Fig. 3: Control Analyses.

Solid colored bars below decoding results indicate significant clusters as identified via permutation tests (see Methods). All results were smoothed using a gaussian filter. Shaded areas represent standard error of mean. a Decoder results when passed preceding trial target location as labels instead of current high-probability target location. Note that this decoding took all 8 locations as factors and included the neutral blocks. This analysis also included trials that could have been at the current high-probability location (32% of total trials). Trial averaging for this analysis was done over five trials (see Decoding methods). b Decoding results when decoders are trained and tested only on ping epochs which followed a trial in which the target was at a HP location (blue) or excluding all such HP trailing epochs (red). Note that due to low trial counts, no trial averaging was possible for HP-only decoding (See Methods). c Eye-density plots showing average eye position in the 600 ms window following pings. Red represents regions with higher densities than combined average of other three conditions. Blue represents regions with lower densities. Black outlines indicate significant clusters identified using permutation tests (see methods). X and Y-axis ticks indicate pixel distance. Color bar indicates proportion out of 1. The fixation dot has been superimposed over the densities for reference. See Supplementary Fig. 4A for an un-baselined towardness analysis of eye data in the intertrial period.