¹ Ferguson wrote the first draft of §§0–5, 7–9, 11.2; Chowdhury wrote the first draft of §§6, 10, 11.1; and both worked through all the material. Ferguson assumes responsibility for the final version.

^1a For further information of the varieties of Bengali and their interrelations (including bibliography of previous work), see E. A. Dimock, Literary and colloquial Bengali in modern prose, and M. Chowdhury, The language situation in East Pakistan, both to appear in Linguistic diversity in South Asia (ed. C. A. Ferguson and J. J. Gumperz).

^1b For a full description of the Bengali writing system and its relation to the pronunciation, see H. M. Lambert, Introduction to the Devanagari script 171–231 (Bengali section in collaboration with G. M. Summers; London, 1953). Throughout the book both ‘systematic transcription’ (= transliteration) and ‘phonetic transcription’ (= modified phonemic transcription) are used.

² Suniti Kumar Chatterji, Bengali phonetics, BSOS 2.1–25 (1921), also reprinted separately for the International Phonetic Association as Brief sketch of Bengali phonetics [hereafter Sketch] (London and Paris, 1921). Chatterji subsequently published other treatments, including sections in his Origin and development of the Bengali language [hereafter ODBL] 267–74 (Calcutta, 1926); A Bengali phonetic reader [hereafter Reader] 7–27 (London, 1928); and Bāṅlā bhāṣā vyākaraṇ ³ 19–78 (Calcutta, 1945). The original Sketch remains the fullest statement, although the subsequent treatments added certain details.

³ M. A. Hai, A phonetic and phonological study of nasals and nasalization in Bengali (unpubl. M.A. thesis, University of London). Ferguson had read this previously, but unfortunately it was not available at the time the present study was written; if it had been, fuller information could have been provided here at certain points.

⁴ C. A. Ferguson, Phonology and morphology of Standard Colloquial Bengali (unpubl. Ph.D. thesis, University of Pennsylvania, 1945).

^4a After this article was completed, Ferguson had access to another study of Bengali phonology: pages 1–15 of E. M. Bykova and I. S. Kolobkov, Kratkie svedenija po fonetike i slovoobrazovaniju bengalʹskogo jazyka, issued separately as a supplement to the Bengali-Russian dictionary of these authors and M. A. Elizarova (Moscow, 1957). It covers many of the main points of Bengali phonology, but offers no new data.

⁵ W. Sutton Page, An introduction to colloquial Bengali [hereafter Introduction] (Cambridge, 1934).

⁶ The four juncture symbols in this section /‖ | ʹ -/ are enclosed in slant lines as part of a phonemic transcription; but in accordance with the usage of Hockett (op. cit. 74–5) as well as most European linguists, the term ‘phoneme’ is restricted to the segmental elements. The juncture symbols were chosen to exhibit maximum similarity with Chatterji's transcriptions in the Reader. The chief differences in the use of the symbols are as follows: Chatterji often transcribed with ‖ what would be regarded here as |; Chatterji used space between orthographic words, which corresponds to /-/ or, more often, to no juncture at all; Chatterji's stress mark corresponds to both /ʹ/ and /ˆ/.

⁷ Cf. N. S. Troubetzkoy, Principes de phonologie 298–305 (Paris, 1949); C. F. Hockett, Manual of phonology 59–64, 167–72 (Baltimore, 1955); D. Jones, The hyphen as a phonetic sign, Zeitschrift für Phonetik 9.99–107 (1956).

⁸ Brief comments on the intonation of SCB have appeared in Chatterji's Sketch (62, 63) and Reader (24–7), and in Page's Introduction (14). Particularly puzzling is Chatterji's assertion that intonation has no distinctive significance in Bengali except for certain one-word sentences. Almost any sentence of SCB may be spoken with several different contours, each of which has an identifiable range of semantic value, and in some cases the contours and their meanings differ strikingly and offer convenient minimal pairs.

⁹ For a similar phenomenon in Marathi, see A. R. Kelkar, The phonology and morphology of Marathi §1.1, §1.2.1 (Ph.D. thesis Cornell Univ., 1958).

¹⁰ Cf. C. A. Ferguson, Word stress in Persian, Lg. 33.132 fn. 23 (1957). This drawl is the ‘phrase-final length’ of C. T. Hodge, Some aspects of Persian style, Lg. 33.359 (1957).

¹¹ In Bengali orthography this interrogative particle -ki is written as a separate word, which makes it and ki ‘what’ heteronyms. Rabindranath Tagore suggested using the ‘short i’ of the Bengali alphabet for -ki and the 'long ī' for ki 'what'.

¹² This pattern—falling intonation with a pitch-stress peak on the interrogative—may possibly be attributed to Proto-Indo-European, since it is found in a number of branches of the family. Cf. Ferguson, Word stress, Lg. 33.133 (1957).

¹³ Two particularly cogent statements are D. Bolinger, Levels versus configurations, Word 7.199–210 (1951), and W. Haas, The identification and description of phonetic elements, TPS 1957. 118–59, esp. 154–9. To their arguments may also be added the fact that children seem to learn American English intonation ‘phonemes’ not in step-by-step development of oppositions as with consonant and vowel phonemes and even with stress levels, but as contours.

In spite of these objections, serious investigation of intonation phenomena in Indo-Aryan languages, as well as many others, very soon shows more contours than pitch levels and ‘terminal junctures’, as well as complexities in the analysis and symbolization of turning points in the contours. Until something better is devised, the pitch-level method of Pike, Wells, Trager-Smith, Hockett seems to be the most promising one available, certainly superior to that of Palmer or of Armstrong-Ward and Jones. Kelkar's highly original analysis of Marathi intonation, although presented in extremely concise fashion, is also very promising.

¹⁴ The aspirated consonants are regarded here as unit phonemes, the digraph symbolization being for typographical convenience. This question is discussed in detail in §8 below.

^14a Instructive analyses of the ḍ-ṛ situation in other Indo-Aryan languages include BSOS 6.673–7 (1931); Indian linguistics 16.120–3 (1955); BSOS 20.7, 8 (1957); Lg. 34.215 1958); Masud Husain, A phonetic and morphological study of the word in Urdu 19–24 (Aligarh, 1953).

¹⁵ Sketch 7.

¹⁶ For a fuller presentation of the East Bengali facts, see M. Chowdhury, op.cit.

¹⁷ The symbol ‘s’ is used to represent the palatal phoneme only for typographical convenience. When there is only one sibilant phoneme in a language it seems simpler to use the letter ‘s’ for it regardless of its phonetic value. For certain purposes it might be more informative to use a symbol such as 'š' or 'ʃ' for the scb/s/, thus identifying its principal allophone.

¹⁸ Chatterji, Sketch 12; improved version, Reader 11.

¹⁹ Cf. Ferguson's thesis §4.21, §4.22, §4.24(5).

²⁰ Cf. ODBL 410–13.

²¹ The ɔ–o–zero problem in relation to orthography is treated at some length by Lambert, op.cit. 183–5, 190–1, 195–7.

²² Sketch 15, Reader 19. Cf. also Ferguson's thesis §4.23.

²³ Sketch 15; Lambert, op.cit. 184, 185, 196.

²⁴ Page, Introduction 6.

²⁵ The examples cited might suggest that the emphatic -i is always added as a separate syllable and that what precedes the final /i/ is the same word minus this suffix. The situation is more complicated than this. For one thing, the emphatic suffix fairly often has the shape /i̯/, especially in certain expressions of very high frequency. Also, certain stem changes occur with the addition of the emphatic suffix. For example, a stem ending in /i̯/ or /e̯/ usually drops this final semivowel. Thus, cai̯ ‘I want’ + emphatic {-i} → cai homophonous with cai ← ca ‘tea’ + emphatic {-i}. It is this kind of phenomenon which provides contrasts of the -ii̯: -ii sort (dii̯ ‘I give’ + {-i} → dii). In a few cases the change may even go further e.g. niscɔe̯ ‘certainly’ + {-i} → niscɔi or niscoi̯.

²⁶ This is weak evidence for an u-u̯ contrast since the different transcriptions ‘ui̯’ and ‘ui’ would satisfactorily ‘take care’ of the contrast, but there is no mistaking the monosyllabic nature of the latter sequence, with the first element the nonsyllabic. Incidentally, the o̯a, o̯e, u̯i sequences in words originally borrowed from Hindustani or English cannot be regarded as non-Bengali sequences in unassimilated loanwords. Many of the words in question have been in use for over a century and are the commonly accepted terms used by monolingual Bengalis. In any case it seems highly unlikely that these three sequences should have been borrowed when many other semivowel sequences of English have been lost or transformed in the borrowing: the Bengali phonology must in some sense have been receptive to these particular sequences.

²⁷ For example, J. R. Firth, Phonological features of some Indian languages, Proceedings of the 2nd Intern. Cong. of Phonetic Sciences 176–82, esp. 180, 181 (Cambridge, 1936); B. N. Prasad, The positions of nasals in the Bhojpuri phonological system, Indian linguistics 17.189–96 (1955–6).

²⁸ Cf. P. B. Pandit, Nasalisation, aspiration and murmur in Gujarati, Indian linguistics 17.165–72 (1955–6); A. R. Kelkar, op.cit. §1.2.1(a) /m/ : /n/ : /ṇ/ : /n/ : /ŋ/.

²⁹ Cf. J. R. Firth, Sounds and prosodies, TPS 1948.132–3; H. M. Hoenigswald, Declension and nasalization in Hindustani, JAOS 68.143–4 (1948). Ferguson's thesis regarded nasality as a ‘component phoneme’, the domain of which was vowels and vowel sequences.

³⁰ A listing of the criteria accepted by various linguists is given in Ferguson's unpublished paper Phoneme, cluster, and accent.

³¹ Cf. fn. 30.

^31a Chatterji gives a nice summary of SCB aspiration, ODBL 442–4. A statement in more structuralistic terms is given in Muhammad Abdul Hai, Aspiration in Standard Bengali, Indian linguistics: Turner jubilee volume 1.142–6 (1958).

³² For phonetic details with kymograph tracings, see A.C. Sen, An experimental study of Bengali occlusives, Proceedings of the 2nd Intern. Cong. of Phonetic Sciences 184–93 (Cambridge, 1936).

³³ Cf. Chatterji, ODBL 558.

³⁴ The dynamic term ‘assimilation’ is used here simply to avoid the clumsiness of purely static terminology for such a limited purpose.

³⁵ BSOS 2.178–9 (1922).

³⁶ R. Wagner, Bengalische Texte in der Aussprache des Standard Colloquial (Berlin and Leipzig, 1933).

³⁷ Pp. 270–4. One other important statistical study of Bengali exists: P. C. D. Chaudhury, Word frequency in Bengali ... (Dacca Univ. Bull. No. 14; Dacca, 1931).

³⁸ G. K. Zipf, The psychobiology of language 68–81 (Boston, 1935).

³⁹ The percentage figure for /i̯/ may be somewhat higher than normal in this count because of the frequent occurrence of mosai̯ ‘Mr., sir’ in one of the stories; this pronunciation of Mr. Mondell's is less common in SCB than mɔsae̯ or mosae̯.

⁴⁰ The two principal references are: R. Jakobson, C. G. M. Fant, M. Halle, Preliminaries to speech analysis [hereafter Preliminaries], 3rd printing (Cambridge, 1955); R. Jakobson and M. Halle, Fundamentals of language 3–51 [hereafter Fundamentals] (The Hague, 1956), which is a slightly expanded version of the same authors' chapter in L. Kaiser (ed.), Manual of phonetics 215–51 (Amsterdam, 1957). The introductory paragraphs of §11.1 are paraphrased from parts of the discussion in the Fundamentals.

⁴¹ The most useful references are probably: C. F. Voegelin, Inductively arrived at models for cross-genetic comparisons of American Indian languages, UCPL 10.27–45 (1954), which offers a kind of forerunner to the system; id., Linear phonemes and additive components, Word 12.429–43 (1956); C. F. Voegelin and J. Yegerlehner, The scope of whole system (‘distinctive feature‘) and subsystem typologies, Word 12.444–53 (1956); C. F. Voegelin, Six statements for a phonemic inventory, IJAL 23.78–84 (1957).

⁴² Manual of phonology 75.

⁴³ It seems unlikely that retroflexion should be equated with labialization and pharyngealization in distinctive feature analysis. These other two examples of ‘flatting’ show similarities of auditory effect and seem generally to be in complementary distribution in that languages with both as distinctive features are probably rare. Accordingly it is reasonable to regard them as the same feature in some sense. But retroflexion is not at all parallel to them. In the first place, labialization or pharyngealization can affect various classes of consonants or vowels while retroflexion can affect only dental (better, apical) consonants and seems always to be allophonic in vowels. In the traditional jargon of phonetics, retroflexion seems to be a ‘position’ rather than a ‘secondary articulation’ like labialization or pharyngealization. In the second place, languages with labialization or pharyngealization tend to show assimilations and other special relationships between homorganic consonants with and without this feature; dentals and retroflexes in South Asia generally do not show relationships of this kind.

⁴⁴ Cf. Preliminaries 29.

⁴⁵ The semivowels of French /j w ɥ/ are regarded (Word 5.154 [1949]; Preliminaries 46) as the lax counterparts of /i u ü/, in spite of the fact that they are shorter in duration and may be more forcefully articulated, on the grounds that their formants show greater deviation from those of the 'neutral position'! The /j/ of Russian is regarded (Lg. 29.38 [1953]) as nonvocalic and nonconsonantal, and Polish /j w/ are regarded (For Roman Jakobson 520 [The Hague, 1957]) as nonvocalic, nonconsonantal, and acute and grave respectively. English /j/ and /w/ are regarded (Preliminaries 20) as nonsyllabic allophones of unstressed /i/ and /u/, thus forcing the analysis of an initial contrast such as that of eat : yip in terms of the position of the onset of stress /'iit : i'ip/.

Thus, in distinctive-feature analysis as currently practiced, Russian /j/ is presumably regarded as homologous with English /h/, since both are analyzed as nonvocalic, non-consonantal; while r/j/ and e/j/, which in fact show some striking phonetic and distributional parallels, are presumably not homologous at all.

In general the tense/lax feature is difficult to understand. As applied to vowels it may mean long vs. short, nonsyllabic vs. syllabic, and higher vs. lower (all these in French: /i : j/, medial /ê : è/, final /é : è/, Word loc.cit.) or full vs. reduced (e.g. Cheremis /i e u ü o ö a/ : /ә/, For Roman Jakobson 431). As applied to consonants it may mean fortis vs. lenis (e.g. French, Word loc.cit.), aspirated vs. nonaspirated (e.g. Chinese, Word 9.268 [1953]), /h/ vs. glottal stop (e.g. Arabic, Studies presented to Joshua Whatmough 112, 113 [The Hague, 1957]), even though the aspirated consonants may be more laxly articulated as in SCB and the glottal stop may be much more strongly articulated than the /h/ as in many Arabic dialects.

⁴⁶ ‘We must determine the minimum set of such features that the listener needs in order to recognize and distinguish all except homonymic morphemes ...‘ Lg. 29.134 (1953).

⁴⁷ ‘The comparative description of the phonemic systems of diverse languages ... gives us important insights into the interrelation and classification of the distinctive features’ (Fundamentals 26, 27). ‘... the application of the dichotomous scale [of distinctive features] makes the stratified structure of the phonemic patterns, their governing laws of implication, and the conclusive typology of language ... transparent ...‘ (Fundamentals 49).