Bayesian Calibration of Simultaneity in Audiovisual Temporal Order Judgments
Figure 1
Two opposing calibrations of simultaneity in audiovisual (B) and tactile (C) temporal order judgments.
(A) Examples of biased distribution of stimulation intervals: Gaussian distributions with positive (red dashed) and negative (blue solid) peaks. Positive interval shows “light first” in audiovisual, and “right hand first” in tactile temporal order judgments. (B,C) Opposing shifts of psychometric functions under the biased distributions in audiovisual (B) and tactile temporal order judgments (C). The probability of “light first” (B) and “right hand first” (C) judgments (ordinate) is plotted against the stimulation interval (abscissa; stimulus onset asynchronies, SOAs). Note that the point of simultaneity, as indicated by the intersection of a psychometric function with P = 0.5, shifted toward the peak of each Gaussian distribution in audiovisual (B, lag adaptation) but away from the peak in tactile temporal order judgments (C, Bayesian calibration). (D) A serial model for lag adaptation and Bayesian calibration. The order of the two systems could be the other way around. A constant time lag between a sound-light pair is adjusted before (or after) the signals enter Bayesian calibration mechanism. (E) Predictions of shifts in the mixed condition (second experiment). Six predictions are shown in a two-by-three factorial manner according to whether lag adaptation was pitch-specific or not (rows), and whether Bayesian calibration existed, existed but was not pitch-specific, or was pitch-specific (columns). Note that the point of simultaneity is expected to move away from the Gaussian peak only when Bayesian calibration is working in a pitch-specific manner, but lag adaptation is not pitch-specific (second row, third column).