Acoustic cues to visual detection: a classification image study.

A non-informative sound is known to improve contrast detection thresholds for a synchronous visual target (M. Lippert, N. K. Logothetis, & C. Kayser, 2007). We investigated the spatio-temporal characteristics of the mechanisms underlying this crossmodal effect by using a classification image paradigm specifically suited to investigate perceptual templates across both space and time (P. Neri & D. J. Heeger, 2002). A bright bar was embedded in 2D (space-time) dynamic noise and observers were asked to detect its presence in both unimodal (only visual) and bimodal (audio-visual) conditions. Classification image analysis was performed and the 1st and 2nd order kernels were derived. Our results show that the cross-modal facilitation of detection consists in a reduction of activity of the early mechanisms elicited by the onset of the stimulation and not directly involved in the identification of the target. In fact, the sound sharpens the 2nd order kernels (involved in target detection) by suppressing the activation preceding the target, whereas it does not influence the 1st order kernels. These data suggest that the sound affects some non-linear process involved with the detection of a visual stimulus by, decreasing the activity of contrast energy filters temporally uncorrelated with the target, hence reducing temporal uncertainty.