Tonal melodies, like segmental melodies, have left and right edges. If autosegmental phonology is right, these edges should coincide with the prosodic domain boundaries known from studying other phonological entities. And this is what we find. For example, across languages, tonal melodies are bounded on either end by the boundaries of the mora, the syllable, the prosodic word, the phonological phrase, and (at least in the case of intonation) the utterance.
This talk examines one predicted entity in particular, the tonal foot. The aim is both to establish its existence and to ascertain its properties. The conclusion is that tonal feet exist and, like their better-known counterparts metrical feet, can be either binary or unbounded and either left- or right-headed.
Tonal feet are harder to document than metrical feet since metrical feet by definition contain only one stress per foot, while the constituency of tonal feet is not normally so neat or obvious. An approach that seems to work is to probe for tonal feet by showing how certain assumptions about their existence and constituency solve descriptive anomalies that have cropped up time and again in the study of lexical tone languages.
We can discern the existence of tonal feet by establishing the occurring and non-occurring melodies that can be characterized in terms of them analogously to the way that melodic possibilities have been found to characterize morphemes in some languages, like Mende (Leben, 1978) and Kukuya (Hyman, 1987). Tonal feet are shown to be binary in Hausa borrowings (Leben, 1997) and in Northern Mande (Creissels and Grégoire 1993, as reanalyzed here) and unbounded in Isixhosa (Cassimjee and Kisseberth 1997) and in other languages analyzed there.
Tonal feet form a part of a general theory of tonal constituents. While focusing here on tonal feet because they are to date the most obscure part of such a theory, the talk suggests what a theory of tonal constituents would look like. Based on the behavior of languages outside of Bantu, some modifications to the optimal domains theory of Cassimjee and Kisseberth 1997 are called for. Finally, the theory is shown to offer new explanations in several instances where similar surface forms exhibit consistently different tonal behavior. Because the differences in tonal behavior arise as a result of distinct representations, rather than, say, through output-output comparisons, this paper may also have implications for how opacity is treated in phonology.
References: