Sixty-two adults with mental retardation of heterogeneous etiology performed four facial emotion discrimination tasks and two facial nonemotion tasks. Staff members familiar with the participants completed measures of social adjustment (the Socialization and Communication domains of the Vineland Adaptive Behavior Scales and the Social Performance Survey Schedule). All facial discrimination tasks had very good reliability (internal consistency), but only some of the tasks correlated with measures of social adjustment. Furthermore, no evidence was found that emotion tasks and nonemotion tasks assessed different social constructs. Emotion tasks in which participants were presented with visual emotion stimuli correlated significantly with prosocial behavior, whereas those with verbal emotion stimuli did not.