How can we characterize Prakrit, as a language and as a literary tradition? The most straightforward answer might be to provide a systematic account of its differences from other languages, and specifically from Sanskrit. For Sanskrit and Prakrit are sister languages: we recognize one by contrast with the other. Prakrit was always represented and imagined through a “schema of co-figuration” with Sanskrit. From a very early period, a comparison between Sanskrit and Prakrit formed the basis of the systematic knowledge of the latter; the forms of the Prakrit language were almost always derived from corresponding Sanskrit forms. There is no doubt that Prakrit was, to a large degree, defined and characterized by contrast with Sanskrit—a contrast that formed the basis of the language order of premodern India. However, this picture is incomplete. It can lead us into thinking about Prakrit in purely structural terms, as if it were constituted entirely by its differences from Sanskrit.1 If Prakrit was a position in the language order from which it was possible to compose literary texts, it was a position of a particular kind. We might say that it had a phenomenology and ask what it was like to occupy this position, to operate in the world of Prakrit textuality. Similarly, we might say that it had an aesthetics and ask what it was about Prakrit itself that contributed to the beauty, or strikingness, of Prakrit texts. Of course, the phenomenology and aesthetics of Prakrit emerge even more clearly when contrasted with those of Sanskrit, but in this chapter I want to examine them for what they are, rather than for what they are not. Similarly, Allison Busch has drawn attention to features of Braj Bhāṣā that made it not simply a vehicle for literary expression but an aesthetic object in its own right. Features of its grammar, its lexicon, and its metrical repertoire combined to give the language a particular “expressive range” that was highly valued in the literary culture of early modern North India.2
The idea that a language has an aesthetics is in some sense familiar from travelers’ clichés. English speakers, for instance, have probably encountered the notion that German is “awful,” angry-sounding, confusingly complex, hyperspecific in some particulars and frustratingly vague in others.3 But I am not talking about a native speaker of one language discovering the “foreignness” of a foreign language, which is the central conceit of these clichés. I am referring to a situation that was common in premodernity but is almost unthinkable now, in which someone chooses to compose in a language not because it is his or her “native” language—for these languages were never anyone’s “native” language—but because it offered specific expressive resources that he or she wanted to make use of. These resources are part of what an earlier generation of scholars meant by the German term Ausbildung, meaning both the historical process of making a language suitable for literary expression and the cumulative result of that process.4
The notion that languages have particular expressive resources is somewhat old-fashioned. Nowadays, one needs to be at least half joking to claim that one language is better than another in any respect. The old prejudices, for example, that one could only philosophize in Greek or in German, have been exposed as prejudices. The background assumption is rather that all languages are created equal, which is, of course, true in a certain sense. The problem occurs when we try to formulate a theory of literary language. Such a theory requires us to understand and explain what it was about a language that made people choose to compose literature in it, and often invest a significant amount of time and effort in mastering it. What they mastered was not “just” the language, but the modes of literary expression associated with it. I say “just” in scare quotes because these modes really were considered to be part of the language rather than external to it.
This is a different approach to literary language from the one literary theorists commonly take. They often take the distinction between “literary” and “non-literary” forms of a language as given, and describe the specific differences of one vis-à-vis the other. This is how Erich Auerbach arrived at his characterization of literary language as being “distinguished from the general language of daily life by its selectivity, homogeneity, and conservatism.”5 This approach, of course, presupposes that both of these forms are actually given. And perhaps it also presupposes a certain ontology of literary language in general, that it exists as a modification of the “general language of daily life.” We might label this second presupposition “homoglossy,” the idea that literary language forms a unity with a corresponding non-literary language. Precisely what kind of unity is meant is not always clear. If, however, we hold Auerbach’s larger argument in mind—that a condition of a thriving literary culture is a literary language that forms a unity with the “mother language” of the community at large—then it becomes clear that homoglossy means that people write literature in a version of the language that they speak in their daily life.
I am very doubtful that either of these presuppositions is met in the case of Prakrit, or for that matter, in the case of many other literary languages. Consider Old Provençal, the language of the troubadours. What, exactly, is the “general language of daily life” that would correspond to it? Presumably a less selective, less homogeneous, and less conservative version of the language of troubadour poetry—a language that is not actually “given,” in the sense of attested to by manuscripts, but postulated on the basis of troubadour poetry itself. But according to the authorities in this field, the Auerbachian presupposition of homoglossy is not met. The earliest troubadour whose works are extant, William of Aquitaine (late eleventh–early twelfth century), probably spoke Old French rather than Old Provençal in his daily life. In several of his poems he addresses a transregional public of troubadour poets, which became more and more transregional in successive generations. Within a century, the language of the troubadours was cultivated across southern France, in Catalonia, in North Italy, and in Sicily. By this point, as Pierre Swiggers has remarked, its public was largely “alloglossic.” The geography of literary languages was clearly different, and bigger, than the geography of the “languages of daily life.” One might insist that homoglossy is still a condition of the origin of literary languages, if not necessarily a condition of their continued use and popularity. Yet here, too, authorities on medieval literature would disagree. “The most recent work on the origin of the poetic languages of the Romance-speaking peoples,” Paul Zumthor writes, “has established... that the languages in question were anything but direct emanations of a given natural dialect; from the very first they bear the mark of at least a potential unity and of artificiality; moreover, in relation to their spoken substrates they show some degree of abstraction.”6
That is also true of Prakrit. Its existence as a literary language is not explained by the existence of another, similar, language of which we have no certain knowledge. Indeed, earlier generations of scholars considered its existence as a literary language to be a “veil” that separates us from its true origins, from the everyday forms of speech in which language “really” consists.7 That is why, in this chapter, I focus on another type of explanation: the expressive resources that Prakrit was believed to offer. For utilizing these resources was, in part, what it meant to compose in Prakrit. I will discuss them on three levels: Prakrit’s “sweet” texture on the level of its phonetics, its “quavering” rhythms on the level of its meter, and its “unbound” character on the level of its poetic compositions. I use quotation marks here to indicate that these are not my own judgments, but characterizations that ancient readers of Prakrit literature, and indeed authors of Prakrit literature, actually supplied.
SWEET SYLLABLES
In a verse from the Brilliance of the Connoisseurs that we have already encountered in the introduction, the poet Vairocana reflects on his chosen medium:
Prakrit poetry is like a beautiful courtesan:
erotic, alluring, full of rasa,
delicate, provoking excitement and desire,
it captivates your heart.8
Much of this verse can be explained by reference to the traditional subjects of Prakrit poetry. Ever since Seven Centuries, Prakrit had been the preferred medium for erotic lyrics. But in what respect is Prakrit “delicate”? We can turn to another reflection on Prakrit for a clue. This one comes from an anthology, called the Vajjālagga, compiled by one Jagadvallabha, which contains an entire section on the gāthā, the Prakrit poem, where the following verse is found:
Interspersed with regional words,
made of sweet syllables put into metrical form,
playful, with meanings plain, powerful, and clear—
Prakrit poetry is fully worth reciting.9
Here we find another set of characteristics, which don’t quite match Vairocana’s, but which are somewhat more specific: Prakrit poetry is “playful,” but it is its meanings that are “plain, powerful, and clear,” and its syllables that are “sweet.”10 These verses highlight a particular feature of how Prakrit sounds, of what we might call its phonic texture, continuing Vairocana’s tactile metaphor, or following the Vajjālagga’s verse into a synesthetic realm, its phonic taste.
The oldest definition of literary “sweetness” relates not specifically to the sound or meaning of a text, but to the general capacity for enjoying it over and over again. The Treatise on Theater of the early centuries CE says that sweetness is “when a text has been heard many times, or spoken again and again, and does not cause annoyance.” Herman Tieken has shown that such a concept was already available to King Aśoka, in the early third century BCE, who invokes it indirectly in his fourteenth Rock Edict.11 This definition operates in the background of more precise and elaborated concepts of sweetness in literature. But I believe we can be more specific regarding what it was that caused people to recognize Prakrit’s syllables as “sweet,” beyond the fact that their repetition was a source of pleasure rather than annoyance. And I think that this quality, which was appreciated by Vairocana and Jagadvallabha, is related to a quality of which other readers of Prakrit were rather more critical.
In his Comparative Grammar of the Modern Languages of India, published in 1872, John Beames made a few observations about the language of Seven Centuries. At the time it was one of the only texts written entirely in Prakrit that was available to European scholars, chiefly through the excerpts that Albrecht Weber had published in the course of preparing the edition of the text that would appear in 1881. Beames jumped to the conclusion that the Prakrit of Seven Centuries was “emasculated stuff”: “the author ruthlessly massacres consonants and long vowels to suit his rhyme or rhythm, or to secure a more harmonious turn to his verse.”12 To Beames, Prakrit had too many “artificial sweeteners.” It was made to sound a certain way by relying on arbitrary and capricious techniques. Prakrit’s artificiality would become a refrain throughout the later nineteenth and early twentieth centuries. Beames connected it, albeit obliquely, with its femininity. I suspect that Prakrit’s long-standing association with the feminine had preconditioned Beames’s judgment, and specifically the fact that female characters were assigned varieties of Prakrit in stage plays, which by Beames’s time had been known to European scholars for over a century, and perhaps also the fact that most of the verses in Seven Centuries were imagined to have been spoken by women, which would have been a more recent discovery. What about the Prakrit of Seven Centuries would have driven Beames to this assessment? And was he right?
Beames alluded to the modification of vowel length. There are certainly cases of shortening and lengthening, but I think these phenomena are hardly indicative of a “modification” of the language for poetic purposes. There are only a handful of words that are subject to these processes, and they seem to be conditioned by phonological factors. The adverbs corresponding to Sanskrit yathā and tathā are one example: each has two variants in Prakrit (jahā/jaha and tahā/taha), but the distribution in Seven Centuries shows that the long-vowel variant is usually conditioned by a preceding ṇa.13 Similarly, almost all of the cases of vowel lengthening involve a preverb, for example, pāaḍa, from prakaṭa, in the above verse from the Vajjālagga. It is likely that the lengthening in such cases is a manifestation of accentual prominence. It does not matter whether Prakrit maintained the mobile accentual system of Vedic, as Richard Pischel maintained, or whether it had Latin-like accentuation rules that fixed the accent two or three syllables from the end of the word, as Hermann Jacobi argued.14 Poets certainly took advantage of this kind of variation, but it is unlikely that they manipulated the length of vowels solely because of the exigencies of meter or rhyme.
What about the “massacre” of consonants? There are a number of phenomena to be noted here. First, Prakrit has a smaller inventory of consonants than Sanskrit as a result of the elimination of place-of-articulation contrasts. This was the most obvious difference between Sanskrit and Prakrit, and was often remarked upon in very early texts.15 Thus there are three sibilants in Sanskrit (ś, ṣ, s), which are articulated in three different places: at the palate, at the palate with a curled tongue, and at the teeth, respectively. In Prakrit, there is only one sibilant (s), which does not contrast in its place of articulation with any other. Similarly, Sanskrit distinguishes dental and retroflex nasals (n, ṇ), even if their occurrence is largely determined by phonological context. In Prakrit, there is no significant contrast between the two.16 Second, Prakrit does not permit combinations of heterorganic consonants, which are consonants articulated at different places in the mouth. This means that all such combinations become homorganic, or articulated at the same place, which includes double consonants (as in uppala from utpala) or combinations with a syllable-final nasal (as in ciṃdha from cihna). Third, single intervocalic consonants are subject to extensive lenition, literally, “softening,” which it is tempting to gloss in this context as “sweetening.” Aspirates are generally reduced to h, losing their place of articulation, and unaspirated stops are generally elided altogether. Cumulatively, these processes often produce forms which are mostly vowels with very few consonants: the word prākṛta itself, which becomes pāua (or pāia or pāaa), is one example.
Taken together, these processes result in two features that we might call musicality and indeterminacy. I don’t mean musicality in the sense of tone or pitch—we know almost nothing about these features—but in the sense that Prakrit, with its high proportion of vowels to consonants, seems especially suitable for continuous and melismatic recitation. It is a phonetic characteristic, having to do with the way that Prakrit sounds, or perhaps even the way that it is pronounced. Prakrit’s high proportion of vowels gives it a more “open” articulation. And the loss of place-of-articulation contrasts often means that the transition from one vowel-sound to another is “smoother,” that is, there are fewer articulatory gestures involved. This quality is reflected especially in the “massacred” consonants that Beames referred to: mṛga “deer,” mṛta “dead,” and mada “lust” all become maa. And the same set of words serves as an example of indeterminacy, which is a semantic rather than a phonetic quality: a single Prakrit word, especially when it represents several different Sanskrit words, can have multiple meanings. Of course, polysemy is a basic fact of any language, and no human languages are completely “determinate” in this sense. Sanskrit, too, has its fair share of polysemous words.17 But the phonology of Prakrit has greatly amplified its indeterminacy relative to Sanskrit.
Both musicality and indeterminacy might be imagined to be as useful in literature and song as they are useless, or even harmful, in other domains of language use: could people really have made themselves understood through forms such as maa? Yet the underlying phonological processes are so well attested across the spectrum of Middle Indic languages, from present-day Afghanistan to Sri Lanka, and are so common among the world’s languages in general, that we should not suspect Prakrit authors of “faking” them. We should rather try to understand what contributions they might have made to Prakrit’s literariness.
We can begin from the theory of alliteration (anuprāsa), the repetition of certain speech-sounds within a given unit of context. Indian literary theorists recognized varieties of alliteration that were distinguished by the character of the speech-sounds that were repeated. Perhaps the earliest such classification is that of Harivṛddha, who distinguished eight bhaṇitis or “modes of speech.” Rudraṭa distinguished six varieties, and Bhoja distinguished twelve.18 The musicality of Prakrit lends itself to some of these and not others: the defining characteristic of what Bhoja calls the “stiff” (kaṭhora), for example, is the combination of r and velar consonants (k, kh, g, gh), which is impossible in Prakrit. Prakrit does indeed lend itself to the varieties called the “sweet” (madhura) and the “delicate” (komala), the words with which Prakrit was described in the verses we examined at the beginning of this section. In Bhoja’s system, these varieties are characterized by the use of a syllable-final nasal (anusvāra) and the use of r and ṇ respectively; Rudraṭa’s “sweet” variety seems to combine both of these characteristics. Here I simply want to highlight Prakrit’s suitability for these types of alliterative compositions.
I also want to draw attention to a type of alliteration that is common in Prakrit but impossible in Sanskrit, and which theorists who operated in Sanskrit seem to have struggled to define: the repetition of nothing. Because of the extensive lenition of intervocalic consonants, Prakrit often has nothing between vowels besides a hiatus, which Sanskrit tolerates in only a handful of rare words.19 To illustrate a type of alliterative composition he called the “powerful” (ojasvin), Bhoja quoted a verse from Rāvaṇa’s Demise, a Prakrit court epic composed by the Vākāṭaka king Pravarasena II around the early fifth century CE:
pattā a sībharāhaa-dhāu-silāala-ṇisaṇṇa-rāia-jalaaṃ |
sajjhaṃ ojjhara-pahasia-dari-muha-ṇikkanta-vaüla-maïrāmoaṃ ||
They reached the Sahya mountain,
where the clouds,
resting on the exposed rocks,
covered them in mist and took on their colors,
and where the laughing of waterfalls
and the wine-like smell of bakura flowers
issued from the mouths of the caves.20
We can detect here a number of alliterative pairs (sajjha/ojjhara), which happen not to alliterate in Sanskrit (sahya/nirjhara), but only one instance of the doubling or repetition of retroflex consonants that Bhoja identifies as the characteristic of “powerful” alliteration.21 This verse does exhibit the density of compound words that characterizes the “powerful” as a compositional quality (guṇa) rather than as a mode of alliteration, and it seems likely that this competing understanding of the “powerful” motivated Bhoja’s choice of this example. But there is an alliterative quality to this verse which Bhoja surely perceived, namely, the density of hiatus, which is in fact only possible in Prakrit poetry.
The aural qualities that distinctively characterize Prakrit are all related to its musicality, the reduction of articulatory gestures and its tendency to openness. The fact that consonants had to combine with themselves or with a placeless nasal, and never with heterorganic consonants, gave it a kind of smoothness—one possible meaning of the key descriptor komala. And the elimination of consonants altogether in certain contexts brought vowels into contact with each other. These qualities, I contend, are what premodern authors had in mind—even if only at the back of their minds—when they described Prakrit poetry in general as sweet, soft, and tender. This feature of Prakrit’s phonic texture or taste might have aligned particularly well with other types of musicality. For, as we will see below, its metrical patterns had their own kind of musicality. And there is some evidence that Prakrit verses were performed with particular melodies, at least in the context of the stage play, which would add another layer of musicality.22
Indeterminacy was put to use in poetry in a variety of ways. We have already encountered verses in Seven Centuries that depend on a single word being understood in two different meanings (e.g., W467, W428, and W364), and in other Prakrit texts there are “apparent contradictions” (virodhābhāsas) that depend upon reading a word in two different senses. These features are of course common in Sanskrit as well. Prakrit merely increases the possibilities for “bitextual” techniques, in which the same sequence of phonemes is productive of different meanings.23 But there are verses called galitakas in which a certain type of “bitextuality” is a constitutive feature of the composition. Since galitakas were only ever composed in Prakrit, these verses might help to make the case that the “sweet syllables” of Prakrit had specific literary purposes.
All of the known examples of galitakas “in the wild” come from Rāvaṇa’s Demise. Writers of metrical handbooks, such as Virahāṅka and Hemacandra, give a few additional varieties. We know that there were additional galitaka compositions in two Prakrit court epics that are now lost, Hari’s Victory and Rāvaṇa’s Victory.24 These verses are characterized by a particular kind of end-rhyme: the exact same syllables are repeated, but each time they must mean something different. This feature, known as yamaka, or “twinning,” is certainly difficult to realize—Daṇḍin discusses it in the “difficult” (duṣkara) chapter of his Mirror on Literature—but Prakrit has the advantage of relative indeterminacy. Here is one example from Rāvaṇa’s Demise:
añjaṇa-rāaeṇa saï dhūsarantaāiṃ
gaṇḍa-alesu khalia-visamosarantaāiṃ |
sura-bandīṇa ṇaaṇa-galiāiṁ aṃsuāiṃ
kappa-laāṇa jattha maïlenti aṃsuāiṃ ||
Always dusky with lamp-black,
trickling down over their cheeks,
the tears from the eyes of the imprisoned nymphs
darkened the garments
on the branches of the kalpa trees.25
As often in these galitaka verses, Pravarasena utilizes the fact that a single Prakrit word, such as aṃsua, might have more than one meaning, corresponding in this case to aśru “tear” and aṃśuka “garment” in Sanskrit. Other strategies for making the rhyme work involve the manipulation of word-boundaries and the use of pleonastic suffixes such as we see in this verse: dhūsarantaāiṃ, osarantaāiṃ, and aṃsuāiṃ all involve the suffix that Sanskrit grammarians call svārthe ka, “pleonastic ka,” which in Prakrit might as well be called svārthe a, since the intervocalic -k- is always lost.
A comparison with Sanskrit offers, by way of a baseline, a convenient way of talking about what was distinctive about Prakrit in terms of the possibilities its musicality and indeterminacy opened up to poets. But these features do not in themselves depend on the comparison with Sanskrit: a word such as aṃsua will have the same semantic range regardless of whether we compare it with a set of corresponding Sanskrit words. This is important, because as much as a text such as Rāvaṇa’s Demise seems to be mediated by Sanskrit—it was, and is, read through Sanskrit commentaries—the text itself does not need to be understood through a layer of Sanskrit meanings that lies underneath the Prakrit surface. Indeed the large number of deśī words, which do not obviously correspond to Sanskrit words, poses a problem for Sanskrit mediation, either as a theory of the text’s composition (i.e., that Pravarasena composed it in Sanskrit and then “sweetened” it by transforming it into Prakrit) or as a theory of the text’s reception (i.e., that readers could only understand it by translating it word-for-word into Sanskrit).26
Some of the representations of Prakrit in Indian literature as soft, delicate, tender, and so on might give us the impression that it was a specialized cant used exclusively for erotic poetry within the broader domain of Sanskrit textuality. This is the impression that scholars of the late nineteenth and twentieth centuries actually had. As we have seen, John Beames thought that Prakrit was an “emasculated” excuse for a language, providing the maximum possible scope to versification and song by suspending all of the rules of grammar. I think this is very far from the truth, but I also think that these critics were onto something. Prakrit does have certain phonic and semantic capacities that poets exploited effectively, capacities that I have been calling musicality and indeterminacy. Their exploitation did not amount to the creation of a language from scratch, but it did result in Prakrit being linked in the literary-cultural imaginary with the features of sweetness and delicacy, not just on the level of what Prakrit poetry was about, but on the level of how Prakrit poetry actually sounded. There was, of course, some interference between the evaluation of the style and content of the poetry—which was correspondingly lyrical, sensitive, and erotic in the main—and the evaluation of its phonic texture. But this is partly what I meant in framing this discussion around Prakrit’s “expressive resources”: the most fundamental features of the Prakrit language, such as its phonetics, become meaningful to its readers and contribute to its aesthetic power.
QUAVERING VERSES
Prakrit is a literature of gāthās. The word gāthā refers both to the most common and characteristic meter of Prakrit poetry and, by extension, to the Prakrit verses composed in that meter. This is clear from the verses in Prakrit anthologies that speak of the beauty of poetry, and in particular of Prakrit poetry: they generally refer to their subjects as gāthās.27
According to its derivation from the verbal root gā, “to sing,” the word gāthā refers to a sung verse. This highlights one of the tensions inherent in Prakrit poetry. Sheldon Pollock has argued that “the realm of the oral, specifically, the sung” lies outside of “the sphere of literary culture.”28 Where are we to place Prakrit gāthās? Are they closer to the songs that one might sing to pass time at the grinding stone, or to the literate productions of professional poets? I have argued in the previous chapter that Prakrit texts helped to establish “the sphere of literary culture” where works of literary art, kāvya, were produced. They are some of the earliest texts to identify themselves as kāvya, and form a crucial part of the genealogy of kāvya. The gāthā, like Prakrit itself, thus seems to stand between two categories that have been essential for conceptualizing and historicizing cultural practices in India: on the one hand, the oral, musical, and sung; on the other, the literate, textual, and recited. In this section I describe what is distinctive about Prakrit versification, and I venture a number of claims about the role of Prakrit versification practices and metrical knowledge in the history of literature and textuality more broadly in India.
Gāthā is an old Indo-European word. Its Avestan cognate (gāϑā), which is probably more widely known, refers to the songs ascribed to Zarathushtra that constitute the oldest and most sacred texts of Zoroastrianism. The earliest attested uses of the word gāthā in India are unsurprisingly connected with the chanting of Vedic hymns. Later Vedic texts cite a number of verses—referred to as ślokas and gāthās—that are unattached to any particular tradition of Vedic recitation.29
None of these earlier traditions exhibit the unique metrical structure that characterizes the Prakrit gāthā. Avestan and Vedic verse are syllable-counting, and it appears that particular forms of syllable-counting verse are an Indo-European inheritance.30 The Prakrit gāthā, however, belongs to a class of verse forms that is regulated by gaṇas rather than by syllables. A gaṇa is a “group” of moras, and a mora is a prosodic unit: it is what a light syllable (⏑) has one of, and what a heavy syllable (–) has two of. Light syllables, for our purposes, are those that contain a short vowel and no final consonants; all other syllables are heavy. On top of a given framework of gaṇas may be overlaid a seemingly endless variety of “surface forms,” consisting of particular syllabic configuration. The basic rule of gaṇa-counting verse is that a heavy syllable, which consists of two moras, must never cross a boundary between gaṇas. These meters, which the tradition generally called jātis, are hence very flexible.31 Fundamental to the entire system of gaṇa-counting verse is the metrical equivalence of two light syllables and one heavy syllable—an underlying prosodic structure that linguists call the moraic trochee. With a few exceptions, this system is absent from earlier traditions of versification in India.
The mora, although it is defined prosodically, could serve as unit of time as well. It is thus a unit of rhythmic equivalence: a gaṇa of four moras, for example, should have the same duration regardless of the particular configuration of syllables in which it is realized. Hence gaṇa-counting meters, in contrast to syllable-counting meters, can be thought of as having an inherent “beat.” A meter that consists of a sequence of four-mora gaṇas can be recited in “common time.”
Most gaṇa-counting meters, and above all the Prakrit gāthā, exhibit additional forms of rhythmic regulation. A gaṇa might be realized with a syncopated or unsyncopated rhythm, that is, with a prominence on the second or first mora of the gaṇa. At this finer level of analysis, “rhythm” does not simply arise from the way light and heavy syllables are strung together, but from the way that syllables are parsed into prosodic feet. The parsing of syllables into prosodic feet is a phonological procedure that Prakrit verse has incorporated into its metrical grammar, and the details of this procedure need not concern us here.32 The upshot of foot-parsing is that word boundaries play an important role in characterizing the rhythm of a gaṇa as syncopated or unsyncopated: thus, for example, the shape ⏑|⏑⏑⏑ patterns with the “syncopated” shape ⏑–⏑, while ⏑⏑|⏑⏑ patterns with the “unsyncopated” shape ––.
The alternation of rhythms is built into the deep structure of the Prakrit gāthā: the odd gaṇas must be unsyncopated, and some but not all of the even gaṇas must be syncopated. But writers on metrics recognized a particular type of gāthā in which this rhythmic alternation appears on the surface. This is the capalā, a “quavering” or “modulating” verse that realizes all of the even gaṇas with the syncopated shape ⏑–⏑, surrounded on either side by a heavy syllable to reinforce the contrast. Writers distinguished variants that were “front-modulating” (mukhacapalā) and “back-modulating” (jaghanacapalā), depending on whether the first or second line exhibited this pattern. Their primarily motive in doing so, however, seems to have been to elicit a pair of double meanings: among the cast of characters in Prakrit erotic poetry are the woman who says just a little too much (mukhacapalā) and the woman who moves her hips just a little too much (jaghanacapalā) to be above suspicion.33 The Prakrit gāthā ends with another built-in syncopation—a single light syllable toward the end of its second line—which is what allows us to recognize the two-line verse as a discrete metrical unit.
To see how this type of versification works, we can take an example from a gāthā about gāthās in Vairocana’s Brilliance of the Connoisseurs:
ekkā vi ittha vivihā samaa-viseseṇa vaṇṇa-bheeṇa |
dīsaï ṇaḍi vva gāhā bhiṇṇa-rasā bhiṇṇa-bhāvā a ||
Though one, it is manifold.
Like an actress
who wears different face paint at different times,
the gāthā,
with different ways of reading its syllables,
expresses different emotional states.34
Determining the weight of each syllable gives us the following pattern:
–– ⏑–⏑ ⏑⏑– ⏑⏑⏑⏑ –– ⏑–⏑ –– ⏑
–⏑⏑ ⏑–⏑ –– –⏑⏑ –– ⏑ –– ⏑
And grouping these syllables into gaṇas gives the following pattern:
––, ⏑–⏑, ⏑⏑–, ⏑⏑⏑⏑, ––, ⏑–⏑,––, ⏑
–⏑⏑, ⏑–⏑, ––, –⏑⏑, ––, ⏑, ––, ⏑
Note the alternation of syncopated gaṇas (in gray) and unsyncopated gaṇas. Note, too, that the gāthā is a “catalectic” meter, which means that both lines leave off the last syllable of the final metrical unit. As noted above, the second line has a shortened sixth gaṇa that syncopates the whole rest of the line, signaling the end of the verse.
These quavering verses, with their endless variety of syllabic patterns and their subtle alternations playing out over a stable rhythmic framework of gaṇas, are the mainstay of Prakrit literature. Gaṇa-counting meters are found in other literatures, and other metrical forms are found in Prakrit. But they are “Prakrit meters” in a sense that goes beyond the fact that they are common in Prakrit. To write in Prakrit was, to a very large extent, to write in gāthās or related gaṇa-counting meters. Less appreciated, but perhaps more historically significant, is the converse: to write in gāthās was to write in Prakrit.
It is well known that there are no traces of gaṇa-counting verse in Vedic literature, or indeed in any Sanskrit texts prior to Patañjali’s Great Commentary (around the second century BCE). These meters occur for the first time in the canonical literature of the Buddhists and the Jains, and hence in the “Middle Indic” languages we call Pali and Ardhamāgadhī. Both canons, however, represent texts that were transmitted orally for centuries before being “committed” to writing. The scare quotes are necessary because, far from fixing the text in a determinate and inalterable shape, the technology of writing introduced completely new possibilities of revision, expansion, and interpolation. Thus, despite containing material that may well go back, in some form, to the time of Buddha and Mahāvīra, and hence to the sixth and fifth centuries BCE, the texts as we have them are products of the early centuries CE. In the case of the Pali canon, an ambitious commentarial enterprise led by the Sri Lankan monk Buddhaghosa represents a snapshot of the textual tradition in the fifth century. And in the case of the Ardhamāgadhī canon, the council of Valabhī, also in the fifth century, represented the end of a long and highly disputed process of canon formation.
Both sets of texts have an internal chronology in which the use of gaṇa-counting meters is centrally implicated. Ludwig Alsdorf has shown that the oldest layers of these texts use the “old āryā,” an archaic version of the gāthā discovered by Hermann Jacobi in 1884. The use of the gāthā in its classical form is limited to chronologically later layers.35 According to the picture sketched by Alsdorf, we have in both canons an “early” layer in which just one gaṇa-counting meter, the old āryā, is used sporadically alongside the more frequent syllable-counting meters such as anuṣṭubh and triṣṭubh, and a “later” layer in which the classical gāthā is found. The classical gāthā is thus a sign of lateness. On this basis, Alsdorf suggested that the “later” layer of the Jain canon, where the gāthā is the preferred verse form, was later than the “later” layer of the Pali canon, where the gāthā is still relatively rare. The Pali canon, he argued, was constituted at a time before the gāthā had become “the metrical fashion of the epoch.” Roy Norman has argued, equivalently, that the Buddhist community ultimately responsible for putting the Pali canon together had moved to South India right around the time when the gāthā was gaining popularity in the North.36
What is the significance of the use of the gāthā in the later portions of the Pali and Ardhamāgadhī canons? The very limited scholarly discussion on this question frames it within the two processes of “development” (or “borrowing”) and “popularization.”37 The first refers to the transformation of existing verse forms into new ones; it is the historical process that “metrical etymology” traverses. According to Hermann Jacobi, and most scholars after him, the gāthā developed from the syllable-counting meters of an earlier metrical repertoire by according greater and greater scope to the techniques of contraction (replacing two light syllables with a single heavy syllable) and resolution (replacing a single heavy syllable with two light syllables) until we can no longer call the meters “syllable-counting” at all. The evidence for such a process comes from “transitional forms” that are partly syllable-counting and partly mora-counting. These include the late Vedic and early Pali/Ardhamāgadhī triṣṭubh, which sometimes employs contraction and resolution; the vaitālīya and aupacchandasika, which are mora-counting at the beginning of the line and syllable-counting at the end of the line; and finally the old āryā, which is mostly mora- or gaṇa-counting but more strictly regulated than the classical gāthā as to its alternating rhythm. According to an alternative hypothesis of George Hart, the gāthā did not develop from the syllable-counting meters we encounter in earlier Sanskrit texts, but was borrowed from a Dravidian tradition of versification. This tradition would have to be old enough for the “early” portions of the Pali and Ardhamāgadhī canons to borrow from it, and thus it would have to be much earlier than the existing corpus of Tamil literature.
These accounts do not explicitly tell us how, much less why, this process of development or borrowing got started. Was there a period of experimentation? Were there influences from other traditions, Dravidian or otherwise, and if so, what was their nature? Or should we assume that traditions are always developing, generating new verse forms and sloughing off old ones? Some of this explanatory work is done, albeit implicitly, by the second process of “popularization.” But this term requires some caution. Being popular in the sense of being frequent within a corpus of texts is very easy to conflate with being popular in the sense of being demotic or current among the common people. There is thus a temptation, most clearly visible in A. K. Warder’s account, to explain gaṇa-counting versification as a popular-demotic movement. And if it is the canonical texts of Buddhism and Jainism where the gāthā and related meters first occur, then that may be because of the willingness of these religions to speak the language of, and sing the songs of, the common man. I think this is highly sentimental. We would, however, expect different systems of versification to be correlated with different forms of life, and perhaps the “Magadhan” culture that lies in the background of Buddhism and Jainism is part of the story of gaṇa-counting versification.38
I would like to offer a different way of thinking about the changes in versification practice from the earlier to the later layers of the Pali and Ardhamāgadhī canons. These traditions were Prakritized. It has long been known that the Pali canon, in particular, was “Sanskritized” over the course of its transmission, and by this word we understand the replacement of earlier Middle Indic forms, whether morphemes such as -ttā or lexemes such as bambhaṇa, with their Sanskrit equivalents (-tvā and brāhmaṇa).39 These replacements indicate that the textual tradition that would later be identified as “Pali” came under the influence of a Sanskrit textual tradition. Although “influence” is a slippery term, we have a close parallel in the tradition that we have come to identify as “Buddhist Hybrid Sanskrit”: texts like the Divine Stories and Extensive Play of the Bodhisattva employ a Middle Indic language that has been Sanskritized to an even greater degree than the Pali canon.40
By Prakritization I mean the transformation of a textual tradition through the language, versification, and aesthetics of Prakrit literature. This process is somewhat more difficult to put into evidence than Sanskritization, but only because our eyes have been trained to the superficially obvious differences between Sanskrit and all varieties of Middle Indic. What if we trained our eyes to the more subtle differences between Prakrit and other kinds of Middle Indic? We have already seen that a distinctively Prakrit kind of versification enters into the Pali and Ardhamāgadhī traditions at some point in their history. We might also see that if they can be assigned a date at all, the texts that prominently feature gaṇa-counting meters date from around the first century CE or later.41
The Jain tradition, at least, provides relatively clear evidence for this sea change in versification practices. Although the new gaṇa-counting meters like the gāthā appear in some canonical texts, most of these texts are rather late (after the first century), and as noted above, Alsdorf showed that the vast majority of gāthā verses in texts such as the Uttarādhyāyana Sūtra that are considered to be earlier are interpolations. But of what period specifically?
The Jain canon is embedded in an extensive exegetical literature, one layer of which—called “explanations” (niryuktis)—is composed entirely in Prakrit gāthās. As we saw in chapter 3, these “explanations” reflect an expansion and transformation of the Jain scriptural tradition associated with the teacher Bhadrabāhu, and dates to around the first century CE. A comparison between the Āvaśyaka Niryukti of the Śvetāmbara Jains, and the Mūlācāra of the Digambaras, two collections of religious stories, shows how this transformation happened: while the two texts contain much material in common, the Mūlācāra, which according to Nalini Balbir is the older version, presents it in anuṣṭubh verses, and the Niryukti presents it in gāthās.42
What else, besides a new kind of versification, betokens the Prakritization of these traditions? The “explanations” are well known to be linguistically distinct from the texts they purport to explain, although the habit of referring to both languages as “Prakrit,” as well as extensive mutual influences over the course of their transmission, have rendered this difference much less conspicuous. Dalsukh Malvania has noted in passing that manuscripts of the Jain scriptures without commentaries look more like Ardhamāgadhī, and manuscripts with commentaries look more like Prakrit (“Jain Mahārāṣṭrī”).43 We may therefore even speak of a double Prakritization. The first phase is the commentarial elaboration of the Jain canon in the language and meters of Prakrit literature, associated with the efforts of Bhadrabāhu. The second is the subsequent conceptual and, to a lesser degree, linguistic redetermination of the canonical texts themselves as Prakrit texts.
We do not encounter such linguistically distinct layers in the Pali canon. But once again, if we look closely, we can see that the use of the gāthā indexes other differences. Take the example of the Songs of the Buddhist Nuns. This is a collection of verses attributed to the first few generations of Buddhist nuns, which has been considered a “precursor” to the Prakrit poetry of Seven Centuries and to the entire tradition of kāvya.44 It is not just a coincidence that the two longest and most expressive poems, those of Isidāsī and Sumedhā, are the only ones to utilize the gāthā. The new verse form betokens a new way of using language, one that is aware of and attentive to its expressive powers. The closest intertext of these poems is not, to my mind, Seven Centuries, but rather Pālitta’s Taraṅgavatī, in which the title character tells the story of her conversion in expressive Prakrit gāthās. The chronological priority of the Buddhist Songs to Pālitta’s Taraṅgavatī is not entirely self-evident; I do not take it for granted, as some scholars do, that the entire Pāli canon was fixed by the second century BCE. But even if no certainty can be reached on this specific point, the later portions of the Pali canon seem to draw from a wider literary discourse in Prakrit that was taking shape around the first century CE.45
The claim that the textual traditions of Buddhism and Jainism were “Prakritized” before they reached their final form does stand in need of further research. It would imply, however, that traditions of versification, just like the languages in which they subsist, do not grow and wither like plants; and that instead of connecting the use of the gāthā in Pali and Ardhamāgadhī texts with a completely hypothetical practice of demotic versification, we might connect it with the actually existing practices of Prakrit literature—which, as I have emphasized at several points, are not necessarily demotic practices. Prakritization is not popularization. My claim here is that the gāthā is not only common in Prakrit texts, but distinctively characterizes Prakrit as a discursive formation. Of course, the gāthā does not exclusively occur in Prakrit, or even “Prakritized” texts: it has a long history of use in technical Sanskrit, from śloka-kārikās in Patañjali’s Great Commentary (second century BCE), to the argumentative verse of Nāgārjuna’s Dispeller of Disputes (second century CE) and Īśvarakṛṣna’s Verses on Sāṃkhya (ca. third century CE). In Sanskrit, however, it was a convenience: its flexibility allowed it to accommodate technical terms, as Helmer Smith argued. In Prakrit, by contrast, it was the default meter.46
The gāthā is the only meter to have entire works written about it: the first, although its date remains uncertain, is the Definition of the Gāthā (Gāthālakṣaṇa) by Nanditāḍhya.47 But other works on metrics—above all Virahāṅka’s Collection of Mora- and Syllable-Counting Meters (ca. eighth century) and Svayambhū’s Meters (ninth century)—provide a glimpse onto a lost world of Prakrit versification that was much more varied than its Sanskrit counterpart. As the title of Virahāṅka’s work suggests, the repertoire included both the syllable-counting meters (vṛttas) that were typically used in Sanskrit literature as well as the mora-counting meters (jātis) that were more often used in Prakrit literature. The most popular of the mora-counting meters, besides the gāthā, was an “acatalectic” variant called the skandhaka, which did not omit the final syllable from the last gaṇa of each line. The skandhaka was employed in Prakrit court epics, such as Hari’s Victory and Rāvaṇa’s Demise. But the category of jāti also included various kinds of rhymed verse, including the galitakas we encountered above and khañjakas we’ll see below. These works defined a large number of strophic forms in which simple verse forms were combined.
These strophic compositions take us back to the theme with which this section began: Prakrit’s dual status as a language of literate textuality of a high order, as well as a language closely associated with musical performance. The few surviving examples of strophic compositions, which come from stage plays, exemplify the ambiguity of this position. Before considering them, it helps to bear in mind a similar ambiguity in the case of stage plays from ancient Greece. These plays were typically performed with choral odes. In earlier plays, such as those of Aristophanes, the text of the odes was transmitted along with the dialogue in manuscripts. In later plays, such as those of Menander, odes were generally not transmitted with the dialogue, although it is usually assumed that they were part of the performance. There is no question that these odes belonged to “the realm of the oral, specifically, the sung,” but the decision of whether they also belong to “the domain of literary culture”—whether they constitute an essential part of the literary work—has been made for us by the manuscript tradition. We might ask whether Prakrit songs, like these choral odes, belong to the play-as-performance or to the play-as-literature.
The Prakrit and Apabhramsha songs that appear in some manuscripts of the fourth act of Kālidāsa’s Urvaśī Won by Valor brings the question into focus. Are they Kālidāsa’s own compositions—which would make them, in the early fifth century CE, the earliest examples of Apabhramsha verse available to us—or were they added in the course of time?48 The stage directions associated with these songs make them out to be dhruvās, a kind of “mood music” that directors may choose to include in their staging of a play. We have plenty of evidence, including from the Treatise on Theater, that Prakrit and Apabhramsha songs were often employed in the play-as-performance, without necessarily constituting part of the play-as-literature. But as the fourth act of Urvaśī Won by Valor shows, the dividing line is not always clear.
The question becomes even more complicated when these Prakrit songs enter into the mimetic world of the stage play. I am referring to situations where characters are represented as singing, or listening to, Prakrit songs. One example could be the verse from the Recognition of Śakuntalā that Śakuntalā intends to send to Duśyanta in a love letter, discussed in the introduction. But let us look at another example, a rare strophic composition found at the beginning of Harṣa’s Ratnāvalī:
kusumāuhapiadūaaṃ maülāvaṃto cūaaṃ |
siḍhiliamāṇaggahaṇao pāaï dāhiṇapavaṇao ||
viasiavaülāsoao icchiapiaamamelao |
palivālaṇaasamatthao tammaï juaīsatthao ||
ia paḍhamaṃ mahumāso jaṇassa hiaāiṁ kuṇaï maüāiṃ |
pacchā viṃdhaï kāmo laddhappasarehiṁ kusumabāṇehiṃ ||
The southern breeze is here, bringing buds to
the mango, the dear messenger of the God of Love,
slackening anger and quarrels,
making the bakula and aśoka trees blossom,
bringing pining lovers together,
while groups of young girls gasp for air,
incapable of waiting any longer.
Thus does the spring month first soften people’s hearts,
then, when his flower-arrows find an opening,
the God of Love pierces them.49
This is a Prakrit song, which Svayambhū identifies for us as a śīrṣaka or strophic composition. It has two parts, and hence is called a dvipadī-khaṇḍa. The first part is a khañjaka, a generic term for a “piece” of a larger strophic composition, which in this case is a avalambaka: two verses made up of quarters of thirteen moras each, with the rhythm ⏔⏑– at the end, and end-rhyme between successive quarters. The second part is a gīti, a verse form very similar to the gāthā but with two lines of equal length, rather than a shortened second line. Both parts exhibit end-rhyme, which is a characteristic of Prakrit khañjakas, and of most Apabhramsha meters, but very rarely figures in the Sanskrit and Prakrit that survives in written form. In this case, the rhyme enhances the musicality of the language, for example in the repetition of the consonant-less sequence -ūaaṃ in the first line, which must have been further enhanced by its musical setting.
We must not forget, however, that this is not just a Prakrit song, but a dramatic representation of a Prakrit song. At this point in the Ratnāvalī, King Udayana comes out to watch the Holi celebrations with his friend Vasantaka, and he sees the two servant girls Madanikā and Cūtalatikā dancing and singing the song quoted above. The king is impressed, and he has Vasantaka go and try to learn it from them. But Vasantaka is a bit clueless, and he mistakes their song for a carcarī, another type of song and dance that was performed at the spring festival. Madanikā tells him that the song was not a carcarī, but a dvipadī-khaṇḍa. By including a dramatization of the spring festival in his play, Harṣa has made the performance of a Prakrit song part of the play-as-literature.
INEXHAUSTIBLE COLLECTIONS
Prakrit is a literature of gāthās, but this latter word does not simply refer to the language’s most popular and most characteristic metrical form. The gāthā is the poem, syntactically and semantically complete on its own, that takes this form: the whole world of the poem must be contained in its two lines. A verse incorporated in the Anuyogadvāra Sūtra, compiled sometime before the fifth century, says that “a soldier is known from his armor, a woman from her outfit, a pot of rice by a grain, and a poet from a single gāthā.”50 The earliest and most influential work of Prakrit literature, Seven Centuries, is made up of such single-verse poems. And it was principally through anthologies such as Seven Centuries that Prakrit literature was known and studied, both in the premodern and the modern world. There were, of course, many other genres. Jain narrative literature in Prakrit, which flourished between the eighth and twelfth centuries, far exceeds anthologies in sheer volume. But the anthology always retained a special connection with Prakrit in the literary imagination.
The anthology is the only Prakrit genre represented by Hindu, Buddhist, and Jain authors. But the sectarian affiliation of the compiler has very little to do with the actual content of the anthology, which is often taken from other poets in any case. The Treasury of Gāthā-Jewels (1194 CE) is a case in point: Jineśvara begins the collection with verses in praise of the Jina, Brahma, Viṣṇu, Śiva, and Sarasvatī taken from earlier literature. This additive and syncretic character is one of the anthology’s key features. We see, in the first few pages of Jineśvara anthology, verses from the Seven Centuries, from Vākpatirāja’s Gauḍa’s Demise (eighth century), from the Vajjālagga, and remarkably, because the original text is completely lost, from Guṇāḍhya’s Great Story (Bṛhatkathā). The anthology is central to Prakrit literature because it defines and presents “Prakrit literature” as a field of intertextuality.
A collection was called a “treasury” (kośa), and the verses contained therein were often likened to gold and jewels.51 Daṇḍin distinguished the “treasury” from “aggregation” (saṅghāta), but it is difficult to tell whether he is following an older tradition.52 The distinction, according to both Ratnaśrījñāna and Vādijaṅghāla (both in the tenth century), is that the treasury features verses on various themes while the aggregation presents verses on a single theme. Vādijaṅghāla offers the Constellation (Tārāgaṇa) of Bappabhaṭṭi, discussed below, as an example of a treasury (along with the Treasury of Gāthās, which likely refers to Seven Centuries, and an otherwise-unknown Spotted Antelope), and the Tamil anthologies (draviḍasaṅghāta) as examples of aggregations.53 According to Taruṇavācaspati, however, the treasury differs from the aggregation in that it contains verses from various authors, and Bhoja also uses the authorship criterion to distinguish the two genres in his Illumination of the Erotic.54
Daṇḍin’s remarks, or rather the various interpretations of his unusually cryptic categorization, raise what I consider to be the two primary issues in the study of anthologies as a genre: their formal organization and their authorship. The history of the genre is another important issue, but it will suffice to note here that the anthology is present from the very beginnings of Prakrit literature—and also of Tamil literature—and that Hari Ram Acharya has traced the influence of the Seven Centuries on later anthologies in Sanskrit.55 This is a major point of difference between Sanskrit and Prakrit as literary traditions. As a literature of gāthās, Prakrit is and always has been a literature of anthologies, many of which precede the earliest anthologies of Sanskrit literature by centuries. When it comes to single-author collections, there are outstanding Sanskrit examples from the middle of the first millennium, such as Bhartṛhari’s Three Centuries, Amaru’s Century, and Ravigupta’s Treasury of Āryās. Or rather, these are traditionally considered to be single-author collections. Daniel Ingalls has judged that Amaru’s Century is actually the work of several poets, and probably carried ascriptions of individual verses to particular poets in the early stages of its manuscript transmission.56 There are a number of single-author collections in Prakrit from roughly the same period, including Bappabhaṭṭi’s Constellation and Vairocana’s Brilliance of the Connoisseurs.
Regarding multiple-author “treasuries,” however, most of the early examples are in Prakrit. Besides Seven Centuries, several collections of Prakrit verse were compiled by Jain monks and laymen. The earliest example—before 1337 CE, and sometime after Vākpatirāja’s composition of Gauḍa’s Demise in the eighth century—is evidently Jagadvallabha’s Vajjālagga. Other examples include Jineśvara’s Treasury of Gāthā-Jewels (1194 CE) and the collections printed with it (Subhāsiyagāhāsaṃgaha and Subhāsiyapajjasaṃgaha). The Verses of the Chappaṇṇayas should be included in this category, too, although the text that survives under this name is almost certainly not the text that authors such as Daṇḍin, Uddyotana, and Abhinavagupta knew. The latter seems to have been the work of a poetic collective, somewhere between the single-author and multiple-author models. From Daṇḍin’s reference to them in the beginning of his Avantisundarī, we know that their Verses were in circulation around the year 700, but I suspect that they, like Seven Centuries, belong to the period of Sātavāhana rule in the first or second century CE.57 And, of course, as Vādijaṅghāla reminds us, the Tamil anthologies (draviḍasaṅghāta) were also in circulation, if only in Tamil Nadu, by the middle of the first millennium. By contrast, the earliest surviving multiple-author “treasury” in Sanskrit, if we do not count Amaru, is the Treasury of Subhāṣita-Jewels (Subhāṣitaratnakośa), compiled just before the twelfth century.
Extent is the most obvious way of characterizing an anthology that has no overall thematic organization, and this is how Seven Centuries received its name. But why are its verses counted in groups of a hundred, and why are there seven of them? S. V. Sohoni suggested that the model was the Bhagavadgītā, which also contains around 700 verses, and that Hāla actually intended it as an anti-Bhagavadgītā. But there is little evidence for this interpretation. Equally unconvincing is Acharya’s suggestion that the phrase “seven centuries” (sattasaī) simply sounds better in Prakrit than other candidates.58 If the element sāta in the names Sātavāhana and Sātakarṇi does in fact derive from sapta “seven,” as S. A. Joglekar has suggested, then the Seven Centuries might be an oblique reference to the name of the patron or his dynasty, but I remain doubtful.59
The commentators on Seven Centuries knew that verses in the anthology sometimes cluster around a given theme or word. Herman Tieken elaborated on this “linking” as an organizational feature, but it is not nearly as systematic as that found, for example, in Kālidāsa’s Cloud Messenger, where almost every verse is linked to the preceding verse by a repetition of a word.60 The verses of each century are, for the most part, “unbound” (anibaddha), as Bhāmaha would call literature of this type.61 They are thus vulnerable to rearrangement. This appears to have happened often in the history of Seven Centuries. Not only are chunks of verses found in different places in different versions of the anthology, but several versions exhibit a complete rearrangement of the verses according to their topic. These topically organized versions include Sādhāraṇadeva’s recension and the “First Telinga Recension,” both studied by Albrecht Weber for his edition of the text, and the Gāthāmuktāvalī described by H. C. Bhayani. The topics are generally referred to by the Prakrit word vajjā, which is etymologically identical to the paryāyas mentioned by Ānandavardhana and Abhinavagupta, or by the Sanskrit word paddhati. Compilers such as Jagadvallabha and Jineśvara would employ this formal device in their Vajjālagga and Treasury of Gāthā-Jewels respectively.62
The arrangement into vajjās seems to be a formalization of a looser thematic grouping evident in earlier collections of verses. Vairocana’s Brilliance of the Connoisseurs, the date of which remains unknown, moves from topic to topic in a natural but not formally explicit sequence: from a reflection on the qualities of good readers, for example, to a reflection on the qualities of good lovers. Bappabhaṭṭi’s Constellation, of the later eighth century, exhibits a similar arrangement. The Constellation was compiled by Bappabhaṭṭi’s friend Śaṅkuka, who composed “index-verses.” Each index-verse names two to five verses by a keyword in each. Often, but not always, Śaṅkuka mentions the theme or topic according to which he has arranged the verses. Here is one example:
Vādin! How can we praise you?
You are the one who praises,
as shown by these five verses:
susiyattaṇa, bahulakkhaya, sirīsa,
jaladugga, and vāraṇārī.63
The five verses whose keywords are mentioned in the index-verse are all eulogies of a king. But the index-verse also serves another important function: it maintains the attribution of the verse to its author.64 The practice of composing index-verses (dvāra-gāthās) is as old as Prakrit textuality itself. In composing their “explanations” (niryuktis) and “discussions” (bhāṣyas) on canonical texts, Jain commentators enumerated topics for discussion in index-verses. This practice was redeployed to strengthen the fragile bond of authorship in Prakrit literary culture. Unbound verses, which collectively represent a great deal of Prakrit literature, are not just unbound from larger structures of meaning, but from the formal and material structures that often served as the locus of attribution. We can think of the anthology not only as a site of collection, where these unbound verses could be integrated into such a structure, but as a site of dispersion: being anthologized in one work or in one manuscript—and it is often impossible to distinguish between the two—was simply a temporary stopover in the life of a Prakrit gāthā.
On this topic, there is a pair of gāthās in the Brilliance of the Connoisseurs that sets out two modes of existence for Prakrit poetry:
suviārasahā vimuhī aṇṇāṇa aṇaṇṇagoarā dāṇi |
kulavālia vva lukkaï gehe ṇiasāmiraṃjiṇī gāhā ||
A gāthā that is very thoughtful
and kept to oneself, away from others,
pleases the one who possesses it,
as a woman in the confines of the family,
prudent, uninterested in and inaccessible to others,
pleases her husband.
kittivaaṃsā vimalā maṇoharā bahuviāraüjjaliā |
aïkkaṃtapiālāvā gāhā savvattha bhamaï vesa vva ||
The more attention is lavished on it,
the more it shines, pure and captivating,
garlanded by wide renown:
the gāthā that goes beyond lovers’ conversations
is to be found everywhere
like a courtesan.65
The first verse seems to recommend the private enjoyment of Prakrit poetry, but this is tempered by the second verse, which recommends, instead, its public circulation. We can note, briefly, that this is how Prakrit gāthās work in general: although they are self-contained and “unbound,” their profusion of meanings depends on a network of prior texts. It is as if every gāthā presupposes every other, each forming a node in a vast intertextual network. Appearing “everywhere” means appearing in an infinitude of contexts, of anthological or performative settings, and hence of new possibilities of contextual meaning. In this case, the meanings of the two verses are not quite complementary but not quite contradictory either; as an ethos of reading, they commend both intimacy and, with a wink, promiscuity.
We are used to distinguishing between a literary work itself and its reception or afterlife, or between an original “meaning” and a “significance” for later readers. But Prakrit gāthās exist entirely in their reception: esse est legeri. The recognition of this fact motivated Śaṅkuka to preserve his friend’s gāthās by anthologizing them, fitting them out with index-verses, and writing them down in manuscript form—by transforming them into structure, we might say, to borrow a phrase of Gadamer’s.66
One example will serve to illustrate the processes of constant recontextualization in which the life of a gāthā consists. The Mirror for Poets is a Prakrit text on metrics of the thirteenth century. In exemplifying some varieties of the gāthā, a commentator on this text, probably not far removed from the time of the Mirror, distinguished the brāhmaṇī variety as having the maximum number of heavy syllables (27). He adduced the following verse:
gajjaṃte khe mehā phullā nīvā paṇacciyā morā |
naṭṭho caṃdulloo vāsāratto halā patto ||
The clouds are thundering in the sky.
The kadamba is in bloom.
The peacocks are dancing.
The moonlight is gone.
The first night of the monsoon is here, my friend.67
This is one of the only verses that the commentator ascribes to a specific author, and that author is Pālitta. Not too long before it was cited in the Mirror, the learned Jain monk Hemacandra cited the first few words of this verse as an illustration of two grammatical rules in his Siddhahemacandra (mid-twelfth century).68 Hemacandra, however, does not identify the author. Neither does Bhoja, one of Hemacandra’s principal sources, who cites the verse on two occasions. First, as an example of the “inferential” kind of reason (jñāpaka-hetu) in his Illumination of the Erotic, and second, as a variety of the “forward-and-backward-looking” kind of inference (sāmānyataḥ) in his Necklace of Sarasvatī (both early eleventh century).69 Here we have three authors citing the same verse: one for its metrical features, one for its grammatical features, and one for its logical features. Yet the verse itself is found in no extant work of Prakrit literature. Where did these authors encounter this verse, and how did the anonymous commentator of the Mirror for Poets know that Pālitta was its author?
I think it is possible that these authors all cited the verse from Pālitta’s now-lost Taraṅgavatī. But if this verse managed to escape oblivion, it is because it was cited; and if it was cited, it is because it was citable. The survival of Pālitta’s poetry, as well as the survival of its attribution to Pālitta, has taken several courses. First, and most obviously, there is the tradition of Taraṅgavatī (including later retellings), to which Pālitta’s name is attached as an author. Yet even here it might be recalled that Pālitta, according to Jain legend, was accused of plagiarizing Taraṅgavatī from one of his colleagues at the Sātavāhana court.70 But there is also the anthology tradition, and further, there are the indirect traditions of “accidental anthologies”: those texts like the Mirror for Poets and Svayambhū’s Meters that, in the course of exemplifying a set of metrical or grammatical phenomena, end up assembling an anthology of verses. Another example is the Explanation of the Suggestion Verses of Ratnākara, which assembles and revises Abhinavagupta’s commentary on the Prakrit verses cited in Ānandavardhana’s Light on Suggestion.
We know very little about the way that anthologies, especially Prakrit anthologies, were produced. The seminal text of this tradition is of course Seven Centuries, but this is a typically problematic case: with our earliest direct witness, the commentator Bhuvanapāla (ca. eleventh century), we intercept the tradition nearly a thousand years into its history. By this time, authors had for hundreds of years been citing verses “from Seven Centuries,” which is to say, verses that are also found in later manuscripts of Seven Centuries. In fact, nobody actually attributes these verses to this work; if the verses are attributed at all, they are attributed to a particular author. Svayambhū’s metrical handbook provides an example: a verse that he attributes to Pālitta is identical to W75 in Seven Centuries, which the commentators on that text likewise attribute to Pālitta. While I do not share the skepticism of earlier scholars regarding these attributions (“worthless” according to A. B. Keith), no serious research has been done on them, and it is not at all clear where they come from.71 Take, as another example, verse W394: “In the spring, the peacock cranes its neck to drink a drop of water from the tip of a blade of grass, as if it were a pearl pierced by an emerald thread.”72 This is a rare case of agreement between the commentators regarding the authorship of the verse: Bhuvanapāla, Ājaḍa, and Pītāmbara all assign it to Pālitta. But how do they know? I speculate that Seven Centuries probably was the source of many of these citations, but that it once circulated with a large complement of intertexts and paratexts—including a list of authors and perhaps collections of the works of individual authors—that has been substantially winnowed over the course of its transmission.
In closing, I would like to return to the larger structures of meaning from which Prakrit gāthās are “unbound.” The great literary theorist Abhinavagupta maintained in the late tenth century that there was a qualitative difference between a large-scale work, in which all of the narrative elements are presented to the reader before his very eyes, and a small-scale work like the single-verse poem, which presents the reader with few or no narrative elements. We aren’t given to know, for example, who is speaking, who is being spoken to, and what has happened prior to the verse being spoken. In order to understand the verse—in other words, to give meaning to it—we must conjecture all of these elements. And while the verse itself might give us some clues, Abhinavagupta makes it clear that only readers who are practiced in the conventions of the relevant kind of poetry can successfully make those conjectures. Such readers can picture the narrative situation as if it were before their very eyes, despite or perhaps due to the fact that they have had to imagine it.73
One difference between the large-scale and the small-scale work thus pertains to reading practices, and indeed to practice in the more common sense: readers of a small-scale work, in the absence of explicit narrative development, need to turn to past experience, to prior texts, which collectively provide the reader with conditions of meaning and interpretation. I know of no better example of this kind of reading practice than Abhinavagupta’s own interpretation of a Prakrit verse (W886) in his commentary on Ānandavardhana’s Light on Suggestion, where he conjectures not one, not two, but eight possible narrative contexts, each with a slightly different meaning.74 In this way, although the Prakrit gāthā is formally “unbound,” it is always reintegrated into a larger structure of meaning—for Abhinavagupta these are primarily narrative structures, but we could also consider figurative or affective structures—that itself depends on a potentially boundless number of intertexts. It is noteworthy that the very narrative elements that Abhinavagupta says the reader must supply, the verse’s “points of attachment” to a structure of narrative meaning such as the speaker and addressee, are usually supplied by the commentaries to Seven Centuries: “a woman says this to her friend,” “a woman says this to her messenger,” and so on. These short introductions serve as paratexts that aid in the understanding of the text. They are strikingly similar to the kiḷavis that are transmitted as paratexts to the Tamil caṅkam poems, which likewise set out the speaker and addressee in certain conventional roles.75
Prakrit gāthās live in the complexities of collection and dispersion, of citation and recontextualization, skipping over and across the transmission histories of individual texts. Within Indian literary culture, their “unbound” character was prized and celebrated, since it allowed individual verses to speak to different purposes from within different texts—but it was also a liability, since it made over to future generations the responsibility of transmitting verses faithfully and preserving their attribution. We might even think of all Prakrit gāthās as fragments: not just the stray verses of now-forgotten poets such as Abhimānacihna that have been preserved in accidental anthologies such as Syavambhū’s Meter, but the verses that are transmitted to us in intentional anthologies as well. For fragments present a shard of meaning that can only be appreciated against a background of intertexts, but this background changes. The conventions that emerge for reading gāthās in one context might change as we move over to another: consider, in this connection, the divergent interpretations of the commentators on Seven Centuries. Prakrit gāthās were characterized by their appearance, and continual reappearance, in various contexts—in performance or in a manuscript, in a topically arranged anthology or cited in a grammatical textbook, introduced by an “index-verse” or by the definition of a poetic figure. This promiscuity was a conspicuous feature of Prakrit’s phenomenology and aesthetics, of what it was like and what attracted people to it.
This chapter has surveyed three kinds of distinctiveness about Prakrit literature: the sweetness of its syllables, which I understood in relation to phonetic characteristics that made the language smooth, open, and musical; the quavering rhythms of its verse, which refers to the special kind of versification associated with Prakrit poetry, which allowed enormous variation over a regular beat with syncopation permitted on the off-beat; and the prevalence of single-verse poems, which is connected with certain forms of textual organization, like the anthology, particular reading practices, and above all with an open-ended “ontology” that not only allowed but positively encouraged poems to circulate promiscuously, to appear in diverse contexts, to mean different things to different people. These features can be said to characterize Prakrit internally, since they are the resources internal to the language and to the tradition that Prakrit poets made conscious use of. In the next chapter, we will turn to the ways in which Prakrit is characterized externally, that is, under a series of contrasts that differentially established its place in the language order of India.