Recent research suggests that small (1:15) classes in the primary grades improve a student's chance to obtain a solid educational foundation. Funding has proved a major hurdle to definitive class size research. Also, many class size studies are inconclusive, since few are longitudinal. Tennessee is weighing its burgeoning welfare and correctional costs against better educational outcomes possible through increased individual attention in smaller classes. In Fall 1984, Tennessee State University initiated a 3-year longitudinal study using an experimental (1:15) group versus a control (1:25) group. The study began in the first grades of two similar schools and followed cohorts of pupils moving through second and third grade. A "blind" control group was matched with the experimental group on five variables. Pupil test scores, self-concept, attendance, behavior, and teacher perceptions were analyzed using analysis of variance and co-variance techniques. In grade 1, important and statistically significant pupil achievement gains favored the experimental group, especially pupils of low socioeconomic status. By grade 2, achievement gains were less pronounced, and by grade 3 had evened out. There were no differences among groups in attendance. Although not statistically significant, self-concept continually improved in the experimental group, but showed no consistent pattern in the control or blind control groups. Behavior favored the experimental group in grade 1 and became significantly better in grades 2 and 3. In the first year, pupils got "up to speed" in smaller classes; over time, gains did not continually increase. The groups evened out by grade 3, but small-class pupils maintained what they had gained. Results clearly identified the need for a large-scale, longitudinal, class size study using the class, rather than the pupil, as the unit of analysis. Included are tables and 13 references. (MLH)