Genetic confounds in the study of sexual orientation: comment on Roberts, Glymour, and Koenen (2014).

Roberts, Glymour, and Koenen (2013) presented evidence that childhood maltreatment is related to adult homosexuality, using an instrumental variables regression analysis. Briefly, several instrumentalvariables—presenceofa stepparent,poverty,parental alcohol abuse, and parental mental illness—were related to adult homosexuality, but these relations were statistically mediated by childhood maltreatment. Roberts et al. concluded that childhood maltreatment causes adult homosexuality. We criticized the statistical approach of Roberts et al. (2013), arguing that the assumptions of both instrumental variables regression and mediation analysis were almost certainly violated (Bailey & Bailey, 2013). The instruments—stepparent presence, poverty, parental alcohol abuse, and parental mental illness—are complex in their causes and effects, and might be correlated with adult sexual orientation, through a multitude of complex pathways that do not require adult sexual orientation to be caused by the instruments via childhood maltreatment. The study falls far short of being a‘‘natural experiment,’’on which the instrumental variables regression approach is designed to capitalize, despite Roberts et al.’s characterization of their study as such. If an unmeasured third variable can cause both the instruments and outcome, the results from instrumental variables regression and mediation analyses cannot be taken as evidence for the causal pathways tested in these models. We proposed that the findings from Roberts et al. (2013) were equally consistent with an alternative causal explanation of adult sexualorientation,wheregenesthat influenceanindividual’ssexual orientation also influence an individual’s personality. Both neuroticism (Zietsch, Verweij, Bailey, Wright, & Martin, 2011) and depression (Zietsch et al., 2012) have been found to correlate at the genetic level with adult sexual orientation. Therefore, parents with these genes may be more likely to divorce (resulting in stepparent presence), live in poverty, abuse alcohol, be diagnosed with mental illness, and to have children who are maltreated. Under this model, Roberts et al.’s instrumental variables regression and mediation analyses would yield apparent evidence for an influence of childhood maltreatment on adult sexual orientation, even though maltreatment does not cause adult sexual orientation. Roberts, Glymour, and Koenen (2014) criticized our reply, stating that‘‘no genetic research supports [the] possibility’’that genes associated with sexual orientation may also be related to parental relationship instability, mental illness, alcohol use, and poverty.Further, theyclaimthat, for thehypothesisweproposed toaccountfor theirfindings,homosexualityandneuroticismmust both be caused by an allele of very large effect (accounting for 14 % of the variance in mother’s neuroticism and 15 % of the variance inchild’s sexualorientation).Robertsetal. asserted that these effects ‘‘are stronger, by an order of magnitude, than any established genetic determinant for any mental health or complex behavioral outcome.’’ Ifweunderstand‘‘establishedgeneticdeterminant’’asasingle gene, then we agree with Roberts et al. (2014). To date, there have been no replicable findings of genes with effects of this magnitude (i.e., 14 % of the variance in neuroticism and 15 % of the variance in sexual orientation). However, our argument does not rely on the assumption that the genetic correlation is due to a D. H. Bailey Department of Psychology, Carnegie Mellon University, Pittsburgh, PA, USA