The multisensory neuron in the superior colliculus (SC) has proved to be an excellent model for understanding how the brain synthesizes information from different senses. Its responses to spatiotemporally concordant cross-modal stimulation are typically greater than that to the most effective of these alone. There is now a large body of information regarding the relationship between the magnitude of the SC neuron’s multisensory response, the physical properties of the stimulus combination driving it, and the particular circuit in the CNS that is activated. The underlying multisensory computation that is engaged during this process appears to contain nonlinearities that are dependent on inputs from cortex. Here we present a neural network model based on simple anatomical and physiological principles that accounts for many of the empirical findings. In the model, spatiotemporally concordant cross-modal stimulation enhances the activity of multisensory SC output neurons by two mechanisms: the clustering of cortically-derived afferents on shared electrotonic compartments, and transient synchronization between the cortical afferents themselves. We review the anatomical and physiological principles upon which the model is founded, the data that it replicates, and offer empirical predictions of the model for future research.