Abstract
Humans integrate signals across the sensory modalities to create multisensory representations of their
environment. For example, speaker in face-to-face communication, we integrate the speech sounds and visible
lip movements to improve our understanding of speech. However, because the brain is constantly exposed to
multisensory signals, it must infer the causal structure of the signals to decide whether to integrate them if
they have a common cause, or segregate them if they have independent causes. In my talk, I will present our
recent studies in which we combined Bayesian causal inference modelling with neuroimaging to show that the
brain represents multisensory causal inferences distributed along the cortical hierarchies: Low-level visual and
auditory sensory cortices encode the unisensory visual and auditory signals, the posterior intraparietal sulcus
(IPS) represents integrated audiovisual signals weighted by their relative precision, and only the anterior IPS
represents integrated audiovisual signals while accounting for their causal structure. We also investigated
whether and how causal inference is altered in schizophrenia, where hallucinations and perceptual
abnormalities are thought to result from imbalanced integration of prior information and multisensory inputs.
Surprisingly, our data show that multisensory causal inference is largely preserved in our sample of post-acute
schizophrenia patients. I will discuss our findings in the context of multisensory perception and the study of
perceptual abnormalities in mental disorders.