The effects of attention on audiovisual integration at an incongruent location was investigated using behavioral measures in humans. A stream of unimodal visual (V), unimodal auditory (A), spatially congruent bimodal audiovisual (AV-Con) and spatially incongruent bimodal audiovisual (AV-Inc) stimuli were randomly presented to the left or right side; subjects covertly attended to all stimuli on one side (left or right) and promptly responded to all target stimuli on that side. The results showed that responses to the bimodal target stimuli were faster and more accurate than those to unimodal visual or atuditory target stimuli when audiovisual stimuli were presented on spatially congruent location. When audiovisual stimuli were presented on spatially incongruent location, responses to visual stimuli that were accompanied by simultaneous task-relevant, but spatially incongruent auditory stimuli were facilitated. However responses to auditory stimuli that were accompanied by simultaneous task-relevant, but spatially incongruent visual stimuli were not facilitated. The results suggested that vision was high preferential processing when auditory and visual information were simultaneously presented on spatially incongruent location.