Figure 2.

Main analysis procedure. a, Theoretical RDMs used in the RSA. Within/Across-D, Within-dimension and across-dimension integration; Single/Sequential R, single response and sequential response. b, RDMs capturing differences in instruction length (number of letters) and RT, included in a multiple regression analysis together with matrices shown in a to control for the effect of these two variables. c, Following a searchlight approach, we extracted the neural RDM at each brain location and compared it (via Spearman correlation) with our three theoretical RDMs. As a result, we obtained three whole-brain correlation maps, one per model. d, To assess the effect of motivation, for each region, significant in c, we extracted the neural RDMs from rewarded (R+) and nonrewarded (NR) trials. To study potential interactions of reward expectation and the corresponding model variable (Hypothesis 1), we averaged the dissimilarity values among same-condition and different-condition trials and tested whether the subtraction among these two values was higher in the rewarded condition (using Wilcoxon signed-rank test). We also checked for a general increase in dissimilarities associated with reward (Hypothesis 2). All matrices in the figure were simplified for visualization purposes by averaging cells within conditions. b, The matrices were further averaged across the sample. d, Matrices display only one task variable (collapsing between the remaining two) to highlight the analysis logic. In all the analyses, however, trialwise and single subject matrices were used.