On the causal interpretation of heritability from a structural causal modeling perspective

Qiaoying Lu*, Pierrick Bourrat

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

1 Citation (Scopus)


Heritability estimated using the analysis of variance (ANOVA) for ascribing causal responsibility to genes for a phenotype has been criticized widely. First, there are problems associated with articulating the exact causal meaning of heritability in the standard model. Second, in conditions of gene–environment interaction or covariation that violate the assumptions made by the standard model, a causal interpretation of heritability is thought to be unwarranted. This paper aims to rethink these ideas and associated disputes from a structural causal modeling (SCM) perspective. Using SCM, we show that, in the standard model, heritability reflects the causal effect of eliminating genotypic differences on the change of phenotypic variance of a population. In the presence of interaction or covariation, heritability is estimated incorrectly using ANOVA. However, SCM can provide the causal effect of genotypes on the phenotypic variance regarding particular interventions. We also show that SCM can identify different types of causal effect and answer individual-level causal questions. We conclude that SCM has the resources to provide a systematic causal interpretation that can supplement traditional heritability estimates via ANOVA and offer a more substantial causal analysis of genetic causation.

Original languageEnglish
Pages (from-to)87-98
Number of pages12
JournalStudies in History and Philosophy of Science
Early online date16 Jun 2022
Publication statusPublished - Aug 2022


  • Causal interpretation
  • Gene–environment covariation
  • Gene–environment interaction
  • Heritability
  • Interventionist account of causation
  • Structural causal modeling


Dive into the research topics of 'On the causal interpretation of heritability from a structural causal modeling perspective'. Together they form a unique fingerprint.

Cite this