Formant frequencies corresponding to different vocal tract shapes

A computer model was used to generate the formant frequencies corresponding to different vocal tract shapes. The degree of lip opening was varied exponentially in five steps. The tongue position was varied in equal steps in accordance with the two factors previously found to underlie English vowels. Sets of approximately 1500 different vocal tract shapes were generated, the exact number in each set depending on how the limits of possible vowels are determined. The formants of these sets of vowels are not evenly distributed throughout the ferment space. A high proportion of them are in the area of the front vowels from [i] to [æ]. In most cases ferment frequencies that differ by small amounts are associated with vocal tract shapes that also differ by small amounts. But there are also cases in which a set of formant frequencies can be produced with very different vocal tract shapes, usually because differences in lip positions compensate for differences in tongue shapes. [Work supported by NSF Grant 75-07158.]