posttest-only design with nonequivalent groups: Here we add a control group to the one-group posttest-only design. this design might be used if a treatment begins before the researcher is consulted so that pretest observations are not available on the scale used at posttest. the possibility of pretest group differences make it very hard to separate treatment from selection effects. a rationale sometimes offered for preferring this weak design is that pretest measurement may sensitize participants and so influence their posttest scores. But when different treatment groups are being compared, testing effects that are constant across groups do not threaten internal validity. only differential testing effects do. they are rare, but they do occur. they can be reduced by administering alternate forms of a test, by using item response theory to calibrate different tests to the same scale, by lengthening the time of interval between pretest and posttest, by using Solomon Four Group Design to assess the presence and impact of such effects, by using unobtrusive measures that are less reactive than self-report, by using retrospective pretests, and by using explicit reference groups or behavioral criteria to anchor responding. Even differential pretest sensitization is a problem, eliminating the pretests can entail much larger costs for detecting selection biases than dealing with them in ways outlined here.