Constraints on Articulatory Variability: Audiovisual Perception of Lip Rounding

Abstract

What are the factors that shape linguistic sound systems? Perceptibility of the acoustic signal has long been argued to play a role in phonological organization. Some theories of historical change (Ohala 1981, 1993; Blevins 2004) argue that sound change results from misperception of the acoustic signal, while teleological models of phonology (Lindblom 1990; Hayes, Kirchner, and Steriade 2004) posit that speech is optimized (in part) for auditory perceptibility. Nevertheless, it is well known that speech perception is influenced by a range of non-auditory cues (McGurk and MacDonald 1976; Gick and Derrick 2009; Mayer et al. 2013).

This dissertation investigates the role of audiovisual perception in constraining patterns of articulatory variation, in which speakers employ differing articulatory strategies to achieve the same acoustic output. Such variation is widely documented for sounds like English /ɹ/ (Delattre and Freeman 1968), but does not arise in some cases where it is hypothetically possible (Harrington, Kleber, and Reubold 2011). Three experiments test the hypothesis that the range of possible variation is constrained by the availability of visual speech cues.

The first experiment considers the case of back vowel fronting in American English, where articulatory variation is predicted to be possible but has not been observed in some other varieties. It is shown that, for speakers from Southern California and South Carolina, /u/ and /o/ have retained their lip rounding as they have undergone fronting. The second experiment focuses on the fronting of /ɑ/ and /ɔ/ associated with the Northern Cities Shift. It is found that Chicago speakers vary in the extent to which these vowels differ acoustically, but that most speakers retain the labial distinction between /ɑ/ and /ɔ/, even if the lingual distinction is lost. The third experiment demonstrates that the visual lip rounding cue enhances perception of the cot-caught contrast, making visibly round variants of /ɔ/ perceptually more robust than unround variants. It is argued that speakers prefer articulatory strategies that are contrastive in both the auditory and visual domains. This preference has typological consequences for phonological systems, such that labial segments tend to retain their labiality in diachronic change.