Audio-visual integration during overt visual attention

Abstract

How do different sources of information arising from different modalities interact to control where we look? To answer this question with respect to real-world operational conditions we presented natural images and spatially localized sounds in (V)isual, Audio-visual (AV) and (A)uditory conditions and measured subjects' eye-movements. Our results demonstrate that eye-movements in AV conditions are spatially biased towards the part of the image corresponding to the sound source. Interestingly, this spatial bias is dependent on the probability of a given image region to be fixated (saliency) in the V condition. This indicates that fixation behaviour during the AV conditions is the result of an integration process. Regression analysis shows that this integration is best accounted for by a linear combination of unimodal saliencies.

Keywords

• eye movements
• attention
• crossmodal integration