Fig. 3: Goal-related baseline vmPFC activity correlates with individual differences in behaviour.

a, Cluster-corrected activity representing goal progress time-locked to the onset of the decision period. b, Cluster-corrected activity representing goal progress time-locked to the intertrial fixation cross (see ‘Whole-brain intertrial analysis’ in Methods). While there was widespread activity in the occipital and parietal areas at decision time (a), the majority of these areas did not track goal progress ‘between’ decisions, where the highest peak was in the vmPFC. c, Time course of vmPFC activity at the onset of option offers, depicting the impact of goal progress (purple), current goal value (blue) and best alternative value (orange) at decision time (beta weight on BOLD activity). Error bars show s.e.m. across participants. We follow previous studies by defining baseline activity as the unconvolved neural activity at offer onset²³, which due to the haemodynamic delay captures ‘pre-decision’ activity rather than decision-related activity (Extended Data Figs. 7 and 8). d, Relationship between baseline goal tracking in the vmPFC and goal-oriented attention (Spearman’s r = 0.48, P = 0.007, 95% CI = (0.15, 0.72)). The baseline measure corresponds to the impact of goal progress on activity in the vmPFC ROI at the moment of choice onset (c). Goal-oriented attention refers to the accuracy advantage for remembering the current goal location compared to alternative goal locations in the spatial attention task (Fig. 2e, right). Note that the attention measure comes from a separate testing session outside the fMRI scanner. e, Relationship between baseline goal tracking in the vmPFC and persistence biases in the decision task (Spearman’s r = 0.46, P = 0.011, 95% CI = (0.12, 0.70)). Notably, the relationships between neural activity and behavioural goal biases are specific to baseline activity in the vmPFC; baseline activity in other regions of interest and vmPFC activity in response to the decision itself are not predictive of behavioural biases (see Extended Data Fig. 8 for control analyses at key timepoints and regions).