As described elsewhere (Boardman et al., 2013), the existing gene-environment interaction typology includes models in which genetic risk may be the most evident in the least risky, the most risky, or the typical environments. Depending on the anticipated GxE relationship and the specific phenotype, environments may either trigger or control genetic expression in a causal manner, or they may simply mask otherwise small genetic associations. Without a representation of the full range of environments, one may conclude that a specific polymorphism is either protective, risky, or not associated with a particular phenotype. Belksy and Pluess (2009) make a very strong case for the differential susceptibility hypothesis that argues that environmentally sensitive loci will be protective in the most enriching environments but deleterious in unhealthy environments. This cross-over association cannot be identified without a representative sample from the full continuum of environments that is, again, why the representativeness of the Add Health study is such an important resource in conjunction with the pairs data.
