Regression Details.

Table 3 shows ROI multiple regression results. In each region, SMARS was entered as the dependent measure and math-cue, math-task, word-cue, and word-task activity (βs for each participant) were entered as predictors. SMARS was chosen as a DV to compare the relative contributions of the various cue and task βs, and in no way implies a causal relation. In each region, only math-cue-activity and word-cue-activity remained significant. These results are summarized in the middle section. Note that the r partial values should not be interpreted as activity levels (i.e., βs relative to baseline) but as partial correlations; see Table 2 for mean βs. For math-activity, the difference in standardized slopes (r partial) between SMARS and cue-activity and SMARS and task-activity was significant in all regions, as assessed by SMARS × 2(Segment: math-cue, math-task) ANCOVAs (the influences of word-cue-activity and word-task-activity were covaried out). F and p values for these interaction terms are shown in the rightmost column. The same was done for word-activity (grey rows), but with math-cue-activity and math-task-activity covaried out.