Neuroimaging-based biotyping of attention-deficit/hyperactivity disorder (ADHD) has advancedrapidly, exemplified by the recent identification of three morphometric similarity network–derivedbiotypes by Pan et al. (2026). However, the normative modeling framework underlying suchapproaches contains systematic methodological blind spots that limit validity and generalizability.Drawing on evidence from neurodiversity research and the clinical literature on twice-exceptional(2E) populations, we identify four interrelated problems: (1) the unexamined variance structure ofnormative baselines, (2) the use of composite cognitive scores that flatten diagnostically meaningfulwithin-individual variance, (3) the failure to account for compensatory masking in cognitively giftedindividuals with ADHD, and (4) the mismatch between the exploratory logic of biotyping designsand the narrow range of phenotypic data typically collected. For each problem, we propose concrete,actionable remedies. We conclude that “typical development,” as operationalized in normativemodeling, functions as a statistical convenience rather than a description of a biologically uniformpopulation, and that making this assumption transparent would strengthen rather than undermine thebiotyping enterprise.
Franny Philos Sophia (Sun,) studied this question.