This study tested the hypothesis that juvenile offenders react differently than nonoffenders by asking the two groups to label nonverbal emotional states while admitting involvement in a crime. Subjects, chosen from a large short-term detention center and an inner city school, were between 15- and 17-years old, and evenly divided between black and white youths. The emotional stimulus material consisted of 32 videotaped vignettes approximately 1 minute long in which male actors portrayed anger, joy, sadness-distress, or a neutral emotional state while describing crimes of theft or vandalism. Ratings for the emotional content of the flms were standardized by showing them to undergraduate students in a psychology class. The videotapes were shown to 20 delinquent and 36 nondelinquent males with the sound track distorted, and the subjects were asked to label the feelings of each actor. The videotapes were also presented to 55 delinquents and 37 nondelinquents with the sound intact, and participants were asked to suggest a punishment for each actor. The verbal intelligence of all subjects was measured by the Quick Test before the other tasks were administered. The delinquents scored significantly lower on both the intelligence and the emotional labeling tests. Analysis of the emotional reaction data, however, showed that the delinquency status had no effect on the results of the test, although the emotional categories affected the decisions of both groups. For example, the harshest punishments were given to actors who portrayed anger. These findings suggest that delinquents may react aggressively in threatening situations because of insufficient verbal resources, not social insensitivity, and that therapeutic programs might focus on improving verbal skills. Notes and a bibliography of 18 references are included.