A quantitative measure of the dorsal constriction in /k/-/t/ contrast in children and adults and its co-articulatory variation

We present a comparative quantification of the primary dorsal gesture in velar stops, comparing the tongue dorsum for /k/ against the tongue shape and location in /t/ – based on mid-sagittal ultrasound tongue images which have been stabilised using a headset. For a given vowel environment, /t/ provides the baseline against which the active, primary constriction undertaken by the tongue dorsum position to achieve /k/ is compared. Here, /k/ and /t/ are measured at a single target, the acoustic burst, though a comparison of the unfolding difference in normalised time differentiation is also possible.This quantification is useful for a number of reasons, including cross-linguistic analysis, the study of co-articulation, or the longitudinal tracking of the acquisition of the /k/-/t/ contrast in a speech and language therapy clinic. It is this latter use which is our main interest (Cleland et al. 2016, 2017). It is relatively common in developmental disorder or delay (and in typical development, in younger children) for /k/ to undergo a “velar fronting” process in which /k/ appears perceptually indistinguishable from /t/, and our aim is to track the dorsal component of the contrast before, during, and after treatment, using an efficient measure related to the functional contrast between /k/ and /t/. Comparison to child (and adult) norms is important, and we report such norms here.Such a measure does not directly examine the planning and execution of the gestural formation of the dorsal constriction directly, but still reveals a great deal about speech production. We discuss both an area measure of the “dorsal crescent” distinguishing /k/ and /t/, which is robust under mid-sagittal rotation and translation, and a radial linear measure, which is quick to measure and robust in cases where a dorsal crescent is incomplete. The two measures are highly correlated.Our main results are from a sample of 30 typically-developing English-speaking children (5;7 to 12;8) each pronouncing one token of /ata/, /iti/, /oto/ and /aka/, /iki/, /oko/, providing three measures of the /k/-/t/ contrast per child. As well as area and radial linear differences, we will report other descriptive measures, like the visible length of tongue surface and the distance from probe to tongue surface, with reference to co-articulaion and cross-sectional age-based developmental trends