Introduction

This work is based on tape recordings made by one of the authors in Tanga, Tanzania, in 1965. The tape referred to throughout as D (Dialogue) is of a conversation between Mr. Hasani Athumani of Bagamoyo (a village near Tanga, not the town further south) and Mr. Mohammed bin Juma of Mwambani. The other tape, referred to as N (Narrative) was made in Tanga Girls' School by teachers there. The principal speaker is Mrs. Abdulla, who described her experiences in hospital. Three other speakers also make short interjections, Mrs. Mbuya, Mrs. Mutayoba and Miss Mhando. We did not include Miss Mhando's brief contributions in our analysis because Swahili was her second language. Each tape recording originally lasted about an hour. We chose passages to analyse starting some way on into the tapes so that the speakers had had time to lose self-consciousness. Moreover the tapes were not the first ones made by these speakers and they were made in an informal situation-tape D on the veranda of a house near the sea and tape N in a classroom after school. Indeed in some cases reality broke in too much, occasionally obscuring the recording. In fact, these speakers seemed remarkably uninhibited by the presence of the microphone and its attendant, even taking into account the fact that the latter had certainly been visiting and recording for several months before these actual tapes were made. The extracts used are each of about 7:Jr2 minutes duration. (N lasts 7 minutes 41 seconds; D lasts 7 minutes 21 seconds). We first listened to them with a loop-repeater, taking down an impressionistic phonetic account of the intonation and of some other features. After having written our individual impressions, we compared our versions of each short stretch, often listening again a number of times until we were reasonably satisfied. At this stage we included pretty well everything we heard, since we could not know in advance what was significant. For example, there were cases of slight falling-off in pitch at the end of a level tone, which might make it seem phonetically somewhat similar to a falling tone. There were, however, other features such as rhythm, loudness and so on which enabled us to distinguish them phonemically later. Although at the phonetic stage of the work we often had considerable discussion before reaching agreement, once the phonemic shape became apparent there seemed no longer any room for disagreement. This was one of the factors that reassured us as to the validity of our work to date. The next stage was to consider the grammar of the passages and see how the intonation correlated. The grammatical analysis used was as put forward in Sentences in Swahili, Maw, 1969. This analysis leans heavily on the early stages of the theoretical work of M. A. K. Halliday. The analysis of the intonation also owes much to Halliday's work on English intonation, at least as far as terminology goes-how far the concepts can be equated when they refer to different languages and function in different ways is a moot

appear in Halliday. (The term was used independently by D. W. Arnott in Sentence intonation in the Combe dialect oj Fula). We are also indebted to earlier phoneticians and linguists for the idea of looking for specific interaction between intonation and grammar. The advantage of using spontaneous peech is that the speakers are in a natural situation, not attempting to produce something especially for the investigator. The disadvantage is that we cannot know how they might have said things in other circumstances. Further, we can only glean the meaning of their intonation from the context, since they do not explain it. (Though it does seem from working with informants in other languages-including English-that the native speaker's explanations are in any case frequently ad hoc). Another drawback is that we only got a few examples of some forms, and indeed there may be other forms which we did not encounter at all. Nevertheless it seemed to us worthwhile to work on natural speech, since we could then see how the intonation worked over long utterances, interruptions, questions and answers, rising tempers and so on. Also because as far as we know it has not previously been attempted for Swahili. One interesting phenomenon which we did not deal with, was what we provisionally deSignated 'register-shift', where the speaker suddenly shifted his entire range up or down-sometimes to falsetto in men. This phenomenon has been noted in other African languages by Guthrie and by Gordon Innes, but the usages they describe do not seem to hold good for Swahili. Nor did we deal with changes in tempo, except where they indicated the end of one tone-group and the beginning of another. But obviously these phenomena have an interesting part to play in the effectiveness of speech. We put forward this description in the hope that it may prove a useful outline, a starting-point for further work, or even a wrong-headed analysis to be overtaken and refuted by a better. We ourselves have some faith in it, mainly because the more we worked at it the more the data seemed to fall into place, as long as we were open-minded enough to allow it. All the same, we might not have found the strength to continue the work to its present conclusion without the interest and encouragement of the late Professor Wilfred H. Whiteley, to whose memory we dedicate it.