Figure 4

Relationship between facial attributes and prosocial behaviour. (a) Association between the scores for the facial attribute ‘generosity’ and the amount donated by the judged subjects in the anonymous version of the ‘Generosity game’ (DG). Linear regression analysis: N = 118, R² = 0.011, B = 0.002, p = 0.254. (b) Association between the scores for the facial attribute ‘trustworthiness’ and the number returned points in the anonymous version of the ‘Trustworthiness game’ (TG2). Linear regression analysis: N = 118, R² = 0.005, B = − 0.001, p = 0.434. (c) Heat map of the p-values of 100 tests addressing the statistical association between each of ten facial scores (columns: the eight facial attributes and the scores for PC1 and PC2) and behaviour in the anonymous version of the ten decision situations in our experiment (rows). The number of significant p-values (three for p < 0.05; four for p < 0.10) does not exceed the number of type II errors to be expected. For continuously scaled games (DG, SD, UG1, UG2, TG2), the scores of the subjects’ facial attributes and their PCs were regressed upon the subjects’ own decisions in the games (in points). For games with binary decisions (TG1, PD, SH, CG, PG) differences were estimated with a Student’s t-test.