Audiovisual cue enhancement in the production and perception of the COT-CAUGHT contrast

Abstract

This paper presents results from an investigation into the articulation and perception of the cot-caught contrast by Chicagoans. Fronted vowels (such as Northern Cities-shifted caught) can remain round as they undergo fronting [1, 2], but some studies have observed interspeaker articu- latory variation in vowel fronting strategy [3]. Articulatory variation has been argued to influence sound change [4, 5], but the factors governing this type of variation are not fully understood. The present study examines the role of visual lip rounding cues in restricting variability in the articula- tion of cot and caught and in enhancing perceptual strength of the contrast. Nineteen Chicago natives (4 men, 15 women, ages 21-70) participated in two experiments. In the first, participants were asked to repeat 112 words containing the vowels /i æ u o ɑ ɔ/ in a variety of phonological contexts. Participants produced three repetitions of each word at a normal speech rate, as well as in a carrier phrase eliciting careful pronunciation. Midsagittal ultrasound tongue contours were captured at 84 frames per second (fps), along with simultaneous audio and lip video (120 fps). Acoustic analysis shows that the strength of the cot-caught contrast varies with age: 50 to 70-year-old participants most strongly exhibit the contrast, with younger speakers exhibiting a weaker contrast or merger. Articulatory data reveal that speakers with a stronger contrast produce cot and caught with differences in both tongue position and lip rounding (Figures 1a and 2a), while speakers with a weaker contrast distinguish the vowels through lip rounding alone (Figures 1b and 2b). In the careful speech task, speakers enhanced the contrast primarily by increasing the degree of lip rounding for caught (Figure 2). For the youngest speakers, who show reversal of the Northern Cities Shift [6–8], cot and caught are merged in both acoustics and articulation. In the second experiment, participants were presented with congruous and incongruous au- diovisual stimuli in which nonce words containing /ɑ/ and /ɔ/ were mismatched to create visually round and unround variants of each vowel. Participants identified the perceived vowel by selecting a rhyming English word (e.g., the choices for auditory [zɑt] were cot and caught). Visually unround variants of /ɔ/ were significantly more likely to be perceived as /ɑ/ (Figure 3), demonstrating that the presence of visible lip rounding aids in listener identification of /ɔ/. In contrast, visually round vari- ants of /ɑ/ were still perceived as /ɑ/, which is perceptually distinct due to its frontedness. Control items containing /i u e o/ were mismatched in height and showed no effect of visual incongruity. These results suggest that the presence of visible lip rounding helps to preserve the cot-caught contrast, and that articulatory strategies in which fronted caught loses its rounding are dispreferred on perceptual grounds. These findings will be discussed in terms of their predictions for ongoing sound changes (including NCS reversal) in Chicago. Consideration of articulatory factors is shown to be crucial to understanding the nature of the cot-caught contrast, revealing patterns that cannot be observed through acoustic study alone.