Classic view of multisensory perception relies on independent processing of sensory inputs in sensory specific cortices, and integration of multiple inputs in a bottom-up, deterministic manner. This rigid, hierarchical structure is being challenged by evidence of multisensory responses in areas which were thought to be sensory specific. However, these evidence show high variability, with some studies reporting crossmodal enhancements, some observe crossmodal inhibition, and others fail to observe any crossmodal effects. Here we seek to examine whether the source of this variability is the dynamic, context dependent nature of multisensory perception. Subjects were scanned during three audiovisual experiments which differ in task, experience and the information conveyed by the auditory input while keeping the stimuli intact. This was enabled by teaching the subjects a novel visual to auditory sensory substitution algorithm (SSA), originally developed as a rehabilitation tool for the blind. Our results reveal two elements of multisensory perception – a stable, deterministic element and a fast, dynamic context dependent element. Sensory cortices did not change their sensory preference throughout the experiments. However, crossmodal attenuations of sensory specific areas were also detected. These changed their direction completely after only a short period of learning SSA, from visual attenuations of primary auditory cortex to auditory attenuations of visual cortex. Outside sensory areas a number of cortical regions, mostly left lateralized, changed their sensory preference, shifting it from one sensory modality to the other, while others demonstrated constant multisensory preference. We therefore suggest that in multisensory perception there are both stable and dynamic processes, allowing stability in one hand and a fast, dynamic tuning of the system when required, and both are taking place together, which may lead to seemingly contradicting evidence. In order to tweeze them apart, and overcome these contradictions, one has to take into account not only on the stimuli themselves, but rather on the context in which they are perceived, and the information they convey.

fMRI; Multisensory Perception; audiovisual