(A) Temporal generalization matrices for a decoder trained on ERFs to distinguish seen from unseen trials in the perception task and tested in the working memory task, either with the same labels (visibility decoder; left) or the unseen correct and incorrect trials (accuracy decoder; right). In each panel, a classifier was trained at every time sample (y-axis) and tested on all other time points (x-axis). The diagonal gray line demarks classifiers trained and tested on the same time sample. Please note the additional event marker: Mean reaction time (target-present trials) for the visibility response is indicated as a horizontal, dotted line. Any classifier beyond this point only reflects post-visibility processes. Time courses of diagonal decoding are shown as black insets. Thick lines indicate significant, above-chance decoding (Wilcoxon signed-rank test across subjects, uncorrected, one-tailed). For display purposes, data were smoothed using a moving average with a window of eight samples. (B) Same as in (A), except that the decoder was trained and tested on average power (relative to baseline) in the alpha band (8–12 Hz). For display purposes, data were smoothed using a moving average with a window of one sample. (C) Same as in (B), except that the decoder was trained and tested on average power (relative to baseline) in the beta band (13–30 Hz). AUC = area under the curve.