Angrier = Blacker?: The influence of emotional expression on the representation of race in faces, measured with serial reproduction

摘要

In principle, race and emotional expression are orthogonal dimensions of face perception. But psychologically, they are intertwined a?? as when racially ambiguous faces are judged to be angrier when categorized as Black than when categorized as White. Does this reflect superficial judgmental biases, or deeper aspects of how faces are perceived and represented? We explored this using the method of serial reproduction, where visual memory for a briefly presented face is passed through 'chains' of many different observers. Here, a single face was presented, with its race selected from a smooth luminance-controlled continuum between White and Black. Each observer then completed a single trial, in which they reproduced that face's racial identity by morphing a test face along the racial continuum. Critically, both the initially presented face and the test face could (independently) have an Angry or Neutral expression, which the observer could not change. Within each chain of observers, these expressions were held constant, while the race of the face initially seen by each observer was determined by the previous observer's response. The chains reliably converged on a region well within White space, even when they started out near (or at) the Black extreme a?? as observers' representations were pulled toward a 'default attractor' in the White region of the face space. Strikingly, however, there was a single situation when this pattern reliably reversed: when observers were shown an Angry face and tested on a Neutral face, chains converged instead on a region well within the Black region. This is exactly the pattern that is predicted if Angry faces are misremembered as Blacker than the equivalent Neutral faces (since the effect cancels out when both faces are Angry). These results illustrate how irrelevant stereotype-consistent information can influence face representations in a deep way, which may have important real-world implications.