Bulletin

Issue: July 2004

Research

I know that face! The eyes have it when it comes to recognition

La Trobe University eye researcher, Dr Adrian Dyer, after a series of innovative experiments, can help tell us how we learn to recognise a face. He has concluded that there is a template in our brain that enables us to imprint a face in our memory.

His experiments with students at La Trobe over the past two years have turned up new evidence that lead him to this conclusion. They have also strongly indicated that when looking at a face, our gaze tends to strongly concentrate on the eyes of that face and this helps the recognition process.

A post doctoral Research Fellow in the School of Orthoptics, Dr Dyer is investigating the role of eye movements in the perception and recognition of faces.

Dr Dyer said that before you can recognise a face, you have to 'learn' that face - have it implanted in a template. 'Once we build up the template, we become really good at recognising faces.'

To understand why humans are so good at recognising faces, he said it was first necessary to understand the process by which we 'learn' faces because we cannot recognise faces that we have not 'learned'.

The first experiments were designed to help him understand the learning process. These were fol-lowed by others designed to throw light on the subsequent recognition process.

To discover how long it took people to 'learn' a face, Dr Dyer recruited La Trobe student volunteers and subjected them to a series of tests. In the first test, each student viewed a computer monitor and a face would appear.

The students looked at the face for as long as they liked and then pushed a button to indicate that they thought they could recognise the face if they saw it again. Another face then appeared and the process was repeated a number of times. On average the students allowed themselves about five seconds to feel they would know the face again.

Students were shown a second series of faces and asked to indicate whether they were among those previously shown. With an average response time of only 1.5 seconds, the students responded the original faces with 93 per cent accuracy. In another experiment, students viewed famous faces presented for a short period of time on the screen. The students were able to identify a person after viewing a face for less than one twentieth of a second.

'We know that it is not possible to move the eye to multiple positions in one twentieth of a second, so we concluded that the eye does not have to move to recognise a face, even if the face is rotated. This indicates strongly that we use a low resolution eidetic - a photographic memory - to recognise a face.

'In another experiment, 25 students were shown a vertical half of a face with a blank space where the other half should have been. Sixty four per cent of the students looked at the blank space which strongly suggests they were influenced to do so by a top down model of processing faces in their brain.

'Another experiment strengthened the template hypothesis. Students were shown a series of pictures each depicting a series of simple line drawings including two circles which changed position in sequence until they formed a face. The students paid no attention to the circles until they finally became eyes in the final image.

'In further tests, students looked at faces broken up into 10 components - forehead, eyes, ears, nose, cheeks, mouth and chin - and we did a recognition test mapping the areas of the face they looked at.

'Those poor at recognising faces spent relatively more time looking at the mouth and lower parts of the face. Those who scored highly spent more time looking at the eyes, although they also looked at all other facial components.

'This indicates that eyes are critically important in the task of learning and subsequently recognising faces. We had suspected this because previous eye movement recordings even from young babies showed that a lot of attention is paid to the eyes.' •

back to top

back contents next