Extended Data Fig. 1: Principles of individual difference analyses.

A) We extract task-based and post-task self-report measures for each participant in the discovery sample. For the behavior, we used epoched task data to regress environment factors (for example reward, predator speed) against task behavior (for example rate of foraging before discovery of the predator in each epoch). To identify unique links between real-life individual differences (questionnaire subscales and gender) and task behavior, we entered the measures as dependent variables in models where the independent variables were clinical scores and demographic measures. To establish hypotheses, we bundled significant findings by concept and arranged them hierarchically (from general to increasingly specific predictions). We expressed these hypotheses as regression models (clinical scores as dependent variables) and pre-registered them. B) To replicate, we collected an independent sample. We computed behavioral measures as for the discovery sample. To predict clinical scores and demographics for each person, we applied the hypothesis regression models fit on the discovery sample to the replication sample. We tested our hypotheses in a hierarchical manner (starting with the most general hypothesis first and proceeding to test its sub-hypotheses only if the general hypothesis was significant) by testing the goodness of our predictions. This meant, for example, that when many individual task measures could have been linked to a questionnaire subscale and were conceptually related to one another, their predictive power was combined in a single regression analysis, thus meaning that a single statistical test was used to test the superordinate hypotheses and only then follow-up tests were used to test subordinate hypotheses.