Abstract
Self-motion perception is a complex multisensory process that relies on the integration of various sensory signals, particularly visual and vestibular inputs, to construct stable and unified perceptions. It is essential for spatial navigation and effective interaction with the environment. This review systematically explores the mechanisms and computational principles underlying visual-vestibular integration in self-motion perception. We first outline the individual contributions of visual and vestibular cues and then introduce Bayesian inference as a normative framework for the quantitative modeling of multisensory integration. We also discuss multisensory recalibration as a critical mechanism in resolving conflicts between sensory inputs and maintaining perceptual stability. Using heading perception as a model system, we further describe the relevant visual and vestibular pathways involved in this process, as well as the brain regions involved. Finally, we discuss the neural mechanisms mediating visual-vestibular interactions through models of the Bayesian optimal integration and divisive normalization.