<<   >>   TOC

6. Conclusion

I thus claim that tonal depression is essentially a nontonal process, i.e. it does not work on a categorial level commonly thought of as tonal. The observed interaction between consonants and H tone melodies is not due to an underlying tonal (i.e. H/L) specification on the relevant consonants, and consonantal features do not have direct access to tonal tiers either. Rather, the interaction takes place when the (in terms of abstractness) higher-level tonal categories are translated into lower-level segmental features in order to become pronounceable.

Previous proposals for integrating tonal categories and segmental feature geometry were inadequate, because categories of different type were mixed: segmental features are articulator-based, while tonal features are function-based. My suggested translating algorithm is a way to overcome this weakness: in the feature geometry, tone is to be represented in articulator-based terms. Once this is achieved, consonant-tone interactions are not only allowed but expected. The reason why depression is crosslinguistically rare is due to the fact that tonal specifications may make use of a set of exchangeable features, and I assume that languages tend to choose features for suprasegmental purposes that are least likely to disrupt the segmental geometry (and vice versa). Zulu exhibits depression because of its wide range of consonantal distinctions, i.e. all features that may be used for the manifestation of tone (i.e. [slack/stiff], [lowerd/raised lx], [tense/lax vf.]) are distinctive on the segmental level. The likelihood of segment-tone interactions in a particular language is thus a function of the complexity of the phonemic inventory.

<<   >>   TOC