A sample of 1755 pupils in 232 elementary-school classes in the State of Tennessee participated in a 3 year longitudinal study of the effects of reduced class size on reading and mathematics performance. Pupils and teachers were randomly assigned to one of three experimental conditions within each participating school: a small class with a median enrollment of 15 students, a regular class with a median enrollment of 22 students, or a regular class with a full-time teacher aide. While no significant effect was found for teacher-aide classes, the small-class setting increased performance significantly in both mathematics and reading. The small-class advantage occurred originally in kindergarten, and was maintained without increasing or decreasing significantly over the following two years. The impact of reduced class size was especially strong in the inner-city schools, with high minority enrollments and performance that was otherwise below that of schools in other locations. These results warrant further study to understand the effects of small classes on teachers' morale, on student-teacher interaction, and on the extent and ways in which students participate in learning activities.