Contribution of visible speech to context effects from speaking rate
Large, Nathan Robert
MetadataShow full item record
A talker's rate of production is one source of variation in the acoustic form of speech sounds. Accordingly, listeners are known to perceive timing-based speech contrasts relative to audible contextual information about speaking rate. Aside from a small number of studies, little is known about the use of speaking rate information from the visual modality, and how visible speaking rate might be perceptually integrated with audible speech. I replicated findings that viewers can perceive speaking rate reliably from visible speech (Green, 1987). Nonetheless, I found little evidence that preceding speech rate information from the visual modality strongly influences fully integrated audiovisual speech perception. Initial testing with visible-only precursor sentence contexts showed no reliable effect of context speaking rate on classification of a subsequent consonant as voiced, based on variation in Voice Onset Time (VOT). I ruled out overt visual inattention as an explanation for this result, using an eye-tracking system. In a subsequent study with different precursor items, a visual-only context presented before an audiovisual target again showed very little effect of speaking rate. In contrast, the audible portions of the same utterances produced robust effects of speaking rate on target classification, and these effects were identical when context or target were presented with or without the visible portion. Speaking rate of visible speech did affect the classification of a subsequent target presented as audible only, a result akin to the subsequent vowel duration effect found by Green and Miller (1985) and Brancazio and Miller (2005). Most likely, the visible duration of a produced syllable can be integrated as one cue to the duration of the corresponding audible syllable. The obtained effect could result from direct integration of duration information across modalities, along with a rate or duration contrast effect within the auditory modality or within an integrated audiovisual process. Supporting this interpretation, the visible rate effect on an audible-only target was reduced when the context and target had mismatched rather than matching vowels. The absence of a true `indirect' context effect of visible speech rate may be due to the relative lesser reliability of visible speech, compared to audible speech, as a source of information about segment durations. An attempt to demonstrate one factor that could contribute to uncertainty about speaking rate for visible speech - `transitional visibility' - was inconclusive, but instructive for future investigations.