Fig 3. Conjunction analyses.

Conjunction analyses reveal, among all participants’ classification images, which facial cues were consistently modulated by a trustworthiness or an attractiveness judgment. For each task, significantly modulated pixels were identified using a cluster-based permutation test (p<0.05, FWER corrected for multiple comparisons). Then, only pixels that were significant for all considered tasks were conserved and the minimum observed z-score were displayed. A positive z-score indicates that the area is consistently lightened (i.e. subjects selected the picture in which this region contains more light pixels) while a negative z-score indicates that the area is consistently darkened (subjects selected the picture in which this region contains more dark pixels). (A) Conjunction analysis between both populations (Chinese, French) and two stimuli types (Asian, European descent) for the trustworthiness (left) and attractiveness (right) judgments. Results showed a clear and significant pattern of classification across populations and faces only for trustworthiness judgments. For this condition, Chinese and French participants darkened selectively the face area within the pupil-iris area of the eyes to make the face more trustworthy (see zoom). No equivalent selective area overlapped for attractiveness judgements. (B) Partial conjunction analyses for in-group (Chines rating Asian faces + French rating European descent faces) and out-group (Chinese rating European descent faces + French rating Asian faces) ratings for a trustworthiness (left) and an attractiveness (right) judgment. For trustworthiness in-group ratings, participants darkened the eyes area; while for out-group rating participants not only darkened this area, but also the extremity of the mouth, as though to make the faces more smiley. This result suggests that the smile is not the most reliable cue for a trustworthiness judgment in our everyday social environment. However, the smile remains an important cue for rapid and automatic evaluation of strangers. Concerning attractiveness, for in-group ratings, participants darkened the upper part of the jaw and lightened the extremity of the mouth, thus forming a chiseled jawline for the attractive face and a more rounded appearance for the non-attractive picture. However, for out-group ratings, no face region emerged as significant.