Ten autistic individuals (mean age: 12.7 years, SD 3.8, range 5.10-16.0), 10 Down individuals (12.3 years, SD 3.0, range 7.1-16.0), and a control group of 10 children with normal development (mean age: 6.3 years, SD 1.6, range 4.0-9.4), matched for verbal mental age, were tested on a delayed-matching task and on a sorting-by-preference task. The first task required subjects to match faces on the basis of the emotion being expressed or on the basis of identity. Different from the typical simultaneous matching procedure the target picture was shortly presented (750 msec) and was not visible when the sample pictures were shown to the subject, thus reducing the possible use of perceptual, piecemeal, processing strategies based on the typical features of the emotional facial expression. In the second task, subjects were required to rate the valence of an isolated stimulus, such as facial expression of emotion or an emotional situation in which no people were represented. The aim of the second task was to compare the autistic and nonautistic children's tendency to judge pleasantness of a face using facial expression of emotion as a meaningful index. Results showed a significantly worse performance in autistic individuals than in both normal and Down subjects on both facial expression of emotion subtasks, although on the identity and emotional situation subtasks there were no significant differences between groups.