Somdev Kar Assistant, Professor, IIT Ropar somdev.kar@iitrpr.ac.in Abstract This paper attempts to address certain glide-formation processes found in the verbal inflections of Bangla using the framework of Distributed Morphology (Halle & Marantz, 1993). Although the morphology in modern Bangla is nonexistent for adjectives and minimal for nouns, it is very productive for verbs (Klaiman, 1987). The most striking parameter for subject-verb agreement in Bangla is the person and the degree of formality attached to the person in this language (Dasgupta, 2003, Kar, 2010). The verbal inflections are possible to categorize in terms of their person and formality levels. However, that does not address certain phenomena of glide-formation among some of the inflected cases. The approach of Distributed Morphology (DM) tries to address this issue through slot-divisions of each inflected case. In a standard verbal inflectional paradigm, every inflectional form of selected verbs is divided into various terminal nodes (slots), which are filled up by either a stem or an affix. These are technically called vocabulary items which are necessarily without phonological features in DM framework. In most of the cases, there are more than one vocabulary item to be inserted in a terminal node. In this structure, such elements gain their eligibility to be inserted in the terminal nodes based on the feature specifications. The DM framework is used here to illustrate such cases where Bangla shows specific instances of glide formation in certain morphological slots. At the initial parts of is work, a comparative account of verbal inflections of English and Bangla in terms of DM analysis is presented. Since, the phonological features are considered late insertions, they are presented in form of formants (using PRAAT) at the final stage of this study. Keywords: Bangla, Bengali, syllable, distributed morphology, glide formation
1. Introduction A high morpheme-to-word ratio is found in different languages of the world where each affix typically represents one unit of meaning (such as number, tense, person etc.). However, these meanings are not always lexical in nature. Bangla, along with some other major Indian languages, shows a certain pattern of morpheme-to-word structure. According to Klaiman (1987), morphology in modern Bangla is non-existent for adjectives it is also observed that nouns in Bangla show a minimal amount of morphology. On the other hand, it is very productive for verbs. The verbal bases found in Bangla can produce an interesting set of inflected forms with a very interesting pattern that gives a new edge in the study of morphology of this language. There are mainly two types of verbal bases present in Bangla, viz., monosyllabic (e.g., /ken-/ buy ) and disyllabic (e.g., /kāmɽaā-/ bite ) with a clear majority of the monosyllabic bases. In this work, two specific monosyllabic verbal bases are taken for analysis, based on their syllabic structure. One of them, /kʰa-/ eat, has an open syllable (vowel-ending) and other one, /ken-/ buy is a consonant-ending verbal base. Dasgupta (2003) describes Bangla as a language without any case or number agreement. Unlike some other Indian languages, such as Hindi, it does not show any kind of grammatical gender phenomena. In Bangla, the subject-verb agreement solely depends on the persons (first, second and third) and the status of the person in terms of the level or degree of formality which is applicable only for the second and third persons. The formality level is shown as F (Formal), P (Polite) and I (Intimate) for the second person and only F and P for the third person. So the personal pronouns are: /apni/ You (F), /tumi/ You (P), /tui/ You (I), /tini/ He/She (F) and /ʃe/ He/She (P). The third person shows only one form for both polite and intimate forms (see Kar, 2010). This type of formality divisions is not present in English, but one can find a similar system in German, where Sie stands for you (Formal) and Du for you (informal). This distinction is also visible in Hindi in terms of /ap/ -- You (F), /tum/ -- You (P) and /tu/ -- You (I). Note that the intimate form (/tui/) may also be used as a derogatory form of second person, which is sometimes used to address a person of lower standard which is purely decided by the speaker. 2
2. Methodology: Distributed Morphology Distributed morphology (Halle and Marantz, 1993) is a framework of morphological analysis which is considered a bridging theory between the lexicalist (Chomsky 1981) and a-morphous (Anderson 1991) approaches. In distributed morphology (hereafter, DM), vocabulary items are the major elements of study, which form the abstract morphemes. Unlike other theories, in DM analysis, the terminal nodes contain three units, namely, semantic, syntactic and morphological features, but the nodes exclude the phonological feature. According to Noyer and Harley (1999), the syntax proper does not manipulate anything resembling lexical items in DM. This, rather, generates structures by combining morphosyntactic features (via Move and Merge) selected from the inventory available, subject to the principles and parameters governing such combination. The major features of DM that helps these operations are Underspecification, Syntactic Hierarchical Structure All the Way Down and Late Insertion (of phonological features). In this connection the same study of Harley and Noyer claims that vocabulary items provide the set of phonological signals available in a language for the expression of abstract morphemes. The phonological content of a vocabulary item may be any phonological string, including zero or null-ø. This scenario could be presented as the following figure. (1) DM structure Semantic features Syntactic features Morphological feature Phonological features In a nutshell, DM allows breaking a morphological form into one or more slots or terminal nodes depending on the morphosyntactic features. However, these slots do not contain any phonological feature until the syntactic operations are completed. Among several candidates (vocabulary items) contesting to be inserted in a terminal slot, only the right one is chosen based on certain features. Noyer and Harley (1999) defines a vocabulary item as a relation between a 3
phonological string or piece and information about where that piece may be inserted. According to them, DM is also considered piece-based in the sense that the elements of both syntax and of morphology are understood as discrete constituents instead of as (the results of) morphophonological processes. According to Kar (2010), the following table represents the ccomplete set of suffixes for the AGR slot in Bangla verbal inflections. These are the vocabulary items contesting for the specific positions respective to the tense and persons. (2) Vocabulary item for the AGR slot SL Affix Feature i. - / [+S] / [-Past, -Future] ii. -am [+S] / [+Past] iii. -o [+S] / [+Future] iv. -(e/u)n [+Formal, -S] v. -(u)k [-Formal, -Intimate, -S] / [-Future, +Imperative] vi. -( )o [-Formal, -Intimate, -S] / [+Future, +Imperative] vii. -o [-Formal, -Intimate, -S, +A] / [-Past] viii. -o [-Formal,-Intimate, -S, -A] / [+Past] ix. -i [+Informal, -S, -A] / [+Past, -Conditional] x. -i [+Informal, -S, -A] / [+Future, -Imperative] xi. -Ø [+Informal, -S, -A] / V-Stem [-Past, -Future] xii. -(i)ʃ [+Informal, -S, -A] xiii. - / ELSEWHERE In addition to the table in (2), let us also consider a set of affixes for the tense slot from the same source. (3) Vocabulary item for the tense slot SL Affix for Tense Feature i. -l- [+Past] ii. -t- [+Past, +Conditional] iii. -( )e- [+Past_Participle] 4
SL Affix for Tense Feature iv. -Ø(cʰ )- [+Present_Participle] v. -b- [+Future] vi. Ø [+Future, +Imperative] / [-Formal, -S, +A] vii. Ø ELSEWHERE From the table in (2) the sixth and thirteenth affixes along with the third affix from the table in (3) will be of interest in this study. In the following sections, we will explore the formation of these affixes in the Bangla verbal morphology. 2. Bangla verbal inflections In Bangla, all the inflected forms (verb) have a general structure in terms of finiteness and DM slots. Every inflected form contains an infinite and a finite part. The basic structural elements like stem, tense, agreement (AGR) and suffixes are placed under each of them. The finite part has all the three elements in the order: stem tense AGR, while the infinite part lacks the tense slot. This structure represents most of the inflected forms of verbs in Bangla. (4) Bangla morpheme (verb inflection) structure Morpheme (Verb) INFINITE FINITE Stem -- Perfective Suffix + Stem -- Tense -- AGR (1) (2) (3) (4) (5) As illustrated in (4), the FINITE part of Morpheme (Verb) is the base structure present in all verbal forms of the inflectional paradigm. However, in certain inflected forms the INFINITE part (the left-hand side of Morpheme) appears as an additional element. A verbal inflectional paradigm of two base forms of Bangla (/kʰa-/ and /ken-/) is illustrated in a tabular form in (5). In this table, every cell of the present perfect (PRP), past perfect (PP) and future imperative (FI) columns contains two inflected forms. The first (upper) 5
one is the inflected form of /kʰa-/ eat. The other (lower) one is for /ken-/ buy and given for an instant comparison. Note that the imperative case is not applicable to the first person (1P); hence, that cell remains empty. Also note that 2P and 3P denotes second person and 3 rd person, respectively, while F, P and I stand for Formal, Polite and Intimate, respectively (Kar 2010). (5) Bangla Verbal Inflections for the base /kʰa-/ and /ken-/ Tense Person 1P F 2P P I F 3P P Present Perfect (PRP) Past Perfect (PP) Future Imperative (FI) kʰ - -ʧʰ-Ø-i kʰ - -ʧʰi-l-am kin-e-ʧʰ-ø-i kin-e-ʧʰi-l-am --- kʰ - -ʧʰ-Ø-en kʰ - -ʧʰi-l-en kʰa-b-en kin-e-ʧʰ-ø-en kin-e-ʧʰi-l-en kin-b-en kʰ - -ʧʰ-Ø-o kʰ - -ʧʰi-l-e kʰ - - o kin-e-ʧʰ-ø-o kin-e-ʧʰi-l-e kin-ø-o kʰ - -ʧʰ-Ø-iʃ kʰ - -ʧʰi-l-i kʰa-ø-ʃ kin-e-ʧʰ-ø-iʃ kin-e-ʧʰi-l-i kin-ø-iʃ kʰ - -ʧʰ-Ø-en kʰ - -ʧʰi-l-en kʰa-b-en kin-e-ʧʰ-ø-en kin-e-ʧʰi-l-en kin-b-en kʰ - -ʧʰ-Ø-e kʰ - -ʧʰi-l-o kʰa-b-e kin-e-ʧʰ-ø-e kin-e-ʧʰi-l-o kin-b-e According to Bhattacharya (1993), the only auxiliary verb in Bangla is /-ʧʰ(ʧʰ)-/ which is a derived form of the main verb /aʧʰ-/ (to be) having the loss of the initial stress. For instance we may consider the past perfect (PP) forms from (5). The third, fourth and fifth (i.e., last three) parts of all the forms in that column show a finite form of / aʧʰ-/ (such as [-ʧʰilam] in the first person). But these are finite forms only when they are taken separately. But, together they constitute the finite form as in [kineʧʰilam] (I) have bought as shown in (6). Bangla has both finite and infinite verb forms as distinct word-forms, but sometimes they appear together as illustrated in (6). In these specific cases, the infinite form acts as the main verb and finite form is an auxiliary verb. 6
(6) Bangla morpheme (verb inflection) structure of [kineʧʰilam] Morpheme (Verb) INFINITE FINITE Stem -- Perfective Suffix + Stem -- Tense -- AGR (1) (2) (3) (4) (5) kin e ʧʰi l am There are clearly five different morphosyntactic slots present in the inflected forms (verb) of Bangla, which are represented by the following terminal nodes in the structure (Kar, 2010). (7) Base1 (Stem) -- Perfective Suffix -- Base2 (Stem) -- Tense -- AGR (1) (2) (3) (4) (5) Some of the inflections, such as the perfect cases, make use of only the last three slots (slot 3, 4 and 5) as described in (7), namely, the Base2, Tense and AGR. In the other forms (as in imperative cases), all the five slots are in use. In Kar (2010), an analysis of the verbal inflections in Bangla marks the affixes of different slots of the inflected forms in the framework of Distributed Morphology. However, as suggested in the DM structure, all the phonological information at every syntactic slot is inserted after the morphological operations (the late insertion rule). Kar (2010) suggests that in some particular cases, the semivowel [e] substitutes the pure vowel [e] at the AGR slot or simply occurs before or after [e] in case of perfective suffixes, when preceded by another vowel. It could be a result of diphthongization or a glideformation depending on the respective case. The null morphs (Ø, as shown in the inflected forms) are omitted in the phonological analysis in the following sections of this study. It does not have any significance in the phonological representations of the morphologically constructed forms. Also, the slot markers (-) are not shown in the phonological forms. Instead, the phonological forms are divided into syllables by using dots (.). 7
3. Glide formation at the perfective suffix slot Perfective suffix occurs at the second slot of a five-slot structure as shown earlier in (4) and (6). This observation is true for both present and past perfect cases in Bangla as illustrated in (5). However, the insertion of [e] does not occur universally and irrespective of the ending of the verbal base. The said insertion occurs in the first, second and third person cases of /kʰa-/ while there is no change in the forms of /ken-/. As of the later cases, the perfective suffix [e] occurs in between two consonants, while in case of /kʰa-/, the perfective suffix occurs after a vowel [e] and followed by the same consonant as the second cases do. Since the other elements of the environment are same throughout the relevant columns, only one case will be taken for the analysis and the outcome will be applied uniformly to all other cases. Let us consider the first person inflected forms for the first person of present perfect (PRP) and past perfect (PP). (8) Present perfect (1P) and past perfect (1P) inflections PRP PP 1P kʰe.ee.ʧʰi.lam ki.ne.ʧʰi.lam kʰe.ee.ʧʰi ki.ne.ʧʰi We will apply the basic rules of syllabification, viz., syllable projection, onset formation and coda formation to these forms before any further rule formation. The remaining cases completely fit into the same moraic structures. (9) Syllable projection, onset and coda formation: [ki.ne.ʧʰi.lam] σ σ σ σ μ μ μ μ μ μ μ μ μ k i n e ʧʰ i l a m k i n e ʧʰ i l a m [ k i n e ʧʰ i l a m ] [ k i n e ʧʰ i l a m ] 8
(10) Syllable projection, onset and coda formation: [ki.ne.ʧʰi] σ σ σ μ μ μ μ μ μ k i n e ʧʰ i k i n e ʧʰ i [ k i n e ʧʰ i ] [ k i n e ʧʰ i ] (11) Syllable projection, onset and coda formation: [kʰ e.ʧʰi.lam] σ σ σ σ μ μ μ μ μ μ μ μ μ kʰ e ʧʰ i l a m kʰ e e ʧʰ i l a m [ kʰ e e ʧʰ i l a m ] [ kʰ e e ʧʰ I l a m ] (12) Syllable projection, onset and coda formation: [kʰ e.ʧʰi] σ σ σ μ μ μ μ μ μ kʰ e ʧʰ i kʰ e e ʧʰ i [ kʰ e ʧʰ i ] [ kʰ e e ʧʰ i ] In the following illustrations, instead of showing each of these operations separately, a unified transition is given to present all three operations shown in (9)-(12). It is evident from the structures shown in (9) through (12) that a non-low and nonsyllabic [e] occurs in between two non-low vowels in the surface form (with phonological 9
features) of these verbal inflections of Bangla. This inventory premises a rule that says a non-low and non-syllabic glide V3 is inserted in a V1V2 sequence where both V1 and V2 are non-low front vowels with the feature [+syll]. This outcome leads to a glide formation rule formed for such cases in Bangla as shown in (13) and the implementation of the same in (14) and (15). (13) Glide-formation rule σ σ μ μ -cons - low - high - back + syll -cons - low - high + syll (14) Syllable projection, onset and coda formation: [kʰ e.ʧʰi.lam] σ σ σ σ μ μ μ μ μ μ μ μ μ kʰ e ʧʰ i l a m kʰ e ʧʰ i l a m [ kʰ e ʧʰ i l a m ] [ kʰ e ʧʰ i l a m ] (15) Syllable projection, onset and coda formation: [kʰ e.ʧʰi] σ σ σ μ μ μ μ μ μ kʰ e e ʧʰ i kʰ e ʧʰ i [ kʰ e e ʧʰ i ] [ kʰ e ʧʰ i ] 10
As illustrated in (14) and (15), this rule indicates that one particular nucleus is shared by the following syllable in the form of the onset. This applies to both the cases where the base form is an open syllable. The same would not be true for the closed syllables as in (9) and (10). The actual derivation is shown in the following illustrations (16) and (17). (16) Derivation 1: UR /ki.ne.ʧʰi/ / kʰ ʧʰi/ Glide-formation ----- kʰ e.ʧʰi PR [ki.ne.ʧʰi] [kʰ e.ʧʰi] (17) Derivation 2: UR /ki.ne.ʧʰi.lam/ / kʰ ʧʰi.lam/ Glide-formation ----- kʰ e.ʧʰi.lam PR [ki.ne.ʧʰi.lam] [kʰ e.ʧʰi.lam] First two slots are the focus of importance here. From the third slot starts a new syllable for both the tense forms, which is irrelevant in this part of the study. The perfective suffix slot (i.e. the second slot) consists of a vowel [e] which can either join the last element of the first slot, i.e., another [e] and form a diphthong or can occur separately and form a hiatus. But neither happens. The sequence [ee] is not a valid diphthong in Bangla (see Sarkar 1985). Therefore, this possibility is ruled out. Then the hiatus option is also not acceptable and hence the glideformation takes place. In the following spectrogram image, two parts are divided medially by a vertical line and those parts represent the spectrograms of [kʰ ʧʰi] and *[kʰ ʧʰi], respectively. In both the spectrograms, some gaps are clearly visible. In [kʰ ʧʰi], there is only one gap before [-ʧʰi]. However, in *[kʰ ʧʰi], there are two gaps, before and after the sound [e]. A crude measurement says that the first gap is of 120 milliseconds (approx.) and the second gap is of 90 milliseconds for the latter. The only gap in [kʰ ʧʰi] gives a measurement of approximately 65 11
milliseconds. The sizes of the second ones are closer than the first one. So, let us consider the second gap as a common phenomenon and omit this from the present analysis. (18) Spectrogram for [kʰ ʧʰi] and *[kʰ ʧʰi] [kʰ.ʧʰi] *[kʰ ʧʰi] The first gap in *[kʰ.ʧʰi] is caused by glottalization which leads to a possible hiatus. In [kʰ ʧʰi] glottalization does not take place, because of glide-formation. Eventually, the hiatus is also resolved by the same operation. 4. Glide-Formation at the AGR slot The future imperative of second person polite form has a very interesting occurrence of a semivowel at the AGR slot (agreement). In addition to the existing suffix [-o], the semivowel [e] is inserted before it. This insertion does not form a diphthong, since the semivowel is posited before the vowel and that is not allowed in the regular diphthongs of Bangla. The relevant cases of future imperative in Bangla from (5) are listed in (19). 12
(19) Second Person Polite Future Imperative (2PP-FI): /kin-o/ [ki.no] /kʰe-o/ [kʰe.eo] In moraic phonology (Hayes, 1989; Hyman, 1986; Prince & McCarthy, 1996), when a sequence of vowel and glide does not represent a diphthong, then the vowel is linked to a mora (as the nucleus) and the glide is directly linked to the syllable as the onset of that syllable. Keeping this formulation in mind, let us apply the first three rules of syllabification to our present cases. (20) Syllable projection, onset and coda formation: [ki.no] σ σ μ μ μ μ k i n o k i n o [ k i n o ] [ k i n o ] (21) Syllable projection, onset and coda formation: [kʰ o] σ σ μ μ μ μ kʰ e o kʰ e o [ kʰ e o ] [ kʰ e o ] Here, in case of [kʰe.eo], the underlying representation does not match with the phonological representation in the above moraic structures. An additional [e] is inserted in order to avoid a possible glottalization and hiatus. Apparently, the Onset Formation rule can add this 13
inserted element in the syllable as the onset. However, it is not possible, because the element [e] is not in the underlying representation. We need a new rule to introduce this element in the syllable structure. From the above discussion it is observed that a non-low and non-syllabic [e] gets inserted between the non-low vowels [e] and [o] in the syllable structure of the verbal inflected forms of Bangla. These observation leads to the argument that a non-low and nonsyllabic glide V3 is inserted in a V1V2 sequence where both V1 and V2 are non-low and syllabic. However, V1 is a front vowel when V2 is back in the present scenario which leads to a generalized rule of glide formation for Bangla. It is illustrated in (22) followed by the implementation of the rule in (23). (22) Glide Formation (generalized) σ σ μ μ - cons - low - high - back + syll - cons - low - high + back + syll (23) Syllable projection, onset and coda formation: [kʰ o] σ σ σ σ μ μ μ μ kʰ e o kʰ e o [ kʰ e o ] [ kʰ e o ] 14
The actual derivation of the generalized glide formation rule is illustrated in (22). (24) Derivation: glide formation of [kʰe.eo] UR /ki.no/ /kʰe.o/ Glide-formation ---- kʰe.eo PR [ki.no] [kʰe.eo] This phonological rule produces the open syllable (/kʰa-/ [kʰe.eo]) in a disyllabic form ([kʰe] and [eo]), while the /kin-/ form ([ki.no]) remains monosyllabic. If the /kʰa-/ form remains monosyllabic and the glide does not occur in the form *[kʰe.o], there would be an unacceptable sequence of [eo], which would lead to a possible hiatus. However, the rules of Bangla phonology do not allow such hiatus. So, it is necessary to bring some element to divide the cluster, which will omit the possible hiatus. According to Chatterji (1921), [e] is a very frequent occurrence between two vowels to avoid a hiatus in a single breath-group and this observation surely validates the rule formulated here. Only the absence of the glide-formation would lead to a syllable having only the nucleus (here, [o]) in the monosyllabic form *[kʰe.o]. In that case, glottalization will take place at the beginning of the syllable as shown in the following spectrogram (25). In is spectrogram (25), the straight line divides the spectrograms into two parts: [khe.eo] and *[khe.o]. A clear gap of approximately 26 milliseconds is visible in the second spectrogram which denotes the glottalization at the beginning of the syllable. It is positioned in between the sounds [e] and [o]. This gap caused a possible hiatus, which is opted out by the insertion of the glide [e] in the first part of the spectrogram. 15
(25) Spectrogram for [khe.eo] and *[khe.o] [kʰ o] *[kʰe.o] Glottalization and glide-formation are found in many other languages like German, Japanese, Italian etc. Van der Veer (2006) and Itô (1986) observed similar instances in Italian and Japanese, respectively. 5. Conclusion This study tried to analyze certain phonological aspects of verbal inflections of Bangla present in the morphological structure. The approach of distributed morphology brings a different angle in the study where the phonological features come very late. We first broke the inflected forms into terminal nodes as suggested by the DM framework and then tries to formulate the rules for the glides that are inserted at specific positions. It is re-established that hiatus is not allowed in Bangla and the specific situation is avoided by glottalization and glide formation. The rule of glide formation is also introduced. 16
Bibliography Bhattacharya, K. (1993). Bengali-Oriya verb morphology. Calcutta: Dasgupta & Co. Pvt. Ltd. Chatterji, S. K. (1921). Bengali phonetics. Bulletin of the School of Oriental Studies, 2(1), 1-25. Dasgupta, P. (2003). Bangla. In G. Cardona & D. Jain (Eds.), The Indo-Aryan Languages (pp. 351-390). London: Routledge. Halle, M. and A. Marantz (1993). Distributed Morphology and the Pieces of Inflection. The View from Building 20. Cambridge, MA: MIT Press. pp. 111 176. Hayes, B. (1989). Compensatory lengthening in moraic phonology. Linguistic Inquiry, 20. pp.253-306. Hyman, L. M. (1986). A theory of phonological weight. Dordrecht: Foris Publication. Itô, J. (1986). Syllable theory in prosodic phonology. University of Massachusetts Amherst, Amherst. Kar, S. (2010). Syllable Structure of Bangla: Optimality-Theoretic Approach. Newcastle: Cambridge Scholars Publishing. Klaiman, M. H. (1987). Bengali. In B. Comrie (Ed.), The World's major languages (pp. 490-513). London & Sydney: Croom Helm. Noyer, R., & Harley, H. (1999). Distributed morphology. Glot International, 4, 3-9. Prince, A. & McCarthy, J. (1996). Prosodic morphology 1986 (No. TR-32). New Brunswick: Rutgers University Center for Cognitive Science. Van der Veer, B. (2006). The Italian "mobile diphthongs": a test case for experimental phonetics and phonological theory. Utrecht: Netherlands Graduate School of Linguistics, LOT Utrecht, Leiden University. 17