Abstract:Discovering the mechanisms by which genetic variation influences phenotypes is integral to understanding life-history evolution. Models describing causal relationships among traits in a developmental hierarchy provide a functional basis for understanding the correlations often observed among life-history traits. In this paper, we evaluate a developmental network model of life-history traits based on the perennial herb Arabidopsis lyrata, evaluate phenotypic, genetic, and environmental covariance matrices obtained under different scenarios of quantitative trait locus (QTL) effects in simulated crosses, test the efficacy of structural equation modeling to identify the correct basis for multiple-trait QTL effects, and compare model predictions with field data. We found that the trait network constrained the phenotypic covariance patterns to varying degrees, depending on which traits were directly affected by QTLs. Genetic and environmental covariance matrices were strongly correlated only when direct QTL effects were spread over many traits. Structural equation models that included all simulated traits correctly identified traits directly affected by QTLs, but heuristic search algorithms found several network structures other than the correct one that also fit the data closely. Estimated correlations among a subset of traits in F2 data from field studies corresponded closely to model predictions when simulated QTLs affected traits known to differ between the parental populations. Our results show that causal trait network models can unify several aspects of quantitative genetic theory with empirical observations on genetic and phenotypic covariance patterns, and that incorporating trait networks into genetic analysis offers promise for elucidating mechanisms of life history evolution.