Why are we so bad at face recognition?
We are remarkably good at recognising our friends and family across a wide range of conditions. However, it seems we overgeneralise this ability - falsely believing ourselves to be good at recognising previously unfamiliar faces.
I will present some experiments supporting these claims, concluding that everyday use of photo-ID is inconsistent with scientific knowledge of human perception. I will argue that the psychological concept 'configural processing' has misled us in our attempts to understand face recognition, and that a more sophisticated representational scheme, based on the statistics of images, will be required to make significant progress in the field.