Experimental evaluations of substance use prevention programs are hamp
ered by high rates of subject attrition. Attrition threatens the inter
nal validity of these studies when attrition is greater in one conditi
on than another. Statistical tests of the effects of attrition are typ
ically lacking. Attrition threatens the external validity of studies b
ecause one may not be able to generalize effects of programs to dropou
ts who are systematically different from those who remain.