Although the role of attention in multisensory integration is not yet clear, a growing amount of evidence highlights their intimate interplay. We addressed how the direction of auditory selective attention modulates brain activity in multisensory integration sites. We used fMRI to measure whole brain BOLD responses while participants heard two auditory speech streams presented simultaneously from the same location but at different pitch. Participants were encouraged to selectively attend one of the two streams, indicated by a pre-trial cue. To promote selection they were tested on memory for words from the cued stream (2AFC) after each trial. Concurrent with the two speech streams, a video-clip of the mouth of a speaker was presented from the centre of the display. The mouth could match the cued audio stream, the uncued one, or neither, with equal probability. The remarkable finding was that, for physically identical conditions, we found a significant BOLD modulation in the Superior Temporal Sulcus (STS) which solely depended on participants’ direction of attention. That is, with the two speech streams present, attending to the one congruent with the mouth enabled STS activity, compared to attending the incongruent one. This result demonstrates that selective attention can strongly determine the physiological expression of audiovisual integration through modulation of the activity in associative areas, such as the STS. The behavioural consequence of this effect was confirmed with another experiment, whereby audiovisual enhancement in speech perception was seen only for attended speech streams thus supporting the significant modulatory role of attention in MSI.

fMRI;attention;speech;audio-visual