A neural network model of multisensory maturation in superior colliculus neurons

Superior colliculus (SC) neurons have the ability to integrate stimulus information from multiple sensory modalities (e.g., visual and auditory) to alter their sensory responses and the behavioral responses they mediate (detection and localization of external events). This characteristic feature of SC neurons is not present at birth and develops gradually with sensory experience. A mathematical model is presented here to describe how the circuitry underlying this capacity uses sensory information to guide its development. The model’s circuitry includes four input areas: two representing unisensory subdivisions of the anterior ectosylvian sulcus (AES) (e.g., visual and auditory), and two representing inputs from these sensory systems not derived from AES. Lateral connections exist within each of these input areas. During early life, the connections from AES to the SC are present but not active and the activity of neurons in the SC is dependent completely on inputs from non-AES sources. Sensory experience is modelled by a “training phase” in which the network is repeatedly exposed to modality-specific and cross-modal stimuli at different locations in space. Cross-modal stimuli are in or out of spatial register during training. Synaptic changes between model neurons are dependent on an algorithm based on the Hebbian rules of synaptic Potentiation and Depression. Prior to the acquisition of sensory experience, and in agreement with empirical findings, individual model neurons respond to multiple sensory modalities but do not integrate cross-modal information. After the acquisition of sensory experience, model neurons respond in an adult-like fashion to modality-specific and cross-modal stimuli in multiple configurations. The receptive fields of model neurons change in a realistic fashion during simulated development and neurons exhibit: i) multisensory enhancement to spatially aligned stimuli; ii) both within-modal and cross-modal depression to spatially disparate stimuli; and iii) the principle of inverse effectiveness. This model codifies much of what is currently known regarding SC physiology in the neonate and the adult, and, using realistic rules of learning establishes a theoretical framework for understanding how sensory experience modifies the underlying circuitry of the SC during maturation.