Skill acquisition by secondary-level severely mentally retarded students was assessed on functional community and living skills task analyses. Students were trained in natural settings according to a behavioral analysis of the curriculum into systematic teaching and error-correction procedures. Classroom training in the district high school supported the teaching that occurred in the community. The evaluation of student performance on specific skills was done using random assignment to training and control conditions in a cross-over design. Students showed significantly greater gains on those tasks for which they were trained than on those for which they were not. In addition to this specific skill training, a daily checklist showed significant progress by all students on a broad array of skill sequences. Discussion stressed the contribution of the project to ideologically appropriate public school programs for severely mentally retarded students, to behaviorally based instructional technology, and to evaluation using true experimental designs.