The first factor, labelled "Positively physically and mentally active" had an eigenvalue of 4.29; for the second factor, labelled "Positive social interactions", the eigenvalue was 3.49; for third factor, labelled "Risky behavior and substance use", the eigenvalue was 2.35; for the fourth factor, labelled "Difficulties in social interaction and impatience" the eigenvalue was 1.56 (see Table 3).
Two within-subjects factors (actual difficulty and labeled difficulty) were used in a 2 × 2 analysis of variance (ANOVA) to assess the effect of labeled difficulty on performance and whether the effect differed depending on actual difficulty.
The ROI beta estimates were subjected to a 2 (actual difficulty) × 2 (labeled difficulty) ANOVA to determine whether sustained activation differed as a function of actual difficulty, labeled difficulty, or an interaction between the two factors.
T-tests were used to test significance of the labeled difficulty effect within each level of actual difficulty.
During training, the labeled difficulty always matched the actual difficulty level, to strengthen the expectation that an "EASY" block would be easier to perform than a "HARD" block.
A 2 × 2 repeated measures ANOVA detected no main effects of either labeled difficulty ("EASY" vs. "HARD") or actual difficulty (low vs. high) on task performance, although a trend emerged toward an interaction of labeled difficulty and actual difficulty, F = 3.54, p =.075.
None of the striatal ROIs exhibited differential activation at the time that the difficulty label was displayed to start each block or any significant differences in sustained activation throughout each block as a function of actual or labeled difficulty.
No other main effects from the event-related GLM reached significance in the three striatal ROIs (see Supplemental Tables 2– 4 for regions outside the striatum exhibiting effects of actual difficulty, labeled difficulty, and the interaction of labeled difficulty and feedback valence).
To explore effects of actual difficulty, instructed difficulty, and feedback valence in the striatum, beta estimates from each ROI were subjected to a 2 (actual difficulty) × 2 (labeled difficulty) × 2 (feedback valence) repeated measures ANOVA.
There were two levels of actual difficulty (high and low), based on the visual similarity of the two families in each block, crossed with two levels of labeled difficulty (labeled "HARD" and "EASY").
As can be seen in Fig. 2a, there was a significant effect of labeled difficulty for the blocks that were low in actual difficulty, where performance was superior during "HARD -labeled blocks, t(19) = 2.17, p =.043 (two-tailed).
