Information integration is critical for the brain to interact effectively with our multisensory environment. Defining the limits and prerequisites of information integration is fundamental for understanding the mechanisms of normal brain functioning and their disintegration in diseases such as neglect & vegetative state. A key question is to what extent multisensory integration (MSI) is automatic or dependent on higher cognitive resources.This proposal combines psychophysics, neuroimaging and Bayesian models to unravel the neural and computational mechanisms of MSI and their dependency on higher cognitive resources in the healthy & diseased brain.First, we manipulate attention and stimulus awareness to dissociate bottom-up automatic from ‘cognitively controlled’ MSI. We hypothesize that automatic MSI relies primarily on feed-forward thalamocortical mechanisms, while ‘controlled’ MSI involves more top-down effects from association areas. Combining concurrent TMS-fMRI & Dynamic Causal Modelling, we will investigate how the network dynamics and integration capacity is affected by perturbations to parietal cortex. This research is complemented with studies in neglect patients to develop a multisensory model and novel MS therapies for neglect.Second, combined fMRI/EEG studies will investigate how sensory inputs are integrated at reduced vigilance during sleep. We hypothesize that MSI is partly preserved in sleep via thalamocortical mechanisms. These paradigms are applied to patients in vegetative state to identify residual MSI functions and develop neural MSI signatures as predictors of recovery.This research characterizes the neural and computational mechanisms of the multifaceted interplay of MSI with attention, awareness & vigilance. It significantly advances our understanding of information integration & segregation in the brain and has important implications for clinical diagnosis and rehabilitation of patients with neglect & vegetative state.