Three paraprofessional staff members and five delinquent girls were observed for 15 days. Behavior was coded according to a rating scale which attempted to summarize each subject's interaction with people and events in the immediate environment. An interaction was defined as a 6-second interval during which stimulus source and behavior and target subject response were recorded. The data indicate that paraprofessional staff and delinquent peers attended positively to a majority of all inmate behavior. While inmates positively attend to staff prosocial talk and passive behavior, they deliver a high rate of nonpositive attention to therapeutic opportunities created by staff. Perhaps these contingencies result in inmate coercion of the staff. Inmates respond positively, provided staff do not attempt to elicit appropriate behavior. The observed low proportion of opportunities created by staff may reflect the success of the inmate coercion process. These findings have implications for the training of paraprofessional staff. To achieve successful modification of behavior, staff must be taught to discriminate between appropriate and inappropriate inmate behaviors and to control their own response in a contingent manner. Psychologists must teach effective tactics to cope with aversive stimuli, and staff need an external form of maintenance which will be more reinforcing than the approval of the inmates for nontherapeutic behaviors. The concept of therapeutic opportunities may prove to be a successful technique for modifying delinquent behavior. Staff could be taught to recognize and increase stimulus events or behavior which will elicit appropriate behavior from the client. In this way, coercive behavior may be overcome in a constructive and efficient method of behavioral control. Reference notes, tables, and 17 references are provided. (Author abstract modified)