(A.B) Across-participant change-change correlations in both data halves after arbitrary split into even trials (A) and odd trials (B). (C – J) Scatter plots of various change-change correlations. See titles and axes labels of each plot for details. Least squares lines (gray) were plotted when the Pearson correlation reached significance (p<0.05). Both the signal SD (panel C) and delta (panel D) were negatively correlated with mMSE, whereas the other canonical frequency bands correlated only weakly (theta, rho = –0.31; alpha, rho = 0.18; beta, rho = 0.26; gamma, rho = 0.29, all p>0.24). Signal SD and delta were also correlated with each other (rho = 0.74, p=0.001, panel E), indicating that delta contributed strongly to overall signal SD, as expected based on the known 1/f characteristic of neural data. Neither liberal–conservative delta (panel F) nor SD (panel G) correlated as strongly with the criterion shift as liberal–conservative mMSE (see Figure 3B). In contrast, the correlation between SD and criterion was not significant when controlling for mMSE (panel J), highlighting the unique ability of mMSE to explain the bias shift.