Functional magnetic resonance imaging (fMRI) investigations are important to understand human multisensory integration; however, the relatively coarse spatial resolution limits the possibility to make inferences about neuronal mechanisms. Here, we explored the fine-scale organization of human multisensory cortex using fMR-adaptation, a method hypothesized to measure functional properties of neuronal populations beyond the resolution of averaged fMRI-signals across voxels. We adapted fMRI responses to repeated audiovisual stimuli and varied the associative relation between the auditory and visual inputs, while keeping unisensory inputs constant. Adaptation in several smaller clusters distributed over superior temporal sulcus and gyrus (STS/STG) showed sensitivity to the audiovisual relation, whereas a larger occipital-temporal network adapted independently of this relation. The differential adaptation effect in STS/STG is unlikely to result from two separate unisensory populations within the identified voxels, since these respond only to one modality and thus should be “blind� to how the different inputs are combined. Our results therefore strongly suggest multisensory integration on the neuronal level in distributed STS/STG clusters, in accordance with previous high-resolution human fMRI and animal electrophysiological recordings. In sum, this study contributes to link invasive animal and human fMRI findings, and thereby offers an important new approach to study human multisensory integration.