This study had three aims: to compare a schizophrenia sample (n = 50) with a substance abuse (n = 25) and normal sample (n = 81) on affect recognition; to compare differences in their performance between positive and negative affect recognition; and to introduce a new videotape method of stimulus presentation. Subjects were asked to identify the predominant affect depicted in 21 5-10-s vignettes containing three trials of seven affect states. Results demonstrate significant group differences: normal subjects scored in the normal or mild range, substance abuse (s/a) subjects scored in the mild and moderate ranges, and the schizophrenia sample scored predominantly in the moderate to severe ranges. Accuracies were 92.3% for the normal sample, 77.2 for the s/a sample and 64.8 for the schizophrenia sample. Response dispersions were 97.6% for the schizophrenia group, 69% for the s/a sample and 38% in the normal sample. A repeated measures ANOVA revealed a group by type of affect interaction with schizophrenia subjects showing far greater differential impairment on negative affect recognition. Difficulty of item did not contribute to this difference. Test-retest reliability at 5 months for this new method was r = 0.76, and stability of categorization was very high over 5 months (weighted kappa = 0.93). These affect recognition deficits in schizophrenia are discussed as they relate to lateralization of brain function, high EE families, social skills impairment and implications for rehabilitation services.