Answers are grounded in this paper and cite their sources.
Structural neuroimaging research is marked by substantial methodological variability, with important implications for inference and reproducibility. A central but often under-justified source for this variability is whether intracranial volume (ICV) is included as a covariate. We argue that ICV adjustment is not a neutral analytical step, but a modeling decision that fundamentally alters the research question by redefining the variance under investigation, and it can even change the very direction of the estimated effect. We evaluate five common research questions in the literature and argue that each is underpinned by distinct assumptions about where ICV sits in the causal structure linking neuroanatomy with other complex behavioral traits. We identify contexts in which ICV adjustment is necessary to define the construct under study, as well as situations in which it removes meaningful variance or introduces interpretational and statistical complications. We argue that ICV adjustment should explicitly be justified, and we provide practical guidance for deciding when such adjustment is appropriate. We hope to stimulate debate and improve consistency, interpretability, and reproducibility of population-level neuroimaging research.
No takes yet. Share an insight, caveat, or question.
Fürtjes et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: