Even though attrition presents a serious threat to the validity of longitudinal research, there is a lack of standardization for reporting it. To examine the effectiveness of four tests for attrition -- two for external validity and two for internal validity -- the tests were applied to two large scale projects: the first, Project HASP, is a smoking, alcohol, and marijuana prevention project implemented with a high school population; the second project (TAPP) is a tobacco and alcohol intervention project conducted with a junior high school population. The first test examined differences in pretest scores for drop-outs and stayers. Results showed that those identified at pretest as using tobacco, alcohol, or marijuana were over represented among drop-outs. Sex, age, and ethnic differences were also found, suggesting questionable external validity for both studies based on this test. The second test addressed attrition among those for whom multiple test data were available. Both projects showed the least likely group to drop out were the continuing non-users. Since treatment occurred between survey times, this finding suggests program generalizability may only apply to those who were relatively stable non-users. The third test assessed differential rates of attrition among conditions. Since attrition for the control group in Project HASP was high, it is possible that attrition masked actual program effects such that significant findings did not emerge. No differential attrition by treatment was observed between conditions in Project TAPP. The final test examined drop-outs' differential rates of behavior among conditions. For Project HASP there was no differential attrition among conditions by any particular substance-using group; in Project TAPP attrition by smokers and drinkers was roughly equally distributed between the experimental and control groups. The absence of a sure correction of attrition artifacts suggests researchers should take positive steps to prevent attrition if possible, and, at a minimum, report potential artifacts as proposed in these tests. The four tests provide a means for the research community to assess the weaknesses and strengths of prevention programs. (JAC)