¹ Particularly the southern variety, as spoken in Saigon.

² Schmidt, W., ‘Die Quantität der Vokale im Khassi’, Wiener Zeitschrift für die Kunde des Morgenlandes, XVII, 1903, 303–22.Google Scholar

³ See Roberts, H., A grammar of the Khassi Language, London, 1891Google Scholar. Schmidt also used two earlier works of Roberts, which I have not seen, namely, his Anglo-Khassi dictionary, Calcutta, 1875Google Scholar, and Khassi primer, Calcutta, 1876.Google Scholar

⁴ Rabel, Lili, Khasi, a language of Assam, Baton Rouge, 1961.Google Scholar

⁵ Compare, for example, in table 1, pp. 581–2, her phonemicization of kti 'hand and kmie ‘mother’ with the forms given by Roberts and Schmidt. See also comments on pp. 578–9 below.

⁶ My informant was Namita Shadap Sen, the daughter of a Bengali father and a Khasi mother. It should be pointed out that despite her parentage she knew no Bengali and had been brought up to speak and write Khasi as her mother tongue.

⁷ Roberts writes (Grammar, p. sly): ‘In this work, the dialect of Cherrapoonjee is taken as the standard, because it is the purest, as universally acknowledged by the natives, besides being more amenable to systematical arrangement than the patois of the smaller villages’.

⁸ For a note on the sonograms in this paper, see p. 567, n. 14.

⁹ ‘Words’ should be taken to include ‘affixes’ here.

¹⁰ Rabel, op. cit., 167–9. Note that in the first paragraph on p. 167 ‘eighteenth century’ should read ‘nineteenth century’

¹¹ Singh, U Nissor, Khasi-English dictionary, Shillong, 1906.Google Scholar

¹² See also pp. 575–6 below.

¹³ See op. cit., 6, 20, 21. It will be seen on p. 6 that ‘patterning and morphophonemic behaviour’ are, however, cited as the justification for treating voiced, as opposed to voiceless, aspirated stops as phoneme clusters rather than unit phonemes.

¹⁴ ‘Sonogram’ is the term commonly used to describe a sound spectrogram produced by the Kay Sonograph, a piece of apparatus manufactured by the Kay Electric Company of Pine Brook, New Jersey. The sonograms presented here were made from tape recordings of N's speech by Mr. A. W. Stone on the Sonograph in the Phonetics Laboratory at the School of Oriental and African Studies. I have not felt it necessary to attempt to take detailed acoustic measurements since comparisons, sufficient for the purposes of this paper, of the duration of given parts of utterances and of the distribution of bands of acoustic energy over the frequency spectrum can be made by eye, with reference to the vertical frequency scale, the unit of which is the Hertz, and to the horizontal time scale, the unit of which is one-tenth of a second. Interested readers who are unacquainted with the considerable literature on the acoustic analysis of speech will find a valuable non-technical introduction to the principles and techniques involved, together with illustrations of the application of these techniques to the analysis of a natural language, in Halle, Morris, The sound pattern of Russian: a linguistic and acoustical investigation, s-Gravenhage, 1959.CrossRefGoogle Scholar

¹⁵ See Grammar, 2.

¹⁶ Ibid., 4.

¹⁷ See pp. 569 and 574.

¹⁸ For the significance of the ‘half-length’ mark in the transcription of N's pronunciation, see below, p. 569.

¹⁹ In an earlier paper, ‘Khasi and the l-clusters in seventeenth century Tonkinese’, Essays offered to G. H. Luce, I (Artibus Asiae, Supplementum XXIII), Ascona, 1966,Google Scholar I transcribed the first of these diphthongs ẹ'∂, believing that the quality of the element was rather closer to the vowel in kmie [kmẹ], lieh [lẹ'?] than that in kti, lih, etc. Measurement of the frequencies of the formants of these vowels has since inclined me to the opposite view.

²⁰ Unfortunately such sonograms in my possession, not having been made with publication in mind, proved unsuitable for reproduction with this paper, and as my informant has since left this country there was no opportunity to repeat them. The sonograms here included were made from tapes recorded before she left.

²¹ See p. 564, n. 2, above.

²² Schmidt, W., Grundzitge einer Lautlehre der Khasi-Sprache in ihren Beziehungen zu derjenigen der Mon-Khmer-Sprachen, Munich, 1904.Google Scholar

²³ op. cit., 712.

²⁴ Schmidt notes, on p. 742: ‘Leider ist nirgendwo etwas gesagt über die richtige Aussprache des eigentlichen Doppelvokals ia. Ich vermute, dass der Ton auf dem i liegt, wobei a verkürzt wird, also ’. This guess was incorrect. As contrasted with the other diphthongs, ia must be regarded as a rising diphthong. Its phonological value is, as Schmidt correctly stated, always short.

²⁵ Roberts (Grammar, p. xiv) deplores spellings such as angew for engow as ‘ugly barbarisms’. N accepted both spellings for this particular word and claimed that although her pronunciation is [sniu] she has heard other speakers say [snou].

²⁶ See, however, 'tangkro = ‘windpipe’ in Nissor Singh's dictionary, and in Leemuel's Anglo-Khasi pocket dictionary, compiled by Diengdoh, A. K., Shillong, 1965.Google Scholar

²⁷ Rabel also recognizes two variants for this word, one with a close and one with an open vowel. See table 1 on pp. 581–2.

²⁸ op. cit., 82. She describes, however, an ‘indefinitely long vowel /ii/’ in this word.

²⁹ See Grundzüge, p. 712, § 85.

³⁰ On the absence of glosses for these forms, see p. 581 below.

³¹ It is interesting to note, however, that Nissor Singh's dictionary contains no form rieh ‘to hide’, but contains instead rih, with the same meaning.

³² In some parts of the Grundzüge Schmidt occasionally leaves other words, such as leh, kyǵeh, and lih (= lieh) without a length mark. This may be a misprint, or it may be because quantity is not the particular point at issue in these passages.

³³ In the transcriptions accompanying the sonograms this off-glide is represented by a small raised ∂.

³⁴ As far as I am aware, there has been little published work so far on the acoustic analysis of such phonation features as ‘glottalization’. Ilse Lehiste has, however, published spectrograms of dysarthric speech in which what she calls ‘laryngealization’ is associated with vertical strictions somewhat similar to those shown here. See Lehiste, I., Some acoustic characteristics of dysarthric speech, Basle, 1965.Google Scholar

³⁵ See pp. 569–70 above.

³⁶ Since drafting this article, I have had the good fortune to meet a second young Khasi, Miss Renee Jyrwa, also a Cherranpunji speaker. I have as yet had no opportunity to compare her pronunciation with that of N except to check the words used for the sonograms. Her pronunciation of these agrees closely with N's, except that she frequently uses not a diphthong in dieng but the same close e as in lieh. My impression (on one hearing only) is that her ẹ in dieng is longer than her I in ding, but this remains to be confirmed.

³⁷ See also p. 577 and n. 46.

³⁸ Rabel does so; op. cit., 82.

³⁹ See pp. 569–70.

⁴⁰ See, for example, Henry Sweet's paper ‘Spoken North Welsh’, TPS, 1882–1884, 412. Khasi roman spelling owes its origin to a succession of Welsh missionary scholars such as T. Jones, W. Pryse, and H. Roberts.Google Scholar

⁴¹ The letters k and g do not enter into the picture, since there is no final velar plosive in native Khasi words. See my earlier paper, ‘Final -k in Khasi: a secondary phonological pattern’, Lingua, XIV, 1965, 459–66.Google Scholar

⁴² When we first began to work together, N herself believed that the -t, -d spellings represented some difference in the pronunciation of the final consonant. She was both surprised and pleased to discover the underlying rule, and said that it would have saved her much trouble at school to have recognized it sooner.

⁴⁸ This feature was noticed in the earlier paper, ‘Final -k in Khasi: a secondary phonological pattern’, already referred to. My auditory impression was almost always of what I should identify as a glottal stop in such cases. It is the spectrographic evidence that has persuaded me to use the less specific term ‘glottal constriction’ here and in what follows.

⁴⁴ This particular type of pronunciation was, curiously, less noticeable with final palatals. From my hearing of the tape, I am doubtful whether the utterance illustrated in sonogram 14 was of this type or not.

⁴⁶ cf. Lehiste, op. cit.

⁴⁴ Perhaps the asymmetrical final -ieng (see p. 575) should be interpreted as filling this gap.

⁴⁷ On p. 742 he describes the ie in thyllieid, etc., as a ‘Doppelvokal’ (= ‘rising diphthong’ here ?) pronounced iē, and not to be confused with īe, which he regards as an occasional spelling for ī.

⁴⁸ cf. har ‘sharp (of large objects)’.

⁴⁹ See Grammar, 34.

⁵⁰ See table 1 on pp. 581–2.

⁵¹ op. cit., 14–15.

⁵² See op. cit., 180, where the author refers to her aim ‘to stay within a five-vowel system’.

⁵³ op. cit., 11.

⁵⁴ Ibid., 12.

⁵⁵ cf. Rabel's observation that open syllables with a short vowel phoneme are ‘chiefly syntactically bound morphemes’, op. cit., 20.

⁵⁶ op. cit., 180.

⁵⁷ See p. 569.

⁵⁸ But see p. 573, n. 32.

⁵⁹ See Rebel, 167–9.

⁶⁰ -ui appears to be restricted to ideophones and exclamatory expressions. I did not elicit any such forms from N, but believe it probable that they exist for her.

⁶¹ On the omission of final -k, see p. 575, n. 41.

⁶² Schmidt appears to have confused the words for ‘night’ and ‘table ’. See two lines below.

⁶³ Rabel cites what may be a rhyming ideophone, /khoop … coop/ ‘to brag’, which implies contrastive length for ɔ before -p as well as -C. See p. 577.