Since its invention almost two decades ago functional magnetic resonance imaging (fMRI) has become the prime research methodology in human neuroscience. Its capabilities continue to evolve based on combined improvements of scanner hardware, experimental designs, and data analysis tools. Within the rapidly growing field of multisensory research the use of noninvasive neuroimaging techniques in general and fMRI in particular is also of increasing relevance. For several years, discussion in the multisensory fMRI community has mainly focused on principles of and statistical criteria for multisensory integration. The recent availability of more sophisticated experimental designs and increasingly sensitive (multivariate) analysis tools allows multisensory researchers to (noninvasively) differentiate between regional and neuronal convergence and to reveal the connectional basis of human multisensory integration.