Fig. 5: Lower goal commitment in patients with vmPFC lesions.

a, Lesion overlap maps of the 23 patients who took part in the study (maximum overlap in a voxel was 10 participants). b, Results from the whole-brain voxelwise analysis. Green shows areas where lesion damage predicts lower persistence biases. Above-threshold t-statistics (t > 2.3 before cluster correction) are displayed for illustrative purposes. We controlled for multiple comparisons by performing cluster correction using FDR described in Methods. Only the vmPFC cluster survived whole-brain cluster correction (cluster threshold t > 2.3 (P < 0.01, one-sided), cluster size = 269 voxels, threshold cluster correction size = 255 voxels, cluster peak = (0,42,−14), t-statistic at cluster peak = 2.74, n = 5 patients with damage within cluster). c, Patients with damage to the vmPFC region identified in the fMRI study show reduced persistence bias. Patients were split into two groups depending on whether they were damaged within a region of interest centred on the peak of BOLD activity tracking goal progress between decisions in healthy participants. This area was damaged in 4 patients, corresponding to 4 out of the 5 patients independently identified in the voxelwise analysis in b. Patients with damage to this region showed lower goal commitment than patients with lesions elsewhere and age-matched controls (one-sided permutation test for lower persistence biases in vmPFC-damaged group compared with other lesion patients: difference in means = 3.79, P = 0.012; lower persistence bias in vmPFC-damaged group compared with age-matched controls: difference in means = 2.97, P = 0.023). Error bars show s.e.m. in each group; green dots depict individual biases. d, Post hoc analysis showing that patients with damage within the identified vmPFC region from the voxelwise analysis shown in b performed better than other lesion patients and no worse than age-matched healthy controls. Performance was measured as the average number of trials to complete a goal, where lower scores correspond to faster goal completion (one-sided permutation test for faster performance in the vmPFC group compared with other lesion patients: difference in means = 0.76, P = 0.015; faster performance in the vmPFC group compared with age-matched controls: difference in means = 0.32, P = 0.190, not significant (NS)).