The pretest-posttest design referred to as Design 2 by Campbell and Stanley (1963) is commonly used in educational research and evaluation. The tenability of the assumption of a zero population difference commonly used with this design is questioned. A nonzero population estimate based on the mean difference observed in test-retest reliability data is recommended. When a control group is available, it is recommended that the pretest-posttest difference for the control group be subtracted from the experimental group difference. This will produce a more accurate estimate of the magnitude of change for the experimental group. (Author)