2010/01/27

Agwuele, A., Sussman, H. M., & Lindblom, B. (2008). The effect of speaking rate on consonant vowel coarticulation. Phonetica, 65, 194–209.

Presentation: Belinda

Summary: Thomas

Consonant-vowel (CV) coarticulation is often characterized by variations of vowels from their baseline positions in F₁/F₂ space. Lindblom et al. (2007), using a modified locus equation (LE) regression metric, aimed to disentangle a possible stress effect on CV coarticulation apart from vowel expansion effects. Using [V1.CV2] sequences, they first used F₂ frequencies of V1 and V2 midpoints to form a model predicting F₂ onsets, and then emphatically stressed V1 and V2 midpoints were inserted to the model to see how well the onsets could be predicted. A poor match would indicate that emphasis, rather than vowel space expansion alone, was contributing to the variation in F₂ onsets. The same procedure was replicated in the present study, but for CVs produced at fast speaking rates. The prediction was lowering of F₂ onsets during fast speech were not solely attributable to the shrinkage of F₁/F₂ space. In the experiment, V.CV sequences embedded in a carrier sentence were pronounced at three different rates, normal, fast, fastest. F₂ values at V1 midpoint, V1 offset, V2 onset, and V2 midpoint, and the duration of V1, C, and V2 were measured. Two LE regression analyses based on these F₂ and duration values were. Two LE regression slopes, the predicted slope of F₂ onsets regression relative to F₂ vowel midpoints and the slope showing observed F₂ onsets were compared across three different stops environments, namely bilabial, alveolar, and velar. Results showed differences between the predicted the observed slopes, most notable for alveolars. The difference can be safely attributed to articulatory adjustments affecting F₂ onset when speaking fast, yet unrelated to vowel reduction influences. This implies both speaking rate and emphatic stress affect coarticulation. Such effects can be disentangled from the effect of vowel space variation by implementing LE analyses.