Figure 3.

(a) Regression coefficients for subject-specific moral sentiment frequency covariate effects and standard errors in septum and subgenual cingulate (BA32) peak voxels. Scatter plot for Z-transformed fMRI effects for (b) pride in the septum and (c) guilt in the subgenual cingulate. There was a significant interaction of condition and effect of subject-specific moral sentiment covariates on Z-transformed fMRI effects within the septum (peak coordinate from Supplementary Table 1, univariate ANOVA, SPSS14, outliers Z-score outside ±2.5 excluded): F_107,7 = 3.18, P = 0.03. There was also a significant main effect of the moral sentiment covariate on the septal signal strength F_114,1 = 4.61, P = 0.03. Unadjusted correlations for septal activity with pride for POS_S-AG: R(29) = 0.46, P = 0.01 (trend for gratitude: R(29) = 0.35, P = 0.06, negative trend for indignation/anger: R(28) = −0.34, P = 0.08, no significant correlations with guilt: R(29) = 0.22, P = 0.25). There was a significant interaction of condition with the moral sentiment covariate effect on the signal within the subgenual PFC: F_112,3 = 2.63, P = 0.05. Subgenual PFC × NEG_S-AG_guilt: R(28) = 0.66, P < 0.0001 (no significant correlations with other moral sentiments: P > 0.30). There was a significant correlation of gratitude with hypothalamic activity (R = 0.43, P = 0.02). However, when adjusting the correlation for the effects of the other conditions, the overall ANOVA shows no significant effect of condition for the strength of correlation between moral sentiment covariates and hypothalamic activation and also no interaction of condition and moral sentiment covariates (at P = 0.05). Thus the effects in the hypothalamus were not robust enough to survive adjustment on the secondary data analysis.