Study limitations are noted. First, studies of a specific gene have the advantage of providing information about that gene, but largely ignore other genes potentially involved (Dick & Kendler, 2012b). However, studying a strong candidate gene provides specific information regarding the mechanism by which a gene influences an outcome. Second, most studies of adult populations use retrospective reporting of adverse childhood events (Hardt & Rutter, 2004). However, self-reports may be unstable over time (Fergusson, Horwood, & Woodward, 2000; Polanczyk et al., 2009), resulting in underestimation of childhood adversity prevalence. Thus, our retrospective items of childhood abuse and neglect are a potential limitation of this study. However, this bias is likely insufficient to invalidate retrospective studies of easily defined major adversities (Hardt & Rutter, 2004) such as parental divorce. Furthermore, as parental divorce showed higher prevalence than abuse or neglect, we examined the association between parental divorce and both alcohol phenotypes. Parental divorce alone showed a weaker relationship to the alcohol phenotypes (β [MaxDrinks]=0.121, p-value<0.01; β [AUD severity]=0.122, p-value<0.01) than the combined abuse/neglect/parental divorce variable (β [MaxDrinks]=0.136, p-value<0.001; β [AUD severity]=0.139),
