Since the late 1950s we have known that the perception of speech is categorical. That is, listeners' perception of speech categories varies non-linearly in response to continuous changes in relevant acoustical quantities. Although there has been a large amount of research investigating this phenomena over the last half century, very little is known about how this perceptual behaviour is represented neurologically. One hypothesis is that this nonlinearity is represented directly as a place-code in auditory cortex. Like other sensory fields, the spatial organization of auditory processing centres is non-arbitrary, and the spatial location of active neural populations seems to covary with the frequency of acoustic stimuli. We hypothesize that this tonotopic map is distorted in the case of speech stimuli. Specifically, if two speech stimuli are from the same phonemic category, they will stimulate relatively near regions of auditory cortex, and if two speech stimuli are from different phonemic categories, they will stimulate relatively distant regions of auditory cortex, even if the acoustic distances are the same across stimulus pairs.

This talk presents an MEG study which is currently in progress, and which tests this hypothesis. An equally-spaced acoustic continuum from /sh/ to /s/ was constructed, and presented to subjects during MEG scanning. Equivalent Current Dipole fitting will be used to localize the neural response to each stimulus. This will allow us to test whether equal distances in acoustic space correspond to equal distances in cortical space. If our hypothesis is correct, we would predict that two different speech signals within a category would correspond to a smaller spatial distance in cortex than two equally spaced speech signals which straddle a category boundary.

Since this is a work in progress, the talk will focus on the experimental design, methods as well as a discussion of categorical perception and sensory cortical maps in general.