A neurocomputational model of multisensory integration in the spatiotemporal domain

We have been fascinated by senses long before we were actually aware of, driven as much by a need to understand how we apprehend the physical world as it has by the need to understand how we become who we are based on those sensory experiences. From the first minutes of our life, our brain learns multisensory processing, and deals simultaneously with information coming from different sensory systems. Countless are the factors that converge in the result- ing perception of objects or external events, as the brain relies, to a varying extent, on previous knowledge as well as on the different features of the sensory stimuli, such as the spatial and temporal proximity between them. Computational models can be developed to disentangle the complexities of multisensory processes, giving researchers the opportunity to express ideas in rigorous quantitative terms, to summarize data into a coherent setting, and to formulate new testable predictions to confirm or reject hypotheses. Therefore, aim of this dissertation is the development of a unifying neurocomputational model for the analyses of multisensory phenomena in varying spatial and temporal conditions, starting from previous biologically-inspired models that handled separately temporal and spatial aspects of multisensoriality. The model has been tested against existing data, proving successful in replicating the behaviour of adult subjects both in reaction time tasks and spatial localisation tasks. Further investigation on the mechanisms implemented in the network confirmed that both feedforward connections from the primary cortices to the associative areas and feedback inhibition are primarily responsible for inter-subject variability in temporal tasks, while additional sensitivity analyses upheld the centrality of cross-modal and lateral intra-area connections in regulating spatial-oriented behaviour. Furthermore, the model makes untested predictions that we are eager to investigate in future behavioural experiments.