Abstract
In the last thirty years, the question of whether face identity and emotional expression are processed independently or interactively has garnered much attention, and a large number of studies have been performed to answer it. Although recent reviews conclude that identity and expression are processed interactively, the behavioral literature is plagued by contradictory results. This may be due in part to the lack of control of stimulus and decisional factors in most studies. An ideal experiment should (at the very least) control for several stimulus factors, including low-level changes correlated with identity (e.g., facial hair, shading, skin color and texture, etc.), strength of the expression shown by different actors, and discriminability of the two dimensions. Additionally, the experimental design and analysis should allow dissociating perceptual from decisional processes. To the best of our knowledge, no previous study has achieved this level of control, and without it any observed interaction could be explained by stimulus and decisional factors, rather than perceptual processing. Here, we created three-dimensional artificial face models and expression pose models (based on face photographs from the KDEF database) that allow for tight control of all the aforementioned stimulus factors. We obtained two sets of four stimuli, each resulting from the combination of two emotional expressions (neutral and angry) and two identities. Using face morphing, and guided by psychometric data from a pilot study, we manipulated the difference between the two levels of emotion and identity so that average discriminability was the same for the two dimensions. Participants were asked to complete an identification task involving the resulting stimulus sets, and the data was analyzed using a general recognition theory model (GRT-wIND) that allows to dissociate perceptual from decisional factors in the study of dimensional interactions. Under such tight experimental control, the dimensions were found to be perceptually separable.
Meeting abstract presented at VSS 2018