Abstract
With a single glance at a face we can typically discern the gender, race, approximate age, attractiveness, and whether we know the individual. At first thought one might assume that these decisions all rely on the same information and processes. However, a prosopagnosic, MJH is almost normal when discriminating gender and emotional expression, but evidences a severe deficit in identifying faces. Is the greater accuracy for categorizing a face compared to individuating it a consequence of a) the stimulus differences being greater for gender and expression than for individuation, or b) a consequence of differences in internal processing? We examined MJH's and normals' performance while carefully controlling the amount of contrast energy available for making three discriminations: expression, gender, and identity. Using a simple match-to-sample task, both MJH and normals showed a marked advantage for differentiating expression and gender over identity, even when both class uncertainty and low level stimulus energies were carefully controlled. Furthermore while MJH approached the lower bound of normal performance for discriminating faces which differed in expression or gender information, his threshold for detecting changes in identity information was several times that of normals. These results indicate that perception of expression and gender changes require less contrast energy than that required for individuation. An ideal observer analysis and an extension of the response classification technique (Ahumada, 1996) provide methods for determining and visualizing what information is available in the stimuli, and what information our observers appear to be using across the three discrimination tasks.