Audio-visual integration in the perception of tap dancing

We measured thresholds for discriminating light-point tap dancing scenes using visual, auditory, and combined information. A dance sequence was filmed with three distinct visual markers positioned on each foot. Both the visual and audio sequences were then thresholded, and the visual markers and auditory taps substituted with circular disks and a stereotyped tap, to produce controlled light-point sequences. Subjects were required to discriminate in forced-choice which of two 3 s noise-embedded intervals contained a light-point dance sequence (rather than a scrambled sequence). Sensitivity was defined as the noise level corresponding to 75% correct discrimination. After equating visual and auditory sensitivity with unimodal measurements, sensitivity was measured bimodally with the auditory sequence in-phase and out-of-phase. The in-phase presentation improved thresholds substantially, by a factor of about 2, while the out-of-phase presentation only by a quarter of this value, more consistent with probability summation. The results suggest that, unlike simple translational motion, biological motion shows complete audio-visual integration.