When two groups, initially dissimilar, undergo different treatments, can subsequent differences be partitioned in such a way that the difference between the two treatments is unbiased? This is the central problem of this paper, and it is confronted by the examination of two levels of information using a Follow Through Evaluation. The first…