Whole-trial decoding of object categories in the three data sets across the broad band and different frequency bands (A) with their Bayesian analyses (B). The results are presented only for features of mean, ERP components, wavelet, and Orig Mag. For full results including other features, see Supplementary Figures 1 and 2. (A) The black horizontal dashed lines on the top panels refer to chance-level decoding. Thick bars show the average decoding across participants (error bars standard error across participants). Bayes factors are shown in the bottom panel of each graph. Filled circles show moderate to strong evidence for either hypothesis, and empty circles indicate insufficient evidence. They show the results of Bayes factor analysis when evaluating the difference from chance-level decoding. (B) Top panel: Bayes matrices compare the decoding results within each frequency band across features separated by data sets. Bottom panel: Bayes matrices compare decoding results across different frequency bands and data set separately. Colors indicate different levels of evidence for existing difference (moderate 3 $<$ BF $<$ 10, orange; strong BF $>$ 10, yellow), no difference (moderate 0.1 $<$ BF $<$ 0.3, light blue; strong BF $<$ 0.1, dark blue) or insufficient evidence (1 $<$ BF $<$ 3 green; 0.3 $<$ BF $<$ 1 cyan) for either hypotheses. For example, for data set 1, there is strong evidence for higher decoding values for the N1 feature in the theta and alpha bands than in the gamma band, as indicated by the red box.