Free Access
Histoire Epistémologie Langage
Volume 39, Number 2, 2017
La grammaire sanskrite étendue
Page(s) 45 - 63
Published online 18 April 2018

© SHESL/EDP Sciences


Referring to the celebrated Tamil grammar Tolkappiyam's absence of Pāṇini's specialized metalinguistic affix-symbols and symbol-clusters to designate the morphophonemic features of Sanskrit, A.C. Burnell muses that “indeed it is impossible to see what use [Pāṇini's technical language] could be in a grammar of one of the so-called Dravidian languages” (Burnell 1976, p. 40). One may feasibly extend this question to query how the adoption of any metalanguage – technical or not – designed for one language significantly benefits the analysis of another language, especially for one that, on the surface, seems unsuited for that kind of analysis. Perhaps in a capacity akin to the actual method adopted to describe a language, a grammatical technical terminology is always tightly linked to a certain conception (more or less sophisticated) of language. Extending, therefore, the metalinguistic conventions from an established grammatical tradition too mechanically and over-projecting similarity between languages may have the unwanted consequence of imprecision, albeit in service of “learning” and teaching the language better in a multilingual or polyglossic context.

The choice of grammatical terms, as with the method of analysis, has the very important function of serving as a code of reference for users of a grammar and strategically facilitates access to extrinsic linguistic resources that inform the text. In a bilingual or diglossic situation, the adoption of grammatical terminology marked to express metalinguistic awareness has an even more pronounced function, namely to illuminate features of linguistic description that might otherwise remain irredeemably occluded or inadequately demonstrated for language learners. It is in these contexts, one might argue, that the text under discussion here – a grammar of Telugu composed entirely in Sanskrit – is illustrative.

For a grammar of Telugu to be composed in Sanskrit, appropriating metalanguage developed to describe Sanskrit, the two languages must have enough in common to make such an exercise possible. Indeed, Telugu shares with Sanskrit several important typological features. In addition to its wide sharing of base lexical forms, like Sanskrit, Telugu has a complex network of sandhi processes (morpho-phonological fusions and transformations within words and between word boundaries). Both languages are also highly inflected, although, unlike Sanskrit, the Telugu adjective is not inflected for number, gender, or case. Telugu also shares with Sanskrit phonological characteristics in the number of nasal sounds that condition consonants. Some notable differences include Telugu's morpho-syntactic features, such as gender agreement with the verb. The language also possesses no discrete feminine gender, as words are described in the grammars as either masculine or non-masculine (which includes feminine and neuter words). Also distinct from Sanskrit and Prakrit, insofar as the letters are concerned, Telugu has a number of sound units, including short diphthong vowels e and o which, in Sanskrit, are always considered long vowels.

These and numerous other relationships and divergences that exist between Telugu, Sanskrit, and Prakrit are described in fine detail in an early Telugu grammar composed entirely in Sanskrit. This work, most commonly referred to as the Āndhraśabdacintāmaṇi (“A treatise [lit. “a wish-fulfilling gem”] on Andhra speech-forms”), is preserved in a single manuscript and printed edition that has been recently translated into English (Sundaram and Patel 2016, p. 1-120). The grammar deserves attention both as a daring creative project of contrastive grammar and as a complex experiment in bilingual mediation among learned users of Sanskrit and Prakrit who also perhaps knew literary Telugu, Tamil, or another Dravidian language such as Kannada or Malayalam.1 Telugu, a major language of Deccan India, has a literary past that extends to, at least, the eleventh century and claims the largest number of speakers of any Dravidian language. To be precise, Telugu is a South-Central Dravidian language that perhaps may have been the earliest language to split from proto-Dravidian (Andronov 2003, p. 492), perhaps two millennia before we have the first non-inscriptional literary work in Telugu, the Mahābhāratamu from poet-saint Nannaya Bhaṭṭu. Nannaya is also known as the traditional author of the Āndhraśabdacintāmaṇi (ĀŚC hereafter), although there are compelling arguments that it may have been actually composed as late as the seventeenth century by Elakuchi Balasaraswati (1590-1670), who has made available a reconstituted Sanskrit text and Telugu commentary (Sundaram and Patel 2016, p. v-vii).2

In the printed version with Balasaraswati's commentary, the anonymous editor has split eighty-two verses in the āryā meter that comprise the ĀŚC into two hundred and seventy four short sūtras.3 While the authorship and chronology of the ĀŚC is obscure, we can be certain that this remarkable text, though not composed in Telugu, has powerfully influenced arguably the most influential Telugu grammar (written in Telugu) still in use today. That work, a nineteenth-century grammar called Bālavyākaraṇamu by Cinnayasūri directly cites or reforms many of the ĀŚC's sūtras.4 Composed not in Telugu, but rather in Sanskrit, the ĀŚC uses an eclectic mélange of pre-Pāṇinian, Pāṇinian, and post-Pāṇinian metalanguage and organizational logic, offering a dramatic example of bold experimentation meeting pragmatic purpose in the field of Indian grammar. Using both ubiquitous technical Sanskrit metalinguistic terms along with the specialized Pāṇinian symbol-based formal system, the ĀŚC's sūtras are organized into five chapters (pariccheda): metalinguistic terminology (saṃjñā); euphonic coalescence (sandhi);5 nominal bases/verbal stems ending in vowels (ajanta); nominal bases/verbal stems ending in consonants (halanta); and verbal action (kriyā). This text's organization, as well as contents, bear striking resemblance to later reformulations of Pāṇini's Aṣṭādhyāyi (“[Grammar in] Eight Chapters”), as will be discussed below.

Metalinguistic Terminology in the Āndhraśabdacintāmaṇi

The first chapter of the ĀŚC has the heading of saṃjñā (lit. a “name”), which, in the context of grammar, may be translated as a linguistic terminology that frames a discussion about a given entity or concept or, more comprehensively, as technical term that facilitates the complete understanding of a given entity or concept. This linguistic terminology either takes the form of natural words meaningfully rerouted to refer to various aspects of syntax, morphology, and phonetics, or to artificial vocabulary rooted in conventionalized symbol-units created to promote convenience of grammatical description. One would expect that while the former could transit from the grammatical analysis of one language to another, the latter would be restricted in jurisdiction to the source language. Therefore, it is unsurprising that certain words found across grammars of South Asian languages, etymologically comprehensible, like svara (vowel), vyañjana (consonant), avyaya (indeclinable) or anunāsika (nasal) occur ubiquitously before and after Pāṇini's time in multiple linguistic contexts extrinsic to Sanskrit. However, Pāṇini's technical language and symbol-clusters, appropriated to languages like Telugu without a hint of improbability, defies expectations. The fact that both types of technical terminologies are used by a work like the ĀŚC (not uniquely in South Asia) leads to reflection not only on the innovative exercise to extend metalinguistic terms by the later tradition but also on the powerful sway held by the metalinguistic technology in the first instance.

The ĀŚC employs both Pāṇinian symbols and their Telugu equivalents, in several cases shifting the significance of a Pāṇinian cluster (as in ac, ‘vowel’ in Pāṇini's grammar, in ĀŚC 1.18, below) to manifest the specificity of Telugu forms. The author juxtaposes Pāṇinian technical language (eṅg [the diphthongs e and o], aic [the diphthongs ai and au], hal [consonants], avyaya [undeclined forms], sup [nominal forms], tiṅ [verbal forms], ktva [gerund], tumun [infinitive]) and pre-Pāṇinian terms (such as nāman [noun/nominal base], pratyaya [suffix], kriyā [verbal action], bindu [a “dot” that graphically represents nasalization]) alongside his own inventions, terms such as vakra (the diphthongs e and o), vakratama (the diphthongs ai and au), prāṇi (consonants), varṇaka (suffix),6 paruṣa (“hard” unvoiced stops), sarala (“soft” voiced stops), sthira (all other “stable” stops), and druta (“dissolution” of n at the end of Telugu words). The term paruṣa (“hard”), for example, is used to identify unvoiced stops (alpa-prāna and aghoṣa are the corresponding Pāṇinian terms) while voiced stops are called sarala (“soft”); all other letters (aspirated, nasals, semi-vowels, and sibilants) are called sthira (“stable”). At other points, the author strategically juxtaposes Pāṇinian and pre-Pāṇinian terms from Sanskrit sources – such as works on phonetics (prātiśākhya) and etymology (nirukta) – that are also found in later works such as the Kātantra, a grammar tradition from the early centuries of the common era traditionally attributed to an author named Sarvavarman. With Pāṇini's sūtras arranged topically for perhaps the first time, the Kātantra introduces various new technical terms, which the ĀŚC freely uses, as in 1.28 where the Kātantra technical term nāman (“name”), signifying a nominal base, is defined with the Pāṇinian corresponding term prātipadika. The ĀSC also clearly echoes a host of grammatical works that include analysis and rubrics of topical organization not found in the Aṣṭādhyāyī. Thus, for instance, the ĀŚC in 1.14 follows Hemacandra's twelfth-century Siddhahemaśabdānuśāna in including a long vocalic while delineating vowels.7 The ĀŚC similarly follows non-Pāṇinian rubrics for organizing content according to topic, first seen in the Kātantra but extending to Vararuci's canonical grammar on Prākrit Prākṛtaprakāśa (probably from the third to fifth centuries CE) and Bhaṭṭoji Dīkṣita's popular early seventeenth-century Siddhāntakaumudī, among other works.

In India, grammars of Sanskrit have been the fountainhead for both methodology and terminology, and the process of language standardization – whether of classical languages, like Prakrit, or local languages from the regions – has taken recourse, in some measure, to the descriptive techniques or metalanguage developed by grammarians of Sanskrit. Thus, the Prākṛtaprakāśa is composed in Sanskrit (as most Prakrit grammars are) and takes Sanskrit as the base language, extending Pāṇinian rules developed to describe Sanskrit to derive Prakrit languages. Additional rules address those aspects in Prakrit that do not correspond with Sanskrit. The general logic of the Prākṛtaprakāśa structures the ĀŚC as well. The typological proximity of both languages certainly facilitated the objectification of one language with the other language's descriptive terminology. Both languages have enough in common to justify such a move. In the case of the ĀŚC, the author places Sanskrit, Prakrit, and Telugu within the same linguistic and cultural domain. The substrata between these three languages should share morphological, structural, phonetic, semantic, lexical and, perhaps even psycholinguistic or cultural features in order to justify the experiment of tying the three together grammatically to promote bilingual pedagogy or, as may be the case, to perform a virtuosic gesture of unexpected comparativism to garner scholastic appreciation. The ĀŚC frames the Telugu language not exactly as a Prakrit but also not necessarily as a distinctive language like Tamil. Rather it marks the boundaries of an identity as a language that is culturally equal to Sanskrit, derived from it, and wholly distinct from it. Each of these positions operates in discrete domains. The problems that might arise in the implementation of a developed metalanguage from the one grammatical tradition to another are not as salient, therefore, in the case of Sanskrit and Telugu as presented by this text.

What this suggests, as the author of the ĀŚC seems to presuppose in the following sūtra, is that Sanskrit, Prakrit, and Telugu operate as structurally similar languages that vary in form. While Telugu word-forms that are derived from (or located in) Sanskrit and Prakrit forms (tadbhava) draw the three languages together, the presence of local speech-forms (deśya) leads the ĀŚC to ambivalently label Telugu as a “distinctive modification,” or vikṛti, of Sanskrit and Prakrit:

The earliest base (prakṛti) is Sanskrit. Then there is Prakrit. This [language, i.e. Telugu] ought to be considered as a distinctive modification (vikṛti) of those two [languages, i.e. Sanskrit and Prakrit].

[ādya-prakṛtiḥ prakṛtiś cādyē ēṣā tayōr bhavēd vikṛtiḥ (1.12)]

This sūtra encapsulates the author's view on Telugu's relation to Sanskrit and Prakrit. The two languages are understood to be sources for Telugu. Indeed, the designation of Telugu as vikṛti is ambiguous, and it is unclear if the author sees the language as derived from Sanskrit and Prakrit or as a distinct evolution of, that is nevertheless located in, the two languages.8

According to the author of the ĀŚC, “established usage of speech is evident from general usage” (siddhir lōkād dṛśyā [1.6]), starkly revealing his stance that the regional language (albeit probably a literary register of it), widely used, qualifies it as siddha (“established”) and, therefore, in the same domain as established languages like Sanskrit and Prakrit. By extension, “established” speech-forms of Sanskrit and Prakrit also have forms “derived from them” (sādhya or sādhyamāna), produced through phonological processes like sandhi or through the addition of various affixes. Hence, grammatical description of established and derived forms as performed in works like the ĀŚC underscores Telugu's proximity to Sanskrit and Prakrit. The use of the term siddha here hearkens back to the Kātantra grammar, which uses the term to describe the established inventory of sound-letters in Sanskrit (siddho varnasamāmnyāyaḥ). Sūtra 1.19 of the ĀŚC elaborates on this term in the context of the nasalization, or anusvāra.9 “Unrefined speech” (apabhraṃśa), in contrast, would be considered “incorrect speech” (yat tv apabhraṃśaḥ [1.10]). Building on sūtra 1.9 (aniyamād grāmyaṃ hi), where the author declares that speech that lacks rules is “unrefined” [lit. “rustic”], sūtra 1.10 implies that grāmya (“rustic”) speech is equivalent to apabhraṃśa. The great Sanskrit grammarian Patañjali (second century BCE) explains apabhraṃśa (lit. “fallen away” or “corrupt”) as ungrammatical usage in Sanskrit. Here, in the Telugu context, a connection is being made between grāmya and apaśabda (“an ungrammatical linguistic usage”). Essentially, therefore, grāmya is an apabhraṃśa of language that is refined and correct (saṃskārārtha).

Analyzing the Technical Terms

The table below summarizes most of the technical terms found in the opening chapter of the Āndhraśabdacintāmaṇi:

The first set (sūtra 1.14) of technical terms speaks to delineating the Telugu alphabet. The ĀŚC follows the Prākṛtaprakāśa, and by extension, the Kātantra in naming fifty varṇa or letters. The difference, of course, is that the fifty varṇa spoken of are not the same.10 The next sūtra (1.18) defines the thirty-six varṇa of the Telugu alphabet with the Pāninian symbols for vowels (ac) and consonants (hal). One noteworthy difference: whereas Pāṇini implicitly regards e and o as long vowels, the ĀŚC theorizes short e and o vowel counterparts to identify a perceived phonetic reality in Telugu. In employing the Pāṇinian technical term it, therefore, the author of the ĀŚC synthesizes Pāṇinian technique with Kātantra theory, which does explain e and o as potentially long and short; thus, ac (here, in the ĀŚC) signifies that i, u, e and o are both long and short (dvividhā). In addition to using the Pāṇinian term for vowel, the ĀŚC introduces a non-technical definition, in sūtra 1.31: “Vowels are called prāṇa or “life-breaths.” A connected definition for consonants occurs in sūtra 1.35: “Consonants are generally called prāṇi, “those possessing breath.”11

The ĀŚC also implies another technical term for vowel, svara, by indicating the term anusvāra (1.32), the “sound which follows the svara,” which marks nasalization. This term is actually encountered first in 1.18 and 1.19, where the author deploys the dual form anusvārau to signify that Telugu, on account of a distinct euphonic coalescence (sandhi) has two anusvāra – full (puṛna) and half (ardha) – which take on different shapes in the orthography of the language.12 This marks an important aspect of the metalanguage of this grammar. What precipitates the use of this terminology is that, in Telugu, anusvāra is theorized as naturally occurring in the word and thus established (siddha), or derived (sādhya) in that it comes resultant of sandhi. When there is sandhi (e.g. cen + tova = ceṃdova), anusvāra, with its graphical mark, conditionally marks the nasal element. The anusvāra primarily appears before consonantal stops. The fact that the graphical symbol of nasalization is here explicitly mentioned suggests the intersection of grammatical thought with writing in Telugu, something that is absent in the Sanskrit grammatical tradition. In fact, sūtra 1.32 defines anusvāra as bindu (“dot”), the script marker for nasalization in the form of a dot, available in old Telugu inscriptions and later replaced by a zero marker.

Also potentially related to script, are the two technical terms – vakra (“bent”) and vakratama (“completely bent”) – respectively defined as names for the diphthongs ai and au in sūtras 1.33 and 1.34. The technical terms for the varṇa could point to their “curved” (vakra) form in the script. Or, as in the case of a similar usage in Tibetan grammar, the implication might be toward the curve made by the tongue within the vocal apparatus to produce the sound. Otherwise, there seems to be no need to differentiate these letters from the others. In some Telugu grammars, short e and short o are included but since here the author resorts to Pāṇinian terminology (eṅ), he cannot represent short e and short o but only the long forms of these vowels.13 The term vakratama (“completely curved”) refers phonetically to the Pāṇinian aic (ai and au). As vakra signifies the first level of vowel gradation Pāṇini calls guṇa (e and o), vakratama, apparently signifies second-level vowel gradation (vṛddhi, i.e. ai and au). In describing the Tibetan grammar, Verhagen (2001, p. 49) describes the technical names given by Tibetan grammarians to capture the phonological system of Sanskrit, including the term bkug-pa (“bent” or “curved”) for i, which is a “bent a”. The diphthong e is “stretched out” (bsgreṅ-ba), while the diphthong ai is “stretched out and bent” (bsgreṅs-te-bkug-pa). Verhagen writes that he has:

[…] not found parallels to this particular nomenclature in the Indic grammatical traditions. It can, in any case, not be related to the traditional classifications introduced in the varṇa-sūtras within the major vyākaraṇa traditions, or – as far as I can tell – to the earlier, Vedic Prātiśākhya materials. The terms appear to be more or less impressionistic indications of the position and form assumed by the tongue (and other relevant parts of the oral cavity?) during the articulation of vowels. (ibid., p.49)

Verhagen also notes that the Smra-sgo describes the phonology of Sanskrit vowels “in terms of the visual form of graphs” (ibid.). Again, if one follows the theory that the technical language is pointing to a grapheme and not the peculiar curvature of the tongue to produce the sound, then the apparent graphical representation of Telugu through the Sanskrit technical terminology would be a novel contribution to Indian linguistics. Unlike other non-Indic linguistic contexts, where the overlay of the graphical representation of one language onto another may lead to distortion (such as applying the Latin alphabet to African languages, for instance), the problem seems not relevant here, as Sanskrit and Prakrit grammars are not explicitly concerned with graphical representation.

In domains shared between Telugu and Sanskrit/Prakrit, on the other hand, the metalinguistic terms are identical, as in sūtra 1.42, where the technical Pāṇinian symbol clusters sup (nouns), tiṅ (verbs), ktvā (adverbial particles), and tumun (infinitives) are imported wholesale from Sanskrit and preserved alongside the specialized word avyaya (indeclinable words). While the actual Telugu nominal, verbal, and participial suffixes obviously differ from the Sanskrit in form, the ĀŚC organizes Telugu suffixes using Pāṇinian symbol clusters as technical terms to suggest conceptual and operational similarities with Pāṇini's scheme for describing suffixes. The author of the ĀŚC then goes on to exemplify this connection. For instance, Pāṇini's term tumun-anta (“ending in tum[un]”) signifies an affix called tum attached to the end of a verb-stem to form the infinitive, as in gantum (“to go”) and paṭhitum (“to read”). The infinitive is called the tumun-anta in Telugu as well, but the affix that is actually attached is ku as in povuṭaku (“to go”) and caduvuṭaku (“to read”). This suggests that the logic of the grammar places the meaning associated with “ending in tumun” (tumun-anta) as an infinitive form to transcend the etymological meaning of “ending in tumun” as an infinitive form that ends in the suffix tum. The situation in sūtra 1.43 is reversed, where the term anukṛta (literally, “imitation” or “following” in Sanskrit) is used to indicate that quotations from Sanskrit sources are treated the same way in Telugu. The peculiar feature here is that a quotation, in Telugu, is followed by the word anu, which corresponds to iti (“thus”) in Sanskrit, used ubiquitously to mark quotation. For example, in Sanskrit, direct discourse followed by iti rāmāvadat (“thus said Rāma”) corresponds to the Telugu form ane rāma (“said Rāma”), where ane is the past-tense form of anu. While the use of the word “anukṛta” is not explicitly being tied to the quotative marker anu, the author is suggesting that the Telugu employment of anu to mark quotation imitates the usage of iti in Sanskrit. A glaring difference between the two markers is that while iti is a particle that does not conjugate along with the changing tense of the verb it accompanies, the Telugu anu does undergo change according to its accompanying verbal form.

The rest of the technical language in Chapter 1 of the ĀŚC consists of a mix of pre-Pāṇinian, and completely new phonological metalanguage. Sūtra 1.28 juxtaposes a Pāṇinian usage with an earlier term for a nominal base: “In Telugu, prātipadika is also called nāma.” Sūtra 1.30 defines a “verbal action” as a dhātu (root). Sūtra 1.29 introduces a new term, varṇaka, in the sense of a suffix (pratyaya) to indicate the common Telugu endings to nominal items (such as first-case endings like ḍu, mu, vu, and lu). The next set of terms is, to my knowledge, unavailable outside of this grammar:

  • 1.36: druta (“disappearing,” “dissolved”)

  • 1.37: paruṣa (“hard”)

  • 1.38: sarala (“soft”)

  • 1.39: sthira (“stable”)

  • 1.44: kalā (“portion”)

The term druta indicates the “disappearance” or “dissolution” of the letter n or na or nu at the end of words. There are three types of n in Telugu: nu, na, and the n that becomes an anusvāra nasal. This last n optionally disappears in Telugu. Unvoiced stops are called paruṣa (“hard”) in Telugu. There are other terms for this feature in Sanskrit that translate to “unaspirated” (alpa-prāṇa) and “unvoiced” (aghōṣa). The term paruṣa, however, applies to the first letters in each consonantal class (varga). In early Prakrit grammars, each are just named according to the varga they inhabit as, for example, in the Prākṛtānusāsana of Puruṣottama and in Ravana's (Laṅkeśvara's) Prākṛtakāmadhēnu. Only the Āndhraśabdacintāmaṇi gives this particular term. By contrast, voiced stops are called sarala (“soft”) in Telugu. Cinnayasūri, mentioned earlier as the author of the influential 1855 Telugu grammar Bālavyākaraṇamu, translates this and the previous sūtra verbatim, using the same terminology.14 All other letters (aspirated, nasals, semi-vowels, and sibilants) are called sthira, as they remain “stable” and unchanged. The author of this grammar may be drawing from the tradition of the Ṛk Prātiśākhya, a pre-Pāṇinian work (c. 600 BCE?) that deals with the phonetics of the Ṛgveda. Ṛk Prātiśākhya 13.9-11 defines a consonantal stop as not fixed (spṛṣṭam asthitam), and therefore, coming and going quickly. As paruṣa and sarala letters are understood as “unstable” (asthira), on account of their transformation in certain phonetic and semantic environments, the logic of this term sthira may stem from the usage found in early texts on phonetics, such as the Ṛk Prātiśākhya. A final technical term – peculiar in its scope and distinctively Telugu – is kalā (“portion”) that, according to sūtra 1.44, denotes: a) all verbs (tiṅ) ending in second person (madhyamā); b) all nouns ending in the sixth case (ṣaṣṭhi), the first case (suḥ), the second case (indicating the object); and c) those words with the meaning of “a little” (kiñcit), “praise” (praśaṃsana), “marvel” (adbhuta), “grief” (santāpa), “at any time whatsoever” (yadātadā), “what?” (kim), “only” (ēva), a “participle” (ktva), “assuredly” (kila), and “indifference” (tūṣṇī).

All “other” words (anyē), according to sūtra 1.45, i.e. those that are not kalā, are subject to being druta. The reason for the technical metalinguistic distinction here, between words that undergo druta (loss of the n, na, or nu) and words that do not (kalā), is not clear. One may surmise that the distinction stems from an anxiety around the loss or “dissolution” (druta) of the n, na, or nu, which occurs either as an augment or, surprisingly, even as an indeclinable (avyaya) component of words that end in n. As this phenomenon is unique to the n, and not applicable to any other augment (āgama) or word-final consonant, the terminological distinction between druta and kalā might exist to mark this special loss of the n. The term kalā (which is tentatively translated here as “portion”) marks those words (or, perhaps, a portion of those words) that especially do not undergo this peculiar druta process. Kalā, as defined in sūtra 1.44, does not encompass all non-druta words, however, so it remains hazy why only these forms have been selected here. The other chapters of the ĀŚC refer to this distinction in giving examples of druta, but the explicit division between words that are druta and kalā occurs only in this opening chapter.

An important set of new terms in the ĀŚC's first chapter refers to prosody. The Telugu metrical gaṇa system is different from the one found in Sanskrit or Prakrit and, therefore, occasions the need for terminological intervention. In Sanskrit, there are no names for the fixed metrical clusters (gaṇa), outside of labels of sets of letters or numbers of syllables. For example, ya-gaṇa signifies a three-syllable cluster of light-heavy-heavy syllables, whereas ma-gana signifies three consecutive heavy syllables, ta-gana would be heavy-heavy-light, and so on; similarly ekākṣara-gana would signify a cluster with only one (eka) syllable, dvyakṣara-gaṇa with two (dvi) syllables, and tryakṣara-gaṇa with three (tri) syllables. In Prakrit, names reflect the number of mora (mātra) in each class. According to sūtra 1.40 of the ĀŚC, however, names for gaṇa are given, such as indra, sūrya, and candra respectively signifying short, moderate, and long units of time wherein specific syllables or morae are recited. In addition to naming the groups, the ĀŚC also provides some distinctive technical terms to describe the specificities of Telugu prosody. For example, it introduces the term vaḷi (a distinctive form of the Sanskrit yati (“caesura”) as well as prāsa, where the second letter in each of the four pādas is identical. Vaḷi, which might indicate a “line” or “fold” that serves as a differentiating mark is a type of caesura distinct from what is meant in Sanskrit prosody. According to sūtra 1.41, the first letter (or a letter related to the first letter of the same class) is called vaḷi (vaḷir ādyam), which is repeated at some prescribed point. Thus, in lieu of a pause (yati) found in Sanskrit, the first letter is repeated further on in the pada or line, largely for aesthetic reasons (perhaps to make it more pleasing). Similarly, the second letter in each of the four lines of a Telugu verse is called prāsa, a characteristic feature of Telugu metrical compositions. Telugu prāsa is not to be confused with the technical term for the Sanskrit trope of alliteration (anuprāsa). Prāsa is not a sound-trope (śabdālaṅkāra), as in Sanskrit poetics, but rather its creative function in Telugu poetic contexts comes from its restrictive aspect. Vaḷi is mandatory in all Telugu verse while prāsa is virtually mandatory in all verse.

The penultimate sūtra (1.46) of this chapter uses metalanguage seen elsewhere in Sanskrit and Prakrit grammars to describe the nature of shared lexicons between Telugu and Sanskrit/Prakrit: tajja (words derived from Sanskrit or Prakrit words),15 tatsama (words that are identical with Sanskrit and Prakrit words), deśya (local Telugu words), and grāmya (rustic usages).16 While derived and identical words have as their substratum (prakṛti) the two classical languages, the “native language”, the author cryptically concludes, “flows along independently” (pravāhinī deśya).

ĀŚC Terms Sūtra Translation
varṇa ādyāyāḥ pañcāśad-varṇāḥ (1.14) “For the earliest [language, i.e. Sanskrit], there are fifty letters (varṇa).”
ac / hal dvividhād-id-ud-ed-ōtō'cō vakratamāv anusvārau. ka-ga-ca-ja-yuga-ṭa-ḍa-ṇa-ta-da-na-pa-ba-ma-ya-ra-la-va-sa-haḻaṃ tu hal-vargaḥ (1.18) “There are two types of vowels (ac) [in Telugu, long and short]: a, ā, i, ī, u, ū, e, ē, o, ō; two types of diphthongs (vakratama) [ai and au]; and two types of representation for nasals (anusvāra) [full (pūrṇa) and half (ardha)]. These are the consonants (hal): ka and ga [unvoiced and voiced velar stops respectively]; ca and ja [palatal and dental stops respectively]; ṭa, ḍa, and ṇa [retroflex stops]; ta, da, and na [dental stops]; and pa, ba, and ma [labial stops]; ya and va, [semi-vowels]; ra [a rhotic flap]; la and ḷa [laterals]; sa and ha [sibilants].
anusvāra siddhas sādhyaś cānusvāraḥ pūrṇārdha-bhēdatō dvividhaḥ (1.19) There are two types of anusvāra – full and half – established naturally [in a word] (siddha) and derived (sādhya).
antaḥstha laghavō 'laghavaś cēti dvēdhāntaḥsthān vadanti śāstāraḥ (1.22) The grammarians say that there are two types of semi-vowels (antaḥstha) [i.e. ya, ra, la, va]: light and not-light.
nāma prātipadikaṃ hi nāma (1.28) The nominal base is called nāma.
varṇaka pratyaya iha varṇakaḥ (1.29) A suffix (pratyaya) is called varṇaka here [in this grammar].
kriyā kriyā dhātuḥ (1.30) The verbal action is called dhātu.
prāṇa prāṇas samāsatō 'caḥ (1.31) Vowels are generally called prāṇa (“life-breaths”).
bindu bindur anusvāraḥ (1.32) The sound which follows the svara (anusvāra) is called bindu (represented by the symbol ‘o’ in early Telugu inscriptions).
vakra eṅg athō vakraḥ (1.33) eṅg (the vowels e and o) is called vakra (“bent”).
vakratama vakratamaḥ aic (1.34) aic (the diphthongs ai and au) is called vakratama (“completely bent”).
prāṇi halas syus samāsataḥ prāṇinaḥ (1.35) Consonants (hal) are generally called prāṇi (“those which are accompanied by vowels or prāṇa”).
druta drutākhyō naḥ (1.36) [The letter] n is called druta (“that which may disappear or dissolve”).
paruṣa ka-ca-ṭa-ta-pāḥ paruṣākhyāḥ (1.37) [The unvoiced letters] ‘ka’, ‘ca’, ‘ṭa’, ‘ta’, and ‘pa’ are called paruṣa (lit., ‘hard’).
sarala ga-ja-ḍa-da-bāstu saralāḥ (1.38) [The [voiced letters] ‘ga’, ‘ja’, ‘ḍa’, ‘da’, and ‘ba’ are called sarala (lit., ‘soft’).
sthira sthirā itarē (1.39) All other letters (other than paruṣa and sarala) are termed as sthira.
gaṇa indrādyā api ca gaṇāḥ (1.40) Moreover, [prosodical groups – such as] ‘Indra’, etc. [‘Sūrya’, ‘Candra’]– are called gaṇas.
vaḷi / prāsa vaḷir ādyam anantarākṣaraṃ prāsaḥ (1.41) The first [letter] is called vaḷi (“caesura”; lit. “a line” or “a fold”) and the succeeding letter is called prāsa (“repetition of consonants”) [in all the lines of a verse].
avyaya / sup / tiṅ / ktvā / tumun na padāny avyaya-sup-tiṅ-ktvā-tumunām (1.42) [In Telugu] words like indeclinables (avyaya), nouns (sup), verbs (tiṅ), adverbial participles (ktvā), and infinitives (tumun) are not [used in the same way as in Sanskrit].
anukṛta anukṛtē 'sti vākyantu (1.43) Let them [Sanskrit sentences] be used in the form of a quotation (anukṛta).
kalā tiṅ-madhya-śeṣa-ṣaṣṭhī-sus-karma-pravacanīyaja-vibhaktīḥ. kiñcit-praśaṃsanādbhuta-santāpa-yadātadā-kim-evāṃś ca. ktvā kila tūṣṇīm-ādyān abhidadhati kalā druto na yeṣāṃ syāt (1.44) Verbs (tiṅ) ending in second person (madhyamā), nouns ending in the sixth case (ṣaṣṭhi), nouns ending in the first case (suḥ), nouns in the second case (indicating the object or karman), and those words with the meaning of ‘a little’ (kiñcit), ‘praise’ (praśaṃsana), ‘marvel’ (adbhuta), ‘grief’ (santāpa), ‘at any time whatsoever’ (yadātadā), ‘what?’ (kim), ‘only’ (ēva), ‘participle’ (ktva), ‘assuredly’ (kila), ‘indifference’ (tūṣṇī), etc. are said to be kalā [by other grammarians], when they are not druta.
tajjā / samā / deśyā /grāmyā taj-jā samā ca dēśyā-grāmyā ceyaṃ catur-vidhā bhavati. prakṛti-dvaya-jā taj-jā tulyā tu samā pravāhinī dēśyā (1.46) [This language] has four kinds of words: taj-ja (i.e. tad-bhava) [words derived from Sanskrit or Prakrit]; tat-sama [words that are the same as (i.e. borrowed from) Sanskrit or Prakrit]; dēśya [local words]; and grāmya [rustic words]. Tad-bhavas are derived from the two original languages (Sanskrit and Prakrit) while the tat-samas are similar to both of them. The native language flows along independently.


On the surface, it is not particularly startling that the Āndhraśabdacintāmaṇi would appropriate and innovate upon its Sanskrit (and Prakrit) predecessors. First, there is the singular fact, demonstrated by the ĀŚC, that Telugu and Sanskrit fundamentally share morpho-phonetic, structural, and lexical qualities that naturally draw them together. Second, relying on a Sanskrit, Prakrit, and Pali tradition that has already worked out over centuries a technology of linguistic description, celebrated for its brevity, comprehensiveness, and rule-ordering logic seems commonsensical, over and above inventing a wholly novel descriptive system. The actual exercise of grafting Pāṇinian techniques developed for describing Sanskrit onto Telugu, however, poses various challenges, as shown above. One intriguing aspect about a text like the ĀŚC, however, is the ways in which inherited techniques of linguistic description integrate with ingenious innovations necessary to explain distinctive features of Telugu phonology, orthography, and in a startling turn, prosody, while also making pointed comments about the special identity of the Telugu language. In this regard, the author confidently asserts in sūtra 1.5 that people

whose minds are passionate toward rasa (the emotional aspect of art and literature) love their own native place, its fashions, and its language. In this world, leaving aside other kinds of poetry (i.e. Sanskrit and Prakrit), they hold in high regards their own poetry.

While acknowledging how special Telugu is and how it requires its own analysis, the author clearly sees the advantages of extending the metalinguistic terminology and methodology from Sanskrit to Telugu. In cases where inherited definitions cannot adequately categorize linguistic units or features, the Telugu grammar either provides a criterion of identification in the form of discrete sūtras unfamiliar to Sanskrit and Prakrit grammar or creates a wholly new definition by either inventing a term or by innovating a sūtra.

One obvious field of analysis from which the ĀŚC profits concerns Sanskrit's comprehensive objectification of phonological processes within words and across word boundaries and the metalanguage that accompanies it. As the author of the ĀŚC intends to standardize a literary Telugu from what is hitherto a standardized spoken Telugu drawing together disparate but related dialects, the appropriations significantly assist the process. Metalinguistic terminology certainly heightens, if not catalyzes, the move toward literary standardization, as the terminology touches on every aspect of literariness, including poetics and orthography. As seen above, the grammar's author seems to reflect on orthography, thus signaling an innovation necessary for Telugu grammarians in a way that it was not necessary for Sanskrit and Prakrit grammarians. Similarly, the ĀŚC accomplishes the description of Telugu phonological features distinctive from Sanskrit and Prakrit as well. The concept of ardhānusvāra, for instance, is described in the ĀŚC using the Sanskrit analytic to explain certain letters (like aspirates) that are in one language but not the other. The terminology has to follow a methodology to demonstrate where one language's feature deviate from another. Thus, the introduction of terms like druta, paruṣa, sarala, bindu, and ardhabindu must have some special purpose for the Telugu grammarian.

In speculating on why such a work – which seems so heavily skewed toward understanding Telugu through Sanskrit – one may proceed in several directions. First, in assuming that the audiences of this grammar knew Sanskrit, one may surmise that the simple goal of the ĀŚC is to make known the structure of the Telugu language to native speakers. Another possibility raises the historical reality of multilingual courts and scholastic contexts, where scholars of Kannada and Tamil, perhaps, would be interacting with Telugu speakers. In this context, shared knowledge of Sanskrit and Sanskrit grammar, especially, could be a valuable instrument to accomplish the understanding of a foreign language in familiar terms; in other words, metalinguistic awareness would be a stepping-stone toward bilingual competency. Some have attributed this logic to the Kātantra grammar, perhaps designed for Prakrit speakers to learn Sanskrit, or the Rūpāvatāram of tenth-century Buddhist grammarian Dharmakīrti, which is the first of the so-called prakriyā works that reorders Pāṇini's sūtras topically, presumably for ease of learning the language. Similar might be the case of the eleventh-century Tamil Vīracōliyam grammar, which has been described, at least in one scholar's opinion, as a “‘contrastive-transfer grammar’ attempting to teach a second language (Tamil) through the application and transfer of categories from the student's first language (Sanskrit)”.17 Another describes some of the Vīracōḻiyam's uniqueness in terms of its complex mediating function, “theorizing Tamil grammar and poetics through lenses first developed for Sanskrit” (Monius 2013, p. 106). Monius sees the act as largely emblematic of processes of vernacularization across South Asia during the medieval period (Monius 2013, p. 115).

Another possibility suggests itself in the ĀŚC itself: that it seeks to elevate Telugu language to the heights achieved by prestigious classical languages and, therefore, by extension, assumes its prestige as well. “Poetry is for universal welfare,” the first sūtra (1.1) asserts, and “poetry without blemishes consists of a combination of speech and meaning” (1.2). Then the author explains that poetry “leads to rasa” (1.3) and that the goal of poets is to “achieve rasa that is brought about in an appropriate way” (1.4). Finally, the author of the ĀŚC makes his strongest statements: poetry is not of one kind nor does it belong to the domain of any single language but only that people “hold in high regard their own poetry” (1.5).

While Sanskrit and Prakrit poetry – never connected to a specific region – is respected and valued, the author explains that it is natural for people to respond to literature that speaks to their place and their local culture. By extension, one may correlate that just as Telugu poetry intimately relates to Sanskrit and Prakrit poetry in shared inheritance of literary topoi and cultural themes, so too should its grammar draw from its inheritance while still identifying those aspects that make it linguistically (phonologically and morphologically) distinctive. To strike a balance in this perceived continuum, however, the Telugu has to destabilize Sanskrit and Prakrit as sole domains of literary expression. The emergence of metalinguistic technologies, whether through appropriation or innovation, often births new grammars that, in turn, simultaneously subject the language to undergo grammaticization and literarization. The ĀŚC's descriptions of Telugu through Sanskrit reveal the language to be both like and unlike Sanskrit and, therefore, as able to function in both a classical and vernacular space.

Significantly, however, the ĀŚC seems to be aware that, while the techniques and presuppositions of Sanskrit and Prakrit grammatical analysis condition its own foundational project, it is by no means necessary that these techniques overwhelm or misrepresent the true of object of the work, which is to describe how Telugu's features operate. The ĀŚC, again, may superficially appear to be offering Sanskrit and Prakrit-based grammatical prescriptions for understanding how Telugu linguistic forms operate, but the plasticity of the sūtras in their formulation and the additions made to accommodate distinctive features of the Telugu seem to quell any suspicions that the Telugu's distinctiveness is being compromised to validate non-Telugu grammatical prescriptions.


  • Ananthanarayana, H.S., 2017. “Pāṇinian Model for Describing Telugu in Bālavyākaraṇamu”, International Journal of Dravidian Linguistics Vol. 46 No. 1, 70-85. [Google Scholar]
  • Andronov, Mikhail S., 2003. A Comparative Grammar of the Dravidian Languages, Wiesbaden, Otto Harrassowitz Verlag. [Google Scholar]
  • Aussant, Émilie, 2012. “Pāṇinian features of the oldest known Malayālam description”, Cardona, George, Aklujkar, Ashok and Ogawa, Hideyo (eds.), Studies in Sanskrit Grammars, Proceedings of the Vyākaraṇa Section of the 14th World Sanskrit Conference, New Delhi, D.K. Printworld, 87-101. [Google Scholar]
  • Burnell, Arthur C., 1976 [1875]. On the Aindra School of Sanskrit Grammarians, Varanasi, Bharat-Bharati. [Google Scholar]
  • Kahrs, Eivind, 1992. “What is a Tadbhava Word”, Indo-Iranian Journal 35, 225-249. [Google Scholar]
  • Monius, Anne, 2013. “‘Sanskrit is the Mother of All Tamil Words’: Further Thoughts on the Vīracōḻiyam and its Commentary”, Buddhism Among Tamils, Part 3, Uppsala, Acta Universitatis Upsaliensis, Historia Religionum 32, 103-129. [Google Scholar]
  • Sundaram, Rallapalli V.S. and Patel, Deven M., 2016. Āndhraśabdacintāmaṇi: A Grammar of the Telugu Language in Sanskrit (Edition and Translation with Notes and Introduction), Mysore, Central Institute of Indian Languages. [Google Scholar]
  • Verhagen, Pieter C., 2001. A History of Sanskrit Grammatical Literature in Tibet Volume 2, Leiden, Brill. [Google Scholar]


There are other such examples of Kannada and Malayalam grammars composed in Sanskrit, the Karṇāṭabhāṣabhūṣanam (12-13th century), a Sanskrit grammar of Kannada, and the Līlātilakam, a Sanskrit grammar of the mediaeval literary language of Kerala. For a technical analysis of Pāṇinian features in the Līlātilakam's description of medieval-era Malayalam, see Aussant 2012, p. 87-101.


Nannaya is considered the first Telugu poet and author of the grand Mahābhāratamu, a version of the first two and a half sections or parva-s of the Sanskrit Mahābhārata. Because of its association with Nannaya, therefore, Telugu tradition offers a series of alternative names of the Āndhraśabdacintāmaṇi that are based on the name or epithets associated with the poet: Nannayabhaṭṭīyamu, Vāganuśāsanīyamu, Śabdānuśāsanīyamu, Prakriyākaumudi, Āndhrakaumudi, and Āndhravyākaraṇamu.


The āryā meter has 12 mātrās, or measures (or “mora”), in the first and third pādas or verse-quarters, 18 mātrās in the second pāda, and 15 mātrās in the fourth pāda.


In a recent article, H.S. Ananthanarayana (2017, p. 70-85) presents Cinnayasūri's various textual iterations for translating Pāṇinian methods to Telugu grammar. Although Ananthanarayana does identify the Āndhraśabdacintāmaṇi as an influential source for Cinnaya's work, he does not identify specific sites of borrowing nor the significance of these borrowings. For example, sūtras 1.37 and 1.38 of the ĀŚC, which introduce the technical terms paruṣa (“hard” unvoiced stops) and sarala (“soft” voiced stops) are reproduced in Telugu virtually verbatim in the Bālavyākaraṇamu. Cinnaya's identification of Telugu as a vikṛti (“modification”) and Sanskrit/Prakrit as prakṛti (“original source”), which Ananthanarayana understands as a mistaken notion on the part of Cinnaya (p. 72), is also a clear borrowing from the ĀŚC's understanding of the relationship between the three languages.


Unlike Sanskrit, all sandhi in Telugu is marked by phonological loss occasioned by such coalescence (lopa-sandhi).


The introduction of the technical term varṇaka (sūtra 1.29) is a distinct innovation of this grammar, referring to suffixes (such as the first-case nominal suffixes −du, −mu, −vu, −lu) and equivalent to the Pāṇinian term pratyaya.


Hemacandra explains that Prākrit has forty letters (Ṥabdānuśāsana 8.1), ten less than the fifty he identifies for Sanskrit (in Ṥabdānuśāsana 1.1.4): ṛḷ-varṇau hrasva-dīrghā-vaijādyāv anunāsikau śa-ṣā vete daśonās syuḥ prākṛtoktiṣu sarvathā. The ten missing letters are: ṛ, ṝ, ḷ, ḹ, ai, au, ṅ, ñ, śa, ṣa.


In an illuminating discussion on the subject, offering an etymological analysis of the term tadbhava, often translated as “derived from it (Sanskrit or, often in the case of regional words, Prakrit),” Kahrs subtly argues that the term tadbhava be understood as meaning “located” in Sanskrit (or Prakrit) and not “derived” from it. See Kahrs 1992, p. 226-237 for specific details of the argument.


For a more elaborate discussion of the term siddha, refer below to the discussion on anusvāra, in the context of sūtra 1.32.


Vararuci's fifty varṇa for Sanskrit and Prakrit [dvidhāka ye cō'nusvārō visargaḥ śōdaśasvarāḥ sparśāntaḥstha-lōṣmānaśca catustriṃśaddhalaḥ smṛtāḥ] are comparable to the Kātantra's list, which does not mention anusvāra, visarga, or ḷa.


This terminology of calling vowels prāṇa and consonants prāṇi is peculiar to Dravidian linguistics. Consonants form syllables only when accompanied with vowels (prāṇa or “life-breaths”); otherwise, they are “lifeless,” as it were.


The ardhānusvāra may or may not have had any phonetic value historically. In modern Telugu, it only appears in the script and is not read with nasalization. Even though it has no phonetic value now, it has historical value because in the ancient epics and in Old Telugu literature, the ardhānusvāra appears to be fixed (see Sundaram and Patel 2016, p. 13-15).


Pāṇini does not make a two-fold delineation (dvividhā) for long and short vowels. For Pāṇini, e and o are long vowels; they have no short counterparts. In this sūtra, the author takes both from Pāṇini (in the use of the it-saṃjñā) and the Kātantra. So, i, u, and e (it, ut, et) are explained as both long and short. There is dvividhā for it and ut but there is no way to explain e and o, as both long and short. Thus the author gives et and ot, signifying that dvividhā applies to both. He takes Pāṇini's technical term (ac) but redeploys it to describe what happens in Telugu.


See above, fn. 4.


While admitting that the only extant usage he can find for the term tajja (lit. “born of it [i.e., Sanskrit]) comes from the twelfth-century Jain author Vāgbhaṭa's poetics text Vāgbhaṭālaṅkāra, Kahrs accepts the term as a synonym for the more common tadbhava (“located in it”), which refers to words that are not identical with corresponding Sanskrit words but rather born from them. See Kahrs 1992, p. 230.


Tātambhaṭṭu's fifteenth-century Kavicintāmaṇi (or, alternatively, Kavilōkacintāmaṇi, Sukavicintāmaṇi, Kāvyānuśāsanamu, and Kāvyacintāmaṇi) acknowledges only the first three types of Telugu words, subsuming grāmya into deśya.


Anne Monius (2013, p. 106, fn. 12) attributes Caṇmukam Piḷḷai (“Vīracōḻiyam: The Earliest Contrastive-Transfer Grammar in Tamil?”) with this suggestion.