ABSTRACT
Recent work has shown that expression recognition shows holistic processing effects much like face recognition (Calder et al., ress). We extend our previous model of facial expression recognition (Dailey et al., 2000) to account for these results. We show that our model, with small modifications to the training procedure, can account for the systematic biases between upper and lower facial expression recognition, and the holistic/configural processing effect. Finally, we show that results that seem to support the idea that separate representations are necessary for emotion and identity processing can be accounted for by a single representation model. This latter effect is demonstrated in subjects by constructing chimeric faces by taking the top half of one face and the bottom half of another.
