Diachronic change in formant dynamics of California low back vowels: an improved analysis method using the Discrete Cosine Transform
Previous work on California English has argued that a low back merger has occurred on the basis of formant point measurements. I add to this literature by investigating whether this pattern of convergence holds for the entire formant trajectory and develop a novel method of formant trajectory comparison to do so. Vowel formants are modeled by generalized additive models with the function terms defined by the discrete cosine transform (DCT-II). The coefficients of these terms are used to represent the vowel as a vector that describes the vowel’s F1 and F2 trajectories. The similarity between any two vowels is evaluated using the 5-dimensional Euclidean distance between their vectors. With data from 438 participants from 5 field sites, I find evidence of whole vowel convergence among white speakers, marginal divergence among multiracial speakers, and no evidence of change among speakers from other racial groups.