Supplementary Figure 7: Experiment 6: The coordinate frame of serial dependence.

We compared the strength of serial dependence in perceived orientation in retinotopic vs. spatiotopic (head-centered) coordinates by varying the stimulus and fixation locations across trials (panel a). On each trial the fixation point appeared in one of two locations, at 6.5° to the left or right of the center of the screen, and the Gabor patch appeared in one of three locations, at 6.5° or 19.4° to the left or right of fixation depending on the location of the fixation point. Trials were paired, and every second (baseline) trial was analyzed. We manipulated the change in fixation-Gabor positioning between the trials in each pair: on ¼ of trial pairs both the spatial and retinal positions stayed the same (“both” trials), on ¼ of trial pairs the spatial position of the Gabor changed while its retinal position remained the same (“retinal” trials), on ¼ of trial pairs the retinal position of the Gabor changed while its spatial position remained the same (“spatial” trials), and on the final ¼ of trial pairs both the retinal and spatial position of the Gabor changed (“neither” trials; hence the trials are named for the aspect of the stimulus position that remained constant between the two successive trials in a pair). We counterbalanced the orientations presented so that subjects saw exactly the same set of orientations on all four of the spatial conditions – the only thing that differed among the spatial conditions was the change in the relative positioning of the fixation point and Gabor between trial pairs. We found significant serial dependence within each of the four conditions (panel b; permutation tests based on 180 trials from each of four subjects in each condition; error bars are ±1 s.d. of the bootstrapped group mean). Serial dependence was significantly stronger in those trials in which the retinal stimulus position was the same as in the previous trial (retinal & both trials) than in trials when the retinal position changed (spatial & neither trials), replicating the tuning of serial dependence to retinal position found in Experiments 4 and 5 (P = 0.019; permutation test based on 360 trials per condition from each of four subjects). At the same time, serial dependence was also stronger when the spatiotopic stimulus position stayed constant between trials (spatial & both trials) than when it changed (retinal & neither trials; P = 0.037; permutation test based on 360 trials per condition from each of four subjects), indicating that serial dependence is tuned to an object's spatiotopic position as well. This tuning to both spatiotopic and retinotopic stimulus position suggests that serial dependence likely operates at multiple levels in the visual processing hierarchy.