Shared EEG activity between the past and the present activation predicts serial dependence effects. Correlations across participants (n = 17) of serial dependence effects. The abscissa plots serial dependence of each participant, calculated as the difference in criterion between trials where the previous response was the same as the current one and those when it was different. The ordinates show the average decoding accuracy over the poststimulus window (0–1800 ms) for different decoding conditions. Statistics shown are two-tailed Pearson's correlation coefficient r and p value against the null hypothesis of no correlation. Line of best fit in blue (dashed for correlations with p < 0.05). a–d, The ordinates show decoding accuracy of current responses trained on current responses (a), decoding of previous responses trained on previous responses (b), cross-condition decoding of current responses trained on previous responses (c), and cross-condition decoding of previous responses trained on current responses (d).