Generalization of affective learning about faces to perceptually similar faces

Different individuals have different (and different-looking) significant others, friends, and foes. The objective of this study was to investigate whether these social face environments can shape individual face preferences. First, participants learned to associate faces with positive, neutral, or negative behaviors. Then, they evaluated morphs combining novel faces with the learned faces. The morphs (65% and 80% novel faces) were within the categorical boundary of the novel faces: They were perceived as those faces in a preliminary study. Moreover, a second preliminary study showed that following the learning, the morphs' categorization as similar to the learned faces was indistinguishable from the categorization of actual novel faces. Nevertheless, in the main experiment, participants evaluated morphs of "positive" faces more positively than morphs of "negative" faces. This learning generalization effect increased as a function of the similarity of the novel faces to the learned faces. The findings suggest that general learning mechanisms based on similarity can account for idiosyncratic face preferences.