Smoothing spline analysis of variance (ANOVA) for tongue curve comparison

Abstract

Ultrasound imaging of the tongue is an increasingly common technique in speech production research. One persistent issue regarding ultrasound data is how to quantify them. Researchers often want to determine whether the tongue shape for an articulation under two different conditions (e.g., consonants in phrase-initial versus phrase-medial position) is the same or different. To address this issue, a method for comparing tongue curves using a smoothing spline ANOVA has been developed (SSANOVA) [Gu, 2002, Smoothing spline ANOVA models]. The SSANOVA is a technique for comparing curve shapes (splines) for two sets of data to determine whether there are significant differences between the curve types. Data sets contain 8–10 repetitions of each of the relevant tongue curves being compared. If the interaction term of the SSANOVA model is statistically significant, then the groups have different shapes. Since the interaction may be significant even if only a small section of the curves is different (i.e., the tongue root is the same, but the tip of one group is raised), Bayesian confidence intervals are used to determine which sections of the curves are statistically different. SSANOVAs are illustrated with some data comparing obstruents produced in word-final and word-medial coda position.