Puṣpikā: Tracing Ancient India Through Texts and Traditions: Contributions to Current Research in Indology, Volume 1

It is perhaps commonplace to say that India is one of the world's richest and most enticing cultures. One thousand years have passed since Albiruni, arguably the first "Indologist", wrote his outsider's account of the subcontinent and two hundred years have passed since the inception of Western Indology. And yet, what this monumental scholarship has achieved is still outweighed by the huge tracts of terra incognita: thousands of works lacking scholarly attention and even more manuscripts which still await careful study whilst decaying in the unforgiving Indian climate. In September 2009 young researchers and graduate students in this field came together to present their cutting-edge work at the first International Indology Graduate Research Symposium, which was held at Oxford University. This volume, the first in a new series which will publish the proceedings of the Symposium, will make important contributions to the study of the classical civilisation of the Indian sub-continent. The series, edited by Nina Mirnig, Péter-Dániel Szántó and Michael Williams, will strive to cover a wide range of subjects reaching from literature, religion, philosophy, ritual and grammar to social history, with the aim that the research published will not only enrich the field of classical Indology but eventually also contribute to the studies of history and anthropology of India and Indianised Central and South-East Asia.

• Language: English
• Publisher: Oxbow Books
• Release date: Dec 23, 2013
• ISBN: 9781782970422

Book preview

One

Defining the Svara Bearing Unit in the śikṣāvedāṅga literature: Unmasking a veiled debate

Giovanni Ciotti*

§ 1 Defining the framework

A long-standing unspoken debate within the Sanskrit grammatical tradition concerns the definition of the nature of the Svara¹ Bearing Unit [SBU], i.e. which part of a word bears the svara. The very label of SBU does not in fact correspond to any specific Sanskrit term, but is a caique - here used for the first time - from the highly controversial concept of Accent Bearing Unit [ABU]. The latter has been at the centre of Western Linguistic speculation at least since the 1930s,² and its elaboration has substantially contributed to the reshaping of the overall interpretation of the architecture of the phonological component of grammar.³ Through the survey of various treatises, it will be shown that the Sanskrit grammarians have also been engaged in an effort to describe the characteristics of a specific set of linguistic entities and their relation to the svaras, a speculation which echoes the quest for the definition of the ABU and which has led to the formulation of different views.

However, before moving any further, it is essential to say a few words about the similarity that is postulated here between the SBU and the ABU. The very nature of the language speculation is in fact intrinsically manifold: it depends on the point of view of its composer, on the tools - both intellectual and technical - which have been used in its formulation, and on the aims which have been pursued.⁴ As a consequence, it should be asked whether the topic that will be treated here represents or does not represent the same problem for the Sanskrit and the Western traditions.

§ 1.1 Sanskrit as a common ground

As with any sort of language speculation, including both the Sanskrit and the Western linguistic traditions, the aim consists in formulating an interpretative and descriptive representation of its object by gathering general observations from a variety of (epi-)phenomena, i.e. by representing its grammar. Since this can be done either through comparisons with different languages, or through the analysis of a single language, although the Sanskrit grammatical tradition has no comparative interests, it is possible to compare its methodologies and achievements with those of the Western Linguistic tradition.⁵

Therefore, Sanskrit itself provides the common ground in which the two traditions can operate: on the one hand describing the characteristics of this language is the very aim of the Indian grammatical tradition and, on the other hand, Sanskrit represents – as any other language might do – a suitable field for testing the validity of the theoretical models provided by Western Linguistics.

Furthermore, although quite problematic in its definition, in particular within the Western tradition, the object under investigation is basically one, namely accentuation. As it will be shown in § 3, it is possible to state that the two traditions have gathered the same general observations while dealing with articulatory phenomena such as the modulation of the vibration of the vocal cords. A remarkable consequence of this common understanding can be seen in the fact that, according to both traditions, it is necessary to split words into minor subunits and to assign to them the property of being bearing units. In fact, just like some strands of the Western Linguistic tradition have divided words into minor units, i.e. units different from morphemes, like segments, syllables, or, more generally, prosodic constituents (see § 4), in the same way, the Sanskrit grammarians have dealt with varṇas and akṣaras in order to understand which part of a pada (word) bears the svaras (see § 5).

§1.2 Phonetics and phonology in the Sanskrit tradition

However, what seems to represent a common problem to which a similar answer has been provided should actually be embedded in two remarkably distinct systems concerning the overall architecture of (Sanskrit) grammar. As a matter of fact, the frameworks in which Sanskrit accentuation, or Sanskrit svaras (cf. § 3), have been analyzed are remarkably different from those of the Western tradition, and it would be rather naive to start our comparison without clarifying this point. In fact, Western Linguistics distinguishes two separate modules concerning how the grammatical architecture of a language deals with the sounds of the language itself, namely the phonetic component and the phonological component,⁶ whereas the Sanskrit tradition never operates – at least manifestly – within such a formal distinction.

Although within Western Linguistics the debate about where to draw the precise limit between phonetics and phonology is still open (cf. Purnell [2009]), it is possible to define phonetics as the study of the articulation and perception of the sounds that human beings use in speech, whereas phonology consists in the study of the sound patterns characterizing the overall properties of contrastive sound inventories, of the distribution of sounds and of their variable realization in different contexts (alternations) (cf. Blevins [2009: 325]). In other words, phonology studies the phenomena concerning the sound repertoire of a language while they are used in speech production.

It is particularly important here to highlight a peculiar characteristic of the phonological constituent as formulated by many strands of the Western Linguistic speculation: phonology may contain abstract entities. The postulation of these entities is justified by the urge to establish an interface with the cognitive apparatus of the brain, along with the principle of descriptive economy which has led to the formalization of the phonetics/phonology dichotomy itself. According to Vaux and Wolfe [2009: 131], by postulating highly abstract formal entities such as syllables […], we begin to bring order to a vast array of seemingly disparate facts that would otherwise remain unconnected and unhelpful.⁷

On the other hand, the Sanskrit grammatical tradition displays, in more or less systematic arrangements according to each treatise, the description of the articulation of sounds and the list of phenomena pertaining to them (in the sense of their contextual distribution, i.e. sandhi). Even when it seems to formalize entities bigger than a single sound (e.g. the akṣara, broadly an entity made of a vowel plus a consonant), it never openly states that they may have no articulatory reality, nor are they said to represent any kind of cognitive entity.⁸

Therefore, it should be expected that any kind of comparison between the speculations made by the Western and the Sanskrit traditions will be affected by this radical difference.⁹

§ 2 Defining the sources

When I refer to the ‘Sanskrit grammatical tradition’, I mean here the huge textual production embedded in the literary label called vedāṅga. This corpus of six ancillary sciences was initially developed in ancient India in order to study and preserve the Vedic scripture and knowledge. Among these six sciences, vyākaraṇa and śikṣā pertain to the speculation concerning the Sanskrit grammar.

The first work testifying to a vyākaraṇa (lit. discrimination) kind of speculation have been lost, and the most ancient known text we possess is Pānini’s Aṣtādhyāyī (The Eight Chapters, ca. 6th-4th century B.C.E.), a sample of an already fully developed degree of grammatical speculation. Roughly, the whole of the subsequent vyākaraṇa literature is constituted by commentaries on the Aṣṭādhyāyī: those produced by the so-called pāṇinīyas, i.e. the grammarians who followed more strictly Pāṇini’s teachings – in particular Kātyāyana (ca. 3rd century B.C.E.) and Patañjali (ca. 2nd century B.C.E.) – and those produced by the non-pāṇinīyas. A definition of vyākaraṇa through Western Linguistic terms would state that it is the science studying the morpho-phonological, morphological, and morphosyntactic components of the language. Included in the pāṇinīya tradition, although not fully in line with it (cf. § 3), we find Śāntanava’s Phiṭsūtra (Aphorisms on [the accentuation of] the nominal bases),¹⁰ a treatise about the principles according to which the positions of the svaras are assigned within the word.

The śikṣāvedāṅga literature includes two categories of texts: the prātiśākhyas ([belonging to] each [Vedic] school), from ca. 500 to 150 B.C.E.,¹¹ and the śikṣās (teachings), ca. 11th-15th century C.E.¹² In Western Linguistic terms, whereas every prātiśākhya deals systematically with what we could define as the phonetic and phonological aspects of specific subsets of the Vedic literature called Saṃhitās, the śikṣās are mostly short monographs about specific phonetic and phonological topics, which seem to pertain to Sanskrit grammar in general rather than to a small literary corpus. Both genres are usually attributed to specific Vedic families or schools, most of which remain almost completely unknown except for their names.¹³

Hereafter, the vyākaraṇa speculation will be investigated to trace a parallel between the definition of svara and that of accent (see § 3), whereas the śikṣāvedāṅga speculation will be used in the discussion concerning the definition of the SBU (see § 5). In particular, the following texts will be referred to:

(a)  Ṛgvedaprātiśākhya: prātiŝākhya of the Ṛgvedasamhitā;

(b)  Vājasaneyiprātiśākhya: prātiŝākhya of the Śuklayajurvedasaṃhitā;

(c)  Śaunakīyā Caturādhyāyikā: prātiŝākhya of the Atharvavedasaṃhitā;

(d)  Pāṇinīyaśikṣā: the most renowned among the śikṣās, traditionally not attached to any specific Vedic school;

(e)  Yājñavalkyaśikṣā: śikṣā of the Śuklayajurvedasamhitā;

(f)  Nāradīyaśikṣā: śikṣā of the Sāmavedasaṃhitā;

(g)  Lomaśīśikṣā: śikṣā of the Sāmavedasaṃhitā.¹⁴

§ 3 Defining a prominence

Providing a definition for the notion of accent has been one of the most challenging tasks Western Linguistic scholars have been engaged in. Here, I will present a concise summary of the most recent ideas concerning this notion (cf., for instance, Cairns [2009], Hayes [1995], and Idsardi [2009]). Nowadays, the majority of scholars believe that accent represents a syntagmatic prominence – realized by various phonetic means –, i.e. a prominence characterizing one out of a certain number of constituents proper to a specific subdivision of the spoken chain, corresponding to the word in the case of the lexical accent.¹⁵

Do the svaras correspond to this notion? There is an overall agreement among the Indian treatises in recognizing three main types of svaras, namely udātta (raised), anudātta ("non-udātta") and svarita (literally sounded)¹⁶ – the latter being defined as the combination of an udātta plus an anudātta.¹⁷

As Pāṇini clearly states, there is only one udātta for each pada (word):

anudāttam padam ekavarjam || 6.1.158 ||

"A word is without udātta with the exception of one."

Every udātta is then followed by a svarita, and these two are surrounded by a variable number of anudāttas. Therefore, the translation of the term svara as accent is rather inaccurate as it does not properly fit the phonological theory developed by Western linguistics. In this sense, only the udātta can be said to be the accent – i.e. the prominence – whereas the anudātta simply represents the absence of accent.

On the other hand, another possible translation of anudātta is unraised so that its meaning would refer to its phonetic characteristics: in this sense the ambiguity of the translation works to our disadvantage, making any classificatory effort concerning the theoretical framework of the Indian tradition difficult to define. In fact, it is not possible to discern whether the terminology concerning the svaras is strictly phonetic-based, or possessed of a ‘phonological flavour’ in the sense that it provides a piece of information concerning the distribution of the accent within each word.¹⁸

Therefore, a suitable translation for svara is pitch modulation, a translation which renders its etymological meaning more closely. It should be noticed that the term svara is used also to indicate the vocalic sounds. The consequences of this ambiguity will be shown in § 5.

However, at present, it is possible to state that both Western Linguistics and the Indian tradition recognize the existence of the same phenomenon – i.e. the prominence – labeling it either as accent or as udātta, respectively. But through which representational model do these two traditions outline the same phenomenon? In other words, how do they define the concept of prominence?

From a phonetic point of view, Western Linguistics describes the nature of this prominence as parasitic (cf. Hayes [1995: 7]): its articulation can in fact be realized through quite a large variety of articulatory means used for other purposes at the same time. Typologically, two kinds of accent have been postulated: stress accent and non-stress accent (cf. Beckman [1986]). The former is realized by a combination of more than one phonetic means, usually a combination of the vibration of the vocal cords (whose perceptual correlate is ‘pitch’), duration, intensity (whose perceptual correlate is ‘loudness’), and variation of segmental quality (for instance, in English), whereas the latter is mainly realized by a faster vibration of the vocal cords (for instance, in Japanese).¹⁹ In this sense, the etymological meanings of the three svara names and their comparison with musical notes – quite common within the Sanskrit grammatical tradition –²⁰ is at the base of the communis opinio that according to the Indian tradition, pitch is the main phonetic characteristic of Vedic accent. Western Linguistics – in particular Historical Linguistics according to some of its strands – reaches the same conclusion through the comparison of the svara system with the accentual characteristics of other Indo-European languages (cf. Clackson [2007: 75-8]) and through a particular interpretation of the nature of the svarita. In fact, a well-known property of tone languages, i.e. languages in which pitch is used for lexical oppositions (for instance, Mandarin Chinese), is that pitch can spread its characteristic either backward, merging then with the preceding low tone and realizing a rising tone, or forwards, merging in this case with the following low tone and producing a falling tone. Once we take the traditional definition of svarita into account as well as the constraint governing its position within the word, it seems therefore plausible to define the svarita as a falling pitch modulation, resulting from the merging of an udātta and an anudātta (cf. Halle [1997: 286-7, fn. 9]).

On the other hand, phonology provides a different approach to the concept of prominence. At present, although Hyman’s [2009] brilliant article demonstrates that our understanding of the phonology of the accent is far from having reached a satisfactory stage,²¹ the two main phonological typologies according to which accent is placed within a word are considered to be the rhythmic accent and the morphological accent (cf. Hayes [1995: 31-3]). The former represents the tendency proper to all languages to arrange its elements in rhythmic patterns, whereas the latter represents the outcome of a complex interaction of phonemic, morphological or syntactical contrasts and structures, commonly known as morphological accent. Vedic accent is usually assigned to the latter typology: its position within the word would depend in fact on the accentual properties of the morphemes composing the word. During the derivational process from the underlying representation and through various cycles, the position of the accent is adjusted according to the specific properties of the morphemes, i.e. for instance accentuated vs. non-accentuated (cf. Halle and Vergnaud [1987b: 57-61]).²² In keeping with this interpretation, Pāṇini assigns to each suffix the property of modifying the position of the accent within the word.²³ Unfortunately, this represents a further complication: from the Western point of view, such operations do not simply happen in the phonological component but in the morphophonological one,²⁴ i.e. they – understood either as rules or as constraints – pertain at the same time to the domains of sounds arrangement and of word formation.

On the other hand, a different explanation for the position of the prominence is provided by Śāntanava’s Phiṭsūtra. According to this treatise, the position of udātta does not depend on the property of the morphemes composing a specific pada, but on the lexical properties of the pada itself, either semantic or phonetic (cf. Devasthali (ed.) [1967]).²⁵ This shows that the same topic can be treated in different ways also within the Indian grammatical tradition. Nevertheless, even in this framework, the position of the udātta depends to some extent on the abstract properties that are attributed to some sounds so that they govern the position of the accent.

The fact that the two Sanskrit grammarians, although from different points of view, provide general observations regarding the position of the accent within words, instead of simply compiling a list of words with the indication of the accent position, is somewhat reminiscent of the representational device of Western Linguistics which is phonology (or morphophonology). If this is admitted, it seems plausible to infer that, although the very idea of phonology is not found in the Sanskrit speculation, nevertheless some of the Sanskrit grammarians moved beyond the simple description of the articulation of sounds towards a certain degree of abstraction.²⁶ At the very basis of this step lies the principle of lāghava (economy) which, although devoid of any cognitive implication, leads the grammarians to provide general rules instead of lists of exceptions. In other words, lāghava is not the core of the (cognitive) plausibility of language description but rather of its effectiveness.²⁷ As a consequence, as Kiparsky [2009: 34] remarks, […] syntax, morphology, and phonology […] are […] emergent constellations of rules rather than predetermined components into which the description [of the grammar] is organized. Therefore, the whole speculation of the Sanskrit tradition is characterized by a tension between the assumptions based on articulatory evidence and the abstractions devised for the effectiveness of the description of the language. On the other hand, the fact that sometimes Western Linguistics and the Indian tradition reach similar conclusions is due to the application of the Ockham’s razor principle by the former for the sake of the cognitive plausibility of language description.

§ 4 Defining the Bearing Unit

Besides the definition of the nature of the accent, Western Linguistics has also been engaged in the definition of what element of the word bears the accent, i.e. occupies the prominent position. This task has turned out to be remarkably challenging, in particular because it implies an interpretative analysis of the available phonetic data characterized by a high level of abstraction.

From a phonetic point of view, accent may generally be defined as multisegmental: specific combinations of sounds within the same syllable allow a better perception, in fact, of its accentual characteristics. For instance, the vibration of the vocal cords, which is the primary characteristic of the Vedic accent, is more easily heard when the consonants following the vowel are sonorants:²⁸ in fact, to each fundamental frequency produced by the vibration of the vocal cords during the articulation of voiced sounds, corresponds a series of harmonic frequencies, which are refractions of the fundamental one within the articulatory channel. The lower harmonics are particularly intense in the sonorant segments, whereas they are absent – since the speaker does not even vibrate the vocal cords – in the articulation of unvoiced segments (cf. Gordon [2006: 85-90]).

On the contrary, the ABU has proved to be a powerful abstraction used to explain various phonological phenomena and whose very definition – still under debate (cf. Cairns [2009] and Raimy and Cairns [2009b]) – has deeply contributed to redefining the shape of the overall architecture of phonology. The ABU is nowadays mainly defined either as a single segment – namely a phoneme or a slot within the word structure, depending on the different theoretical frameworks – or as a syllable.²⁹

In the first case (cf. Halle and Vergnaud [1987a]), syllable and accent information belong on orthogonal planes where segments are arrayed along the line defined by their intersection. The syllabic plane contains the syllables, which are the domains for specific phonological operations, whereas the metrical plane contains the so-called metrical feet, i.e. sequences of segments, whose shapes depend on the specific accentuation properties of each language. In other words, the metrical plane represents the locus for the application of rhythmic principles either with or without morphological constraints.

In the second case (cf. Nespor and Vogel [1986]), the ABU is usually the syllable, understood as the prosodic constituent – phonological subunit – of the word. Each word is in fact said to possess a prosodic structure, i.e. a phonological structure different from the morphological one, having the form of a node-tree, where lower constituents are governed by higher ones: segments are governed by syllables, syllables by metrical feet, and so on. Accentuation operates on the tree constituents – i.e. the syllables – along with other phonological operations. The syllabic option for the definition of the ABU is supported by a long tradition (cf. Hayes [1995: 49-50]), in particular because it leads to more straightforward generalizations as it allows one to account for the influence of the syllabic weight on the accent position within words.³⁰

To summarize, the ABU can be either a single segment of a word or some sort of sequence of its segments (e.g. a syllable) and it can occupy either an independent plane from the locus where the segments are arrayed to form a word, i.e. the metrical plane, or the same locus, i.e. the prosodic tree:³¹ does this complex multi-linear³² representation find any parallel within the Sanskrit tradition?

Because of the absence of a formalized distinction between phonetics and phonology, the Sanskrit tradition does not question the mono-linear approach to the sounds phenomena, i.e. it does not introduce different planes in the architecture of the grammar in which these phenomena operate. Nonetheless, the idea of singling out different non-morphological subunits of a word seems to find a straightforward parallel in the Sanskrit tradition. According to the common interpretation of the śikṣāvedāṅga views, in fact, varṇas and akṣaras are seen as subunits of a pada (word).³³ While the former are well-known entities corresponding to the speech-sounds, whether vowels or consonants,³⁴ akṣaras (literally imperishable, indivisible) are indeed more ambiguous entities. Allen [1953: 80] presents some instances from the primary literature in order to show the reasons which support an interpretation of the term akṣara as syllable, adding that it has also been ambiguously extended to indicate a simple vowel.

However, taking a closer look at the definitions provided for akṣara, it appears that often this entity corresponds to a simple vowel, either with or without consonant(s) (see § 5).³⁵ In other words, it seems that the notion of akṣara depends on a property of vowels, namely their being independently articulated sounds, contrary to the consonants that cannot be articulated without the support of a vowel.³⁶ This interpretation of the term akṣara seems to be remarkably consistent with the phonetic grounding which seems to be characteristic of most terms in the śikṣā tradition. Therefore, it also prevents the interpretation of akṣara as syllable.³⁷ Unfortunately, an exhaustive survey concerning the definition of the term akṣara is still missing: there are in fact contexts in which this term seems to be used in a sense close to that of syllable, in particular if one takes the Indian graphic systems into account (cf. Allen [1953: 80]).

Nonetheless, certain treatises define the SBU as a combination, to different extents, of either a vowel or an akṣara with one or more consonants. In this sense, like the ABU, also the SBU corresponds to a sequence of sounds within a given word.

Finally, it should be remarked that the Sanskrit treatises refer to the svaras (pitch modulations) and not only to the udātta when defining the accent bearing unit: this surely represents a more phonetically based approach to the language, focused on providing a definition of the pitch modulation bearers rather than based on the phonological notion of prominence. Unfortunately (see § 6), since the phonetic foundations for this speculation are not spelled out, a real understanding of the nature and aims of the Sanskrit speculation remains quite unclear.

§ 5 Defining the Svara Bearing Unit

It is now time to look at the various definitions of the SBU. Here, it should be noted that it is impossible to state which definition is the most widespread or authoritative, since a comprehensive survey of the topic throughout the śikṣāvedāṅga literature is still a desideratum.

§ 5.1 Pāṇinīyaśikṣā

In the following verses, the Pāninīyaśikṣā seems to define the SBU as a vowel:

so ’dīrṇo³⁸ mūrdhny abhihato vaktram āpadya mārutaḥ |

varṇāñ janayate teṣāṃ vibhāgaḥ pañcadhā sṃrtaḥ || 9 ||

svarataḥ kālataḥ sthānāt prayatnānupradānataḥ |

iti varṇavidaḥ prāhur nipuṇaṃ taṃ nibodhata || 10 ||

udāttaś cānudāttas ca svaritaś ca svarās trayaḥ |

hrasvo dīrghaḥ pluta iti kālato niyamā aci || 11 ||

"Once this breath, which has been upraised and has thumped at the head, enters the mouth, it produces the speech-sounds, whose classification is transmitted as fivefold (9), according to svara (pitch modulation), length, place of articulation, prayatna (primary effort), and anupradāna (additional effort). So say those who are learned in the sounds: listen to this carefully (10). The restrictions [i.e. the qualifications] concerning the vowels are the three svaras (pitch modulations), namely udātta, anudātta, and svarita, and [those] based on [their] length, namely short, long, and protracted (11)."³⁹

This translation follows Allen’s [1953: 83] interpretation: in his opinion, in fact, the three svaras (pitch modulations) seem to be characterizations of the vowels. Allen seems to read aci in connection to both svarāḥ (pitch modulations) and kālato (based on [their] length).⁴⁰ On the other hand, if one divides verse 11 into two independent sentences, the following translation would be possible: "The svaras (pitch modulations) are three: udātta, anudātta, and svarita. Short, long, and protracted are restrictions [i.e. qualifications] concerning the vowels, based on [their] length." According to this interpretation, the Pāṇimyaśikṣā would simply present a list containing the three svaras (pitch modulations), without giving any specific indication about the nature of their bearer.

Because of the ambiguity of the term svara, once divided into two parts, its first half could also be read as follows: "The svaras (vowels) are three: [articulated with] udātta, anudātta, and svaritaľ Although this interpretation seems to be the less plausible one – the definition of svara as vowel is already given in verse 4 – it does not really invalidate the meaning of the verse as given in the first translation above.

§ 5.2 Ṛgvedaprātiśākhya

By employing the notion of akṣara, the Ṛgvedaprātiśākhya exemplifies a different definition of the SBU:

udāttaś cānudāttaś ca svaritaś ca trayaḥ svarāḥ |

āyāmaviśrambhākṣapais ta ucyante || 3.1 ||

akṣarāśrayāḥ || 3.2 ||

"The svaras (pitch modulations) are three: udātta, anudātta, and svarita. They are articulated through stretching, relaxing, and carrying across (3.1). They are based on the akṣara (3.2)."

In this case, it seems reasonable to translate svaras as pitch modulations since sūtra 3.2 specifies that they are to be understood as characteristics of the akṣara which, in turn, is elsewhere defined as follows:

aṣṭau samānākṣarāṇy āditah || 1.1 ||

tataś catvāri saṃdhyakṣarāṇy uttarāṇi || 1.2 ||

ete svarāḥ || 1.3 || […]

anusvāro vyañjanaṃ cākṣarāṅgam || 1.22 ||

"The first eight [sounds of the varṇasamāmnāya] are simple akṣaras (1.1). Then [there are] four more combinatory akṣaras (1.2). These [akṣaras] are the svaras (vowels) (1.3).⁴¹ […] Both an anusvāra and a consonant are [to be considered as] an appendix to akṣara (1.22)."

In these sūtras, the SBU is depicted as an akṣara not in the sense of a syllable, i.e. a sequence of sounds in a word, but in the sense of a vowel, depicted as an independent articulatory sound, i.e. either with or without consonants attached to it.

§ 5.3 Nāradīyaśikṣā, Vājasaneyiprātiśākhya, and Yājñavalkyaśikṣā

Other treatises define the SBU in connection with the combination of svaras (vowels) and vyañjanas (consonants), without an explicit use of the term akṣara. For instance, the Nāradīyaśikṣā states:

svara uccaiḥ svaro nīcaḥ svaraḥ svarita eva ca |

vyañjanāny anuvartante yatra tiṣṭhati saḥ svaraḥ || 2.5.2 ||

svarapradhānaṃ traisvaryam ācāryāḥ pratijānate |

maṇivad vyañjanaṃ vidyāt sūtravac ca svaraṃ viduḥ || 2.5.3 ||

"A svara (pitch modulation) can be uccaiḥ (high), nīca (low), or svarita. Where there is that pitch, the consonants imitates [it] (2.5.2). The teachers assert that the traisvarya (triplet of pitch-modulated vowels) has the svara (vowel) as [its] basis. They believe that the consonant is like a bead, while the svara (vowel) is the string (2.5.3)."

Similarly, the Vājasaneyiprātiśākhya describes the synergy of vowels and consonants in the articulation of the pitch modulation:

vyañjanaṃ svareṇa sasvaram || 1.107 ||

"The consonant shares the svara (pitch modulation) with the svara (vowel)."

Likewise, the Yājñavalkyaśikṣā is also explicit in stating that:

svara uccaḥ svaro nīcaḥ svaraḥ svarita eva ca |

svarapradhānaṃ traisvaryyaṃ vyañjanaṃ tena sasvaram || 118 ||

"A svara (pitch modulation) can be high, low, and svarita. The traisvarya (triplet of pitch-modulated vowels) has the svara (vowel) [as its] basis. The consonant shares the svara (pitch modulation) with it."

Although, in other passages, both the Vājasaneyiprātiśākhya and the Yājñavalkyaśikṣā employ the term akṣara in connection with the term svara (pitch modulation), the remark on the capacity of consonants to be part of the SBU is made in connection with the vowels (svareṇa) and not the akṣaras.⁴² Therefore, it is possible to state that these three treatises provide an interpretation of the nature of the SBU as consisting of ’vowel plus consonant(s)’, i.e. as a sequence of sounds within the word.

§ 5.4 Śaunakīyā Caturādhyāyikā

As far as the definition of the SBU is concerned, the Śaunakīyā Caturādhyāyikā does not provide a final answer but presents two different views on the topic: on the one hand, a view corresponding to an akṣara-only solution and, on the other hand, a view according to which akṣaras and consonants bear the svaras (pitch modulations) together, i.e. a combination of the two solutions shown in §§ 5.2 and 5.3. First of all, in fact, this treatise states:

samānayame ’kṣaram uccair udāttam nīcair anudāttam ākṣiptaṃ svaritam || 1.1.16 || […] svaro ’kṣaram || 1.4.2 ||

"In the same vocal range, a high-pitched akṣara is udātta, a low-pitched is anudātta, and an ākṣipta (cast down from high to low) is svarita. (1.1.16) […] A vowel is an akṣara. (1.4.2)"

Thereafter, the Śaunakīyā Caturādhyāyikā presents two different opinions on the role consonants may have in bearing the svaras, but it does not state which one should be chosen:

asvarāṇi vyañjanāni || 3.3.26 ||

svaravantīty ānyatareyaḥ || 3.3.27 ||

"Consonants have no svaras (pitch modulations) (3.3.26). Ānyatareya says that [consonants] bear svaras. (3.3.27)"⁴³

Moreover, an interesting discussion – although not devoid of ambiguities – is reported in the following sūtras, where the Śaunakīyā Caturādhyāyikā agrees with the grammarian Sāṅkhamiti:

kim akṣarasya svaryamāṇasya svaryate || 3.3.31 ||

ardhaṃ hrasvasya pado dīrghasyety eke || 3.3.32 ||

sarvam iti śāṅkhamitri || 3.3.33 ||

akṣarasyaiṣā vidhā na vidyate yad dvisvarībhāvaḥ || 3.3.34 ||

"Of an akṣara having a svarya ("svarita"), which part bears it? (3.3.31)⁴⁴ Some say that it is [the first] half of a short vowel, and the [first] quarter of a long vowel (3.3.32). Śāṅkhamitri says that the entire [akṣara bears it] (3.3.33), [since] it is not the pattern of an akṣara to bear two svaras (pitch modulations) (3.3.34)."

In § 5.3, svarya has been translated as svara (pitch modulation) but, according to Deshpande [1997: 448-50], it means svarita.⁴⁵ This interpretation implies that this group of sūtras is to be connected to the well-known debate concerning which part of the S’BU is to be considered high-pitched and which part low-pitched in case the svara is a svarita. However, one should not overlook the fact that here the terms hrasva and dīrgha are mentioned in connection with akṣara, i.e. the SBU: these terms are usually employed to indicate the length of a vowel, so that akṣara here would simply mean vowel. Furthermore, if one agrees with Ānyatareya (see sūtra 3.3.27), the definition of the SBU advanced by the Śaunakīyā Caturādhyāyikā seems to be closer to the one ‘vowel+consonant(s)’ found in the treatises discussed above (see § 5.3).

§ 5.5 Lomaśīśikṣā

The last treatise that will be considered is the Lomaśīśikṣā which presents a couple of verses where the term akṣara is not mentioned but – unusually for the patterns of the śikṣāvedāṅga speculation – a pseudo-syllabic definition of the SBU seems to be upheld:

naitat svarati pūrvāṅge na parāṅge kadācana |

na vyañjane na mātrāyāṃ kathaṃ yogo vidhīyate || 3.1 ||

svarasyaiva tu pwrvārddhe vyañjanārddhārddhapaścime |

tayor arddhārddhasaṃyoge svaraṃ kuryād vicakṣaṇaḥ || 3.2 ||

"It [i.e. the svara (pitch modulation)] does not resound in the preceding part (pūrvāṅge) [of the vowel], nor in the following part (parāṅge) [of the vowel (?)],⁴⁶ nor in a [single] consonant, nor in a [single] mora: how is its production (yoga) [then] prescribed? (3.1)

The learned should pronounce the svara (pitch modulation) at the connection of the two halves of these two [sounds], [that is] the first half of the svara (vowel), [and] the last half (arddha-paścima) of the [preceding] half consonant (vyañjana-arddha). (3.2)"

The interpretation of the expression vyañjanārddhārddhapaścime is rather dubious: the proposed translation does not seem to match with the next half-verse, where we find the expression tayor arddhārddhasamyoge (the connection of the two halves of these two [sounds]). This, in fact, implies that the svara is found at the juncture between the two halves of the two sounds, rather than between the quarter of a consonant and the half of a vowel.⁴⁷ Regardless of the correct interpretation of this passage, what matters here in connection with the definition of the SBU is the term saṃyoga: the pitch modulation, in fact, emerges as something pertaining not only to a single sound, but to a combination of more than one sound. Therefore, the Lomaśīśikṣā seems to be in favour of a syllabic definition of the SBU, although not in the sense – proper to the Western Linguistic speculation – of a sequence of sounds, but of a sequence of two parts of two contiguous sounds.

According to two manuscripts preserved at the University Library of Cambridge,⁴⁸ svaraṃ in verse 3.2 is to be read as svāraṃ, usually a synonym for svarita. If this variant were the correct one, it would mean that this couple of verses do not refer to all the three svaras but only to the svarita. In this case, the Lomaśīśikṣā would be presenting a view on what part of a word bears the high-pitched part of the svarita that is close to the Śaunakīyā Caturādhyāyikā (see § 5.4). In any case, this would not invalidate my interpretation of the verses since the notion of consonant is used in order to define the SBU.

§ 6 Open conclusions

The rich variety of viewpoints on the definition of the SBU expressed by the different treatises bears witness to ‘a lively speculation’ on this topic among different grammarians. To wit, the Śaunakīyā Caturādhyāyikā openly quotes two proper names, i.e. Ānyatareya and Śāṅkhamitri, whereas both the Pāṇinīyaśikṣā and the Nāradīyaśikṣā respectively speak of varṇavidaḥ (those who are learned in the sounds) and ācāryāḥ (teachers) using the plural.⁴⁹

However, the reasons behind such a variety of views remain unfortunately unknown, since the treatises simply state what part of the word bears the svaras without providing any kind of explanation. In particular, those treatises according to which the SBU consists in a sequence of sounds are the most intriguing from a theoretical point of view: they in fact presuppose some sort of multi-linear speculation. Two plausible motivations for the formulation of this last approach are here listed:

(a)  according to a strict phonetically-oriented explanation, it is possible that the grammarians understood that the articulation of the svaras, in particular of the udātta, were more audible when accompanied by voiced consonants (see § 3), and therefore extended this observation to consonants in general;

(b)  to define a sound as akṣara implies that a fundamental characteristic of the sound has been singled out, i.e. its independence. Therefore, it is possible that some grammarians thought that it was plausible to attribute to a sound seen in such a perspective other characteristics, such as the fact of bearing the svaras, whereas other grammarians simply attributed it to the vowels.

However, the main reasons preventing us to reach any definitive, or even plausible, conclusion concerning the motivations behind this speculation are that, on the one hand, the śikṣāvedāṅga genre – just as the vedāṅga literature in general – possesses an assertive character, and very little space is given to explanations and, on the other hand, this tradition lacks a well-developed commentarial literature.⁵⁰

Bibliography

Allen, William Sidney [1953] Phonetics in Ancient India, London-New York: Oxford University Press.

Beckman, Mary E. [1992] Evidence for Speech Rhythm across Languages in Y. Tohkura, E. Vatikiotis-Bateson & Y. Sagisaka (eds.) Speech Perception, Production and Linguistic Structure, pp. 457-463, Amsterdam: IOS Press.

Beckman, Mary E. [1986] Stress and Non-Stress Accent, Dordrecht: Foris.

Bhise, Usha R. (ed.) [1986] Nāradīyā śikṣā: With the Commentary of Bhatta Śobhākara, Poona: Bhandarkar Oriental Research Institute.

Blevins, Juliette [2009] Phonetic Explanations for Recurrent Sound Patterns: Diachronic or Synchronic? in Raimy, Eric and Charles E. Cairns (eds.) [2009a], pp. 325-36.

Cairns, Charles E. [2009] Phonological Representation and the Vaux-Wolfe Proposal in Raimy, Eric and Charles E. Cairns (eds.) [2009a], pp. 145-64.

Cardona, George [1976] Pāṇini: A Survey of Research, The Hague-Paris: Mouton.

Cardona, George [1988] Pāṇini: His Work and its Tradition, Delhi : Motilal Banarsidass.

Chomsky, Noam and Morris Halle [1968] The Sound Pattern of English, New York: Harper & Row.

Clackson, James [2007] Indo-European Linguistics: An Introduction, Cambridge: Cambridge University Press.

Clements, G.N. [2009] Does Sonority have a Phonetic Basis? in Raimy, Eric and Charles E. Cairns (eds.) [2009a], pp. 165-75.

De Lacy, Paul (ed.) [2007] The Cambridge Handbook of Phonology, Cambridge: Cambridge University Press.

Deshpande, Madhav M. (ed.) [1997] Śaunakīyā Caturādhyāyikā: A Prātiśākhya of the Śaunakīya Atharvaveda, with commentaries Caturādhyāyībhāṣya, Bhārgava-Bhāskara-Vṛtti and Pañcasandhi, London: Harvard University Press.

Deshpande, Madhav M. [2006] The role of linguistics in India society and education in Sylvain Auroux, E.F.K. Koerner, Hans-Josef Niederehe, and Kees Versteegh (ed. by) History of the Language Science: an International Handbook from the Beginnings to the Present, Vol. 1, pp. 173-177, Berlin-New York: Walter de Gruyter.

Devasthali, Govind Vinayak [1967] Phiṭsūtras of Śāntanava, Poona: University of Poona.

Garde, Paul [1968] L’Accent Paris: Presses universitaires de France.

Ghosh, Manmohan (ed.) [1938] Pāṇinīyaśikṣā, Calcutta: University of Calcutta. (Reprint: [1986])

Gordon, Matthew Kelly [2006] Syllable weight: phonetics, phonology, typology, New York-London: Routledge.

Hall, T.A. [2007] Segmental features in De Lacy, Paul (ed.) The Cambridge Handbook of Phonology, pp. 311-34, Cambridge: Cambridge University Press.

Halle, Morris [1997] On Stress and Accent in Indo-European in Language 73, pp. 275-313.

Halle, Morris and Jean-Roger Vergnaud [1987a] An Essay on Stress, Cambridge: MIT Press.

Halle, Morris and Jean-Roger Vergnaud [1987b] Stress and the Cycle in Linguistic Inquiry 18, pp. 45-84.

Harris, John [2007] Representation in De Lacy, Paul (ed.) The Cambridge Handbook of Phonology, pp. 119-37, Cambridge: Cambridge University Press.

Hayes, Bruce [1995] Metrical Stress Theory, Chicago-London: The University of Chicago Press.

Hyman, Larry M. [2009] How (not) to do phonological typology: the case of pitch-accent in Language Sciences 31: 213-38.

Jakobson, Roman [1931] Die Betonung und ihre Rolle in der Wort- und Syntagmaphonologie in Travaux du Cercle Linguistique de Prague 4, pp. 164-82, Prague: Jednota československych matematikû a fysikû.

Kanta, Surya (ed.) [1970] Ṛktantram: A Prātiśākhya of the Sāmaveda, Delhi: Meharchand Lachhmandas.

Kiparsky, Paul [2007] Pāṇini is slick, but he isn’t mean in Nagoya Studies in Indian Culture and Buddhism: Saṃbhāṣā 26, pp. 1-28.

Kiparsky, Paul [2009] On the Architecture of Pāṇini’s Grammar in Huet, Gérard, Amba Kulkarni, and Peter Scharf (eds.) Sanskrit Computational Linguistics: First and Second International Symposia, pp. 33-94, Berlin-New York: Springer.

Idsardi, William J. [2009] Calculating Metrical Structure in Raimy, Eric and Charles E. Cairns (eds.) [2009a], pp. 191-211.

Laver, John [1994] Principles of Phonetics, Cambridge: Cambridge University Press.

MacDonell, Arthur Anthony [1916] A Vedic Grammar for Students, Oxford: Oxford University Press.

Mishra, Vidhata [1972] A Critical Study of Sanskrit Phonetics, Varanasi: The Chowkhamba Sanskrit Series Office.

Mohanan, K.P. [1995] The Organization of the Grammar in Goldsmith, John (ed.) The Handbook of Phonological Theory, pp. 24-69, Blackwell: Cambridge (Mass.)-Oxford.

Nespor, Marina and Irene Vogel [1986] Prosodic Phonology, Dordrecht: Foris.

Parameswara Aithal, K. [1991] Veda-Lakṣaṇa Vedic Ancillary Literature: A Descriptive Bibliography, Stuttgart: Franz Steiner Verlag.

Pāṭhaka, Yugalakiṣora Vyasa (ed.) [1889-93] Śikṣasaṃgrahaḥ, A collection of Sikshás by Yájñavalkya and others, Varanasi: B.B. Das & Co.

Purnell, Thomas [2009] Phonetic Influence on Phonological Operations in Raimy, Eric and Charles E. Cairns (eds.) [2009a], pp. 337-54.

Raimy, Eric and Charles E. Cairns (eds.) [2009a] Contemporary Views on Architecture and Representations in Phonology, Cambridge: MIT Press.

Raimy, Eric and Charles E. Cairns [2009b] Architecture and Representation in Phonology in Raimy, Eric and Charles E. Cairns (eds.) [2009a], pp. 1-16.

Rastogī, Indu (ed.) [1967] Kātyāyanācārya-praṇītaṃ Śuklayajuḥ-prātiśākhyam: athavā, Vājasaneyi-prātiśākhyam, Varanasi: Caukhambā Saṃskrta Āphisa.

Scharfe, Helmut [1977] Grammatical Literature Vol. V, Fasc. 2, in Jan Gonda (ed.) History of Indian Literature, Wiesbaden: Otto Harrassowitz.

Shastri, Mangal Deva [1937] The Ṛgveda-prātiśākhya with the commentary of Uvata, Punjab Oriental Series 24, Lahore: Moti Lai Banarsi Das.

Trubetzkoy, Nikolai Sergeyevich [1939] Grundzuge der Phonologie in Travaux du Cercle Linguistique de Prague 7, Prague: [s.n.].

Varma, Siddheshvar [1929] Critical Studies in the Phonetic Observation of Indian Grammarians, London: Royal Asiatic Society.

Varma, Virendra Kumar (ed.) [1970] Ṛgveda-prātiśākhya: Uvaṭa-bhāshya-sahitam, The Banaras Hindu University Sanskrit Series 5, Varanasi: Kāśī Hindu Viśvavidyālaya Śodha-Prakāśana.

Vaux, Bert and Andrew Wolfe [2009] The Appendix in Raimy, Eric and Charles E. Cairns (eds.) [2009a], pp. 101-43.

*My thanks to Dr. Vincenzo Vergiani and Mr. Alastair Gornall for their valuable comments on earlier drafts. All faults remain mine alone.

¹On the definition of the term svara, see § 3.

²From Jakobson [1931] and Trubetzkoy [1939], through Garde [1968], to Chomsky and Halle [1968], accentuation has been dealt with by many scholars and by many different theoretical frameworks within the modern Western Linguistic tradition. In this article (see §§ 3 and 4), the label ‘Western Linguistics’ refers exclusively to the post-Sound Pattern of English (Chomsky and Halle [1968]) generative strands which are characterized, on the one hand, by the metrical and prosodic approaches to the treatment of accentuation and, on the other hand, by a modular approach to the architectural representation of the language. Cf. Beckman [1986], Hayes [1995], and Part III: Metrical Structure in Raimy and Cairns [2009a].

³lt should be remembered that, in Western Linguistic terms, ‘grammar’ means ‘Universal Grammar’, i.e. the universal cognitive mechanism which, according to the setting of its parameters, can produce all the possible shapes by which the architecture of a language can be characterized.

⁴Suffice to say that the Sanskrit grammatical tradition seems to have aimed, at the beginning of its history, to provide the Vedas with ancillary sciences (cf. § 2) used for preventing them from undergoing modifications, and, in a second stage, to provide the intellectual and political establishment with an unchanging symbol of prestige (communis opinio), along with a valid tool for the teaching of Sanskrit (cf. Deshpande [2006] and Kiparsky [2007]). However, Western Linguistics is interested in providing an elegant – in the sense that it should comply with the Ockham’s razor principle - representation of the functioning of the cognitive processes whose perceptible manifestation is language itself and whose parameters are the same for all human beings (Universal Grammar).

⁵In this sense, the difference between ‘linguistics’ and ‘grammar’ can be overcome.

⁶In this article the term ‘sound’ refers to the phonetic domain, whereas the term ‘segment’ – i.e. the representational device where a certain number of the articulatory features are said to coexist – refers to the phonological domain (cf. Harris [2007: 124-31]). The latter is preferred to the term ‘phoneme’.

⁷Commenting on Vaux and Wolfe [2009], Clements [2009: 165] states: […] not all phonological concepts have phonetic correlates. For example, though the syllable is an essential unit in phonology (and underlies many aspects of phonetic and prosodic patterning as well), it has no universally valid phonetic definition. This fact is not surprising once we recognize that the syllable is primarily a phonological construct, defined over sequences of discrete phonological segments rather than over phonetic primes as such. At this level of abstraction (which includes most of phonology), few constructs have direct phonetic definitions. Vaux and Wolfe rightly emphasize that the ultimate justification for such concepts depends on their success in bringing order to a vast array of seemingly disparate facts. The syllable does just that. About the importance of the syllable in connection with accentuation, see § 4.

⁸For the implications of the concept of lāghava (economy) in the Sanskrit tradition, see § 3.

⁹This field surely deserves to be widened. In particular, a comparison between pre-Saussurean Western Linguistics, i.e. pre-langue/parole dichotomy (e.g. the works of the first scholars engaged in the reconstruction of Proto-Indo-European) and the Sanskrit grammatical tradition could prove to be remarkably interesting, since the former still operates in a framework in which phonology has not yet been defined as a distinct component of grammar.

¹⁰Although dating the Phiṭsūtra is almost impossible, a sensible guess places it at the end of the first millennium B.C.E. (cf. Cardona [1976: 175-7]).

¹¹Also known as pārṣadas ([belonging to] an assembly). The two names of this genre clearly refer to the strict relation between these texts and the sacerdotal families, de facto the Vedic schools, in charge of preserving the Vedic texts.

¹²Any attempt to date these texts is necessarily tentative: for a detailed discussion on this topic cf. Scharfe [1977: 127-34, 176-7], who says The dating of most of these texts is next to impossible [p. 176], and Varma [1929]. The communis opinio that the prātiśākhyas are older than the śikṣās is based on the fact that the former are composed in sūtras – a style typical of the most ancient vedāṅga treatises –, whereas the latter are composed in verse or in prose.

¹³In a later stage (cf. Parameswara Aithal [1991]), the śikṣāvedāṅga texts are classified under the label of lakṣaṇa, i.e. the literary genre which comprises all the texts containing information about and instructions for the Vedic recitation.

¹⁴For the attribution of these treatises to their respective Vedic texts, cf. Mishra [1972].

¹⁵The definition of the notion of ‘word’ is not treated here. For a definition of the term ‘word’ in phonology, cf., for instance, Nespor and Vogel [1986: 109-144].

¹⁶It should be noted here that, although presented in the treatises as compulsory for the correct pronunciation of the language, the svaras have been probably lost in the translation from Vedic to Classical Sanskrit within an overall change of the accentuation pattern. Additionally, according to some treatises, a few more svara types should be added to this list, but they are probably variations of the main three, pertaining to specific recitational styles (cf. Deshpande [1997: 438]), or particular combinations of the basic svaras.

¹⁷Aṣṭādhyāyī 1.2.31 says samāhāraḥ svaritaḥ "svarita is the samāhāra (combination) [of high pitch and low pitch]."

¹⁸The translation of Aṣṭādhyāyī 6.1.158 should then be changed to: [the svaras] of a word are non-raised with the exception of one. The problem of the double interpretation of the term anudātta is already discussed in Patañjali’s Mahābhāṣya (cf. Cardona [1988: 442]).

¹⁹In addition, the accent can be enhanced by the fact that it triggers some phonological phenomena. A well-known example is the so-called ‘Verner law’ (cf. Clackson [2007: 75-6]).

²⁰For instance, cf. Pāṇinīyaśikṣa 12 and Nāradīśikṣa 1.8.8.

²¹Presumably, Hyman’s article will have a profound influence in reshaping the common knowledge Western Linguistics has developed in interpreting the phonetic correlates of accentuation. Since it is impossible to forecast the results of this reshaping, here are presented the most accepted phonetic and phonological frameworks Western Linguistics has developed so far.

²²In particular it is here the phonological metric module which interacts with morphology. For the implications of these model in the overall architecture of the linguistic representation, cf. Cairns [2009].

²³lt should be remembered that Pāṇini does not provide rules for attributing the accent to the underlying representation of nominal and verbal bases.

²⁴The definition of morpho-phonology within the Western linguistic tradition is not uncontroversial (cf., for instance, Mohanan [1995]).

²⁵For instance, it is not the neuter suffix which determines the position of the accent but it is the property of the word of being neuter that implies a particular position for the accent: athādiḥ prāk śakaṭeḥ || 2.1 || […] nabviṣayasyānisantasya || 2.3 ||"From here the initial syllable (is udātta) prior to śakaṭi. (2.1) […] Of (a word) necessarily neuter in gender (and) not ending in is, (the initial syllable is udātta) (2.3)" (Devasthali’s translation [1967: 56]). According to Śāntanava, it is not the suffix to be neuter but the lexical item, i.e. gender is not a morphological property.

²⁶Sandhi phenomena, which are not treated here, are encompassed within the same theoretical horizon.

²⁷In this sense, the fact that sometimes the grammarians do provide lists of exceptions should be seen as more in keeping with the idea of lāghava – namely more economical – than providing extremely complex derivational rules. For instance, cf. Aṣṭādhyāyī 2.2.38, 3.1.101, etc.

²⁸Sonorants are those sounds whose articulation is voiced, like vowels, nasals, liquids, etc. (cf. Hall [2007: 314–6]).

²⁹A more flexible approach to the definition of the ABU is found in Halle and Vergnaud [1987a]. According to them, the ABU is an ‘element X’, i.e. an element which can change from language to language: segments (both vowels and [+sonorant] consonants in particular positions), morae, segments in the rhyme, syllables, or lexically designated segments.

³⁰In case the accent is mainly characterized by the pitch, the mora has been said to be the ABU since, in certain languages (e.g. Lithuanian), accent can create oppositions within the same syllable. As Hayes [1995: 49-50] remarks, defining the ABU as the syllable works in particular for the so-called stress languages, whereas […] pitch languages […] in the generative phonology they can be treated as involving tonal representation, either in addition to or instead of a metrical representation, but cf. Hyman [2009] for the fact that ‘pitch language’ represents an improper phonological category.

³¹The nub of the issue is whether or not the syllable is the inviolable unit for bearing stress (e.g. Hayes 1995 and much recent works). If it were, one would expect syllables to nest neatly within feet; but if some languages were to employ vowels as the stress-bearing unit, then the possibility would exist that the constituents needed for stress might conflict with those needed for syllabic facts (Raimy and Cairns [2009b: 4]).

³²Borrowing a Saussurean and then Firthian terminology (cf. Garde [1968]), a multilinear representation of the language corresponds to the necessity of explaining the syntagmatic – vs. paradigmatic – properties of some phonological entities.

³³Although extremely controversial, mātrās (morae) could be added to this list, but they are not relevant to the present discussion.

³⁴Usually, after a general statement about the overall number of varṇas, which varies according to the various treatises, the Indian grammarians possibly list them according to their articulatory characteristics.

³⁵At least, as it will be shown in § 5, in connection with the definition of the SBU. akṣaras are in fact also classified as either laghu (light) or guru (heavy), i.e. according to a terminology which echoes the Western Linguistic phonological notion of syllabic weight (cf. Allen [1953: 85-7]).

³⁶This refers in particular to the stops: nasals and sibilants for instance can be articulated independently, but cannot be, in Sanskrit, syllable nuclei, i.e. they occupy a secondary place in the sonority scale (cf. Clements [2009]).

³⁷According to Western Linguistics, the phonetic definition of a syllable concerns mostly the internal distribution of its sonority, whereas its phonological definition mostly depends on which sound can occupy which position in the sonority scale. This is also the main reason for not translating akṣara as syllabic nucleus, i.e. the syllabic segment characterized by the higher degree of sonority.

³⁸For the peculiar sandhi of so ’dirṇo, cf. Ghosh [1938: 55].

³⁹All translations are mine.

⁴⁰Ac is a pratyāhāra (abbreviation) used in the pāṇinīya tradition to indicate the vowels.

⁴¹The varṇasamāmnāya (collection of sounds) – which is not presented in the Rgvedaprātiśākhya – contains the list of sounds whose articulation and properties are described in the text. The first twelve are the vowels, namely: a, ā, i, ī, u, ū, ṛ, and ṝ followed by e, ai, o, and au (the sound ḷ is probably regarded as an allophone of ṛ). Shastri [1937: 4] translates samānākṣara as monophthong and saṃdhyakṣara as diphthong.

⁴²The Vājasaneyiprātiśākhya in fact states that: svaro ’kṣaram || 1.99 \\ sahādyair vyañjanaiḥ || 1.100 || uttaraiś cāvasitaiḥ || 1.101 ||"An akṣara is a svara (vowel) (1.99), [and it is so] also with the preceding consonants (1.100) and the following ones in pause (1.101)." The Yājñavalkyaśikṣā also employs the term akṣara more than once, at least in one case in direct connection with the term udātta: udāttākṣarayor madhye bhaven nīcas tv avagrahaḥ ||84.b ||"[As in the previous examples,] if [it occurs] between two high-pitched akṣaras, the avagraha (?) should be low-pitched" (here the sense of the term avagraha is not at all clear).

⁴³For the interpretation of these sūtras, cf. Deshpande [1997: 447-8].

⁴⁴Here, it is worth noting that the verb svaryate – here translated as it bears – literally means it resounds or even, in this context, it is accented. This shows how, unlike Western Linguistics, each Sanskrit notion and term in the field of śikṣā tends to be grounded on articulation.

⁴⁵Here, if one took svarya as a synonym of svara (pitch accent), it would be difficult to find a plausible interpretation for these sūtras.

⁴⁶The term aṅga is rather obscure: the tentative translation proposed here (part [of the vowel]) echoes the debate