Gene-environment interaction (G x E) has been treated as both a statistical phenomenon and a biological reality. It is argued that, although there are important statistical issues that need to be considered, the focus has to be on the biological implications of G x E. Four reports of G x E deriving from the Dunedin longitudinal study are used as exemplars of the biological considerations that should lead to an hypothesis-driven choice of the specific genetic polymorphisms and the specific environmental influence to be investigated. The same four studies are used to discuss how the assessment of internal and external validity can be undertaken and how experimental approaches in humans and with animal models may be informative in the elucidation of the relevant operative biological mechanisms.