A randomized evaluation of a 9-month comprehensive math program for preprimary Bangladeshi school children addressed skills such as numbers, measurement, shapes, patterns, and space. Nine schools were randomly selected for the Intervention group and nine for the Control group, with 12 children randomly selected from each to participate in testing. Pre- and posttested scores on each unit's skills and a cumulative test at the end were subjected to analyses of covariance. Intervention children doubled their scores on almost all tests while control children increased only slightly. Reasons for the program's effectiveness were discussed. (Contains 2 tables.)