The present chapter is an attempt to summarize the material discussed in all the preceding chapters one by one.

CHAPTER 1

i) The linguistic characterization of Bangla syllables (cf. chap. 2),

ii) the identification of syllables as a functional unit in the verse language of Bangla (cf. chap. 4), and

iii) the limitations of segmental treatment in the field of Bangla phonology (cf. chap. 3)

- all these three factors provided the groundwork for a new framework, viz. the metrical framework, to enter the field of Bangla phonology.

Bangla, one of the eighteen national languages of India and the sole national language of Bangladesh, because of a wide geographical distribution, political and religious diversity, a rich literary tradition, a good speaker-strength and their deep attachment to the language, presents a wide range of dialectal variations and parallel standardization in Kolkata and Dhaka.

The present work deals with the Kolkata Standard Colloquial Bangla (KSCB), which has often had to face some sort of identity problem because since 1946 down to this day it had to experience the continuous and overwhelming political migration from Bangladesh to India. To avoid any language variation problems, which seem to be irrelevant to the main theme of this book, I restrict myself to the non-controversial data of KSCB.

Three models, viz. the concrete model of Kiparsky (1968b), the metrical model of Hogg and McCully (1987), both belonging to the generative tradition, and the Sen school model (developed from 1922 to 1986), have been used here.

The concrete model is one of the many offshoots of the post-SPE period questioning the abstractness at the underlying level of the classical theory of SPE.

The metrical model belongs to the non-linear sector of the post-SPE developments and it questions the segmental treatment of English stress as proposed in SPE.

The Sen school model presents a systematic metrical theory of Bangla verse language and is chronologically much older than the western theory of metrical phonology.

The present work follows the transcription convention of Ray et al. (1966).

CHAPTER 2

This chapter presents a treatment of the Bangla syllable as a phonetic, i.e. performance, unit as well as a phonological, i.e. structural, unit, rather than a mere intuitive unit.

On the basis of the two descriptive studies, viz. Sarkar (1986) and Mallik (1960) on Bangla syllables and Bangla consonant sequences respectively, we present here some 10 rules of syllabification and list some 16 canonical patterns resulting from them. Following the metrical conventions we divide these 16 canonical patterns into two types of syllable, viz. the heavy syllable with branching rime and the light syllable with non-branching rime, relevant for our purpose.

The quantity of these heavy and light syllables has been measured on two different levels, viz. phonetic and phonological.

On the one hand, two instrumental phonetic studies, viz. the Kymograph tracing by Hai (1964) and the digital Sonagraph reading by us, show that in Bangla the heavy syllables are longer than the light ones.

On the other hand, the evidence from the field of Bangla metrics (Sen 1974, 1986) shows that the phonological quantity of the structurally light syllables is 1, which is fixed as well as discrete; while that of the heavy syllables oscillates between 1 and 2, via 1½, a non-discrete reading, also.

An attempt to correlate these phonetic, i.e. physical, and phonological, i.e. psychological, realizations of syllable quantity raises a few theoretical issues, which are left open here, regarding i) the psychologically real moric phonology of the heavy syllables, ii) the impressionistic grouping of languages, viz. the syllable-timed and the stress-timed, and iii) the metrical style vs. the normal speed of speech in Bangla.

We then treat the predictable prominence or stress, another prosodic aspect of Bangla syllable, in accordance with the principles of metrical phonology (Hogg and McCully 1987). On the basis of the two very familiar phonological processes of Bangla, viz. vowel harmony and O-o alternation, we assume that word medially Oào in prosodically weak syllables. On the basis of the above plausible assumption and in terms of the three basic concepts of metrical phonology, viz. adjacent, alternating, and clashing, we formulate four rules to account for Bangla word stress rhythm.

Our formulation, on the one hand, agrees with a long attested process of Bangla pronunciation called bimorism / dimetrism / bisyllabism / loss of second vowel; and on the other hand, feeds the higher level phrasal intonational phonology of Hayes and Lahiri (1991).

CHAPTER 3

This chapter presents an analysis of the segmental phonology of verb morphology, a representative and substantial domain of standard colloquial Bangla that has drawn the notice of several phonologists in the linguistic literature. The current analysis is directly motivated by two earlier studies, viz. Dasgupta (1982) - the latest one done in the SPE framework, and Paul (1985) - a reaction to Dasgupta (1982).

The methodological framework of our analysis is that of Kiparsky (1968b), i.e. the concrete framework, which is a version of the standard theory of SPE modified in the following respects:

i) Neutralization device should obey either 'weak' or 'strong' alternation condition.

ii) Evaluation measure should be considered in terms of synchronic motivation and psychological reality.

iii) Exceptions should be treated in terms of rule-feature analysis.

Though Paul (1985) too claims to take the position of Kiparsky (1968b), her mechanisms result in two major types of inadequacy, viz. derivational inadequacy - as they fail to generate the Bangla verbal forms correctly, and theoretical inadequacy - as they deviate from the theoretical constraints.

Our mechanism is divided into two portions, viz. the underlying phonological forms and the rule component.

The underlying forms contain stems and suffixes, and are postulated mainly on the basis of Dasgupta (1982) and Paul (1985). The stems are chiefly of two types - non-causative and causative - both of which have monosyllabic verb roots with low or mid vowels. The non-causative stems are formed with or without the stem formative a; whereas the causative stems are formed with oa. In terms of their phonological shapes the verb stems are divided into 8 groups, viz. CV, CVC - where V=a; (C)V1CV, (C)V1CCV - where V1=a; CV, CVC - where V?a; and (C)V1CV, (C)V1CCV - where V1?a.

The underlying forms of 53 inflectional suffixes are listed in a tabular fashion.

The rule component consists of 8 phonological rules, viz. a-Mutation, Vocoid Raising, Suffix Truncation, Degemination, Minor i-Deletion, Semivowel Formation, Semivowel Deletion, and Glide Assimilation, and the ordering restrictions among them.

The validity of our mechanism has been established in terms of the derivation of actual verbal forms of some eight verb stems with eight different phonological shapes.

The handful of exceptions to the proposed mechanism are accounted for in terms of a few natural concepts, viz. word boundary, stem allomorphy, and suppletion, along with factual justification.

CHAPTER 4

This chapter reports one indigenous methodology of Bangla metrical analysis, viz. the Sen school methodology, mainly on the basis of the first chapter of Sen (1986), which, according to Sen himself, is a phonological characterization of Bangla metrical foot formation.

Each verse foot of Bangla starts with a stress and ends in a pause - of these two, pause plays a more important role. Grade 1, grade 2, grade 3, grade 4, and grade 5 - these five types of pause are available in Bangla poetry. They are arranged hierarchically and they indicate the boundaries of the syllable, the sub-foot, the foot, the clause, and the verse respectively.

Bangla verse metre presents two grades of stress, viz. the primary and the secondary, indicating the commencement of the initial and non-initial sub-feet of a foot respectively.

The elision of both the pause and the stress at expected places is treated as a rule exception. The stress may be elided depending on the elision of the immediately preceding pause.

The three well-defined metrical units of Bangla, viz. the syllable, the foot, and the verse, are identified in terms of the existence and elision of pause and stress.

Depending on the two types of segments, viz. the independent type consisting of complete vowels, and the dependent type consisting of one of the four vowel particles and any consonant, Bangla shows two types of syllable - the open syllables and the closed syllables ending in independent and dependent segments respectively.

Each metrical unit is ultimately quantified in terms of syllables. And the syllables are measured in terms of two types of unit, viz. the syllabic unit and the moric unit.

The distribution of these two types of quantity-unit results in three different metrical patterns in Bangla as follows:

i) Syllabic pattern: In accordance with the syllabic unit, in this pattern, each syllable is considered as one unit of a foot.

ii) Simple moric pattern: According to the moric unit, in this pattern, the open and the closed syllables of a foot are considered as monomoric and dimoric respectively.

iii) Composite pattern: This pattern employs the moric unit in its own characteristic way. Here the open syllables and the non-final closed syllables are considered as monomoric; whereas the final as well as the isolated closed syllables are considered as dimoric.

Tagore, in his voluminous poetic works, established the above three as the three distinctive metrical patterns in Bangla literature. Exceptions to these three metrical patterns are also evident in the Bangla verse literature.

CHAPTER 5

In terms of the domains of the predictable stress, verb morphology, and indigenous metrical studies the syllables appear to be quite versatile micro-units of Bangla phonology. Even the cross-linguistic evidence from English, Greek, Latin, German, Kannada etc. has established the syllable as a persistent natural class in the fields of phonology, morphology, and metrics.

This chapter presents an analysis of the structure of Bangla syllables in terms of the tools of the current western theory of metrical phonology, viz. sonority value, sonority hierarchy, syllable template, and ambisyllabicity.

In accordance with the sonority scale, as proposed in Hogg and McCully (1987), Bangla vowels, the more sonorous segments, are assigned the higher sonority values (8-10), and the other Bangla segments, the less sonorous ones, are assigned the lower sonority values (1-7).

In terms of the sonority hierarchy Bangla syllable structures obey the Sonority Sequencing Generalization (SSG) of Selkirk (1984b). It says that in a sequence of segments preceding or following a sonority peak, the sonority values increase as one nears the peak, and decrease as one moves away from the peak.

The syllable template (i.e. the abstract tree structure which all syllables have to fit in in order to be recognized as acceptable syllables in a particular language) for Bangla monosyllables is CCVVC. It may be arranged as follows:

s

O R

O2 O1 R1 R2 R3

C C V V C

Along with the sonority hierarchy and SSG, this structure obeys at least five general and nine specific template conditions in Bangla.

In polysyllabic words this syllable template determines the number of syllables within a word quite efficiently, though the determination of syllable boundary remains a problem as the intervocalic semivowels, word internal sth sequences etc. often appear to be ambiguous between two syllables. The problem is taken care of in terms of the rule of ambisyllabicity.

Agreeing with the phonotactic approach of syllable boundary determination that accommodates ambiguity and specific phonetic information, we accept the Principle of Maximal Onsets to be the underlying one as it agrees with the intrinsic spirit of Bangla syllabification quite satisfactorily. Next we assume that the rule of ambisyllabicity applies to these underlying structures and assigns the problem segments the membership of another, viz. the preceding, syllable also.

Such a treatment is justified in terms of the existence of ambisyllabic segments in Bangla.

The macro-units beyond the syllable, viz. the foot, the mot, and the word are not dealt with in the present study.