You catch a glimpse of someone you think you know, but who you haven't seen for years. You're not sure it's them, then something - a raise of the eyebrow, a toss of the head - gives them away.

Without any other information, we can recognize and sex individuals from how they move their heads and faces, researchers now report1. The finding could improve face-recognition security and help to humanize animated 'synthespians'.

Harold Hill and Alan Johnston of University College London made a computer model of an average head, based on 200 men and women. They used motion-capture, like that used to animate computer games characters or in movie special effects, to record the facial movements of individuals telling a joke. They then recreated this motion in the average head, thus retaining movement information but not features or gender.

Observers shown only rigid head movements - nodding, shaking or tilting - without the flexible changes in expression were best at identifying individuals. Non-rigid movements alone did not seem to be enough to distinguish identity.

Hill suspects that these features, which are typically used to communicate emotion, or in speech, need to be the same despite individual differences. "Rigid head movements can be more idiosyncratic," he says.

Whether or not we're designed to exploit motion information to recognize familiar people in every day life "is arguable", says Tim Valentine, a psychologist at Goldsmiths College, London. "But we can do it when we have to." Motion might be a back-up to identification when other cues are unreliable, such as in poor lighting, Valentine suggests.

Hill and Johnston also asked their subjects to determine the joke-tellers' sex. This time non-rigid movements seemed to be the give away, suggesting that we have learnt to recognize systematic differences in the facial expressions used by men and women. The evolutionary biologist William Hamilton noted that women pull fewer lopsided faces than men.

Turning the animations upside down hampered people's judgement: "It's almost as if it turns off the face processing system," says Hill. This suggests that motion feeds into the brain's specialized ability to recognize faces.

Computer security systems that recognize faces could be programmed to take account of the individuality of motion, says Michael Lewis, a psychologist at the University of Cardiff, Wales. These currently use static facial images, which, he points out "could be fooled by a photograph".

Movement idiosyncrasies could also be used to enhance virtual actors' performances. Says Valentine: "If you were creating a digital stuntman, you'd want to capture individual, rather than generic movements."