Autism is a rare but severe psychiatric disorder, including deficiencies in social and communicative behaviour. One of the more striking abnormalities observed in autistic children is their abnormal looking behaviour toward other persons in their environment. It has been suggested that these autistic children are unable to process facial information correctly. This notion has been supported by psychological experiments designed to test the ability of autistic subjects to recognise, match and identify faces.

These experiments showed that autistic children were less able to recognise, match and identify faces on basis of emotional expression and identity than normal children. Langdell (1978, Journal of Child Psychology and Psychiatry and Allied Disciplines, 19: 255-268) and Tantam et al. (1989, Journal of Child Psychology and Psychiatry, 30: 623-630) reported that autistic children performed worse in these tasks for upright faces, but that for faces which were turned upside down their performance was equal to or even better than in normal children. These results suggested that autistic children are less influenced by the turning of the human face, whereas the face-processing of normal children is hampered by this inversion effect. This disruption can also be seen in the 'Thatcher- illusion' where the upright orientation of the eyes and mouth in an upside-down face is hardly noticed.

These results led Langdell to suggest that the scanning of faces in autistic children differs from normal children. Autistic children would rely less on the natural configuration of the facial element and, hence, would be less influenced by the distortion of this configuration. We tested this hypothesis in an experiment where autistic and normal children looked at 12 upright oriented faces and twelve upside-down oriented faces. We recorded their eye movements and analysed their fixation behaviour for these faces.

Our results showed that normal children spent more time looking at upright oriented faces and spent less time looking at upside-down oriented faces than the autistic children. The looking times of autistic children were similar in the two conditions.

From this we can conclude that there is indeed a difference in looking time and possibly scanning strategies between autistic and normal children for human faces. The difference in scanning time could account for the differences in performances in psychological tasks with upright and upside-down faces. The equal performance for upright and upside-down faces as observed in autistic children is in concordance with the same amount of time these children spent looking at these two types of faces. The decrease in performance as observed in normal children is also in concordance with their decrease in looking time for upside-down faces. Further analysis and experiments are necessary to elaborate these results.