As good as it gets — testing a bayesian model of transsaccadic change blindness

Abstract

During fast or saccadic eye movements human perception of stimulus shifts is poor. This phenomenon is called “transsaccadic change blindness for displacement”. Employing neural network simulations, we have recently shown that even if the brain optimally used all signals available for shift perception it would exhibit change blindness. This concept predicts that the accuracy of the nonvisual eye position signal should inversely influence change blindness. To test our hypothesis, we asked seven subjects to perform horizontal and vertical saccades toward target stimuli that shifted either parallel or orthogonal during the saccade. As will be presented at the meeting, our results conformed the predictions of the model. This corroborates the view that change blindness reflects a mechanism of Bayesian inference from the noisy signals available in the brain and from the probabilities of the events underlying these signals.

Footnotes

 Supported by CIHR.