(a) Encoded variables used to regress EEG signals. The encoded orientation (, left) and encoded momentary decision update (, middle) were dependent on the orientation presented to the observer. The encoded accumulated evidence () varied over all presented orientations in a trial, the figure on the right shows only one example. (b) Representation precision of encoding variables using different low-pass filters. (c) Cross correlation between encoding variables over consecutive samples. (d) Temporal generalisation of representations: the regression weights were calculated on EEG signals at each time point and precision was tested across time. Colour scales are relative to the maximal precision, with zero precision in white and negative in grey (a sign flip of the regression weights). (e) Representation precision of the accumulated evidence for the first (left) and last (right) four stimuli of the Less and More conditions. Shaded error bars show the 95% within subject confidence intervals, red horizontal bars mark cluster corrected significant differences between conditions. (f) Representation precision of the previous (n-1), current (n) and future (n+1) accumulated evidence, based on the EEG signals locked to the current epoch. (g) Representation precision of the momentary decision update (top) and the accumulated evidence (bottom) for epochs separated by the timing of the subsequent stimulus, shown in coloured bars (317 ms, red, left; 333 ms, green, middle; and 350 ms blue, right).