Despite the considerable focus on coarticulation in speech perception research over the past few decades, researchers still know relatively little about the time course of perception of coarticulated speech. We know, for example, that appropriate coarticulatory information speeds listeners' reaction times relative to inappropriate information, but how do listeners use coarticulatory information over time? That is, what is the nature of the moment-by-moment processing that takes place as such information unfolds and becomes available to the perceiver? In conjunction with the Psycholinguistics Laboratory at the University of Michigan, and with colleagues Julie Boland and Andries Coetzee, we are currently using eye-tracking techniques to explore these questions for coarticulatory vowel nasalization.
Our approach builds on our previous studies of the acoustics of
coarticulation in CVNC sequences, which show that nasal consonants are
shorter before voiceless than before voiced consonants, and that vowels
are correspondingly more nasalized (i.e., nasalization begins earlier)
when followed by NCvoiceless than by NCvoiced. (See
"The interplay between segmental timing and coarticulation: acoustic, perceptual, and phonological investigations.") By monitoring listeners'
eye movements to visual representations during auditory presentations of
CVC and C
NC stimuli, we hope to learn when listeners use coarticulatory
vowel nasalization in online word processing (e.g., in selecting between
bent and bet or bend and bed), and whether this time course differs for
VNCvoiceless and VNCvoiced stimuli. We will then align the perceptual
time course with the dynamic acoustic patterns, as well with aerodynamic
(nasal to oral airflow) measures of coarticulatory nasalization.