^{page 157 note 1} Brief reading conventions are as follows:— i, e are front spread vowels, close and mid respectively; u, o are back rounded vowels, close and mid respectively; a is an open vowel midway between front and back. Geminated vowels count as disyllabic for purposes of tonal analysis.

The voiceless plosives p, t, k are unaspirated, and t is dental; d and l are alveolar; v is a weak voiced bilabial fricative; x is a voiceless uvular fricative-trill; ∫ is a voiceless palato–alveolar sibilant fricative; ny is a palatal nasal; intervocalic m, n, ŋ are bilabial, alveolar, and velar nasals respectively.

All ‘ sequences of consonants ’ (‘ consonant ’ here means anything except a, e, i, o, u) are, like all simple consonants, homosyllabic with the following vowel, except that (a) the first consonant of a gemination is syllabic, e.g. ∫a-l-lo-la, k-kyo, m-mwa, and (b) nasals are syllabic before voiceless consonants, with which they are then homorganic, e.g. n-su, m-fu, a-ŋ-xe-ka. All homosyllabic ‘ sequences ’ arise from the linear representation of the prosodies of nasalization, labiovelarization, and yotization: this ‘ orthographic ’ transcription has been adopted here in order that attention may be more readily concentrated on the subject of the paper. Thus, etc., where the underlined ‘ sequences ’ all function as C in a syllable-structure CV.

^{page 157 note 2} For some account of the operation of grammatical concord in Bantu languages see Gutbrie, Malcolm, ‘ Gender, Number and Person in Bantu Languages’, BS0AS., 12, pp. 847–856Google Scholar: and for the prefixes of the Machame noun see Meinhof, 's introduction to Pfarrer Emil Müller's Wörterbuch der Djaga-Sprache (Madjame-Mundart), Hamburg, 1947, especially p. 39*.Google Scholar

^{page 158 note 1} This correlation is observable only where a noun occurs in conjunction with certain kinds of concordial elements, and must be considered at least as much a feature of those elements, which form part of other words, as of the noun itself. Within these limitations, Meinhof classes 1, 4, and 9 in some cases, class 1 alone in others, stand in opposition to the remaining classes.

^{page 163 note 1} In Frames I and J the concordial prefixes of the words of the frames themselves are similarly left unindicated, as their form naturally depends on the grammatical sub-category of the particular noun inserted into the frames. Thus for Frame J we have, e.g.

mwana ŋka∫a aŋxeka

ŋumbe ∫ika∫a yaŋxeka, etc.

As far as the frames given are concerned, this alternation of concordial prefixes is entirely without consequence for the tonal analysis.

^{page 164 note 1} The terms contonation and contonational were first used by Professor J. R. Firth in the Staff seminar of the Department of Phonetics and Linguistics at the School during discussions on pitch, tone, and intonation. Professor Firth has sent me the following amplificatory note. ‘ In pursuance of my theory of levels of analysis first outlined in “The Technique of Semantics”, TPS., 1935 (see especially p. 52 for intonation), I suggest that the general word “ intonation ” be used as at present to refer to the “ tunes ”, “ contours ”, or relative pitch patterns considered as some sort of “ music ” or speech melody to which pieces or sentences are, so to speak, “ sung ”. D. Jones, Ida Ward and others use such expressions as “falling intonation”, “rising intonation ”, “ Tune 1 ” and “ Tune 2 ”. These “ tunes ” have sometimes been loosely associated with other classifications of the text, such as emphatic and unemphatic, interrogation and affirmation. In accordance with this view, attempts have been made from time to time to relate “ sentence intonation ” to syllable tones, word tones, “ basic ” tones, “ inherent” tones, and other types of lexical tone pattern. The terms “ contonation ” and “ contonational ” are intended to refer to “ intonational ” and prosodic patterns abstracted from and correlated closely with formally established grammatical structures, colligations, and collocations.’

^{page 164 note 2} Contrast the technique of Longacre, Robert E., who states in ‘ Five Phonemic Pitch Levels in Trique ’ (§ 2.1, p. 67) in Acta Linguistica, vol. 7, fascicule 1–2 (1952)CrossRefGoogle Scholar: ‘ Tonemic contrasts may be established among the pitches of items of a given substitution list relative to the pitches of the frame syllables, if it can be shown that (a) the tones of the frame itself remain unchanged regardless of which item of that substitution list is employed; …’

^{page 165 note 1} See Robins, R. H., ‘ Phonology of the Nasalized Verbal Forms in Sundanese ’, BSOAS., 15, p. 140, footnote 3.Google Scholar

^{page 166 note 1} See footnote o, p. 168.

^{page 167 note 1} The change from 1st to 2nd person singular in the verb has no tonal significance.

^{page 167 note 2} ↑ is used here at the phonetic level to indicate a high voice-register.

^{page 167 note 3} See Pike, Kenneth L., Tone Languages, University of Michigan Press, 1948, p. 3.Google Scholar

^{page 168 note 1} And less well adapted than is generally supposed for use as an orthography.

^{page 168 note 2} See Longacre, loc. cit., § 3, for the asymmetrical nature of the tonal‘ system ’ of Trique.

^{page 168 note 3} See Firth, J. R., ‘ General Linguistics and Descriptive Grammar ’, TPS., 1951, p. 85.Google Scholar

^{page 168 note 4} No such attempt will be made here, but it may be of interest to note the serious disequilibrium between the tonal possibilities in utterance-final position and those in other positions (see, e.g., Frame B). Reference to what was said earlier about the cases where the differentiation of patterns depended on a difference in the pitch of a frame-syllable, and also to those cases where the pitch-indication of the first syllable of a noun is constant throughout a frame (e.g. Frames B and F) suggests that this disequilibrium may be resolvable by regarding the tonal features of final syllables as the cumulate exponents of two tonal abstractions. Compare in this connexion the phenomena described for Kikuyu by Harries, Lyndon (‘ Some Tonal Principles of the Kikuyu Language ’, Word, vol. 8, no. 2, 1952)CrossRefGoogle Scholar, who states (§ 3), ‘ the tone actually heard on a syllable (the speech tone) may be the shifted basic tone of the preceding syllable ’— and see footnote 2, p. 169.

^{page 168 note 5} It may be objected that all that is required is to select a tonemic form from which all the others are predictable. It not infrequently happens, however, that no tonemic variant whatever will serve as a basis for prediction. The retention of Category 6 in our material helps to illustrate this state of affairs: for although the forms of Frame C would serve as bases for prediction for all the contonational categories other than 6 and 9, these latter would still have to be differentiated by other means.

^{page 168 note 6} See Firth, J. R., ‘ Sounds and Prosodies ’, TPS., 1948.Google Scholar

^{page 169 note 1} See Pike, op. cit., p. 25.

^{page 169 note 2} See footnote 4, p. 168. We do not need the device of ‘ shifted tones ’ because our contonational formulæ need have no more implication of linear extension in time than we care to give them.

^{page 169 note 3} For a reference to these, see Buiskool, H. E., The Tripādī, Being an abridged English recast of Pūrvatrāsiddham (An analytical-synthetical inquiry into the system of the last three chapters of Pāṇini's Aṣṭādhyāyī), Leiden, 1939, chap, i, § 3, A.Google Scholar I am indebted for this reference to my colleague, Dr. W. S. Allen.

^{page 169 note 4} See footnote 5, p. 168, and also the discussion of the difficulties experienced by lexicographers when marking the stress features of English words in Pike, K. L., The Intonation of American English, Univ. of Michigan Press, 1946, p. 84.Google Scholar The parallel with English stress patterns is instructive throughout.

^{page 169 note 5} There is an obvious parallel here with the variable domain of intonational sentence-prosodies. For the particular relevance of the point at issue to the present material, see what was said earlier about the related forms ukwa/ŋgwa, mbivi/uvivi. In all such cases, the contonational category of either member of the correlative pair is predictable from that of the other.

^{page 169 note 6} Beach, D. M. cf. (‘ The Science of Tonetics and Bantu Languages ’, Bantu Studies, vol. 2 (1923–6), p. 104, § 48)CrossRefGoogle Scholar, ‘ But in view of the great importance of tone in all languages, the term tone-language is peculiarly inappropriate and should be replaced by some other term …’ His suggested alternative—‘ semantic tone language ’—is, however, scarcely less inappropriate.