The target undershoot model of vowel production predicts that vowels are characterized by formant values called targets that are not reached under all circumstances. It is assumed that these targets are fixed with respect to speaking rate and stress pattern. In the work described here we attempted to investigate this issue. We dynamically and cinematically modelled formant time series to estimate unreached targets; then we compared the results of a speaking-rate and stress-pattern variance analysis of the first two formants of vowels [a] and [c] to the results of a two-factor variance analysis of the corresponding formant target estimates. The two vowels were produced by two speakers in a [iVi] context at several speaking rate and stress pattern combinations. The corpus was perceptually evaluated to spot uncertain data. Part of our results gave support to the target undershoot model, but they were not sufficiently convincing to back it unambiguously.
Bibliographic reference. Pitermann, M. / Schoentgen, Jean (1995): "Experimental study of the target theory of vowel production", In EUROSPEECH-1995, 1917-1920.