How robust is familiar face recognition? A repeat detection study of more than 1000 faces

Recent theories suggest that familiar faces have a robust representation in memory because they have been encountered over a wide variety of contexts and image changes (e.g. lighting, viewpoint and expression). By contrast, unfamiliar faces are encountered only once, and so they do not benefit from such richness of experience and are represented based on image-specific details. In this registered report, we used a repeat detection task to test whether familiar faces are recognized better than unfamiliar faces across image changes. Participants viewed a stream of more than 1000 celebrity face images for 0.5 s each, any of which might be repeated at a later point and has to be detected. Some participants saw the same image at repeats, while others saw a different image of the same face. A post-experimental familiarity check allowed us to determine which celebrities were and were not familiar to each participant. We had three predictions: (i) detection would be better for familiar than unfamiliar faces, (ii) detection would be better across same rather than different images, and (iii) detection of familiar faces would be comparable across same and different images, but detection of unfamiliar faces would be poorer across different images. We obtained support for the first two predictions but not the last. Instead, we found that repeat detection of faces, regardless of familiarity, was poorer across different images. Our study suggests that the robustness of familiar face recognition may have limits, and that under some conditions, familiar face recognition can be just as influenced by image changes as unfamiliar face recognition.