhyāyī, with the goal of bringing to order a wider variety of language practices.25 This movement, which propelled Sanskrit beyond its ritual confines into its new role as a language of power, started with Kaumāralāta and Kātantra, both composed in the early centuries of the common era.26 Light on Prakrit’s debts to the tradition of Kātantra have been overlooked, perhaps because they are obvious. Besides some overlap in their technical terminology, the sūtras of both works, unlike those of Aṣṭādhyāyī, are arranged topically. Light also puts its very brief treatment of nominal suffixes at the end of a chapter on “miscellaneous rules,” and the section on nominal suffixes in Kātantra is believed to be a secondary addition by none other than Vararuci-Kātyāyana. Perhaps because of what many perceived to be his critical attitude toward Pāṇini in his vārttikas, Vararuci-Kātyāyana was the go-to sage for authorizing additions and interventions in these new non-Pāṇinian systems.27
The Light that Vararuci, as we may continue to call him, shone on Prakrit came from the Sanskrit grammatical tradition. His use of Sanskrit as a metalanguage, of concise transformational rules, and of technical terms and abbreviations sets Light far apart from the general descriptions of Prakrit contained in the floating Prakrit verses discussed above. It became the most popular and most widely circulated grammar of Prakrit, used directly or indirectly as a source by every single subsequent grammar.28
What did Light shine on exactly? It has repeatedly and rightly been emphasized that Light is not a grammar of Prakrit in the broad sense of “Middle Indic.” The language it defines, as scholars were quick to notice, is substantially similar to the language of the Prakrit literary tradition, represented above all by Seven Centuries. Nitti-Dolci in particular insisted that Light is not general or extensive enough to serve as a grammar of a language, but must instead be seen as a grammar of a text. She speculated that Vararuci sought to describe the language of an anthology that was similar to, but not identical with, Seven Centuries as it has been transmitted to us. Its purpose, she claimed, was to assist people who already knew Sanskrit to compose verses in Prakrit like those found in that anthology.29
Light is a grammar of a literary language, but the crucial question, which Nitti-Dolci glosses over with her assumption of a text “similar to but different from” Seven Centuries, is: exactly what literature was composed in the language that Light describes? Against the common equation of “literary Prakrit” with “grammatical Prakrit,” there stands the fact that many forms either directly mentioned in or presupposed by Light are not attested in the extant classics of Prakrit literature such as Seven Centuries. This in itself is not surprising, because much of this literature has been lost. More striking is the fact that some forms taught by Vararuci have turned up only in quite early Jain texts. The best example is the past tense in -īa, which appears in Light but which was not noted in any literary texts prior to 1936, when Ludwig Alsdorf found it in Wanderings of Vasudeva.30 Another example is the locative singular form of the first-person pronoun mae, which is likewise mentioned in Light, but which Anna Aurelia Esposito has only recently spotted “in the wild”—again, in Wanderings of Vasudeva.31
It seems very plausible to me that Light on Prakrit was composed with such texts in mind—not just Wanderings of Vasudeva, but romances in verse like Taraṅgavatī. It has often been remarked (starting with Hermann Jacobi) that Jain texts in Prakrit deviate from the rules established by grammars like Vararuci’s, and this deviation licenses us to speak of “Jain Prakrit” (or “Jain Mahārāṣṭrī”) as distinct from the language Vararuci sought to describe.32 This label, which Jacobi originally based on Sanskritizing features of relatively late Jain commentaries and narrative literature, has since been applied to any form of Prakrit written by Jains. But as I noted in chapter 3, we need to be careful of overstating the continuities within the use of Prakrit by Jains and understating its continuities with its use by non-Jains. Forms taught by Vararuci that occur in Jain literature and nowhere else have greater weight in regard to the question of the grammar’s target language than forms occurring in Jain literature and nowhere else that are not taught by Vararuci. It may even be possible that Light on Prakrit was composed by a Jain author in a Jain literary milieu, and like Trivikrama’s transformation into Vālmīki, non-Jain authors found it necessary to reattribute the text to Vararuci-Kātyāyana.
Little can be said with certainty about Light’s textual history. Nitti-Dolci died soon after publishing her study, and her call for a “critical edition of Vararuci based on all the commentators and all the grammarians who have drawn materials from his work” has gone unheeded.33 I doubt very much that Bhāmaha, the author of the popular Manoramā commentary on Light, is identical to the scholar who wrote Ornament of Literature. Vīrasena and Jinasena in the ninth century do not seem to have been aware of Light. Abhinavagupta, in the eleventh century, does refer to Light in a little-known passage where he glosses “half-Sanskrit” by mentioning the opinion of others that it refers to “Prakrit itself, defined in accordance with the rules pronounced by Vararuci and so on, and distinct from the regional languages such as Śaurasenī.”34 This is, to my knowledge, the earliest datable reference to the text, along with quotations of Light in the commentaries of Bhuvanapāla on Seven Centuries and Harṣapāla on Rāvaṇa’s Demise (both eleventh century). Despite his reference to Vararuci, Abhinavagupta himself seems to have been more familiar with a lost work called Illustration of Prakrit (Prākṛtadīpikā) and Utpaladeva’s commentary thereon, which he recommends to his readers. One might have expected Abhinavagupta to have known the Manoramā commentary on Light if it was really composed by the well-known scholar of poetics.35
One event in Light‘s textual history, however, is worth remarking upon, since it signals a fundamental shift in the orientation of Prakrit knowledge. As Nitti-Dolci demonstrated, the “Prakrit” that Vararuci’s Light originally illuminated was singular. At some point, however, chapters were added to describe Paiśācī, Māgadhī, and Śaurasenī. These additional chapters represent a pluralization of the category of “Prakrit.” Previously, knowledge of Prakrit meant knowledge of the grammar, lexicon, and metrical forms of Prakrit literature. This was “literature heard” (śravyakāvya), poetry such as Seven Centuries and Rāvaṇa’s Demise. The languages used on the stage, of “literature seen” (dṛśyakāvya), were similar enough to this unitary kind of Prakrit to have been considered variants or ectypes of it, and hence they never formed the primary object of systematic knowledge in contradistinction to the Prakrit of “literature heard.” At first, we might interpret Daṇḍin’s declaration that the languages of the stage should be considered Prakrits (discussed in chapter 5) as an affirmation a centuries-old approach that awarded conceptual and analytic primacy to Prakrit as the language of “literature heard,” and in which the languages of the stage were somewhat of an afterthought. But we can also see it as his idiosyncratic solution to the problem of whether literary Prakrit, used in “literature heard,” could be identified in some sense with the languages of “literature seen,” and thus whether Prakrit was a species or a genus. The difference is that genera do not have specific characteristics, and in this case, they do not have grammars. The redactors of Light on Prakrit clearly considered it a genus. What had earlier been “Prakrit” was reconfigured, in accordance with the logic of regional specificity that governed the languages of the stage, as the species “Mahārāṣṭrī”: crucially, the word appears in the expanded version of Vararuci’s Light, but not the older version. Pluralization meant that Prakrit, now Mahārāṣṭrī, no longer stood above the other languages, but alongside them.
The languages added to Light confirm that the pluralization of Prakrit implied thereby is the exact same pluralization evident in Rudraṭa’s expansion of the archetypal schema from three to six languages, which, as noted in chapter 5, attends a shift in analytical focus from monoglossic to polyglossic forms. From this point on, knowledge of Prakrit had a very different shape. It was, first of all, knowledge of “the Prakrits”; second, it was primarily but not exclusively oriented toward the theater; third, it formed part of an increasingly large and interconnected body of literary-cultural knowledge, at the apex of which was poetics (alaṅkāraśāstra).
It was in this context that Hemacandra compiled his grammar of the “six languages” around the middle of the twelfth century. To understand Hemacandra’s position in the history of Prakrit grammar, it is useful to pair him with another twelfth-century scholar, Puruṣottamadeva. Hemacandra was a Śvetāmbara Jain monk who spent most of his career at the Cāḷukya court of Aṇahilavāda, in the north of today’s Gujarat, patronized first by Jayasiṃha and then by Kumārapāla. His works span, and in many ways define the boundaries of, the totality of literary-cultural knowledge; he is known as kalikālasarvajña, “an omniscient of the Kali age.” And he was, according to George Grierson, the founding figure of the “Western School” of Prakrit grammar. Puruṣottamadeva represents the “Eastern School,” which Grierson traces back to Vararuci. He was a Buddhist from eastern India. Besides his Grammar of Prakrit, he wrote a large number of Sanskrit lexicons and a commentary on Pāṇini’s Aṣṭādhyāyī.36
For both Hemacandra and Puruṣottamadeva, the care of Prakrit was part of the care of language, and this care in turn had much stronger links to a cosmopolitan literary and intellectual culture than it did to the particular religious traditions with which Hemacandra and Puruṣottama were affiliated. Hemacandra offers only a few comments about the specific features of the language of Jain scriptures—ārṣa Prakrit, as he calls it—in comparison to the language of poetry, which he quotes in abundance.37
Scholars have justly criticized Grierson’s idea that there existed two separate “schools” of Prakrit grammar, one prevalent in the east and one in the west.38 The curious persistence of Grierson’s historiography warrants a longer critique, but three main problems can be summarized here. The first is the very idea of a “school.” If it means a fixed set of core doctrines that are elaborated and defended by its members, and if belonging to a school means self-consciously identifying with it to the exclusion of other schools, then there have never been “schools” of Prakrit grammar. Grierson’s “schools” are made up of authors who tend to rely on common sources, and thus a more appropriate term—although still problematic for reasons discussed below—is “traditions.” The second is the idea that these schools were regional. For Grierson, the regionality of these schools was not simply a question of where their authors are located on a map, but a promise, which turned out to be false, that these schools would address the linguistic particularities of their respective regions. Besides this false equivalence between an author’s regionality and the regionality of the language he describes, Grierson also constructed a false equivalence between the regionality of a tradition and the regionality of its sources. There are authors whose works are transmitted only in eastern India, among them Puruṣottama, Rāmaśarman, and Mārkaṇḍeya. But this does not imply that their principal source, Vararuci, came from eastern India as well, since his work was known everywhere from Kashmir to Kerala. The final problem is use of the figure of “two schools” to structure the history of Prakrit grammar. This figure creates the false impression that two schools developed in parallel and in isolation from each other. But all of the “western” grammarians discussed by Grierson relied directly or indirectly upon the “eastern” Light on Prakrit, and “eastern” writers like Mārkaṇḍeya relied heavily on the “western” Hemacandra. The differences between the “western” Hemacandra and the “eastern” Puruṣottama, for example, largely reflect differences in how this source material has been refashioned; they do not do not amount to a radically different theories of Prakrit or radically different descriptions of the language.
In defense of Grierson’s theory, however, it must be admitted that Puruṣottama, Rāmaśarman, and Mārkaṇḍeya constitute a somewhat separate and localized tradition. They were much more concerned with the languages used on the stage, and although they incorporate Vararuci’s grammar in its entirety, they appear to have utilized a larger body of early material on this subject than Hemacandra or his followers had access to. All of them operate with a top-level classification of bhāṣās, vibhāṣās, apabhraṃśas, and paiśācikas that appears to be an elaboration (by Kohala?) of the schema we find in Bharata’s Treatise on Theater. But they also refer to authors, foremost among whom is Śākalya or Śākalya-Māṇḍavya, whose account was closely related to the one given in Treatise on Theater.39
The history I have reconstructed for the systematic knowledge of Prakrit prior to Hemacandra can be articulated into three phases. In the final phase, Prakrit and Sanskrit are both objects of the same systematic knowledge. Prakrit needs to be accessed through Sanskrit: in the case of Hemacandra’s grammar, this literally meant getting through seven books of Sanskrit grammar for the treatment of Prakrit in the eighth. In this phase Prakrit is a container and template for a multiplicity of languages that occur in the domain of theater or “literature seen,” where these languages co-occur with Sanskrit.
In the preceding phase, Prakrit and Sanskrit exist in their respective traditions of “literature heard,” and they are each objects of separate discourses of knowledge. These discourses themselves, however, are articulated in Sanskrit through the conventions of the Sanskrit grammatical tradition. This is the phase in which Sanskrit forms of knowledge are deployed in order to fully account for Prakrit difference, and it is best represented by the original version of Light on Prakrit.
In the earliest recoverable phase, knowledge of Prakrit is articulated in Prakrit and without much reference to Sanskrit forms of knowledge. As an example, sometimes the same metrical forms that are used in Sanskrit and treated in Sanskrit metrical treatises are defined somewhat differently in Prakrit metrical treatises. It was in this phase that Prakrit difference was first enunciated under the category of “the regional” (deśī), and knowledge of Prakrit was thus articulated under this name (deśīśāstra). A fitting representative of this phase is Harivṛddha, but it encompasses almost the entire discourse of metrics (Virahāṅka, Svayambhū) and lexicography (Dhanapāla) prior to Hemacandra.
These phases do not, of course, divide the history of Prakrit knowledge into discrete and non-overlapping segments. Instead they represent different ways of constituting Prakrit as an object of knowledge. The logic of one phase can, and often does, continue into subsequent phases: this is exemplified by the chapters added to Light on Prakrit, or by the stray rules in Caṇḍa’s Definition of Prakrit that brusquely characterize other varieties of Prakrit. These “phases” might even be differentiated more by audience than by time: as Nitti-Dolci emphasized, works like Light were intended for an audience whose knowledge of Prakrit was mediated by Sanskrit, whereas the works that I assign to the first phase were largely intended for people who read and engaged with Prakrit literature without the mediation of Sanskrit. By describing them as “phases,” I mean to evoke a model of additive development, in which knowledge is received, revised, and reenunciated, rather than the Griersonian model of spontaneous generation, in which the entirety of a tradition’s content and principles are present at the moment of its foundation.40 An important feature of my additive model is that the concepts of the earlier phase are foundational concepts upon which the whole subsequent history of the discourse depends.
GRAMMAR, METAGRAMMAR AND THE REGIONAL
One of these foundational concepts is the division of Prakrit into three categories. The earliest discussions of such a division occur in Bharata’s Treatise on Theater and in Daṇḍin’s Mirror of Literature, and luckily Daṇḍin’s tenth-century commentator Ratnaśrījñāna quotes several passages from Harivṛddha on the subject.41 All of these discussions imply what Ratnaśrījñāna makes explicit: under this analysis, Sanskrit is singular, and Prakrit is plural. Its plurality, however, does not consist in the plurality of Prakrit languages such as Śaurasenī and Māgadhī, but in the plurality of its “modes” (prakāra), the aspects in which Prakrit appears in relation to Sanskrit. This point bears emphasis, because it might at first appear that Prakrit’s plurality makes it an open-ended category for an endless variety of language practices, whereas in my view it has the exact opposite effect: it is a precondition for its precise grammatical description.42 “Sanskrit-identical” Prakrit (Daṇḍin’s tatsama) appears identical to Sanskrit. “Sanskrit-derived” Prakrit (Daṇḍin’s tadbhava) can be understood as a systematic modification of Sanskrit. Finally, “Regional” Prakrit (Daṇḍin’s deśī), has no perceptible relation to Sanskrit at all.43
These three categories refer, in all of these discussions, to the Prakrit language. Ratnaśrījñāna reproduces Harivṛddha’s examples: hari- “Viṣṇu,” hara- “Śiva,” and kamalā- “Lakṣmī” are identical in both Sanskrit and Prakrit, allowing for some differences in their case-endings; mahinda- “Indra,” sindhava- “of Sindh,” and bahira- “deaf” can be thought of as “derived” from the corresponding Sanskrit forms (mahendra-, saindhava-, and badhira-); bokkaṇa- “crow,” kaṃkelli- “Aśoka tree,” ciriḍḍihilla- “curds,” and sitthā- “bow-string” have no apparent relation to the Sanskrit words that are current in those meanings. These categories, however, are not limited to the analysis of lexical units. In principle, they apply to “all aspects of the structure” of the language.44 I would press this point further: the paradigmatic status of language meant that the categories developed for language could apply to a wide range of other practices, and the threefold analytic could—and in limited ways did—function as a general analytic of culture.
A closer look at these categories shows how they are indebted to the analysis of language but not confined to it. One function that they perform is comparing two forms and converting the difference between them into one of three values. Crucially, however, the differences between individual forms are a function of the global differences between the domains from which these forms are drawn. They are structural. In Harivṛddha’s examples, the different phonological systems of Sanskrit and Prakrit are what generate the particular differences between selected lexical forms. This analysis is exhaustive and non-overlapping: every single Prakrit word can be brought under one, and only one, of these three categories. The analysis can therefore be thought of as a way of characterizing the relation between a given Sanskrit “input” and a desired Prakrit “output,” provided that exactly the same rules—in this case the rules of Prakrit phonology—apply equally to all inputs. “Sanskrit-identical” are forms to which the rules apply vacuously. “Sanskrit-derived” are forms in which the input and output differ, but in which those differences can be brought under a regular description. “Regional” are forms in which the input–output relation is opaque.
The three categories thus serve as what I call a metagrammar: a figure that simultaneously delineates the domains in which the rules can apply non-vacuously and characterizes the rules themselves as derivational.45 A metagrammar presents something to us as an object of grammatical knowledge and tells us, in very broad terms, what that knowledge consists of and how it is to be applied. In the case of Prakrit, this tripartite figure programmatically lays out the shape that knowledge of Prakrit in fact took. Whatever was “Sanskrit-identical” was to be passed over, since it was already targeted by other knowledge systems. The goals of grammar and lexicography were to relate Prakrit forms to Sanskrit forms in those cases where the relation was not already transparent.
The original metagrammatical usage of these categories is very different from the merely descriptive usage that George Grierson and his students introduced in the late nineteenth century. Grierson used tatsama to refer to any word, in any early modern or modern Indian language, that had more or less the same form as the Sanskrit word, and tadbhava to refer to those words that had undergone some kind of phonological transformation. Because of the continuous reintroduction and retransformation of Sanskrit words, however, new categories such as semi-tatsama and semi-tadbhava had to be invented. The same language—indeed the same speaker—could use a tatsama form such as bhakt, a tadbhava form such as bhāt, and a semi-tadbhava form such as bhagat, each with a specialized semantic value.46 In Harivṛddha’s system, however, the rules apply without exception, and the only possible “output” in Prakrit of the Sanskrit word bhakta- would be the “Sanskrit-derived” form bhatta-.
The role of history is another important difference between the premodern and modern use of these terms. For Grierson, a tadbhava word was one that had undergone change with respect to its Sanskrit original, and this kind of change took place in history. The process that transformed bhakta- into bhatta- and then bhāt is the inexorable progression of the Indic languages from “Old” to “Middle” to “New.” For the Prakrit grammarians, however, the three categories of course constituted a single synchronic system. The “derivation” of Prakrit forms from Sanskrit forms, too, was primarily thought of as an analytic procedure, with absolutely no reference to the historicity of either Sanskrit or Prakrit: these were emphatically not historical forms of knowledge.47 The decision to make Sanskrit the fixed point of reference for the analysis of Prakrit had nothing to do with the priority, either in historical or axiological terms, of the former to the latter. It seems to have been motivated, instead, by the very grammatical principle of lāghava, or economy: if 50, or 90, or 95 percent of the derivation of a word can be accomplished by referring to knowledge systems that already exist, why duplicate the effort?
This is not to say that premodern Indians were incapable of thinking about their language practices in historical terms, as some have argued.48 In a famous passage, Namisādhu declares that Prakrit is prāk-kṛta, “fashioned first,” and that the prakṛti or “original” from which it derives is not Sanskrit but “the innate faculty of speech of all living beings without being refined by grammar and so on.”49 Hemacandra, too, refers to Prakrit as “without a beginning.”50 Yet both authors happily define Prakrit and its subvarieties in reference to Sanskrit.51 Hemacandra makes it clear that his analysis of Prakrit starts from Sanskrit at the beginning of the Prakrit section of his grammar:52
The original [prakṛti] is Sanskrit, and Prakrit is so called because it either “originates in” or “comes from” Sanskrit.53 Prakrit is introduced as a topic immediately after Sanskrit. And providing rules for Prakrit immediately after Sanskrit has the purpose of indicating that the rules given here pertain only to Prakrit that has its origin [yoni] in Sanskrit words, which are either fully formed or not, and not to Regional Prakrit. Sanskrit-identical Prakrit, however, is already known from the rules on Sanskrit. Further, the stems, affixes, genders, case assignments, ways of forming compounds, technical terms, and so on are the same for Prakrit as they are for Sanskrit.
Hemacandra saw no contradiction between his belief in the eternality of Prakrit and his use of metagrammatical categories that made Sanskrit the standard of comparison. These categories allowed him to systematically divide up the realm of Prakrit knowledge more than any previous author had. He treats of “Sanskrit-derived” words in his grammar and generally defines “Regional” words in a separate lexicon, the Garland of Regional Nouns.
Such an approach requires comparison between two linguistic domains, but one of them, the “original,” is named in the very categories, while the other, Prakrit, is merely implied. But the metagrammatical categories did serve to characterize Prakrit as a language, insofar as it was distinguished from Sanskrit both by its transformational rules and by the mysterious category of the “regional.” Prakrit knowledge, too, was distinctively constituted by its concern with regional practices. An important rule of Vararuci’s Light on Prakrit introduces certain words as whole-cloth substitutes for Sanskrit words. When commenting on this rule, Vasantarāja notes an alternative classification of Prakrit words into “imitations” (anukārin) and “transformations” (vikārin) of the corresponding Sanskrit words, which roughly map onto the categories of “Sanskrit-identical” and “Sanskrit-derived.” Vasantarāja rejects this classification precisely because it fails to account for those words which are “known with utter certainty to be Prakrit” but are neither identical with nor derived from Sanskrit words.54
The regional came to characterize Prakrit and its forms of knowledge in two different ways, to the mild confusion and frustration of modern scholars.55
On the one hand, “the regional” is a purely negative concept: it is what is left over when the Sanskrit-identical and Sanskrit-derived portions of the lexicon are sifted out. This is the concept that underlies Hemacandra’s Garland of Regional Nouns (Deśīnāmamālā), which organizes and defines the words that are left over (avaśiṣyante) because they cannot be properly formed by the rules enunciated in his grammar.56 This does not mean that all of the words collected in Hemacandra’s lexicon cannot, in principle or in practice, be derived from Sanskrit words. The lexicography of the regional was emphatically not etymology, in the modern sense of tracing words to their historical roots. There are many words in Hemacandra’s lexicon that can easily be traced to an Old Indic root.57 What matters to Hemacandra is whether the corresponding word actually exists in Sanskrit as he knew it, and further, whether it is current in the same sense in which the Prakrit word is used. Further, many words have been excluded from Hemacandra’s lexicon simply because he chose to include them in his grammar instead.58 The significance of the regional as a negative concept for Hemacandra was precisely that the words included under this category were excluded from the positive space occupied by Sanskrit and Sanskrit-derived Prakrit.
On the other hand, “the regional” is a positive concept. It refers to the practices of a region, regardless of or prior to the analysis of those practices in relation to others. “The regional is defined,” according to a verse attributed to Bhoja by Mārkaṇḍeya, “by what occurs in each particular region of kings and peoples.”59 This positive sense is more expansive, in that it should include forms that are identical to or derived from Sanskrit forms, since after all these forms too have their place in the practices of a region. Prakrit knowledge was knowledge of the regional, and it seems to have been the first branch of knowledge that defined itself by and concerned itself with regional practices.60 Hemacandra refers to earlier works on Prakrit as deśīśāstras, and his predecessor Dhanapāla referred to his own Prakrit lexicon as a deśī; similarly Pṛthvīdhara refers to a work called Light on the Regional (Deśīprakāśa) when commenting on the Prakrit of Little Clay Cart.61
With what particular region was “the regional,” as the distinctive element of Prakrit and its forms of knowledge, associated? All early authorities agree that it was Mahārāṣṭra that gave content to the regional as a category: “the regional is defined,” Harivṛddha said, “by those words whose meanings are conventionally known in the region of Mahārāṣṭra.”62 On this vision, which very likely represents the way that the pioneers of Prakrit literature thought about their own practices, the regionality of Prakrit refers to its connection with Mahārāṣṭra in particular, and not to a general connection with one of any number of regions. This vision did not recognize parallel “dialects” of Prakrit, each associated with its own region. Or rather—as we will see below—it recognized such dialects but did not place them on the same level with Prakrit properly speaking. As we see from Harivṛddha’s definition, the regional is defined by the conventional acceptance of words, or potentially any kind of practice, within that region.63 Regional knowledge, in other words, has a distinct modality: it works by convention (prasiddhi), whereas Sanskrit knowledge works by derivation (siddhi). That is, rather than locating forms within a derivational matrix that lies outside of space and time, it locates them within a temporally and geographically bounded field of practice.
Prakrit is often called Māhārāṣṭrī in modern scholarship, and it is widely and mostly correctly thought of as a linguistic precursor to Marathi.64 The territorial limits of Mahārāṣṭra as a “region” in premodern India were no doubt different, and of a different nature, than the limits of the modern state of Mahārāṣṭra. But even if we accept that Prakrit and Marathi are associated with the same region, the nature of that association is different. It does not seem possible to think of Prakrit and Marathi as situated on a single historical continuum. One of the unique aspects of Prakrit, which at the same time makes it difficult to fit into existing typologies of language, is that it was regional without being vernacular.
There are two senses of “vernacular” which it helps to distinguish here, and neither of them apply to Prakrit.65 The first is a language practice that has an exclusive connection with a regional imaginary, which in turn serves as the basis for a cultural, social, or political identity. This way of thinking about the regional is deeply ingrained in the discourse of language in modern India, but it is almost completely absent throughout the period in which Prakrit literature first took shape. And it is particularly absent from Mahārāṣṭra, which was a cover-term for a number of smaller regions such as Vidarbha, Ṛṣika, Aśmaka, and Kuntala that had long been more salient, culturally and politically, than the macroregion that they constituted. Although the Cāḷukya king Pulakeśin II, in the early seventh century, could be described as “king of the Mahārāṣṭras,” it was not until the Yādavas in the twelfth and thirteenth centuries that Mahārāṣṭra formed the basis of a vernacular polity in this sense.66 The Sātavāhanas, who presided over the political integration of this region, never used the term Mahārāṣṭra, although the title mahāraṭṭhi in Sātavāhana-era inscriptions refers to a high-ranking official who administered a relatively large region on behalf of the Sātavāhanas, and this arrangement of shared sovereignty was probably the precursor to the territorial notion of Mahārāṣṭra or “the Mahārāṣṭras” that we encounter later on. But as far as I can tell, Prakrit was never thought of as a marker of identity, regional or otherwise, and hence it does not have the element of political salience that is so important to modern vernacular languages.
This, of course, raises the question of why Prakrit was defined in relation to Mahārāṣṭra in the first place, especially if this relation conferred no obvious benefits or consequences. I can only guess that, around the time when Prakrit was theorized, Mahārāṣṭra was one of those spaces—like the “Northern Cities” of the United States—which is defined in the present by shared linguistic phenomena that are presumably explained by shared social, cultural, or economic determinants in the past. The linguistic landscape of the Deccan must have been very diverse in the first few centuries CE, but the space between the Vindhyas and the Bhīma river might have formed a linguistic area with sufficiently self-similar patterns of speech, at least among people of a particular social background—let us say, suggestively, the mahāraṭṭhi elite that are so well represented in inscriptions.
The etymology of “vernacular” furnishes a second sense: the untutored language of the household slave, and thus a language practice that is natural, common, and prior to grammatical discipline. Clearly Prakrit, as the language of courtly literature and the object of an appreciable body of articulated knowledge, does not fit very well into this category. Many scholars, however, follow Namisādhu in arguing that Prakrit must once have been a “vernacular” in this sense, before courtly literature and its forms of knowledge arrested its natural development. In the introduction I stated my insistence on viewing Prakrit as a cultural practice rather than as a natural phenomenon, and here I can add a further argument for distinguishing Prakrit from the natural phenomenon of vernacular speech. The first person (so far as we know) to theorize Prakrit’s regionality, Harivṛddha, clearly maintained that this regionality did not make it into a “common” language, since that was a different category of language use altogether.
To the standard three categories of analysis—Sanskrit-identical, Sanskrit-derived, and Regional—Harivṛddha added a fourth, which he called “common” (sāmaṇṇa).67 A “common” language, on this schema, is the language of everyday conversation. This, at any rate, is what Bhuvanapāla means when he explains a word in Seven Centuries “by recourse to the Common,” since he appeals to the practices of everyday people.68 The idea seems to have been that the first three categories constituted “Prakrit” within a single system of literary practice, whereas the fourth category could be called “Prakrit” only within a different system. Consonant with Harivṛddha’s distinction is Daṇḍin’s statement that certain languages are considered Prakrit when they are used to represent conversation in plays.69 The implication is that conversational language is not considered Prakrit outside the confines of this genre. Within the tradition constituted by Seven Centuries and Rāvaṇa’s Demise, Prakrit is not a “common” language that represents conversation, but the primary language of the literary work. This interpretation is corroborated by the fact that several vernacular grammars that adapt the classification of Prakrit include alongside the traditional three categories a fourth category of grāmya, meaning vulgar or unsophisticated, which seems to reflect the earlier category of “common” (see below).
The regionality of Prakrit is thus quite different from the regionality of a vernacular, either in the sense of a vehicle of regional identity or in the sense of a common language of conversation. It can be seen as a kind of regionality that is self-undermining for the following reason. The regionality of Prakrit is a site of impermeability to a general approach by which language practices are understood in relation to a given model: what you cannot understand by comparison with a model based on Sanskrit is, by definition, regional. This very impermeability, however, is the raison d’être of the systematic knowledge of Prakrit. Making regional forms an object of systematic knowledge, however, renders them intelligible outside of the region in which they are “conventionally recognized” (saṃketita, prasiddha). If Prakrit was in any sense based on the regional language of Mahārāṣṭra in the first few centuries CE, the literature and its forms of knowledge quickly became almost as transregional as Sanskrit itself. Light on Prakrit exemplifies this point, both in its distribution (it was studied throughout the entire subcontinent) and in the purposes that it serves: namely, to allow people to read, understand, and compose Prakrit literature, whether or not they were familiar with the regional language practices of Mahārāṣṭra.
This sketch of the tripartite and quadripartite divisions of Prakrit helps to explain the shape that knowledge of Prakrit actually took. The objects of systematic knowledge of the regional (deśīśāstras) were the Sanskrit-derived and Regional aspects of Prakrit. Less obvious, but no less important, is the fundamentally supplemental, practical, and instrumental character of this knowledge. When Trivikrama began his influential grammar in the thirteenth century with the principle that “the formation of Prakrit should also be known from actual practice,” he was simply making explicit a principle that had guided the enterprise of Prakrit grammar from its beginnings. “Actual practice,” as Appayya Dīkṣita III would later make clear in his commentary on Trivikrama’s grammar, did not mean the language of casual conversation, but “the usage of literary authorities.”70
The “founding of grammatical norms on literary practices” in Prakrit knowledge, as Sheldon Pollock has noted in connection with vernacular knowledge, is the very opposite of the priority of theory to practice in Sanskrit literary culture.71 This empirical approach, as well as the categories that Prakrit grammar provided, would have profound effects on the self-theorization of vernacular literary culture. But in order to understand these effects, we need to understand what motivated the theorists of Prakrit to give priority to literary practice, and what the theoretical implications of this commitment were for the knowledge which they were giving shape to.
Early attempts to articulate knowledge of Prakrit were wildly unsystematic, including such rules as “vowels are sometimes substituted for other vowels.” Even Vararuci’s Light on Prakrit, despite its thematic organization, is more or less a list of Prakrit equivalents for Sanskrit forms. Nitti-Dolci hesitated even to call it a “grammar,” since, in contrast to Sanskrit grammars such as the Aṣṭādhyāyī or even the Kātantra, it did not build up a coherent system from general principles: it outsourced the general principles to Sanskrit grammar (“the rest comes from Sanskrit” is the last rule of Light on Prakrit) and confined itself to a sketch of Prakrit’s deviations.72
The rules that Prakrit grammar did provide were, of course, thought to be correct and authoritative—otherwise there would be no point in enunciating them—as shown by Mārkaṇḍeya’s corrections to the text of Rājaśekhara’s Karpūramañjarī, and Ghanaśyāma’s tireless criticism of alleged mistakes in Kālidāsa’s Prakrit, both on the basis of Prakrit grammar.73 But the rules were not exhaustive. The conjuring word of Prakrit grammar is bahulam, “variously,” which allows forms not otherwise derived by the grammar to be admitted as correct. Hemacandra begins his discussion of Prakrit with this word. In Vararuci’s Light on Prakrit, it appears in a list of substitutes. Although in principle many of these words could be derived from a corresponding Sanskrit word (e.g., dāḍhā from daṃṣṭra), in practice it would have been tedious—even by the standards of Prakrit grammar—to do so. The eighteenth-century commentator Rāma Pāṇivāda remarkably proposes to split the rule into two, a trick of the Sanskrit grammatical tradition called yogavibhāga, and produces a rule that simply reads bahulam. He is quite upfront about the implications of this strategy:
How then is the following usage possible: “then the Pauravas listened to Nārāyaṇa, who was standing nearby”?—Our answer: because the rule has exceptions.—You keep shouting “exceptions! exceptions!” for every rule. I don’t know what your authority is for that.—That’s true. But later we will see the rule dāḍhādayo bahulam, and there I will split up the rule, with the result that that the rule “with exceptions” [bahulam] is construed with every single operation. Taking usage as our guide, we can understand the words “with exceptions,” and the grammar can derive anything that we want it to.74
The status of Prakrit grammar can be summarized as follows. It sketched out the basic forms which one was likely to encounter in Prakrit literature, even if “Prakrit literature” was somewhat of a moving target, and was “empirical” to the extent that it followed literary practice (prayogānusāreṇa). It could be used in a regulative capacity, to show that certain forms were incorrect, or to correct a transmitted text. It was not, however, held to characterize all of the forms that could possibly be encountered in literature exhaustively. Thus its regulative authority was founded on that of the literature on which it was putatively based. The resulting form of knowledge suffered, in comparison to Sanskrit grammar, from a “lack of rigor,” as scholars were eager to note. But the comparison is misplaced, since Sanskrit and Prakrit grammar were different enterprises—vyākaraṇa, or “language analysis,” almost never being used to describe Prakrit grammar—that were motivated in very different ways and sought to define very different fields of language use.75
PRAKRIT IN THE VERNACULAR
As I argue in the following chapter, Prakrit receded into the background over the course of the second millennium, and its obsolescence is directly related to the emergence of vernacular textuality. We can say that the regional vernaculars occupied much of the same space in the language order that Prakrit had previously occupied. There are perhaps functional reasons for this replacement: if Prakrit had executed some of the functions of a vernacular within the classical language order—as a counterpractice to Sanskrit, for example—then true vernaculars, once literized and literarized, could perform those functions just as well or better. But such an approach to the problem would need a much more detailed account of the functions that the languages performed, and even then I doubt it would be entirely convincing. What I will focus on here, instead, are the genealogical reasons, that is, the influence that Prakrit forms of knowledge had on the self-theorization of vernacular literary culture. This influence was profound, and it has gone almost entirely unrecognized.
To put the argument in a stronger way: the concepts provided by Prakrit forms of knowledge, and the particular relationship to literary practice embodied in it, were some of the conceptual conditions for the emergence of vernacular literature in South Asia. It is not that vernacular literature would never have existed without Prakrit—indeed an argument could be made that Prakrit delayed the emergence of vernacular literature by several centuries—but that Prakrit provided the conceptual foundations for these new literary practices, including the concept of “the regional” itself.
There are three general types of relationship that emergent vernacular literatures had to Prakrit. These relationships seem to depend both on the region and the linguistic distance, in Heinz Kloss’s sense of Abstand, between Prakrit and the vernacular in question. The first relationship obtained in North India, where vernacular languages were more or less closely related to Prakrit and Apabhramsha. Here, the vernaculars were largely thought of as a further iteration of Apabhramsha, which was itself conceived of as a kind of iteration of Prakrit. The early history of literary vernaculars in North India is a very complex topic, in part because these vernaculars do not identify themselves in the way that makes them easily recognizable as “early” forms of modern vernacular languages. As is well known, this literature generally identifies its language either as a form of Apabhramsha (avahaṭṭha), or simply as vernacular speech (bhāṣā), or, particularly but not exclusively among Muslim authors, as “Indian” speech (hiṃdavī).76 Making these literary languages into protoforms of languages that came to be known, named, taught, classified and described under the epistemic regimes of European colonialism has quite a few liabilities.77 I will only mention one: this project puts a lot of emphasis on the “forward” connections, and very little on the “backward” connections. Thus Apabhramsha works are sometimes taken to represent “Old Hindi,” whereas the vernacular poems of Vidyāpati are often claimed for “Old Bengali” or “Old Maithili,” and the rāsos of Rajasthan and Gujarat are variously identified as “Old Rajasthani” or “Old Gujarati.”78
Useful as these identifications may be for some purposes, they obscure the “backward” connections that these literatures make, often explicitly and deliberately, to foregoing traditions of literature in Prakrit and Apabhramsha. They also obscure the connections across these literatures, not only through their Prakrit and Apabhramsha models, but in terms of the circulation of textual material across linguistic boundaries. Within the region of North India, where Apabhramsha and early vernacular literatures shade into each other, Prakrit was available as a model of literary language distinct from Sanskrit, but this model was never invoked to produce grammars of the literary vernaculars. The only precolonial grammar of a North Indian literary vernacular is Mīrzā Khān’s grammar of Braj Bhāṣā, written in Persian in 1676, with which this book began.
By contrast, the South Indian literary vernaculars—Tamil, Kannada, Telugu, and Malayalam—were described in grammars from a very early period. This difference may be due in part to the influence of the Tamil grammatical tradition, represented above all by the Tolkāppiyam. But in the case of the earliest grammars of Kannada and Telugu, the model was not Tamil grammar but Prakrit grammar. The categories of Prakrit grammar provided a way of organizing knowledge about languages like Kannada and Telugu that had come to incorporate a large number of Sanskrit lexemes but still included elements that were not derived from Sanskrit. We will see how vernacular grammars redeployed these categories. In the South, the vernaculars did not represent themselves as continuous with Prakrit, as in the North, but in place of Prakrit: the “regional” (deśī) was no longer a category of Prakrit knowledge, but of vernacular knowledge.
The third region was Southeast Asia, where, much as in South India, the regional vernaculars were completely unrelated to Sanskrit and Prakrit in terms of their structure, but had incorporated a large amount of their vocabularies. Here I will confine my observations to Java, since this is the only part of the region where we have some idea of the kind of cultural work that Prakrit, or rather the idea of Prakrit, performed. As in North India, no precolonial grammars of Javanese, or any other regional vernacular in Southeast Asia, were ever produced. But we know that Prakrit provided a general model of a literary language that was not Sanskrit. And it is relatively clear that Javanese poets thought of their literary language as a kind of Prakrit. They describe the translation of a text from Sanskrit into Old Javanese as both Javanization and Prakritization. Both occur in the preface to the Virāṭaparvan, which was performed in 996 CE at the court of Dharmavaṃśa Təguḥ: the king “partook of the auspicious beginnings of Javanizing the work of Vyāsa,” which was also the “auspicious beginnings of composing the Prakrit version of the present story of the Virāṭaparvan.”79 The use of the word “Prakrit” to refer to Old Javanese is relatively widespread. One text, in outlining the norms of poetic composition, states axiomatically that “language is Sanskrit and Prakrit,” where the latter clearly refers to Old Javanese.80
One other region that was undoubtedly transformed by the culture of reading and writing in Sanskrit was the land to the north of India, including modern Tibet and China’s Xinjiang province. I will skip over a discussion of how, if at all, Prakrit might have affected the course of vernacularization in this area, but of course vernacularization did proceed very differently here than in the other three regions noted above.
In the remainder of this chapter we can examine more closely the ways in which Prakrit forms of knowledge provided a model for understanding the emergent literary vernaculars. These forms of knowledge first of all addressed the foundational question of how regularity, systematicity, and grammaticality can exist outside of the paradigm of Sanskrit. We saw in chapter 5 that Abhinavagupta’s pointed question “What regularity can a degraded practice have?” was answered in the context of the Treatise on Theater by a short overview of Prakrit grammar. And there we also saw that Kumārila Bhaṭṭa was able to criticize the Buddhist scriptures as “not even Prakrit” because Prakrit provided the model for a practice that was regular in its own way despite its deviation from Sanskrit. Secondly, Prakrit forms of knowledge supplied an analytic for the systematic comparison of Sanskrit and its others. Vernacular languages had no choice but to retrace these two major theoretical steps, and retrace them—rather than blaze a new theoretical trail—is precisely what they did.81
Vernacular knowledge takes its major categories of analysis from Prakrit knowledge: Sanskrit-identical, Sanskrit-derived, Regional, and in some cases, Common. As I have argued above, these categories are not simply descriptive. Just as in the case of Prakrit, they simultaneously define the domains and the character of vernacular knowledge. In Prakrit grammar, in an important sense, these domains were “given”: a word’s belonging to one or another of them was a brute fact, not a parameter that could be manipulated. In vernacular grammars, however, the differentiation of these domains had consequences for literary practice, in that an author could choose a word from one category rather than another in order to achieve certain goals.
One of the best examples for the reuse of these categories comes from Jewel-Mirror of Language of Keśava, composed in 1260 CE. The only two languages under discussion are Sanskrit and Kannada. Kannada can be mixed with Sanskrit, or it can be “pure Kannada” (accagannaḍaṃ). The latter can be analyzed, however, into Sanskrit-identical (tatsamaṃ), Sanskrit-derived (tadbhavaṃ), and Regional (dēśīyaṃ) components, an analysis that clearly demonstrates the “absent presence” of Prakrit grammar. Just as in Prakrit grammar, Sanskrit-identical words are a small subset of Sanskrit words to which the rules of “pure Kannada” apply vacuously, and Sanskrit-derived are those that can be related to corresponding Sanskrit words by means of transformational rules. Regional are those words that modern linguists would classify as having “Dravidian” rather than “Indic” roots; in any case they cannot be derived in a stepwise fashion from Sanskrit words. Keśava’s discussion of these three categories relates to the conditions under which Sanskrit and Kannada words can co-occur. Jewel-Mirror notes that Sanskrit and Kannada words generally cannot join to form compound words.82 These restrictions are not new in Keśava; similar guidelines can be found in earlier works of Kannada literary theory, including Way of the Poet-King and Analysis of Literature.83
Such restrictions were not based on a proto-nationalist ideology of linguistic purism, but on the recognition that the phonological systems of Sanskrit and Kannada are different. The underlying principle is that the same phonological constraints should apply throughout a word, including throughout each constituent of a compound word. Otherwise, the compound is “contradictory” (viruddhaṃ); it is, in other words, a constraint against word-level macaronism. But this constraint only applies to “unmodified Sanskrit” stems (samasaṃskṛtaṃ). If a stem is either Sanskrit-identical or Sanskrit-derived, it can be used freely with Regional words. In effect, a poet can use any Sanskrit word he wishes, so long as he follows Keśava’s guidance, in the seventh chapter of Jewel-Mirror, in transforming them into words of “pure Kannada.”84 This chapter provides rules that are similar to, and must have been modeled on, the rules of Prakrit grammar that take Sanskrit forms as input and yield Prakrit forms as output.85 Using such procedures, authors could mix Sanskrit and Kannada in a way that was validated by general linguistic and aesthetic principles. In order to constitute Kannada as a language categorically distinct from Sanskrit, but at the same time capable of absorbing its lexical resources, Keśava theorized it in exactly the same way that earlier scholars had theorized Prakrit.
Prakrit served Keśava and other vernacular intellectuals as a model of a counterpractice to Sanskrit: one that basically mirrored Sanskrit practices, but at the same time transmuted them into something different, and included within this difference sites of analytical impermeability or resistance that were gathered under the category of the regional. This final category, which constituted the exceptions to the rules in Prakrit grammar, became the principal target of the rules in vernacular grammars. Keśava’s discussion of Sanskrit-identical and Sanskrit-derived words in the seventh chapter of Jewel-Mirror makes it clear that he understands the rest of the vocabulary of “pure Kannada” to be regional.
Around the same time as Keśava, Ketana produced Ornament of the Āndhra Language, likely the earliest grammar of Telugu.86 Ketana invokes the same three categories, with the addition of a fourth, the Vulgar or Common (grāmya). His examples make it clear that Common words are not “obscene” words, as some scholars have maintained, but rather colloquial forms not preferred in poetry. The category is thus parallel to Harivṛddha’s “common” (sāmaṇṇa). It is quite possible that Ketana actually took this classification from Prakrit grammars now lost to us, since he refers to such works—albeit vaguely—in his introduction.87 Whereas Keśava’s “pure Kannada” (accagannaḍaṃ) is a cover term for Sanskrit-identical, Sanskrit-derived, and Regional words, Ketana numbers “pure Telugu” (accatenugu) as a fifth category alongside the inherited four—but only to include the other categories, “excluding Sanskrit-identical words,” under “pure Telugu” as a larger category.88 And although Ketana gives examples of “pure Telugu” words separately from the other categories, it is unclear exactly what makes these words different from “Regional” words.89
Ketana appears to have understood by “Sanskrit-identical” any Sanskrit words not accommodated into the phonological system of Telugu; he collapses the distinction that Keśava had observed between “Sanskrit-identical” (tatsama), referring to small class of Sanskrit words that already conform to the phonology of Kannada and therefore do not require further transformation, and “Sanskrit” plain and simple (samasaṃskṛta). Whereas Keśava’s “pure Kannada” includes “Sanskrit-identical” words, Ketana’s “pure Telugu” does not. The Wishing-Stone of the Āndhra Language, ascribed to the eleventh-century poet Nannaya, but only “rediscovered” by Appakavi in the mid-seventeenth century, also uses the fourfold distinction between Sanskrit-identical, Sanskrit-derived, Regional, and Vulgar words. On the basis of this text, Appakavi defines “pure Telugu” (accatelugu) as consisting of Sanskrit-derived and Regional words without any mixture of Sanskrit words. For him, the regional is defined by what the Āndhra people actually speak, and can thus be further divided into two categories: “pure Āndhra words” (śuddhāndhram), presumably those spoken in Āndhra itself, and “Āndhra words of foreign origin” (anyadeśajāndhram), presumably words of other regional vernaculars that had taken hold in Āndhra.90
The strategy of reappropriating existing categories to create new spaces for analysis would not work for vernacular metrics. Vernacular metrics defined itself against a single but bifurcated tradition: Nāgavarman’s tenth-century Ocean of Meters begins with the meters of “the two languages,” Sanskrit and Prakrit, which are used “in all regions,” before discussing the meters used “in the language of the region of Karnataka.”91 In fact the division is not as neat as Nāgavarman makes it out to be. The last section involves a completely different system of prosody, and consequently some of the meters that are particular to Kannada literature but nevertheless use the same system of prosody as Sanskrit and Prakrit meters—such as the ragaḷe—are treated in the earlier section. Nāgavarman’s combination of two prosodic theories in one treatise is iconic of the “cosmopolitan vernacular” he is concerned to theorize, which combines the literary resources of both traditions.92
But there were certain features of the discourse of Sanskrit and Prakrit metrics that were conducive to Nāgavarman’s intervention. It was modular from the beginning, in the sense that it accommodated two different systems of prosody, one that counted by syllables (vṛtta) and one that counted by moras (jāti). Although syllable-counting meters were widely associated with Sanskrit, and mora-counting meters with Prakrit, both types occur in both languages, and treatises on metrics in Sanskrit and Prakrit differ primarily with regard to the detail they go into for each class.93 Nāgavarman seems to have considered the Kannada meters, which consist of “blocks” (aṃśas) that count moras but in a different way than Prakrit jātis, as a subclass of jāti meters.
There is, moreover, a close relationship—perhaps but not self-evidently one of influence or descent from a common ancestor—between the jāti meters of Prakrit and the jāti meters of the Dravidian languages.94 These meters, in contrast to Sanskrit vṛttas, are typically composed of underlying rhythmic structures that can each be realized by any number of combinations of light and heavy syllables. The internal structure of these structures in Prakrit and Kannada is very similar, and the major difference between them is just that the former and not the latter have a fixed number of moras. In view of these similarities, the opposition between Kannada, on the one hand, and Sanskrit and Prakrit, on the other, has much more to do with the regionality or transregionality of their respective literatures, as Nāgavarman himself makes clear, than with the underlying principles of verse construction. But if we were to categorize meters according to their underlying principles, we would probably see a larger category of “regional” versification that includes Prakrit, the original and archetypal deśī tradition, alongside a range of vernaculars. This category would owe its existence, first of all, to the structural similarities between Middle Indic and Dravidian prosody, as well as to historical processes of “Prakritization” in the early phases of vernacular textuality. The kanda, the most popular meter of early Kannada literature, is an example of the latter, as it derives transparently from the Prakrit skandhaka. The ragaḷe, strongly reminiscent of Apabhramsha meters, may be an example of the first, unless it is actually derived from Apabhramsha models.
By way of summary, we may say that the metagrammatical categories so widely invoked in the enterprise of vernacular self-theorization were borrowed from Prakrit, and that this borrowing is one of the most important ways in which the Prakrit tradition, as a tertium quid, mediated between an established Sanskrit tradition and an emergent vernacular tradition. Since my primary goal in this chapter is a history of effects of Prakrit forms of knowledge, my focus has been on the conceptual relations between these traditions; much more could be said about the historical processes by which these concepts were transmitted.
What does it mean for vernacular knowledge to be mediated by Prakrit knowledge? It is not simply that the latter was a condition of historical possibility for the former, but that vernacular knowledge is essentially defined by a mediation between Sanskrit and vernacular forms. The primary site of this mediation is the domain called “pure Kannada,” or “pure Telugu.” The concept of purity is bound up in the modern world with concepts of genealogical descent that are not only absent from these domains but fundamentally incompatible with them: both “pure Kannada” and “pure Telugu,” according to their earliest definitions, admitted words originating in Sanskrit, namely, Sanskrit-identical and Sanskrit-derived. Their “purity” consisted, rather, in the fact that they were brought under a single linguistic description. Words of any origin could be integrated into a “pure” vernacular through the mediation of a transformational grammar. Prakrit, I have argued, provided the model for this mediation, but Prakrit was not itself a participant in it: it served as a catalyst, and then receded into the background.
Prakrit’s absent presence in vernacular forms of knowledge has become a simple absence in modern scholarship. One example is Lisa Mitchell’s sketch of premodern grammarians of Telugu against the background of what she calls “the Sanskrit vyākaraṇa tradition.” By this latter term, however, she really means “the Prakrit grammatical tradition,” since the categories she describes are the three categories discussed above that constitutively and contrastively define the field of Prakrit grammatical knowledge and never had anything to do with the analysis of Sanskrit or the discourse of vyākaraṇa in which that analysis was undertaken. Sheldon Pollock similarly classed Prakrit with Sanskrit as part of a “cosmopolitan” tradition, in dialectical opposition to which vernacular forms of knowledge developed. And it is very true that Sanskrit forms of knowledge were much more important to this process than Prakrit forms of knowledge. The concepts and terminology borrowed from Sanskrit grammar in Keśava, Ketana, and Appakavi are all much conspicuous than those borrowed from Prakrit grammar.95 But the specific connections between Prakrit and vernacular forms of knowledge have dropped out, and as a result, the latter are invested with a somewhat illusory newness. And while Prakrit was, in many relevant senses, “cosmopolitan,” it also provided a template—one that was followed again and again—for constructing systematic knowledge of regional practices (deśīśāstras).
The metagrammatical categories, and particularly that of the regional, were crucially important to the self-theorization of vernacular literature in Kannada and Telugu. But the effects of Prakrit knowledge on vernacularization were hardly limited to these categories. The notion of a mixed language was important to several vernacular traditions, above all Malayalam.96 To all appearances, the earliest actual practice of composing in a mixed language in South Asia, and certainly the earliest theoretical reflection on the practice, is the combination of Sanskrit and Prakrit in Jain commentarial culture of the mid-first millennium CE. Jinasena describes the mixture of Sanskrit and Prakrit in his Jayadhavalā commentary (completed in 837 CE) as maṇipravāla, a mixture of rubies and red coral.97 In explaining the word “half-Sanskrit” (ardhasaṃskṛta) in Treatise on Theater, Abhinavagupta suggests that it is a combination of Sanskrit with a regional language and refers to “maṇipravāla in the South” and “śāṭakuta in Kashmir,” and in the same breath mentions the possibility that it is simply Prakrit.98
The case of maṇipravāḷa is a straightforward instance, but not the only one, of Prakrit creating a space that vernacular languages would fill, thus seemingly creating the conditions for its own obsolescence. This has led, in the scholarly world as well as in popular narratives, to the erasure of Prakrit from the history of language in South Asia, which is commonly told through the oppositional categories of Sanskrit and regional language, cosmopolitan and vernacular. What I have tried to show in this chapter is that Prakrit forms of knowledge formed the background for vernacular forms of knowledge. Similarly, Prakrit grammar has long been seen as a half-baked and flawed enterprise, falling far short of the theoretical economy and sophistication of Sanskrit grammar. I have argued here that many of its perceived failures can be explained by the purposes it served, its relation to other discourses, and the way in which it was elaborated over the centuries. Further, these theoretical and methodological deviations from Sanskrit grammar are precisely where Prakrit grammar, along with Prakrit metrics and lexicography, had the longest and most important history of effects: its concern with practice, its orientation toward existing bodies of literature, and the concepts devised for shuttling between Sanskrit universality and Prakrit particularity.