Weekly observations of direct-care staff in a facility for persons with brain injury yielded less than optimal interactional style with facility residents. Following an observational baseline, staff were asked to self-rate a 15-min video sample of their interaction behavior with participants on their unit. They were then asked to compare their self-ratings to those of a supervisor, as well as view a video exemplar of appropriate positive interaction behavior. Elements of their interactional style were highlighted and specific feedback was provided on how to improve their performance. Interaction style was then reevaluated via an unobtrusive observer and yielded positive gains for all participants. Subsequent on-the-job feedback sustained performance gains. (Contains 1 figure.)