3.0. Introduction

The aim of the present chapter is to present a brief and partial sketch of the segmental phonology of Standard Colloquial Bangla. For that purpose we choose the area of verb morphology in Bangla because this particular sector occupies a major portion of colloquial Bangla phonology; and thus appears to be the most promising as well as representative domain for phonological investigations. This is not just our view; this particular domain has often drawn the attention of other phonologists also as the most interesting field of research. An earlier version of the observations of this chapter was published as a paper called Bangla Verb Morphology: the actual derivation in 1998 in Indian Linguistics vol. 59.

In the literature Bangla verb morphology has been discussed by several scholars and in different frameworks. Mention can be made of Chatterji (1926), Basu (1962), Beams (1970), Dasgupta (1970), Sarkar (1976), Dey (1979), Singh (1980), Dasgupta (1982), Paul (1985, 1986), and Bhattacharya (1993) among others.

As far as the methodological approach is concerned, Chatterji (1926), and Beams (1970) belong to the comparative philological period of linguistic research, while the works of Basu (1962), Dasgupta (1970), and Bhattacharya (1993) use the structuralist framework and those of Sarkar (1976), Dey (1979), Singh (1980), Dasgupta (1982), and Paul (1985, 1986) use the generative framework. I shall no more mention Paul (1986), i.e. Paul's paper on verbal morphology separately as it is identical with chapter 7 of her dissertation (1985).

Among these, Chatterji (1926), Basu (1962), Dasgupta (1970), and Bhattacharya (1993) have listed the verbal roots, classifying them into different classes, and have also listed the suffixes that are attached to verbal roots. They have not, however, dealt with the exact derivation of the verbal forms. The analyses done in the generative framework, especially Dey (1979), Singh (1980), Dasgupta (1982), and Paul (1985) go beyond traditional research and aim to formulate mechanisms for the derivation of the Bangla verbal forms.

Since my primary concern here is to investigate the processes of the native phonology of Standard Colloquial Bangla, in the present chapter I shall concentrate on the works done in the generative framework as such writings have discussed in considerable detail the phonological processes and constraints involved in the derivation of Bangla verb forms.

Among four such analyses a clear cut distinction can be made between Paul (1985) and the other authors in terms of their methodological positions. The works of Dey, Singh, and Dasgupta are based on the standard theory of phonology - that of SPE. The basis of Paul's analysis is a modified version of the standard theory of SPE; her modifications are in what she calls a 'concrete' direction.

The three analyses in the SPE framework are interrelated: Singh's paper is a critique and extension of Dey's work, and Dasgupta's article, responding to Singh, sets forth a tentative analysis as a basis for critical study.

As a reaction to Dasgupta (1982), Paul (1985), working in a somewhat different perspective, criticizes him and formulates her views on Bangla phonology.

However, none of these writings explicitly exhibits the derivation of Bangla verb forms in actual cases; they do not show how the mechanisms they propose really work. Though they explain their mechanisms in considerable detail, their failure to supplement these explanations with actual derivations is a major gap in the arguments for their mechanisms. Although Dasgupta (1980) did exhibit derivations of many verb forms to illustrate his proposals, the differences between his 1980 rule system and his 1982 rule system make it difficult to use his 1980 derivations when we seek to evaluate his 1982 mechanisms.

Accordingly, we must set ourselves, afresh, the tasks of

i) Formulating the necessary mechanisms of Bangla phonology, and

ii) Explicitly deriving actual verbal forms by exploiting the proposed mechanisms and thus establishing the validity of these mechanisms in a verifiable manner.

Section 1 will describe our methodological framework; section 2 will present the phonological mechanisms; section 3 will point out the differences between Paul (1985) and the current analysis; the derivation of the verbal forms of some 8 verb stems will be exhibited in section 4; exceptions will be dealt with in section 5, which will be followed by the concluding section.

3.1. Methodological Framework

Like Paul (1985), this chapter also takes the position of Kiparsky (1968b), which is a modified version of the standard theory of SPE in terms of the abstractness of the underlying representations.

The standard theory did not impose any restriction on the degree of abstractness of the underlying representations compared to the surface forms. Later, the question of abstractness vs. concreteness of underlying phonological structure became a major issue in the literature. On the other hand, there were natural generative phonologists like Stampe (1969), Vennemann (1971, 1972, 1974), Hooper (1976) etc. who argued against allowing any degree of abstractness in the underlying representation or UR. They rejected the entire model of standard theory, claiming it to be too powerful; instead, they advocated a heavily constrained theory which they said offered a more realistic model for natural languages. In practice, however, instead of yielding a concrete phonological system, the proponents of natural generative phonology often ended up merely listing some surface data, opening themselves to the charge of forgetting the classical arguments against structuralism.

Kiparsky (1968b, 1975) offered a via media in the debate of abstractness vs. concreteness. He proposed to modify the standard theory by supplementing it with constraints such as the following:

1. The device of neutralization should be used only in accordance with one of the two alternation conditions ('strong' and 'weak').

2. The evaluation measure should treat synchronic motivation rather than formal simplicity as the principal operating factor.

3. Exceptions to a rule should be treated in terms of rule-feature analysis, viz. marking them as [-rule n], a process that is as natural as the existence of exceptions in the language, instead of positing underlying segments of diacritic features and making the rule sensitive to them.

4. The psychological reality of the processes involved should be taken into account.

Given the above formulation of Kiparsky's framework, we may now ask: where do the works of Dasgupta (1982) and Paul (1985) stand? The rest of this section will discuss this issue briefly.

3.1.1. Dasgupta (1982)

As mentioned in the introduction, Dasgupta (1982) belongs to the standard theory framework. He deals with Bangla verb morphology in terms of the following assumptions:

1. Abstract segments in the UR, which often involve absolute neutralization.

Dasgupta prefers to account for the minority group of verbs, e.g. ga 'to sing' etc. which seem to be more directly describable as rule exceptions, in terms of abstract underlying segments and rule sensitivity to such segments. For example, Dasgupta (1982: 22) postulates gah etc. underlyingly on the basis of the forms like dohon 'milking'etc., which are from outside the domain of verbs. But in Bangla verbs, in fact, systematically reject any kind of intervocalic and final h-sounds, though VhV and Vh sequences are tolerated outside this particular domain, e.g. bah 'interjection', Dahuk 'a kind of bird', Saha 'a surname' etc.

Thus, in terms of the particular domain of verbs in Bangla the underlying h may be called an abstract segment that involves absolute neutralization, though the ramifications of such abstractness are yet to be decided in the theoretical literature.

2. Dasgupta handles the verbal derivations in terms of 17 rules as well as the ordering restrictions among them, as a result of which computation becomes longer. Such lengthy derivations, however, overestimate the mental abilities of the speaker and thus the psychological reality of a grammar involving such lengthy procedure becomes dubious.

However, none of the above procedures is allowed in the modified standard theory of Kiparsky.

3.1.2. Paul (1985)

Though it claims to take the concrete position of Kiparsky (1968b) and argues on the one hand against the standard theory approach (chap. 1, 1.2) and on the other hand against the NGP approach (chap. 1, 1.3), in its approach it does deviate from the principles of concrete approach, especially in the area of verb morphology, in following respects:

1. It posits abstract underlying segment. Paul (1985: 183) accepts Dasgupta's proposal of postulating a final h for the minority group of verbs, e.g. ga 'to sing' etc. which violates the constraints of concrete framework.

2. It claims that, with a shorter length of derivation of the verbal forms in terms of a 4-way rule component, this system achieves enormous amount of psychological reality. But in section 3.3.1. I shall show that such shorter derivations in terms of Paul's 4-rule mechanism fail to generate all the verbal forms of Bangla. Longer computations in terms of additional rules are required to handle the data - a fact that obviously weakens her claim of greater psychological reality.

3. Even her treating the evidence from the high/ Sadhu variety of Bangla as synchronic support is a dubious move. In section 3.3.3.3.2. I shall argue that forms belonging to the high variety of Bangla are as unusable as diachronic evidence in terms of the basic logic of the concrete approach. An argument along this line obviously fails to support her claim of synchronic motivation.

Where Dasgupta provides a basic layout of the phonological mechanism for Bangla verb morphology, Paul retains only those portions of Dasgupta's system that are in accordance with the logic of her concrete approach (e.g. underlying forms of suffixes) and rejects the rest (e.g. the rules). Paul completes the system by adding some machinery of her own.

In the following section, I shall outline the details of a concrete phonological analysis in order to handle verb morphology of Bangla, borrowing certain tools from both the analyses, viz. Dasgupta (1982) and Paul (1985).

3.2. Mechanism

Bangla verbs, unlike those in certain other languages, form a closed though large class. Except through onomatopoeia, no new verb stems can be formed. Innovation is restricted to the creation of new 'composite verbs' of the form NOUN DO, ADJECTIVE BE etc. Dasgupta (1990) provides a comprehensive description of Bangla composite verbs and argues (1990: 5) against the term 'conjunct verb' used by certain earlier authors.

The present section will describe the morphology of this closed class of Bangla. Subsection 3.2.1 deals with underlying phonological forms. Subsection 3.2.2 presents rules converting them into phonetic representations.

3.2.1. Underlying phonological forms

This sub-section discusses the underlying phonological forms of the verb stems and inflectional suffixes and classifies the two types of verb stems, viz. causative and non-causative, on the basis of their phonological shapes.

3.2.1.1. Verb stems

In Bangla, monosyllabic verb roots, depending on their different morphological constructions, show systematic alternations of vowel heights, e.g. kin~ken 'buy', dekh~dEkh 'see', bujh~bojh 'understand', kor~kOr 'do', Dek~Dak 'call' etc.

In accordance with the assumption that phonologically Bangla shows a three way distinction of vowel height (Paul, 1985: 176; Sarkar, 1987; Nath, 1997: 23), the above vowel alternations in the verbal roots may be characterized in terms of a one-step difference between the alternants. There are two ways to describe this state of affairs:

A. Postulating verb roots with low and mid vowels as basic forms and formulating a phonological rule of vowel raising;

B. Postulating verb roots with mid and high vowels as basic forms and formulating a rule of vowel lowering.

In other words underlying forms with low and mid vowels plus vowel raising and underlying forms with mid and high vowels plus vowel lowering are apparently the two options to be considered. Dasgupta and Paul have opted for solutions B and A respectively. Strictly speaking, Dasgupta chooses kin , dekh, and bujh as basic in the first three types of alternations (i/e, e/E, u/o) but a as basic in a/e (e.g. kha~khe 'eat'), leaving the O/o case (i.e. kOr~kor) open. We may take this to be a variant of solution B. He argues for this on the basis of the different sorts of evidence available that bear on these alternations.

Following Paul, I shall go in for system A; I postulate the forms with low and mid vowels, i.e. with e, o, E, O, and a, as underlying. In other words verbal roots like ken, lekh, bojh, So, dEkh, kOr, aMk, Dak etc., in the present system, are treated as underlying while the forms like kin, likh, bujh, Su, dekh, kor, eMk, Dek etc. are treated as alternants rule-derived from the above underlying forms respectively.

Paul (1985: 162) has already provided some arguments - some of which are obviously valid - for this choice. In addition to Paul's arguments, I would like to add an independent argument with reference to the deverbal forms, especially the action nominals. Sarkar (1976: 287) and Dan (1989: 23) discuss about Bangla action nominals.

In Bangla at least 6 types of action nominals may be attested which involve verb roots with low and high vowels:

a) without any phonological suffix

khoMj 'trace'
Dak 'call'
cOl 'practice' etc.

b) with oa/no

khEla 'play'
pORa 'study' etc.

c) reduplicated forms

dhOra-dhori 'seeking favour'
lekha-likhi repeated correspondence' etc.

d) compounding

dEkha-Sona 'looking after'
lekha-pORa 'education' etc.

e) echo words

khaWa-daWa 'meal' etc.

f) with on

cOlon 'walking'
kOron 'performing' etc.

Besides the second member of the forms of (c), the rest of the forms involve verb roots with low or mid vowels. On the basis of the forms of (c), someone may construct an argument against my position. The argument would run:

Verb roots with high and mid vowels are the basic forms and those with mid and low ones are derived from them in terms of a vowel lowering rule; the conditioning factor of which is the following low vowel a. Moreover, such an argument will be supported by the forms of (b), (d), and (e) too, where a follows the low and mid vowels of the verb root.

However, the forms of (a) and (f) will immediately reject such a claim, because

Firstly, the forms of (a) containing base verb roots show low and mid vowels not preceded or followed by any conditioning sound, and secondly, the forms of (f), instead of a following a, contain a following o, a mid vowel, though they show verb roots with low and mid vowels.

Hence the postulation of verb roots with low and mid vowels as basic forms and those with mid and high vowels as derived ones is well justified. A few counterexamples here are the forms like

Obujh 'unreasonable'
Ocin 'unknown'
ghurghur 'continuous prying movement'
Dub 'a dip'
mil 'similarity'
buli 'speech'
hOrbola 'mimic' etc.

Among these Obujh and Ocin are noted in Dasgupta (1982); he drew my attention to ghurghur, Dub, buli, and mil also afterwards.

The rest of the present subsection will draw a brief sketch of the stem formation processes and classification of stems in Bangla.

Bangla exhibits two types of verbal stems, viz. causative stems and non-causative stems; of these the causative stem formation is comparatively simpler, and thus requires less discussion than the other type.

Thus, we shall take up the non-causative stems before the causative ones.

3.2.1.1.1. Non-causative stems

Non-causative stems are of two sorts.

A. The bare verb roots may serve as stems, verb roots without any stem formative. This subtype comprises monosyllabic verb stems like kin 'to buy', dekh 'to see', So 'to lie down', pOR 'to read' kaT 'to cut', kha 'to eat', etc. to which inflectional suffixes are attached directly.

B. Some stems consist of a root plus an empty stem-formative (whose typical form is a). In this subtype, the root may be monosyllabic or reduplicated (and thus disyllabic); correspondingly, the stem is either disyllabic or trisyllabic. For example, laph-a 'to jump', tak-a 'to look', olT-a 'to turn over', SaMtr-a 'to swim', daMR-a 'to stand', kamR-a 'to bite' etc.; kOnkOn-a 'to ache', hOnhOn-a 'to speed up (said of walking)' etc.

A few observations regarding non-causative stems of type B may be useful at this point.

a) Sometimes instead of -a, one finds -o as the phonetic form of the stem-formative, e.g. cib-o 'to chew', ghum-o 'to sleep', bil-o 'to distribute', bul-o 'to stroke' etc. At this point one feels tempted to account for this a~o variation in terms of a progressive vowel raising rule, saying that the preceding high vowel raises the following low vowel a by one step to the mid vowel o. But some counterexamples make it difficult to accept at face value the simplest form of this idea. In the case of forms like eg-o 'to proceed', pech-o 'to retreat', ber-o 'to go out', per-o 'to cross' etc. the stem-formative -o is not preceded by a high vowel, but by a mid vowel, and mid vowels do not trigger raising.

This a~o variation, related to some phenomena shown in the preceding chapter, appears to be a potential area of extensive research in the overall picture of Bangla phonology. As such extensive investigation is tangential to my concerns and beyond the scope of the present book here I leave the question open for future research and assume an underlying -a for my purpose, hoping that future research will somehow be able to derive the surface -o cases from this -a.

b) The stems that take the stem-formative -a belong to various categories, e.g.,

i. Deverbal: peT-a 'to beat', bhaN-a 'to break' etc.

ii. Denominal: chobl-a 'to bite (said of snakes)', daMR-a 'to stand', cabk-a 'to whip', Saml-a 'to manage', ghum-o 'to sleep', SaMtr-a 'to swim' etc. associated with nouns like chobol 'snake bite', daMR 'stand (said of birds)', cabuk 'whip', Samal 'control', ghum 'sleep', SaMtar 'swimming' respectively.

iii. Deadjectival: olT-a 'to turn over', muT-o 'to become fat' associated with ulTo 'up side down', moTa 'fat' respectively.

iv. Depostpositional: eg-o 'to proceed', pech-o 'to retreat' from age 'front' and pechone 'behind' respectively.

v. Onomatopoeic: kOnkOn-a 'to ache', hOnhOn-a 'to speed up (said of walking)' etc.

vi. Unclassified: goN-a 'to groan', cib-o 'to chew', SaMtl-a 'to season' etc.

Except for i, all the other stems may be grouped under a cover term, viz. denominal verbal stems.

c) In those cases where a nonverbal 'cognate' is identifiable, it is possible to regard the root to which the stem-formative -a is attached as a free morpheme. A clear case is ghum-o 'to sleep' (cf. the noun ghum 'sleep'). We may extend this treatment to SaMtr-a 'to swim', cabk-a 'to whip' etc. at the cost of doing something (like postulating a deletion rule) to handle the differences between the reduced shape here and the full-bodied variant in the nouns like SaMtar 'swimming', cabuk 'whip' etc. However, there are many cases where the root to which -a is attached must be treated as a bound root (e.g. the cases of B vi); when in doubt, this treatment is to be preferred (e.g. cases like the verb ThEN-a 'to thrash' and the noun ThEN 'leg' exhibit insufficient semantic affinity and are thus doubtful cases; thus the root in ThEN-a is a bound root).

d) Even the roots with the stem-formative -a show some sort of vowel alternation, though in a very irregular fashion. For example, stems with u irregularly show vowel alternation in cases like bola~bulo; but goNa~*guNo, ghumo~*ghoma etc. In contrast, the stems with i never show any kind of alternation, e.g. cibo~*ceba, bilo~*bela etc.

The above observations provide quite a few open areas of research.

3.2.1.1.1. Causative stems

In Bangla causative stems are very regularly formed by adding the causative stem-formative -oa to monosyllabic roots. The causative stem-formative -oa is phonetically realized as either -Wa or -a, depending on the phonological shape of the verb root. To be more specific, vowel ending roots take -Wa, whereas the consonant ending roots take -a, e.g. kha-Wa 'to feed', So-Wa 'to put to sleep', pOR-a 'to teach', kOr-a 'to get something done' etc. In fact, in Bangla the causative stem-formative -oa and the gerund suffix -oa exhibit the same phonological conditioning of the distribution of the allomorphs -a and -Wa.

3.2.1.1.2. Classification of stems

The verb stems, both causative and non-causative, may be classified according to their phonological shapes into the following groups and sub-groups:

Verb stems
_______________________________

monosyllabic disyllabic
__________ _______________

a b c d
(v?a) (v=a) (v1?a) (v1=a)
______ ______ _________ _________

i ii i ii i ii i ii
/cvc/ /cv/ /cvc/ /cv/ /(c)vcv/ /(c)vccv/ /(c)vcv/ /(c)vccv/
pOR So pat kha ghumo olTa lapha Samla
bojh hO kaT pa Sekha cOTka khaWa palTa
dEkh choM Dak ja kOra chobla daMRa kamRa
lekh etc. aMk etc. cibo etc. etc. etc.
etc. etc. etc.

One notable point here is that the above classification of verb stems is an informal one whose sole purpose is to enable us to provide the phonological rules with non-technical paraphrases for ease of reading.

In other words, the above classifications do not represent any kind of diacritical feature. The class membership of the verb stems is not a 'feature' of any sort, being fully predictable from the phonological shapes of the stems as follows: verbs are primarily classified into two groups, viz. monosyllabic stems and disyllabic stems. The monosyllabic group is further classified into two groups depending on the fact whether the syllabic nucleus of the root is a, the low vowel, or not. Each of these a and non-a group is further classified into two sub-groups on the basis of whether the stems contain light syllables or heavy syllables, i.e. cvc and cv. Likewise the disyllabic group too is classified into two groups depending on whether the initial syllables contain a or non-a. And each of these two groups is further divided into two sub-groups on the basis of whether the verb stems under them contain two light syllables or one heavy syllable followed by a light one.

3.2.1.2. Underlying forms of inflectional suffixes

A B C D E
1. e en o iS i
2. icche icchen iccho icchiS icchi
3. ieche iechen iecho iechiS iechi
4. uk un o Ø no form
5. ilo ilen ile ili ilam
6. ito iten ite iti(S) itam
7. icchilo icchilen icchile icchili icchilam
8. iechilo iechilen iechile iechili iechilam
9. ibe iben ibe ibi ibo
10. ibe iben io iS no form
11. ite ie ile (i)ba oa/no

3.2.2. Phonological rules

I shall postulate 8 rules and state the ordering restrictions among them.

3.2.2.1. The rules

R-1 a-Mutation (a-M)

-low +high
/ [+syll] [-syll]1 ___ [+syll]
-back -cons

+high
[+cons]
-cons
-round [-back] / ___ +high [+syll]
+low à ( [+cons] )
+back -cons

+high
[+round] / [+cons]0 ___
-cons

This rule applies to the a of the stems under groups b, c, and d as follows:

i. In the case of stems under groups c and d it changes the stem final a to e if the suffix initial i is followed by another vowel, e.g. Sekha-ie à Sekhe-ie; olTa-io à olTe-io; lapha-io à laphe-io; Samla-ie àSamle-ie etc.

ii. In the case of stems under group bi it changes the a to E if the suffix initial i is followed by another vowel, e.g. kaT-ie à kET-ie; pat-io à pEt-io etc.

iii. In the cases of stems under group bii it changes the stem final a to E if the suffix initial i is followed by either a vowel or a non-final consonant, e.g. kha-ie àkhE-ie; kha-ile à khE-ile; kha-ite à khE-ite etc.

iv. In the cases of stems under group c it changes the stem final a, which has not yet been altered, to O if the preceding syllable contains a high vowel, e.g. ghuma-e àghumO-e; ghuma-icche à ghumO-icche etc.

R-2 Vocoid Raising (VR)

-cons
aback -ßhigh +high
around à /_____ [+cons]0
ßlow -low -cons

By this rule a high vowel raises its preceding vowels, viz. e, o, E, O, by one step to i, u, e, o respectively in spite of any intervening consonant or consonants.

One notable characteristics of the vocoid raising rule is that it reapplies to its own output, affecting each relevant segment of the string only once, e.g. Sekhe-ie àSekhi-ie à Sikhi-ie; Samle-ie àSamli-ie; olTe-ie à olTi-ie à ulTi-ie; kET-ie à keT-ie; khE-ie à khe-ie; So-ie à Su-ie; kOr-ie àkor-ie etc.

Such reapplication of rules, however, is not allowed in the strict SPE framework. But the current framework, which stands somewhere in between the SPE model and the metrical treatment, demands such a relaxation of the theory. Even Dey (1979), Singh (1980), and Dasgupta (1982), done in the SPE framework, allowed a similar provision.

R-3 Suffix-Truncation (ST)

[+syll] à Ø / [-syll]+___([-syll]) [+syll]

This rule deletes the initial vowel of a polysyllabic suffix after a stem ending in a consonant (or a semivowel, see section 3.5.3), e.g. Sikh-icchilo àSikh-cchilo; kor-ilo à korlo; Sekh-oa àSekh-a etc.

R-4 Degemination (DG)

aSyll asyll
à Ø / [-syll] ___ [+seg]
ßFF ßFF

This rule deletes the first member of a geminate consonant sequence or of a sequence of two identical vowels when immediately preceded by a non-syllabic segment and followed by any segment, e.g. kor-cche à korche; gay-cche à gayche; Sikhi-ie à Sikh-ie; ulTi-ie à ulT-ie etc.; but no effect in cases like di-i, where no segment follows the identical VV sequence.

R-5 Minor i-Deletion (Mi-D)

+high -high
-back à Ø / [+cons] [+cons] ____ -back
+syll +syll

This rule deletes i if preceded by a consonant sequence and followed by e, e.g. ulT-ie à ulTe; Saml-ie à Samle etc.

In some dialect as mine this Mi-D is a regular process as it applies whenever lexical entry is specified as [-syll] [-syll] + stemformative.

But in other dialects Mi-D is a true 'Minor' rule as it has to consider grammatical information there: it is sensitive to the information about causativity and idiosyncratic properties of certain lexical items.

R-6 Semivowel Formation (SVF)

-cons -cons
-low à [-syll] / _____
<-high> <-high>

This rule weakens a nonlow V2 in any V1V2 sequence where V2 is no lower than V1, e.g. hO-o à hOW etc. Note that V here stands for vocoid: khaw-ie à khaw-ye.

R-7 Semivowel Deletion (SVD)

-syll
àØ / [+syll] + ____ [-syll]
-cons

This rule deletes postvocalic preconsonantal semivowels in suffix initial position, e.g. kha-wn à khan; pala-yS à palaS; but no effect on dowR-e, gay-che, khaw-ye.

R-8 Glide Assimilation (GA)

-back
+high
[-back] / ____ -cons
-syll à -syll
-cons
-cons -cons
[-high] / ____
-high -high

This rule applies to a semivowel and assimilates it either to the following semivowel y, e.g. khawy-e à khayye etc., or to the height level of the surrounding vowels, e.g. ho-yo à hoYo etc.

3.2.2.2. Ordering restrictions

The rules a-M, VR, ST, DG, in that order, precede all other rules, viz. Mi-D, SVF, SVD, and GA; of these, SVF crucially precedes SVD and GA; there are no other extrinsic ordering restrictions.

A late rule, viz. O to o, and syllable readjustment rules operate on the final output of these rules and result in PR.

O to o, as mentioned in chapter 2, applies to O at unstressed positions and changes it to o, e.g. ghumOY à ghumoY etc.

A sketch of the syllable readjustment rules in Bangla will be given in chapter 5.

3.3. Difference between Paul (1985) and the current approach

The present section will point out the relevant differences between Paul (1985) and the present approach, both of which are done in the modified generative, i.e. the concrete framework.

The differences between these two analyses occurred mainly because of the inadequacy of Paul's phonological mechanisms in either of the two respects, viz. either they are inadequate in terms of practical derivation, i.e. they fail to generate the verbal forms of Bangla correctly, or they are inadequate in terms of theoretical considerations, i.e. they deviate from the theoretical constraints.

3.3.1. Differences

The differences between the two approaches are listed as follows:

1. Paul claims to shorten the length of verbal derivations in terms of her 4-rule mechanisms, whereas the present analysis posits 8 rules, sometimes involving longer computations also. In fact, Paul's 4-rule mechanism is derivationally inadequate.

2. In the underlying forms of the stems and inflectional suffixes the present analysis differs from Paul's analysis in the following respects:

a) Paul posits underlying stem final h for minor group of verbs like ga etc. whereas the present analysis does not posit any such abstract segment.

b) Paul posits ch underlyingly, whereas like Dasgupta (1982), the present analysis too posits cch.

c) Paul posits ia for 11B, whereas the present analysis posits ie.

d) Paul posits a for 11E, whereas the present analysis posits oa.

In fact, except for 2(d), which too results in derivational inadequacy, the postulations of all the above underlying forms of Paul (1985) deviates from the theoretical constraints.

The rest of the section will deal with these two types of inadequacies in terms of the differences mentioned above.

3.3.2. Inadequacy of practical derivation

Paul's system derives unacceptable forms as follows:

i. Paul classifies the verb roots into 3 groups, viz. a) cv(c), in which v is any vowel other than a, b) cv(c), in which v is a, and c) cvc(c)v. She formulates 4 rules, viz. Final Vowel Deletion (FVD), Vowel Raising (VR), i-Assimilation (i-A), and i-Deletion (i-D).

The VR, as Paul formulates it, refers to a 3-way classification of verb roots as follows:

"In case of roots under gr.b the suffix initial i raises the root vowel a to e if the suffix initial i is followed by another vowel", e.g. pat-ie à pete, but pat-iten à patten.

This rule will also generate kha-iten à *khaten; kha-ilen à *khalen etc. instead of kheten, khelen etc. Paul (1985) lists such forms under exceptions.

This problem is handled in the present analysis by (a) differentiating the vowel ending roots with a (cf. gr. bii) from the consonant ending roots with a (cf. gr. bi), and (b) formulating two rules, viz. a-M and VR as follows:

a-M applies to a of the stems under gr. bii and changes it to E even if the suffix initial i is followed by a non-final consonant, whereas it does not apply to the stems under gr. bi under similar conditions. And the rule VR, in the present system, is made insensitive to a. Thus the system correctly generates kha-ie àkhE-ie àkhe-ie à khe-ye à kheYe; kha-ite à khE-ite à khe-ite à khe-yte à khete; pat-ie à pEt-ie à pet-ie à pete; but pat-ite à patte etc.

ii. i-D in Paul's system deletes the suffix initial i if (a) the root ends in a consonant and i is followed by a vowel or a non-final consonant, and (b) the roots end in a vowel and the i is followed by a vowel or a non-final [-palatal] consonant. Moreover, i-D has to consider one piece of grammatical information, viz. whether the roots form a causative - non-causative pair or it is a lone non-causative form. i-D does not apply in cases of the causative member of the pair and the lone non-causatives, e.g.

Causative Non-causative Lone non-causative
kOra-ie àkorie kOr-ie à kore pala-ie àpalie etc.

This will also generate Samla-ie à *Samlie; bigRo-ie à*bigRie etc. instead of Samle, bigRe etc. Generally Samlie, bigRie etc. are not acceptable in Kolkata Standard Bangla, though Paul accepts them in her system.

This problem too is handled in our system by (a) differentiating the roots with cvccv (cf. grs. cii, dii) from those with cvcv (cf. grs. ci, di), and (b) formulating a minor i-Deletion rule which applies as a redundancy rule in some dialects and as a 'true' minor rule in some other. For example, Samla-ie à Samle-ie àSamli-ie àSaml-ie à Samle etc.

iii. Paul's system consists of 4 rules as mentioned earlier, whereas the present system introduces an 8-rule mechanism. The rules here are a-M, VR, ST, DG, Mi-D, SVF, SVD, and GA.

Let me illustrate the derivational capacity of Paul's 4-rule system.

Paul (1985: 190) claims that the derivations of the forms of the non-causative bisyllabic stems and those of the causative bisyllabic stems are the same.

Paul's system poses a problem for the Wa ending stems:

khaWa-ieche SoWa-io
FVD khaW-ieche SoW-io
VR - SuW-io
i-A - -
i-D *khaWeche *SuWo

instead of khayyeche and Suyyo respectively.

In order to generate the above forms correctly, apart from modifying VR and i-D (which becomes ST in our system), the postulation of a-M, SVF, and GA is also necessary.

Paul's system poses a problem for the a ending bisyllabic stems:

kOra-e pala-e dEkha-o khaWa-o
FVD - - - -
VR - - - -
i-A - - - -
i-D - - - -
*kOrae *palae *dEkhao *khaWao

instead of kOraY, palaY, dEkhaW, khaWaW respectively.

In fact, not only in case of a ending bisyllabic stems but also in case of monosyllabic stems containing light syllables Paul's system poses a problem, for example, it generates forms like *khae, *Soe from kha and So respectively instead of khaY, SoY.

In order to generate the above correct forms, the postulation of SVF is necessary.

Now let me consider the following derivations of vowel ending roots:

kha-en So-iS kOra-uk
FVD - - -
VR - Su-iS -
i-A - - -
i-D - - -
*khaen *SuiS *kOrauk

instead of khan, SuS, and kOrak.

In order to obtain the correct derivations of the above forms the postulation of SVF and SVD is necessary.

iv. For 11E, i.e. for gerunds, Paul posits a underlyingly and inserts W after vowel ending stems, e.g. kOr-a, but kha-Wa etc.

The present system posits oa in UR; the rule SVF turns this oa into Wa in cases such as kha-Wa, whereas the rule ST deletes the o when preceded by a consonant.

Though the effect of both the systems apparently amounts to the same, in terms of the glide formation habit of Bangla Paul's insertion rule results in counterintuitive forms as follows:

With an underlying gerund suffix a the gerund form of the stem ja 'to go' must be assumed to have the underlying representation ja-a. This ja-a would be expected to become *jaYa rather than jaWa, as in Bangla speech and songs the insertion of Y glide is the commonest phenomenon, e.g. hoYo, kheYo etc.

In other words, with an underlying a, in Paul's system, there is no way to predict the colour of the glide to be inserted.

Hence it is wise to postulate oa and convert o to its [-syll] counterpart W after a vowel and delete the o after a consonant. In other words, here deletion, rather than insertion, can take care of the derivational inadequacy of Paul's system in a more justified way.

One question may arise here regarding the abstractness of o of the gerund suffix as it surfaces as W.

But we observe alternation between o and W elsewhere in the paradigm; jan-o 'you know', ja-W (from underlying /ja-o/) 'you go'. It makes no sense to postulate an unmotivated /W/ in Bangla if one can eliminate this semivowel from the underlying phonological system.

3.3.3. Inadequacy in terms of theoretical considerations

As is mentioned earlier, the differences under 2, i.e. the differences in the underlying forms of the stems and suffixes, come under this category.

3.3.3.1. Underlying h
Though Paul argues against positing any abstract underlying forms, 0strangely enough, she prefers to accept the only abstract underlying segment of Dasgupta (1982), viz. the underlying stem final h for minor group of verbs, and thus deviates from the constraint of the framework.

However, in terms of a true concrete approach the present system rejects such abstract underlying h, which never surfaces.

3.3.3.2. ch vs. cch

Paul posits ch underlyingly and in terms of the rule i-A, i.e. the geminate being a result of the assimilation of the preceding segment, i.e. i, to ch, derives cch from UR ch. i-A lacks motivation as it is not a functional rule in Bangla phonology.

The present system, however, posits cch underlyingly and derives ch from it in terms of DG, a rule that gains motivation as it reflects the phonotactic constraint of Bangla.

3.3.3.3. ia vs. ie in 11B

Paul posits the underlying form of 11B as ia, whereas the present system posits it as ie. To be more precise, the differences here is of the final segment, viz. a vs. e.

Let us consider the consequences of Paul's proposal one by one.

3.3.3.3.1. If ia is posited underlyingly Paul's analysis will require an extra rule, viz. Contiguous Progressive Fronting (CPF) of Sarkar (1976: 279). With CPF the underlying a becomes e because of the high vocoid next to it, e.g. kOr-ia àkore.

However, Dasgupta (1982: 19) rejects this CPF on the following grounds:

a) As CPF does not crucially interact with anything else in the grammar so it is likely that children acquiring Bangla never have a reason to postulate CPF, and that therefore CPF does not even exist to a Bangla speaker. In other words, CPF lacks psychological reality.

b) In order to obtain rule viability Dasgupta extends the notion of CPF and formulates Progressive Fronting-Backing (PFB), that covers both contiguous fronting and backing and accounts for the forms like cuRo from cuRa 'peak' etc. But at the same time Dasgupta shows that such an extension, i.e. rule like PFB will enable the learner to set up the underlying forms of verb endings as wa (10C), yta (11A), yla (11C) etc. Dasgupta (1982: 18) posits yo (10C), yte (11A), yle (11C). However, the present section is concerned about the differences in the final segments of the forms not the initial ones.

Dasgupta (1982) says surely these perverse hypothesis must not be made available and therefore the final vowel in 11B should be e, rather than a, in both UR and PR.

Paul (1985: 171) mentions Dasgupta's first argument and rejects it on the ground of motivation for CPF, a point which I am coming to.

Paul (1985), however, overlooks Dasgupta's second argument.

3.3.3.3.2. Paul (1985: 171) claims that the formulation of CPF is well-motivated because, i) it accounts for the derivational relationship between the forms of the formal and the colloquial varieties of Bangla, e.g. koria ~ kore etc., and ii) ia surfaces in the forms of the formal variety. In other words Paul considers the forms of the formal variety as synchronic evidence, a point on which I like to differ from her on the following grounds:

a) Theory of diglossia says that the formal or the high variety is never used as anyone's speech. The high variety has to be learnt through formal education. Bangla being a language with diglossia situation, is no exception to the theory of diglossia. Hence it could be argued that unless and until a child acquires the high variety through formal education, the grammar of the high variety is just non-existent to him, which he can use as language internal evidence.

b) Illiterates as well as Bengali children going to some non-Bangla schools do not possess the high variety grammar at all. So for them the question of seeking support form the high variety just does not arise.

c) Even after formal schooling people do make mistakes in using the high variety forms, though they never make any mistake in their speech - a fact that stands against any synchronic derivational relationship between the forms of high and low varieties of Bangla.

d) Moreover, deriving the forms of the colloquial variety from those of the formal variety goes against Paul's own principle of deriving the inflected forms from the basic ones. From a strict synchronic point of view kore appears to be more basic than koria. As far as their frequency in speech is concerned koria has zero; even in the current literature, koria will have fewer occurrences compared to that of kore. If one is compelled to choose one between the two as the basic form, no doubt, he will choose kore without hesitation.

e) In Bangla the high variety is marked for i) verbal forms, ii) pronominal forms, and iii) tatsama words. While treating the status of semivowels Paul (1985: 29) herself submits that the pairs like SONbad~SOmbad 'news' etc. occur in Sanskrit and standard Bangla has adopted both the members for its own different registers. The forms with N are used in the formal variety while those with the homorganic nasals are used in the colloquial variety.

Moreover, Paul (1985: 31) rejects the assumption of the standard theory that variations in the different dialects are the result of differences in their rule component and the underlying forms normally remain stable and unchanged. She treats such pairs as 'doublets' in accordance with the assumption that the lexicon undergoes restructuring from generation to generation.

However, treating some portions, viz. tatsama words, of the high variety as derivationally unrelated doublets, while some other portion, viz. the verbal forms, of the same variety as derivationally related synchronic evidence results in inconsistency in the theory.

On the basis of the above arguments the conclusion may be drawn that the evidence from the high variety can not be used as synchronic explanation. So, with the high variety being non-existent as language internal facts, the postulation of ia underlyingly will result in absolute neutralization, which makes the theory costly. Therefore, I think, 11B should be ie in both UR and PR.

However, with the above differences I formulated a modified version of Paul (1985), in terms of which I shall show the actual derivations of at least eight verb roots in the following section.

3.4. Actual derivations

3.4.1. kOr 'to do' of group ai

A B C D E
kOr-e kOr-en kOr-o kOr-iS kOr-i
VR - - - kor-iS kor-i
kOre kOren kOro koriS kori

kOr-icche kOr-icchen kOr-iccho kOr-icchiS kOr-icchi
VR kor-icche kor-icchen kor-iccho kor-icchiS kor-icchi
ST kor-cche kor-cchen kor-ccho kor-cchiS kor-cchi
DG korche korchen korcho korchiS korchi

kOr-ieche kOr-iechen kOr-iecho kOr-iechiS kOr-iechi
VR kor-ieche kor-iechen kor-iecho kor-iechiS kor-iechi
ST koreche korechen korecho korechiS korechi

kOr-uk kOr-un kOr-o kOr-Ø -
VR koruk korun - -
- - kOro kOr

kOr-ilo kOr-ilen kOr-ile kOr-ili kOr-ilam
VR kor-ilo kor-ilen kor-ile kor-ili kor-ilam
ST korlo korlen korle korli korlam

kOr-ito kOr-iten kOr-ite kOr-itiS kOr-itam
VR kor-ito kor-iten kor-ite kor-itiS kor-itam
ST korto korten korte kortiS kortam

kOr-icchilo kOr-icchilen kOr-icchile kOr-icchili kOr-icchilam
VR kor-icchilo kor-icchilen kor-icchile kor-icchili kor-icchilam
ST kor-cchilo kor-cchilen kor-cchile kor-cchili kor-cchilam
DG korchilo korchilen korchile korchili korchilam

kOr-iechilo kOr-iechilen kOr-iechile kOr-iechili kOr-iechilam
VR kor-iechilo kor-iechilen kor-iechile kor-iechili kor-iechilam
ST korechilo korechilen korechile korechili korechilam

kOr-ibe kOr-iben kOr-ibe kOr-ibi kOr-ibo
VR kor-ibe kor-iben kor-ibe kor-ibi kor-ibo
ST korbe korben korbe korbi korbo

kOr-ibe kOr-iben kOr-io kOr-iS -
VR kor-ibe kor-iben kor-io kor-iS -
ST korbe korben koro koriS -

kOr-ite kOr-ie kOr-ile kOr-ba kOr-oa
VR kor-ite kor-ie kor-ile - -
ST korte kore korle - kOra
kOrba

3.4.2. hO 'to be' of group aii

hO-e hO-en hO-o hO-iS hO-i
VR - - - ho-iS ho-i
SVF hOY hO-Yn hOW ho-yS hoy
SVD - hOn - hoS -

hO-icche hO-icchen hO-iccho hO-icchiS hO-icchi
VR ho-icche ho-icchen ho-iccho ho-icchiS ho-icchi
SVF ho-ycche ho-ycchen ho-yccho ho-ycchiS ho-ycchi
SVD hocche hocchen hoccho hocchiS hocchi

hO-ieche hO-iechen hO-iecho hO-iechiS hO-iechi
VR ho-ieche ho-iechen ho-iecho ho-iechiS ho-iechi
SVF ho-yeche ho-yechen ho-yecho ho-yechiS ho-yechi
GA hoYeche hoYechen hoYecho hoYechiS hoYechi

hO-uk hO-un hO-o hO-Ø -
VR ho-uk ho-un - - -
SVF ho-wk ho-wn hOW - -
SVD hok hon - - -
hO

hO-ilo hO-ilen hO-ile hO-ili hO-ilam
VR ho-ilo ho-ilen ho-ile ho-ili ho-ilam
SVF ho-ylo ho-ylen ho-yle ho-yli ho-ylam
SVD holo holen hole holi holam

hO-ito hO-iten hO-ite hO-itiS hO-itam
VR ho-ito ho-iten ho-ite ho-itiS ho-itam
SVF ho-yto ho-yten ho-yte ho-ytiS ho-ytam
SVD hoto hoten hote hotiS hotam

hO-icchilo hO-icchilen hO-icchile hO-icchili hO-icchilam
VR ho-icchilo ho-icchilen ho-icchile ho-icchili ho-icchilam
SVF ho-ycchilo ho-ycchilen ho-ycchile ho-ycchili ho-ycchilam
SVD hocchilo hocchilen hocchile hocchili hocchilam

hO-iechilo hO-iechilen hO-iechile hO-iechili hO-iechilam
VR ho-iechilo ho-iechilen ho-iechile ho-iechili ho-iechilam
SVF ho-yechilo ho-yechilen ho-yechile ho-yechili ho-yechilam
GA hoYechilo hoYechilen hoYechile hoYechili hoYechilam

hO-ibe hO-iben hO-ibe hO-ibi hO-ibo
VR ho-ibe ho-iben ho-ibe ho-ibi ho-ibo
SVF ho-ybe ho-yben ho-ybe ho-ybi ho-ybo
SVD *hobe *hoben *hobe hobi *hobo

hO-ibe hO-iben hO-io hO-iS -
VR ho-ibe ho-iben ho-io ho-iS -
SVF ho-ybe ho-yben ho-yo ho-yS -
SVD *hobe *hoben - hoS -
GA - - hoYo - -

hO-ite hO-ie hO-ile hO-ba hO-oa
VR ho-ite ho-ie ho-ile - -
SVF ho-yte ho-ye ho-yle - hOWa
SVD hote - hole - -
GA - hoYe - - -
hOba

3.4.3. kaT 'to cut' of group bi

kaT-e kaT-en kaT-o kaT-iS kaT-i
kaTe kaTen kaTo kaTiS kaTi

kaT-icche kaT-icchen kaT-iccho kaT-icchiS kaT-icchi
ST kaT-cche kaT-cchen kaT-ccho kaT-cchiS kaT-cchi
DG kaTche kaTchen kaTcho kaTchiS kaTchi

kaT-ieche kaT-iechen kaT-iecho kaT-iechiS kaT-iechi
a-M kET-ieche kET-iechen kET-iecho kET-iechiS kET-iechi
VR keT-ieche keT-iechen keT-iecho keT-iechiS keT-iechi
ST keTeche keTechen keTecho keTechiS keTechi

kaT-uk kaT-un kaT-o kaT-Ø -
kaTuk kaTun kaTo kaT -

kaT-ilo kaT-ilen kaT-ile kaT-ili kaT-ilam
ST kaTlo kaTlen kaTle kaTli kaTlam

kaT-ito kaT-iten kaT-ite kaT-itiS kaT-itam
ST kaTto kaTten kaTte kaTtiS kaTtam

kaT-icchilo kaT-icchilen kaT-icchile kaT-icchili kaT-icchilam
ST kaT-cchilo kaT-cchilen kaT-cchile kaT-cchili kaT-cchilam
DG kaTchilo kaTchilen kaTchile kaTchili kaTchilam

kaT-iechilo kaT-iechilen kaT-iechile kaT-iechili kaT-iechilam
a-M kET-iechilo kET-iechilen kET-iechile kET-iechili kET-iechilam
VR keT-iechilo keT-iechilen keT-iechile keT-iechili keT-iechilam
ST keTechilo keTechilen keTechile keTechili keTechilam

kaT-ibe kaT-iben kaT-ibe kaT-ibi kaT-ibo
ST kaTbe kaTben kaTbe kaTbi kaTbo

kaT-ibe kaT-iben kaT-io kaT-iS -
a-M - - kET-io - -
VR - - keT-io - -
ST kaTbe kaTben keTo - -
kaTiS

kaT-ite kaT-ie kaT-ile kaT-ba kaT-oa
a-M - kET-ie - - -
VR - keT-ie - - -
ST kaTte keTe kaTle - kaTa
kaTba

3.4.4. kha 'to eat' of group bii

kha-e kha-en kha-o kha-iS kha-i
SVF khaY kha-Yn khaW kha-yS khay
SVD - khan - khaS -

kha-icche kha-icchen kha-iccho kha-icchiS kha-icchi
SVF kha-ycche kha-ycchen kha-yccho kha-ycchiS kha-ycchi
SVD khacche khacchen khaccho khacchiS khacchi

kha-ieche kha-iechen kha-iecho kha-iechiS kha-iechi
a-M khE-ieche khE-iechen khE-iecho khE-iechiS khE-iechi
VR khe-ieche khe-iechen khe-iecho khe-iechiS khe-iechi
SVF khe-yeche khe-yechen khe-yecho khe-yechiS khe-yechi
GA kheYeche kheYechen kheYecho kheYechiS kheYechi

kha-uk kha-un kha-o kha-Ø -
SVF kha-wk kha-wn khaW - -
SVD khak khan - - -
kha

kha-ilo kha-ilen kha-ile kha-ili kha-ilam
a-M khE-ilo khE-ilen khE-ile khE-ili khE-ilam
VR khe-ilo khe-ilen khe-ile khe-ili khe-ilam
SVF khe-ylo khe-ylen khe-yle khe-yli khe-ylam
SVD khelo khelen khele kheli khelam

kha-ito kha-iten kha-ite kha-itiS kha-itam
a-M khE-ito khE-iten khE-ite khE-itiS khE-itam
VR khe-ito khe-iten khe-ite khe-itiS khe-itam
SVF khe-yto khe-yten khe-yte khe-ytiS khe-ytam
SVD kheto kheten khete khetiS khetam

kha-icchilo kha-icchilen kha-icchile kha-icchili kha-icchilam
SVF kha-ycchilo kha-ycchilen kha-ycchile kha-ycchili kha-ycchilam
SVD khacchilo khacchilen khacchile khacchili khacchilam

kha-iechilo kha-iechilen kha-iechile kha-iechili kha-iechilam
a-M khE-iechilo khE-iechilen khE-iechile khE-iechili khE-iechilam
VR khe-iechilo khe-iechilen khe-iechile khe-iechili khe-iechilam
SVF khe-yechilo khe-yechilen khe-yechile khe-yechili khe-yechilam
GA kheYechilo kheYechilen kheYechile kheYechili kheYechilam

kha-ibe kha-iben kha-ibe kha-ibi kha-ibo
a-M khE-ibe khE-iben khE-ibe khE-ibi khE-ibo
VR khe-ibe khe-iben khe-ibe khe-ibi khe-ibo
SVF khe-ybe khe-yben khe-ybe khe-ybi khe-ybo
SVD *khebe *kheben *khebe *khebi *khebo

kha-ibe kha-iben kha-io kha-iS -
a-M khE-ibe khE-iben khE-io - -
VR khe-ibe khe-iben khe-io - -
SVF khe-ybe khe-yben khe-yo kha-yS -
SVD *khebe *kheben - khaS -
GA - - kheYo - -

kha-ite kha-ie kha-ile kha-ba kha-oa
a-M khE-ite khE-ie khE-ile - -
VR khe-ite khe-ie khe-ile - -
SVF khe-yte khe-ye khe-yle - khaWa
SVD khete - khele - -
GA - kheYe - - -
khaba

3.4.5. Sekha 'to teach' of group ci

Sekha-e Sekha-en Sekha-o Sekha-iS Sekha-i
SVF SekhaY Sekha-Yn SekhaW Sekha-yS Sekhay
SVD - Sekhan - SekhaS -

Sekha-icche Sekha-icchen Sekha-iccho Sekha-icchiS Sekha-icchi
SVF Sekha-ycche Sekha-ycchen Sekha-yccho Sekha-ycchiS Sekha-ycchi
SVD Sekhacche Sekhacchen Sekhaccho SekhacchiS Sekhacchi

Sekha-ieche Sekha-iechen Sekha-iecho Sekha-iechiS Sekha-iechi
a-M Sekhe-ieche Sekhe-iechen Sekhe-iecho Sekhe-iechiS Sekhe-iechi
VR Sikhi-ieche Sikhi-iechen Sikhi-iecho Sikhi-iechiS Sikhi-iechi
DG Sikhieche Sikhiechen Sikhiecho SikhiechiS Sikhiechi

Sekha-uk Sekha-un Sekha-o Sekha-Ø -
SVF Sekha-wk Sekha-wn SekhaW - -
SVD Sekhak Sekhan - - -
Sekha

Sekha-ilo Sekha-ilen Sekha-ile Sekha-ili Sekha-ilam
SVF Sekha-ylo Sekha-ylen Sekha-yle Sekha-yli Sekha-ylam
SVD Sekhalo Sekhalen Sekhale Sekhali Sekhalam

Sekha-ito Sekha-iten Sekha-ite Sekha-itiS Sekha-itam
SVF Sekha-yto Sekha-yten Sekha-yte Sekha-ytiS Sekha-ytam
SVD Sekhato Sekhaten Sekhate SekhatiS Sekhatam

Sekha-icchilo Sekha-icchilen Sekha-icchile Sekha-icchili Sekha-icchilam
SVF Sekha-ycchilo Sekha-ycchilen Sekha-ycchile Sekha-ycchili Sekha-ycchilam
SVD Sekhacchilo Sekhacchilen Sekhacchile Sekhacchili Sekhacchilam

Sekha-iechilo Sekha-iechilen Sekha-iechile Sekha-iechili Sekha-iechilam
a-M Sekhe-iechilo Sekhe-iechilen Sekhe-iechile Sekhe-iechili Sekhe-iechilam
VR Sikhi-iechilo Sikhi-iechilen Sikhi-iechile Sikhi-iechili Sikhi-iechilam
DG Sikhiechilo Sikhiechilen Sikhiechile Sikhiechili Sikhiechilam

Sekha-ibe Sekha-iben Sekha-ibe Sekha-ibi Sekha-ibo
SVF Sekha-ybe Sekha-yben Sekha-ybe Sekha-ybi Sekha-ybo
SVD Sekhabe Sekhaben Sekhabe Sekhabi Sekhabo

Sekha-ibe Sekha-iben Sekha-io Sekha-iS -
a-M - - Sekhe-io - -
VR - - Sikhi-io - -
DG - - Sikhio - -
SVF Sekha-ybe Sekha-yben - Sekha-yS -
SVD Sekhabe Sekhaben - SekhaS -

Sekha-ite Sekha-ie Sekha-ile Sekha-ba Sekha-no
a-M - Sekhe-ie - - -
VR - Sikhi-ie - - -
DG - Sikhie - - -
SVF Sekha-yte - Sekha-yle - -
SVD Sekhate - Sekhale - -
Sekhaba Sekhano

3.4.6. olTa 'to turn over' of group cii

olTa-e olTa-en olTa-o olTa-iS olTa-i
SVF olTaY olTa-Yn olTaW olTa-yS olTay
SVD - olTan - olTaS -

olTa-icche olTa-icchen olTa-iccho olTa-icchiS olTa-icchi
SVF olTa-ycche olTa-ycchen olTa-yccho olTa-ycchiS olTa-ycchi
SVD olTacche olTacchen olTaccho olTacchiS olTacchi

olTa-ieche olTa-iechen olTa-iecho olTa-iechiS olTa-iechi
a-M olTe-ieche olTe-iechen olTe-iecho olTe-iechiS olTe-iechi
VR ulTi-ieche ulTi-iechen ulTi-iecho ulTi-iechiS ulTi-iechi
DG ulT-ieche ulT-iechen ulT-iecho ulT-iechiS ulT-iechi
Mi-D ulTeche ulTechen ulTecho ulTechiS ulTechi

olTa-uk olTa-un olTa-o olTa-Ø -
SVF olTa-wk olTa-wn olTaW - -
SVD olTak olTan - - -
olTa

olTa-ilo olTa-ilen olTa-ile olTa-ili olTa-ilam
SVF olTa-ylo olTa-ylen olTa-yle olTa-yli olTa-ylam
SVD olTalo olTalen olTale olTali olTalam

olTa-ito olTa-iten olTa-ite olTa-itiS olTa-itam
SVF olTa-yto olTa-yten olTa-yte olTa-ytiS olTa-ytam
SVD olTato olTaten olTate olTatiS olTatam

olTa-icchilo olTa-icchilen olTa-icchile olTa-icchili olTa-icchilam
SVF olTa-ycchilo olTa-ycchilen olTa-ycchile olTa-ycchili olTa-ycchilam
SVD olTacchilo olTacchilen olTacchile olTacchili olTacchilam

olTa-iechilo olTa-iechilen olTa-iechile olTa-iechili olTa-iechilam
a-M olTe-iechilo olTe-iechilen olTe-iechile olTe-iechili olTe-iechilam
VR ulTi-iechilo ulTi-iechilen ulTi-iechile ulTi-iechili ulTi-iechilam
DG ulT-iechilo ulT-iechilen ulT-iechile ulT-iechili ulT-iechilam
Mi-D ulTechilo ulTechilen ulTechile ulTechili ulTechilam

olTa-ibe olTa-iben olTa-ibe olTa-ibi olTa-ibo
SVF olTa-ybe olTa-yben olTa-ybe olTa-ybi olTa-ybo
SVD olTabe olTaben olTabe olTabi olTabo

olTa-ibe olTa-iben olTa-io olTa-iS -
a-M - - olTe-io - -
VR - - ulTi-io - -
DG - - ulTio - -
SVF olTa-ybe olTa-yben - olTa-yS -
SVD olTabe olTaben - olTaS -

olTa-ite olTa-ie olTa-ile olTa-ba olTa-no
a-M - olTe-ie - - -
VR - ulTi-ie - - -
DG - ulT-ie - - -
Mi-D - ulTe - - -
SVF olTa-yte - olTa-yle - -
SVD olTate - olTale - -
olTaba olTano

3.4.7. lapha 'to jump' of group di

lapha-e lapha-en lapha-o lapha-iS lapha-i
SVF laphaY lapha-Yn laphaW lapha-yS laphay
SVD - laphan - laphaS -

lapha-icche lapha-icchen lapha-iccho lapha-icchiS lapha-icchilam
SVF lapha-ycche lapha-ycchen lapha-yccho lapha-ycchiS lapha-ycchilam
SVD laphacche laphacchen laphaccho laphacchiS laphacchilam

lapha-ieche lapha-iechen lapha-iecho lapha-iechiS lapha-iechi
a-M laphe-ieche laphe-iechen laphe-iecho laphe-iechiS laphe-iechi
VR laphi-ieche laphi-iechen laphi-iecho laphi-iechiS laphi-iechi
DG laphieche laphiechen laphiecho laphiechiS laphiechi

lapha-uk lapha-un lapha-o lapha-Ø -
SVF lapha-wk lapha-wn laphaW - -
SVD laphak laphan - - -
lapha

lapha-ilo lapha-ilen lapha-ile lapha-ili lapha-ilam
SVF lapha-ylo lapha-ylen lapha-yle lapha-yli lapha-ylam
SVD laphalo laphalen laphale laphali laphalam

lapha-ito lapha-iten lapha-ite lapha-itiS lapha-itam
SVF lapha-yto lapha-yten lapha-yte lapha-ytiS lapha-ytam
SVD laphato laphaten laphate laphatiS laphatam

lapha-icchilo lapha-icchilen lapha-icchile lapha-icchili lapha-icchilam
SVF lapha-ycchilo lapha-ycchilen lapha-ycchile lapha-ycchili lapha-ycchilam
SVD laphacchilo laphacchilen laphacchile laphacchili laphacchilam

lapha-iechilo lapha-iechilen lapha-iechile lapha-iechili lapha-iechilam
a-M laphe-iechilo laphe-iechilen laphe-iechile laphe-iechili laphe-iechilam
VR laphi-iechilo laphi-iechilen laphi-iechile laphi-iechili laphi-iechilam
DG laphiechilo laphiechilen laphiechile laphiechili laphiechilam

lapha-ibe lapha-iben lapha-ibe lapha-ibi lapha-ibo
SVF lapha-ybe lapha-yben lapha-ybe lapha-ybi lapha-ybo
SVD laphabe laphaben laphabe laphabi laphabo

lapha-ibe lapha-iben lapha-io lapha-iS -
a-M - - laphe-io - -
VR - - laphi-io - -
DG - - laphio - -
SVF lapha-ybe lapha-yben - lapha-yS -
SVD laphabe laphaben - laphaS -

lapha-ite lapha-ie lapha-ile lapha-ba lapha-no
a-M - laphe-ie - - -
VR - laphi-ie - - -
DG - laphie - - -
SVF lapha-yte - lapha-yle - -
SVD laphate - laphale - -
laphaba laphano

3.4.8. Samla 'to manage' of group dii

Samla-e Samla-en Samla-o Samla-iS Samla-i
SVF SamlaY Samla-Yn SamlaW Samla-yS Samlay
SVD - Samlan - SamlaS -

Samla-icche Samla-icchen Samla-iccho Samla-icchiS Samla-icchi
SVF Samla-ycche Samla-ycchen Samla-yccho Samla-ycchiS Samla-ycchi
SVD Samlacche Samlacchen Samlaccho SamlacchiS Samlacchi

Samla-ieche Samla-iechen Samla-iecho Samla-iechiS Samla-iechi
a-M Samle-ieche Samle-iechen Samle-iecho Samle-iechiS Samle-iechi
VR Samli-ieche Samli-iechen Samli-iecho Samli-iechiS Samli-iechi
DG Saml-ieche Saml-iechen Saml-iecho Saml-iechiS Saml-iechi
Mi-D Samleche Samlechen Samlecho SamlechiS Samlechi

Samla-uk Samla-un Samla-o Samla-Ø -
SVF Samla-wk Samla-wn SamlaW - -
SVD Samlak Samlan - - -
Samla

Samla-ilo Samla-ilen Samla-ile Samla-ili Samla-ilam
SVF Samla-ylo Samla-ylen Samla-yle Samla-yli Samla-ylam
SVD Samlalo Samlalen Samlale Samlali Samlalam

Samla-ito Samla-iten Samla-ite Samla-itiS Samla-itam
SVF Samla-yto Samla-yten Samla-yte Samla-ytiS Samla-ytam
SVD Samlato Samlaten Samlate SamlatiS Samlatam

Samla-icchilo Samla-icchilen Samla-icchile Samla-icchili Samla-icchilam
SVF Samla-ycchilo Samla-ycchilen Samla-ycchile Samla-ycchili Samla-ycchilam
SVD Samlacchilo Samlacchilen Samlacchile Samlacchili Samlacchilam

Samla-iechilo Samla-iechilen Samla-iechile Samla-iechili Samla-iechilam
a-M Samle-iechilo Samle-iechilen Samle-iechile Samle-iechili Samle-iechilam
VR Samli-iechilo Samli-iechilen Samli-iechile Samli-iechili Samli-iechilam
DG Saml-iechilo Saml-iechilen Saml-iechile Saml-iechili Saml-iechilam
Mi-D Samlechilo Samlechilen Samlechile Samlechili Samlechilam

Samla-ibe Samla-iben Samla-ibe Samla-ibi Samla-ibo
SVF Samla-ybe Samla-yben Samla-ybe Samla-ybi Samla-ybo
SVD Samlabe Samlaben Samlabe Samlabi Samlabo

Samla-ibe Samla-iben Samla-io Samla-iS -
a-M - - Samle-io - -
VR - - Samli-io - -
DG - - Samlio - -
SVF Samla-ybe Samla-yben - Samla-yS -
SVD Samlabe Samlaben - SamlaS -

Samla-ite Samla-ie Samla-ile Samla-ba Samla-no
a-M - Samle-ie - - -
VR - Samli-ie - - -
DG - Saml-ie - - -
Mi-D - Samle - - -
SVF Samla-yte - Samla-yle - -
SVD Samlate - Samlale - -
Samlaba Samlano

3.5. Exceptions

Within the closed class of verb morphology there remain a few exceptions which apparently fall beyond the scope of the proposed mechanism here. The existence of such exceptions in the language, as is said in section 3.1., is a perfectly natural phenomenon and these are treated in terms of equally natural concepts, viz. word boundary, stem allomorphy, suffix allomorphy, and suppletion, the treatment that are factually justified also as follows:

3.5.1. Perfect forms of section 3.4

The perfect forms, i.e. those of 3 and 8, in terms of VR may generate forms like *korichi, *bolichiS, *korichilam etc., a problem that remains unaccounted for in both the previous analysis, viz. in Dasgupta (1982) and Paul (1985).

The present system accounts for them by postulating a word boundary immediately after the perfect suffix e in the UR, a remedy that stops the VR from applying to e across the word boundary. Hence for example, the underlying form of korechi is kOr-ie#chi, and the VR applies only to the O of the stem.

The forms like kore-i-chi, and kore-o-chi may justify the existence of such word boundary, where the emphatic suffix i or the inclusive suffix o is allowed to occur after e, i.e. precisely after a word boundary.

3.5.2. Forms in b

In this subsection mention can be made of the forms of 9A, B, C, E; 10A, B of stems under group aii, and 9A, B, C, D, E; 10A, B of stems under group bii, for example, *hobe, *hoben, *hobo, *khebe, *kheben, *khebi, *khebo for hObe, hOben, hObo, khabe, khaben, khabi, and khabo respectively.

These forms, as far as the environments specified for the a-M and VR are concerned, should undergo them, but they do not.

These exceptions may be accounted for in terms of a pair of phonologically conditioned allomorphs, viz. ib and b for the future suffix morpheme. The distributions of ib and b are as follows:

ib occurs after consonant stems, e.g. kaT-ibo à kaTbo, kOr-ibe à korbe etc.; whereas b occurs after vowel ending stems, e.g. kha-bo à khabo, So-be à Sobe etc.

It should be noted, however, that in the case of polysyllabic stems (which in Bangla always end in a vowel), there is no evidence to show whether the allomorph is b or ib.

That the postulation of these phonologically conditioned allomorphs b and ib is the right way to approach the problem is also suggested by the behaviour of the forms I have called "second gerunds" in my M. Phil work (De 1984). Consonant ending stems in second gerunds in some people's speech show variations like kOrbar~korbar (under certain socio-linguistic conditions which have not yet been studied). We would imagine that a "free variation" between ba and iba is the best description of the variation.

Hence in the cases of the stems under groups aii and bii b is the form of the suffix; hence VR is not triggered.

3.5.3. ga, na, SO, bO, do, ro

The above roots, viz. ga 'to sing', na 'to bathe', SO 'to endure', bO 'to carry', do 'to milk', ro 'to sow', pose problem for all the three systems, viz. Dasgupta (1982), Paul (1985), and the present one.

Dasgupta (1982: 22) handles the problem by positing a final -h in the UR of these roots and making the rules sensitive to it. Thus according to Dasgupta underlyingly these roots are gah, nah, SOh, bOh, doh, and roh respectively.

Paul (1985: 183) too posits the final -h in the underlying representations of these roots and thus groups them under one specific class, and marks this class as

- i Deletion
- i Assimilation

In the present system these roots result into unacceptable forms in some 7 out of 11 categories, viz. 2, 5, 6, 7, 9, 10, and 11. In other words, forms like boyche, gayche, boylo, gaylo, boyto, gayto, boychilo, gaychilo, boybe, gaybe etc. pose problem for the present system. These forms, though they apparently show the environment specified for ST and VR they often do not undergo these two rules.

The present system, however, rejects the idea of positing an abstract -h underlyingly and accounts for such problematic forms in terms of a pair of morphologically conditioned allomorphs for each stem, e.g. ga~gay, na~nay, SO~SOy, bO~bOy, do~doy, ro~roy etc., between which the member with the final -y occurs before a few specific inflectional suffixes, e.g. ib, it, il, icche etc.

Thus we could do away with abstract segments and obey the constraints satisfactorily.

3.5.4. Irregular verb roots

Two types of irregular verb roots are as follows:

3.5.4.1. ach, ja, aS

ach 'to be' / 'has', ja 'to go', aS 'to come' are three well-known irregular verb roots of Bangla, involving suppletive allomorphy.

In order to account for the exceptional forms of these stems, at places I postulate full suppletive forms in the lexicon, which, in terms of the blocking principle of Aronoff (1976) stop the otherwise expected but unacceptable forms, e.g. gElo, aY etc.; and at other places I prefer to postulate pairs of morphologically conditioned allomorphs, e.g. ja~ge, ach~thak etc.

3.5.4.2. de, ne

Roots de 'to give', and ne 'to take' are irregular in the sense that they do not show the usual one degree vowel height alternation, viz. mid and high vowels, rather they show four alternate forms involving all the three vowel heights, e.g. di 'I give', de 'you (inf.) give', dEn 'he (hon.) gives', daW 'you (non-hon.) give' etc. Among these four alternants the one containing the high vowel, e.g. i, appears at the places specified for the roots with high vowel, e.g. diS, dik, nilen, nitiS etc. (cf. SuS, Suk, Sulen, SutiS etc.); and the other three alternants share the places specified for the roots with mid vowels, e.g. ne, neba (cf. So, Soba); dEY, dEn (cf. SoY, Son); daW (cf. SoW) etc.

These forms too could be taken care of in terms of postulating stem allomorphy.

3.5.5. A few deverbal forms

Forms like Obujh, Ocin, ghurghur, Dub, hOrbola, buli etc. (cf. section 3.2.) come under this category and they may be considered as exceptional forms of quite different sort.

Such deverbal forms contain high and mid vowels, whereas their corresponding basic verb roots, in accordance with the present hypothesis, contain mid and low vowels respectively, e.g. bojh 'to understand', cen 'to recognize', ghor 'to roam', Dob 'to sink', bOl 'to say'. Thus these forms, as I mentioned in section 3.2., stand as counterexamples to the claim of the present chapter that the verb roots with low and mid vowels are the basic ones and those with mid and high vowels are derived from them.

However, these few forms do not really stand as very strong evidence against the claim of the present chapter because of the following reasons:

Firstly, these are just a few isolated forms, i.e. they do not represent any particular functional class of the language.

Secondly, none of these forms is the result of any productive phonological or morphological processes of the language, rather, as has been said before, they are just a few scattered exceptions.

In fact, the actual status of these isolated forms would better be judged in terms of the gradiential view of productivity of Mark Aronoff, as has been discussed and extended in Dasgupta (1983: 122). Dasgupta prefers to extend Aronoff's gradiential view of morphological productivity even to the phonological sector of the language and thus to measure the degree of 'livingness' of different synchronic patterns.

In accordance with the above treatment such isolated forms, as they do not form any 'pattern' as such, would not be considered as quite 'alive' forms, or even the results of some 'alive' synchronic process. Rather parallel forms like Ocena 'unknown', ghoraghuri 'continuous prying movement' etc. would be considered as more 'alive' compared to the exceptional forms. The exceptional forms might be considered as a few residual forms of 'something else'. Such conclusions, which seem plausible, make it clear that the facts do not seriously threaten our theory.

3.6. Conclusion

Above is the analysis of the segmental phonology of a quite representative domain, viz. the verb morphology, of Bangla. As is said before the framework used here belongs to the generative tradition and the analysis is done in accordance with the principles of a modified version of the approaches of SPE, viz. the concrete approach as propounded by Kiparsky (1968b).

However, segmental phonemes are the smallest phonological units in a language. In order to go beyond the segmental aspect of phonology we need to investigate the larger phonological units, viz. syllable and foot, through the approaches of metrical phonology. The following chapters will deal with the larger phonological units of Bangla in terms of the principles of metrical phonology.