Fig. 1: Behavioral task and choice history biases.

a Behavioral task. b Probability of repeating the previous stimulus category (left) or choice (right). c Correlation between individual P(repeat) in the first and second MEG sessions (Pearson’s r = 0.5134, p = 0.00003). d Choice repetition separately for alternators (N = 25, orange circles) and repeaters (N = 34, purple triangles). One observer had a repetition probability of exactly 0.5, and was excluded from subgroup analyses. Large markers indicate those individuals whose block-wise repetition probability was significantly different from 0.5, as determined by a t test. e Impact (regression weights) of several previous choices on current bias. T test against 0 within each subgroup and lag: significant for repeaters on lags 1–7; significant for alternators on lag 1. f Build-up of history bias across multi-trial choice streaks (i.e., successive repetitions of same choice), for both groups. Sequences were separated by whether they end in a repeating (red purple) or an alternating (black) choice. “Repeating bias” was quantified as shift in decision criterion into the direction of final choice (i.e., “X“ for red sequences, “Y” for black sequences). Main effects (Sequence length: p = 4.14e−8, sequence end: p = 1.71e−7) and interaction (p = 1.14e−25) from a repeated-measures ANOVA (see “Methods”). Data are shown as mean +/− 95% bootstrapped confidence intervals (n = 60).