### Figure 6-1

*A*, *B*, Weighted discrimination indices for valence ratings (*A*) and arousal ratings (*B*). Valence and arousal ratings were first *z*-transformed within each participant using the mean and SD of all ratings of valence and arousal of a participant, respectively. With the *z*-transformed data, we computed a weighted discrimination index per learning phase as the difference between the reinforced 45° (CS^{+}) grating and the weighted average of the four CS^{–} gratings. Weights for the CS^{–} correspond to the angular difference in orientation between the four CS^{–} orientations (25°, 35°, 55°, 65°) and the CS^{+} orientation (45°): the two more similar CS^{–} orientations (±10° to the CS^{+}) were weighted with 0.33[…], while the more dissimilar orientations (±20° to the CS^{+}) were weighted with 0.166. Data and effect sizes are shown as a Cumming estimation plot (http://www.estimationstats.com). See the legend of Extended Data Figure 5-1 for a detailed description of a Cumming estimation plot. For valence data (*A*), the unpaired Hedge’s *g*: for habituation: –0.039 [95.0% CI, –0.680, 0.568], *p =* 0.896; for acquisition: –0.660 [95% CI, –1.219, 0.048], *p =* 0.0372; for extinction: –0.291 [95% CI, –0.925, 0.354], *p =* 0.3522; and delayed recall: –0.218 [95% CI, –0.832, 0.423], *p* = 0.4848. For arousal data (*B*), the unpaired Hedge’s *g*: for habituation: –0.296 [95% CI, –0.914, 0.386], *p =* 0.3372; for acquisition: –0.877 [95% CI, –1.459, 0.302], *p =* 0.0074; for extinction: –0.382 [95% CI, –1.020, 0.273], *p = *.2216, and for delayed recall, –0.142 [95% CI, –0.778, 0.510], *p =* 0.6472. The 5000 bootstrap samples were taken for CI estimation; the CI is bias corrected and accelerated. The two-sided *p* values are the likelihoods of observing the effect sizes, if the null hypothesis of zero difference is true. For each permutation *p* value, the 5000 reshuffles of the group labels were performed. Download Figure 6-1, TIF file.