BEHAVIOURAL AND COMPUTATIONAL MODELS OF VISUAL-VESTIBULAR SELF-MOTION PERCEPTION WITH TEMPORAL DELAYS

Behavioural and neurophysiological studies show that visual and vestibular cues to self-motion can be optimally integrated, improving both response speed and response accuracy. However, this integration degrades under cue conflict, such as temporal delays between modalities. Here, we use a computational model motivated by behavioural data to investigate how temporal misalignment disrupts multisensory integration.
We extend the Wong-Wang two-variable decision-making model to simulate visual, vestibular, and visual-vestibular neural processing and behavioural responses during a heading discrimination task, with and without temporal delays between sensory signals. Three candidate mechanisms of multisensory integration were tested, ranging from no integration, through linear summation, to late-stage multisensory processing.
Of the three models the linear summation model was most consistent with behavioural and neurophysiological recordings. It reproduced behavioural accuracy and reaction times in unisensory conditions and showed optimal integration for temporally aligned multisensory trials. Crucially, under visual-vestibular temporal delays, the model captured the breakdown of integration observed behaviourally, with performance depending on both the length of the delay and the leading sensory modality . Reaction times increased and accuracy decreased with longer temporal delays, and responses were significantly faster when the visual signal began 500 ms before the vestibular signal than when it began 500 ms after. In addition, multisensory neural dynamics produced by the model were consistent with neurophysiological recordings. Together, these results demonstrate that this modelling framework provides a flexible tool for probing the neural dynamics underlying multisensory integration and its failure under temporal conflict.