Tone, Utterance Length and F0 Scaling

One important strand of research into the scaling of fundamental frequency has been concerned with the hypothesis that utterance length is a determining factor of initial pitch height, with longer utterances involving a higher initial F0 than that found in shorter utterances. Most work in this vein has been done on ‘intonation’ languages, with very little investigation having been carried out on tone languages, and the effect utterance length might have on the scaling of tones. This study investigates the relationship between utterance length and tone in Mambila. Previous work revealed that, for sentences comprised of all High tones, or of all Low tones, utterance length in Mambila had no effect on the scaling of either initial or final tones. Here utterances of alternating Hs and Ls are investigated. Results confirm those found earlier, that utterance does not determine the scaling of F0 in Mambila. These results are discussed from the point of view of utterance planning and tonal specification. 1. F0 scaling and utterance length Declination is an area of research into fundamental frequency that has received a considerable amount of attention. Among studies of declination, one strand of research pertains to the effect of utterance length on rate of declination or, alternatively, the scaling of initial F0 peaks. The basic hypothesis in this work is that utterance length is a determining factor of initial (and/or final) pitch height. Longer utterances will involve either a higher initial F0 or a lower final F0 than is found in shorter utterances. The first of these two possibilities, higher initial F0, suggests a ‘look-ahead’ or preplanning mechanism, by which utterance initial F0 values are raised proportionate to utterance length. The second possibility indicates, rather than preplanning, that adjustment may be made ‘on-the-fly’, and given that F0 range is finite its bottom may be reached before utterance end. Most work in this vein has been done on non-tone languages and has produced mixed results. One finding has been that initial pitch height does increase with sentence length [1, 2], while others have found no such increase [3, 4]. Ladd & Johnson [5], however, propose that constituent structure and related metrical factors, rather than utterance length, have the greatest effect on the scaling of utterance initial F0, and argue that the mixed results found across other studies have to do with the failure to control for such factors. 2. F0 scaling and utterance length in tone languages Very little investigation of this issue has been carried out on tone languages, though they provide an obvious test bed for such hypotheses. Interestingly, the only two known studies on tone languages done from this perspective also offer contradictory findings. Snider [6] for Chumburung reports the height of initial High, but not Low tones in Chumburung to be positively correlated with utterance length, and argues that speakers operate with initial and final F0 targets in mind for L, while for H only a final F0 target is necessary. Lindau [7] found no such effect on initial Highs in Hausa ( with the methodology Lindau used it was not possible to investigate Low tone in Hausa with natural sentences). However, despite the difference in results, like Snider, she assumes fixed targets, or anchors, for initial tones, but that in Hausa it applies also to initial Hs. Both final Hs and Ls in Hausa showed more variability, though Lindau doesn’t relate this directly to utterance length. Both Chumburung and Hausa have only two tones, and both show a degree of declination across the duration of utterances regardless of tonal structure. 3. F0 scaling and utterance length in Mambila The present study examines Mambila, a Bantoid language spoken in the Nigeria-Cameroon borderland. Mambila has four level lexical tones High (H), Upper Mid (UM), Lower Mid (LM), and Low (L); previous work on tone in Mambila [8, 9, 10, 11] has shown that declination (in like-tone sequences) only reliably occurs with L. Preliminary work on F0 scaling in Mambila is reported in [11]. This work showed no correlation between the height of F0 and utterance length for either H or L tones. The present study continues this strand or research, testing this hypothesis that utterance length can effect initial F0 in another environment in Mambila where a downtrend is found, in this downdrift in sentences with alternating H and L tones. Downdrift differs from declination in that it is a cumulative result of a local interaction, while declination is considered to be a global effect. A correlation between initial F0 and utterance length here would provide strong evidence for preplanning in sentence production.