Modeling the subglottal space for American English /r/

One challenge in modeling the acoustics of American English /r/ is the fact that F3 may vary from 1300–2000 Hz in different contexts, even in the same speaker. Previous models have accounted for F3 lowering to the upper portion of this range, but have not accounted for the lower F3 range. In previous work, acoustical models of /r/ were tested utilizing the Maeda computer model (Maeda, 1982) and vocal tract dimensions for /r/ derived from MRI studies of two American English speakers. It was concluded that the F3 range exhibited by the speakers could be modeled successfully only by including the sublingual space. In the study presented here, two methods were examined for incorporating the sublingual space, (i) as an extra cavity resonator, and (ii) as an increment to the length and volume of the front 190 cavity. Using a recently developed adaptation of the Maeda model that incorporates a side branch from the front cavity, theoretical projections were compared against model results with and without the sublingual space resonator. It was concluded that an adequately low F3 can be generated using either method of incorporation, as long as certain conditions apply.