Computerized interactive videotapes were used to train college students to use a 10-second partial-interval observational recording system. Students viewed videotapes and scored response occurrences on a computer keyboard. Incorrect scoring resulted in immediate computerized feedback and rescoring.Two independent variables, three versus seven target behaviors, and maintenance (thinning feedback) versus nonmaintenance (no feedback) training were compared. Average accuracy values across all target behaviors were consistently above 90%, and the lowest accuracy values obtained for most target responses typically met current standards for interobserver agreement. Data indicate the standard practice of occasional observer agreement checks may be inadequate.