In principle, these different types of gene-environment interaction effects should be easily distinguishable, but in practice they are not. One can imagine many variations on these two clear-cut models, including variations in whether there are main effects of genotype at each end of the environmental continuum, and the extent to which the slope for each genotype differs as a function of the environment (see Kendler, in press, for further elaboration of these models and Belsky, Bakermans-Kranenburg, & van IJzendoorn, 2007, for steps to formally test for differential susceptibility). Our data are illustrative of these complexities: Although the shape of the interactions clearly suggests a crossover effect, the confidence intervals at the environmental ends of the distributions are large and overlapping; accordingly, the mean differences in externalizing behavior across genotypes are nonsignificant at either end. Thus, there is insufficient power to conclude that the genotypic effects exhibited significant crossing. This is a limitation frequently encountered in research at the intersection of genetics and psychology: Exploring nuanced questions about the effects associated with identified genes requires samples with detailed phenotypic and environmental
