Previous research on class size has serious limitations because factors such as the abilities of pupils or the experience and teaching style of teachers were not controlled, and because of inconsistent criteria for large and small classes and for criteria of success. A major two-year study was mounted with four class sizes, random assignment of teachers and pupils to class sizes, and a variety of measures--observation, achievement, and attitudinal. The study was successfully implemented and completed because of the process of consultation and involvement followed. The paper details the design and the process. (Author)