Using the Genetic Analysis Workshop 12 data, we applied case-control and case-case approaches to study the effects of a major gene and its interaction with sex on the disease liability. Although no joint additive effect was simulated, the case-case approach detected a small but significant multiplicative interaction effect, which could not be explained by the effect of random error. Given that analyses of "real" data will not be made with the knowledge of the true effects a priori, this study showed that the measure of gene x environment interaction is critical and the definition of interaction should be explicit.