Our results should be interpreted within the context of the following limitations. First, COGA is a high-risk sample and findings may not generalize to other populations. Second, the measure of parental relationship discord is retrospective and reported from the child’s perspective. Others report moderate correlations between children’s and parents’ perceptions of relationship discord52. Additionally, and as shown in Supplementary Table 12, mean parental relationship discord was higher among subjects who provided retrospective reports in SSAGA-II (age 6–13) compared to subjects who provided reports in SSAGA-IV (age 12–17), although the differences were modest (e.g., 1.74 vs. 1.19 in EAs, and 2.02 vs. 1.44 in AAs). Third, strict interpretations of mediated effects require that the predictor precedes the mediator in time. In this study, we modeled lifetime phenotypes and our results are not meant to be interpreted causally. Fourth, our conceptual model emphasizes a direction of influence from parents→offspring; however, we recognize that there are offspring effects on parents53, 54. If the offspring phenotype directly influences the parental phenotype, transmitted parental alleles and non-transmitted parental alleles provide quantitatively different contributions to the
