This talk summarizes our efforts to identify the spatio-temporal brain dynamics and behavioural relevance of auditory-visual multisensory interactions in humans and the consequence such has had on our understanding of the functional selectivity of low-level cortices. Across studies we have used combinations of psychophysics, ERPs, fMRI and TMS. The collective data support several general conclusions about auditory-visual interactions in humans. First, primary cortices are loci of multisensory convergence and interactions. Second, these effects occur at early latencies (i.e. < 100 ms post-stimulus onset). Third, these effects directly impact behaviour and perception. Finally, multisensory interactions are context-contingent. One the one hand, they affect not only current stimulus processing, but also later unisensory recognition. Current unisensory (auditory or visual) object recognition and brain activity are incidentally affected by prior single-trial multisensory experiences; the efficacy of which is predictable from an individual’s spatio-temporal dynamics of multisensory interactions. On the other hand, cross-modal activations of visual cortices by sounds critically depend on whether the sound location is (un)predictable. Together, these data underscore how multisensory research is changing long-held models of functional brain organization and perception.