Fig. 3.

Representative unimodal task data. A, In each session, subjects localized remembered auditory and visual targets presented in isolation. Both modalities were subject to idiosyncratic inaccuracies, with visual localization generally being more precise and accurate than auditory localization. Linear fits to localization error provided estimates of offset (intercept, uniform errors in azimuth) and spatial gain (slope, tendency to overestimate or underestimate target azimuth), which were used to correct target locations in the bimodal task. The labeled visual and auditory corrections were applied to the highlighted point in Figure 1 for this session's bimodal data analysis. B, Unimodal responses were used to estimate perceived target locations in bimodal tasks. Physical target locations were corrected for spatial gain and offset in each sense, and corrected values were used to estimate corrected audio-visual disparity. Corrected disparity was used as the basis for bimodal data analysis.