A BayesianModel ofMultisensory Integration in Peripersonal Space

Thumbnail Image
Issue Date
Journal Title
Journal ISSN
Volume Title
The environment is filled with a plethora of multisensory stimuli. In order to process these signals correctly, your brain must deduce the origin. For example, if you want to swat a fly off your arm, you can use the visual and tactile information to deduce where the target is. This multisensory integration is sensitive to spatial and temporal discrepancies. In order for visual and tactile information to be integrated into a single representation, the two cues have to happen shortly after each other and the visual cue has to be located near the location of touch, in peripersonal space (i.e. the space around a body part). When the two cues are in spatio-temporal register, there is a higher chance they are caused by the same source. The exact process of inferring the causal structure of multisensory stimuli is unknown, though evidence suggests multisensory integration follows the principles of Bayesian inference. This project will attempt to answer the question: ‘To what extent can the integration of multisensory stimuli in peripersonal space be approximated using Bayesian multisensory visual-tactile integration?’ In this project, I built a Bayesian model of multisensory integration to recover the underlying parameters and approximate such an integration. Additionally, I designed a localisation task to measure multisensory integration within peripersonal space. The majority of participants integrated the visual and tactile stimuli throughout the extent of peripersonal space. They did so following the precision-control principle in order to reduce uncertainty. A minority of participants ignored the visual stimulus and just localised the touch. The recovered visual variances were higher than measured, suggesting that participants used attentional mechanisms, such as precision control, to make them noisier. The proposedmodels were a good fit to the behavioral data. This suggests that spatial localisation within peripersonal space follows the principles of Bayesian integration. As the spatial localisation was mostly multisensory, we infer that Bayesian inference can be used to approximate visual-tactile localisation in peripersonal space.
Faculteit der Sociale Wetenschappen