What can we learn from twin studies? A comprehensive evaluation of the equal environments assumption

•The goal was a comprehensive test of the equal environments assumption (EEA).•I reanalyzed data from a frequently cited study, Loehlin and Nichols (1976).•I also analyzed data on twins from the Midlife Development in the United States survey.•Results suggest a middle ground: the EEA is not strictly valid, but bias is modest.•Controls for environmental similarity significantly reduced heritability of neuroticism.

Twin studies are a major source of information about genetic effects on behavior, but they depend on a controversial assumption known as the equal environments assumption (EEA): that similarity in co-twins’ environments is not predictive of similarity in co-twin outcomes. Although evidence has largely supported the EEA, critics have claimed that environmental similarity has not been measured well, and most studies of the EEA have focused on outcomes related to health and psychology. This article addresses these limitations through (1) a reanalysis of data from the most cited study of the EEA, Loehlin and Nichols (1976), using better measures, and through (2) an analysis of nationally representative twin data from MIDUS using more comprehensive controls on a wider variety of outcomes than previous studies. Results support a middle ground position; it is likely that the EEA is not strictly valid for most outcomes, but the resulting bias is likely modest.