Now consider a binary environmental exposure such that the risk ratio for the causal variant on the exposure is 1.5. No main effect of the exposure is assumed, although this does not matter in this example. Assuming that the quantities in Table 1 apply to unexposed subjects, some further calculations give the risk ratio for the marker as 1.6 in the exposed and 1.5 in the unexposed subjects (Table 2 and Appendix A). This reveals an interaction between the marker and the exposure on the risk of disease, although there is none for the causal variant. We regard this interaction as spurious, because it does not correspond to an interaction at the causal variant.