Homeostatic brain network mechanisms facilitate perceptual binding of audio-visual speech
In daily lives, speech perception requires binding of spatiotemporally disjoint auditory and visual cues. On the other hand, functional segregation and integration are the two complementary mechanisms that capture brain information processing. Here, we demonstrate using fMRI recordings that subjective perceptual experience of multisensory speech stimuli is dependent on a homeostatic balance of segregation and integration mechanisms. Previous reports conceptualized posterior superior temporal sulcus as the key brain region for binding signals from multiple sensory streams. However, we report an enhancement of segregated information processing in distributed brain regions, defined as the perceptual binding network. The seed-based whole brain functional connectivity of each node in this network was anti-correlated with higher propensity for illusory perception. Interestingly, the perceptual binding network was anti-correlated with other intrinsic brain networks, such as dorsal attention and default mode networks during cross-modal perception. The pattern disappeared for people who rarely reported the illusory perception, further strengthening the hypothesis of homeostatic balance. The cognitive theories of Bayesian causal inference and predictive coding hypothesis could explain the balance of segregative and integrative mechanisms during cross-modal perception.