Humans are well able to operate tools whereby their hand movement is linked, via a kinematic transformation, to a spatially distant object moving in a separate plane of motion. An everyday example is controlling a cursor on a computer monitor. Despite these separate reference frames, the perceived positions of the hand and the object were found to be biased towards each other. We propose that this perceptual attraction is based on the principles by which the brain integrates redundant sensory information of single objects or events, known as optimal multisensory integration. That is, (1) sensory information about the hand and the tool are weighted according to their relative reliability (i.e., inverse variances), and (2) the unisensory reliabilities sum up in the integrated estimate. We assessed whether perceptual attraction is consistent with optimal multisensory integration model predictions. We used an everyday tool-use task - controlling a cursor on a computer monitor - and manipulated the relative reliability of the unisensory hand and cursor position estimates. The perceptual biases shifted according to these relative reliabilities, with an additional bias due to contextual factors that were present in Experiment 1 but not in Experiment 2. The biased position judgments' variances were, however, systematically larger than the predicted optimal variances. Our findings suggest that the perceptual attraction in tool-use results from a reliability-based weighting mechanism similar to optimal multisensory integration, but that certain boundary conditions for optimality might not be satisfied.
- multisensory integration
- bayesian integration
- Copyright © 2016, Journal of Neurophysiology