Abstract
When asked to judge an unknown face on social traits such as trustworthiness and dominance, a high level of agreement is found among people, suggesting that some visual information in faces correlates with these judgements (Oosterhof & Todorov, 2008). We used the Bubbles technique (Gosselin & Schyns, 2001) to reveal the visual information used to judge the trustworthiness (Exp. 1) and dominance (Exp. 2) of 300 faces. Participants (N=50 for each experiment) were presented with either bubblized faces (phase 1) or fully visible faces (phase 2) and were asked to judge the level of trustworthiness or of dominance of the stimuli using a nine-level Likert scale. The number of bubbles was kept constant (i.e., 65 bubbles) across participants and trials. The judgements obtained for each fully visible face were considered "accurate" for a given participant. The judgements obtained with the same bubblized faces were possibly influenced by the available information. Thus, the following analyses allowed us to verify how judgements were influenced by the visual information available for the task. For each experiment, a classification image showing which visual information favoured the percept of trust or of dominance was computed by performing a multiple linear regression on the bubbles' location and on the difference (i.e., transformed into z-scores) between the bubbles judgements and the accurate judgements. For trust judgements, the areas used were: the eyes in the spatial frequency (SF) bands ranging from 21 to 84 cycles per face (cpf); the mouth in the SF bands ranging from 42 to 84 cpf and from 10 to 21 cpf; and most of the face in the two lowest SF bands. The dominance judgement was based mostly on the utilization of the eyebrows area in mid-to-high SFs. Interestingly, these two judgements were based on orthogonal visual information.
Meeting abstract presented at VSS 2013