Since the mean score for a sample composed of several subgroups can be viewed as the sum of the mean of each subgroup weighted by the proportional size of the subgroup, then the mean change in a time period--in this case, from 1972 to 1980--is the sum of the differences between the means for each subgroup, with each mean weighted by its proportional size. Thus, the mean change for the total sample is influenced both by mean change in the score for each group--in this case, a test score--and by change in the proportional size of the group. A decline or gain in mean test scores has both a subgroup mean component and a population shift component. In addition, how much a particular subgroup may contribute to the total change can be shown to depend on what might be called its weighted distance from the original total mean. How these simple concepts can be used to decompose score gains or declines is demonstrated by means of data from the seniors in the National Longitudinal Study (NLS) High School Class of 1972 and High School and Beyond (HSB) 1980. The presentation should be of both methodological and substantive interest. (Author/LMO)