The connection between smiling and GOAT fronting: Embodied affect in sociophonetic variation

This study examines the effect of smiling on GOAT fronting, a sound change common to many varieties of American English. The data are audiovisual recordings of ten speakers of American English recorded in dyadic conversations in an interactional sociophonetics laboratory. We applied an existing computer vision algorithm for smile detection to the video recordings to identify smiling intervals. A mixed-effects linear regression reveals that higher F2 (i.e., auditorily fronter GOAT) positively correlates with whether speakers are smiling while articulating the vowel and their self-reported comfort levels in the interaction. The latter factor does not correlate with whether vowels were smiled. Together, the findings suggest that GOAT fronting is not only a phonetic consequence of smiling, but also serves an affective, interactional function. While sociophonetic studies typically analyze audio recordings alone, patterns of variation are better explained by also attending to embodied practices observable only in the visual domain.