Differential achievement growth patterns between comparison groups is a problem associated with data analysis in compensatory education programs. Children in greatest need of additional assistance, are usually assigned to the program rather than to an alternative treatment so that the comparison groups may vary in several ways, in addition to the effect of the treatment each group receives. A gain score procedure is recommended as a solution, using an approach that is based on the regression of achievement on time and requires the availability of pretreatment achievement data from two points in time prior to the beginning of the program. It is shown that the effect estimated is unbiased for any linear model of academic growth, including the fan spread hypothesis. Finally, the precision of the effect estimated compares favorably with competing analytic strategies. (Author/CP)