A single multisensory neuron model simulates both enhancement between modalities and suppression within modalities

The deep layers of the superior colliculus (DSC) are the first site of major convergence of multisensory input in the mammalian brain. Individual DSC neurons can receive sensory input of more than one modality, and the receptive fields of the same DSC neuron for different modalities are large and spatially coincident. Multimodal DSC neurons exhibit cross modal enhancement (CME), in which the response to input of one modality is augmented by input of another modality. DSC neurons may also exhibit modality-specific suppression (NISS), in which the response to input of one modality is suppressed by input of the same modality presented within the neuron´s receptive field. Findings on CME are consistent with the hypothesis that the response of a DSC neuron is proportional to the probability that a target has appeared in its receptive field. Counter-intuitively, findings on NISS are also consistent with the same hypotheses. Both CME and NISS can be simulated using the same nonlinear model neuron that computes target probability using cross-modal and modality-specific inputs, which are represented as multivariate Gaussian with plausible values for mean, variance, and covariance.