How does representation of faces change with increasing familiarity?

It is well established that familiar and unfamiliar faces are processed differently, in terms of both behavioural and neural responses. Although these differences indicate that representation of faces changes when faces become more and more familiar, how familiar and unfamiliar faces are represented in memory and how increasing familiarity changes face representation remain unclear. Here we investigated how faces are represented when participants saw them for the first time (i.e., unfamiliar faces), when participants had developed visual familiarity with them (i.e., visually familiar faces), and when participants knew the faces very well (i.e., personally familiar faces). Participants were shown an array of facial morphs varying from highly anti-caricatured (-50) to highly caricatured (+50), and were asked to select one matching the original face of the person best. The results showed that, for unfamiliar faces, participants’ response frequency was highest for the highly anti-caricatured faces (-50), and decreased when faces became more and more caricatured. For personally familiar faces, participants selected the original faces most often, and the further away a facial morph was from the original face, the less likely it was selected. Responses to visually familiar faces differed from responses to both unfamiliar and very familiar faces and can be described as a transition between them. These results indicate that unfamiliar faces are initially represented with a bias towards the facial norm, whereas personally familiar faces are represented veridically. Importantly, these data demonstrate that face representation shifts in face space with increasing familiarity. The more familiar a face is, the more its corresponding representation shifts from a location near the facial norm (i.e. anti-caricatured) to its veridical location in face space