¹ The work resulting in this publication was carried on under a Ford Foundation fellowship. The data were collected during a 20–month stay in the area. I am indebted to Morris E. Opler and the staff of the Cornell University India Project for providing quarters in Rankhandi and transport for the field trips. Special thanks are due to Gordon Fairbanks for his aid in formulating the problem, and to Charles A. Ferguson for many helpful comments.

² The problem was first discussed by N. S. Troubetzkoy in a paper Phonologie et geographie linguistique, reprinted in his Principes de phonologie 343 ff. (transl. J. Cantineau; Paris, 1949). Some recent studies: George L. Trager and Henry Lee Smith Jr., An outline of English structure 27–9 (Norman, Oklahoma, 1951); Uriel Weinreich, Is a structural dialectology possible?, Word 10.388 ff. (1954); Edward Stankiewicz, The phonemic patterns of the Polish dialects, For Roman Jakobson 518 ff. (The Hague, 1956). A number of dissertations dealing with phonological differences are based on field records collected for the Linguistic Atlas of the United States: Yakira H. Frank, The speech of New York City (University of Michigan diss. 1948); David DeCamp, The pronunciation of English in San Francisco (University of California diss. 1953); Virginia McDavid, Regional and social differences in the grammar of American English (University of Minnesota diss. 1956; not available in time for consultation).

³ See for example Henry Lee Smith's review of Kurath, A word geography of the eastern United States, in SIL 9.7–12. [A notable exception to the statement in the text is the Linguistic Survey of Scotland, directed by Angus McIntosh, which is structurally oriented in every phase.—BB]

⁴ Sir George Abraham Grierson, Linguistic survey of India 9:1 (Calcutta, 1916).

⁵ For a definition of this term, see M. N. Srinivas, The social system of a Mysore village, Village India 17 ff. (ed. McKim Marriott; American Anthropological Association, 1955). It is important in India to control the factor of caste in sampling for a dialect study, since there are many speech differences between different castes living in the same area.

⁶ A detailed discussion of the phonology of the Rankhandi dialect was given by Gumperz, The phonology of a North Indian village dialect, S. K. Chatterji Jubilee Volume, Indian Linguistics (Calcutta, 1955). The present paper will deal only with those features that have been revised and those that are relevant to the comparative statement.

⁷ The following symbols are used, aside from those defined in the text: R Rankhandi; S Sivaya; K Kalraun; V any vowel; V¹ class I vowel; V² class II vowel; D diphthong; VV dissyllabic vowel sequence; C any consonant; CC geminate cluster; C¹C² dissimilar consonant cluster; N any nasal; NC nasal plus homorganic consonant. The term ‘initial’ refers to position after juncture or pause; ‘medial’, before and after consonants or vowels; ‘final’, before juncture or pause. All examples, unless otherwise identified, are transcribed in their Rankhandi form. In examples from Rankhandi and Sivaya, accent is not marked in monosyllables. In the area phonemic inventory, contrasts marked with a superscript s do not occur in Sivaya, those marked with sr occur only in Kalraun.

⁸ The phonemes of each dialect were determined separately. The three systems were then collapsed into a single statement for the area phonemic inventory, in order to provide a uniform terminology and consistent statements in cases where more than one phonemic solution is possible. In the original analysis, for example, the S dialect was found to have phonemic extra-loudness of the type usually referred to as stress; K has variations in pitch of the type usually called tone; in R the phoneme /h/ in certain environments is always preceded by a pitch glide, which can be analyzed as part of the /h/ or as part of the syllabic vowel or as an allophone of stress. The revised statement covers all these phenomena in terms of a single accentual framework.

⁹ There is also some historical justification for this analysis. Middle Indo-Aryan h and aspiration after voiced consonants are both reflected by glide accent in village K. For a discussion of the history of h in related Punjabi dialects see Benarsi Das Jain, A phonology of Punjabi 29–33 (Lahore, 1934).

¹⁰ These geminates are reflexes of the Middle Indo-Aryan geminates which in other New Indo-Aryan dialects were reduced to single consonants, with compensatory lengthening of preceding short vowels. Short vowels before medial nasals plus homorganic consonants shared the same development. See S. K. Chatterji, Origin and development of the Bengali language 1.19 (Calcutta, 1926); R. L. Turner, The position of Romani in Indo-Aryan, Gypsy Lore Society monographs 4.25 (London, 1927).

¹¹ In the earlier study, R vowels were described in terms of seven simple vowels /i, e, æ, a, ɔ, o, u/ and a phoneme of length. This seems forced, however, since /a/ would be the only simple vowel occurring in all positions, /i/ and /u/ would be found in a few restricted environments only, and /e, æ, o, ɔ/ would occur only with length. Charles A. Ferguson has suggested setting up a phoneme of vowel shortness as a way of reducing the number of vowel phonemes. Objections similar to those above apply also to Henry M. Hoenigswald's use of the length phoneme for Standard Hindi. See his Declension and nasalization in Hindustani, Journal of the American Oriental Society 68.139 ff. (1948).

¹² There is some evidence that utterance stress coincides with the center of an intonation pattern; but the data are not sufficient for a complete statement. Phonemic word stress has not been noted previously in Hindi dialects. It also occurs in Standard Hindi; the contrast ‘address’, /bətā/ ‘tell’ is also found there.

¹³ Similar tonal features have been reported for many dialects of Punjabi. A recent descriptive statement indicates that the dialect of Amritsar has three tonal contrasts; see Kali Charan Bahl, Tones in Punjabi, Taraporewala Memorial Volume, Indian Linguistics 17.139–43 (1957). A listing of other pertinent literature is given in Jain 29 ff. and 167 ff.

¹⁴ See Gumperz 286. For a discussion of similar junctural problems see Charles F. Hockett, A manual of phonology 167 (Baltimore, 1955).

¹⁵ See Troubetzkoy 343 ff. Frank and DeCamp (cf. fn. 2) also give classifications of dialect differences; but these are based on occurrence or nonoccurrence of certain features in particular words rather than in the phonological system.

¹⁶ This use of ‘etymological’ is different from Troubetzkoy's différences étymologiques.

¹⁷ Cognates are used to show correspondences. Their use is not necessary in the categories of §2.1 and §2.3.

¹⁸ For a historical treatment see Turner 16.

¹⁹ See Turner 24.

²⁰ Troubetzkoy 344.

²¹ Speakers in R show dialect difference with respect to these items.

²² Isoglosses stated in typological terms need not always be convertible into historical terms, since historical reconstruction does not operate with loanwords. If we compare the dialect of Sivaya with the Standard Hindi of educated persons, we find that the former has only /ḍ/, the latter has /ṛ/ and /ḍ/. In Hindi, however, /ḍ/ occurs only in certain loanwords from English which are not found in S. The typological difference in phonemic inventory is not statable in historical terms, since there are no correspondences.

²³ Bahl, ibid.

²⁴ R. L. Turner, Dictionary of the Nepali language, Introduction 13 (London, 1931).

²⁵ Troubetzkoy 345.

²⁶ Similar observations were made by William McCormack and H. A. Gleason for the contrast between the vowels /ə/ and /Λ/ in the Dharwar dialect of Kannada.