In light of the ongoing debate on multisensory processing in the human brain, we employed spatial self-organizing group independent component analysis (sogICA) on an fMRI data set acquired during a passive audio-visual (AV) experiment with common object stimuli. Two independent components (ICs) were initially designated as unisensory auditory and visual, respectively, based on their spatial layout and their activation time course. However, in addition to their main clusters both ICs also comprised regions traditionally considered to belong to the other sensory modality. To explicitly test these regions-of-interest (ROIs) for their potential involvement in object-related AV processing, we conducted a second AV experiment. We found object-related AV integration effects (MAX-criterion: 0<A<AV>V>0) in left occipito-temporal cortex, right posterior superior temporal sulcus (pSTS), and bilateral inferior frontal cortex (IFC). The additional manipulation of crossmodal congruency in this experiment revealed a parametric preference for incongruency (congruent < incongruent within categories < incongruent between categories) in the same IFC ROIs as well as in bilateral posterior parietal cortex. This clearly demonstrates the particular value of IC maps in the context of crossmodal object processing, as they represent the underlying functional connectivity, which is of prime importance especially for higher-order processing along the multisensory hierarchy.