To examine the effects of varied roles, forms of feedback, and error spacing on subject attitudes and learning in programed instruction (PI), 125 students in an introductory psychology course were randomly assigned to treatment groups differentiated by three ways of spacing error-producing frames (spaced, semi-massed, and massed), two forms of feedback (knowledge of results or explanation), and three subject roles (teacher, pupil, or individual). Over 12 days, subjects completed a pretest, an attitude toward instructional media scale, study of the course material, a posttest, a computer-based testing attitude scale, and another posttest to assess the effects of explanation on learning. Results of analyses of variance of the data showed significant learning of course material for all students between pre- and posttest scores, that individual role subjects performed better than teacher and pupil role subjects, that feedback in the form of explanation was more effective than knowledge of results, and that the method of spacing error-producing frames did not adversely affect student attitudes toward PI. In addition, it appears that attitudes toward PI and computer-based testing were generally dependent on the number of errors made on the tests, and that student attitudes toward technology did not affect performance. (SP)