In adoption studies, the possibility of inadvertent matching between biological and adoptive parents for some environmental variable (known or unknown) correlated with illness must be considered. We examine such bias quantitatively and show how a genetic effect can be simulated. Existence of a genetic effect which is independent of environmental correlation can be accepted, when the frequency of a disease in adoptees who have a biological parent affected and no adoptive parents affected is significantly greater than the frequency of the disease in adoptees who have an adopted parent affected and no biological parents affected. The published data on schizophrenia, alcoholism, and criminality do not exclude the possibility of undetected environmental correlations simulating a genetic effect, according to this direct criterion.