Abstract
Humans are highly accurate at recognising familiar faces, even from highly variable and completely novel images. This remarkable ability is based on robust visual representations that are activated by nearly any image of a known face. However, relatively little is known about the specific characteristics of such face representations. To investigate some of their key properties, the present experiments tested how much idiosyncratic identity information and how much time to extract such information from a face are necessary to activate them. We analysed the so-called N250r, which is an event-related brain potential correlate of immediate repetition priming, and reflects the activation of an (image-independent) familiar face representation. The effect consists of more negative amplitudes at occipito-temporal electrodes between 200-400ms after target onset when said target is preceded by the face of the same relative to a different identity. In Experiment 1, we varied the amount of identity information available in the prime (100%, 75%, 50%, 25%) by morphing it towards an average face. We found that only full identity primes elicited a reliable N250r. In Experiment 2, we then varied the duration of the prime (33ms, 67ms, 100ms) to test how much time is needed to extract the relevant identity information. Here, we observed a significant N250r in all conditions, including the shortest presentation time. Together, these findings demonstrate that familiar face representations are at the same time highly sensitive to the resemblance of a particular identity and highly efficiently activated if a matching identity is presented. These properties allow for both effective and efficient face recognition, as they prevent erroneous activation of a specific representation by a similar-looking but non-matching person while also enabling fast identification of the correct face.