This study examined whether the efficacy of ROOTS, a 50-lesson mathematics intervention program focused on whole number concepts for at-risk kindergarten students, differed by group size and whether initial skill moderated intervention effects by group size. The study utilized a randomized block design with at-risk students (n = 1,251) within classrooms (n = 138) randomly assigned to one of two treatment conditions (a small group of two or five students) or the control condition. Proximal and distal measures were collected in the fall (pretest), spring (posttest), and winter of first grade (follow-up). Results indicated that students who participated in ROOTS performed better at posttest than control students (Hedges's "g" from 0.09 to 0.81), that impact did not vary by group size, and that initial skill moderated the impact of ROOTS compared with control student outcomes but likely was not due to differences in group size.