Supplementary Figure 1: Measuring perceptual serial dependence.

To measure serial dependence in orientation perception, we presented subjects with an oriented grating (Gabor) on each trial and asked subjects to report the orientation of the grating by adjusting a bar using the arrow keys (panel a). Of interest was whether the reported orientations were systematically dependent on the stimulus orientations seen in previous trials. We constructed a plot capturing the relationship between the perceptual error on each trial and the orientation presented on the previous trial (panel b). For each trial we computed the relative orientation of the previous stimulus to the present one (e.g., by how much was the previously seen orientation clockwise or counter-clockwise of the present orientation; indicated by the blue box) with positive values indicating that the orientation seen on the previous trial was clockwise of the orientation seen on the present trial. We also computed the error on the present trial with positive values indicating perceptual errors in the clockwise direction (indicated by the red box). Plotting error as a function of the relative orientation of the previous trial allowed us to examine whether subjects' errors were drawn in the direction of the previous stimulus (shaded region on the plot), or repelled away from the orientation of the previous stimulus as would be found for traditional negative aftereffects at short inter-stimulus intervals (unshaded region on the plot). A preponderance of points within the shaded region of the plot, as was the case in the actual data (see Fig. 2a), indicates that perceived orientation was attracted toward the orientation seen on the previous trial.