Do acoustics predict tongue shape in American English rhotics?

American English /ɹ/ is one of the most common and treatment-resistant speech deviations in children with residual speech sound disorder (RSSD; Flipsen, 2015) and has also been the focus of considerable basic science research on articulatory-acoustic relations (Zhou et al., 2008). English /ɹ/ can be produced using multiple tongue shapes that are commonly grouped into the broad categories bunched/tip-down and retroflex/tip-up. These shapes have a similar acoustic signature across the first three formants (Mielke et al., 2016) and are not readily perceptually distinguished (Twist etal., 2007). Previous work in adults has demonstrated that these shapes are acoustically distinct at the level of the 4th and 5th formants (F4 and F5; Espy-Wilson & Boyce, 1994; Zhou et al., 2008). The current study will investigate these acoustic-articulatory relationships using ultrasound and audio data from 36 children with no history of speech-language-hearing challenges and 30 children with RSSD affecting /ɹ/ who completed an ultrasound biofeedback treatment study. The /ɹ/ interval was segmented from the acoustic signal of each utterance and ultrasound frames in the rhotic interval were coded in a customized program. This provides ground truth articulatory labels for the data, thereby allowing for examination of relationships between F4 and F5 frequency values and bunched and retroflex tongue shapes.