Postdiscrimination Gradients With Familiar and Unfamiliar Faces

Discrimination training is often used to improve accuracy in stimulus recognition. However, intradimensional discrimination training with a single positive exemplar and a single negative exemplar may reduce accuracy during a generalization test. More specifically, discrimination training can cause a shift in responding from the positive exemplar to stimuli unlike the negative exemplar. Previous studies have shown that the length of discrimination training affects the degree to which a shift occurs. We sought to determine whether pre-experimental exposure (“familiarity”) to the stimuli would have a similar effect. Twenty participants were trained and tested with two different sets of facial images. Each set of images consisted of stimuli that ranged from the naturally asymmetrical version of the face to a symmetrical version of the face. In one condition, the images were based on the participant’s own face; in the second condition, the images were based on the face of a stranger. In both cases, the median face served as the positive exemplar. In half of the conditions the naturally asymmetrical face served as the negative exemplar, and in the other half, the symmetrical face served as the negative exemplar. Images based on a stranger’s face produced greater generalization than images based on the participant’s own face. A significant gradient shift occurred only with familiar faces.