2050 words

Introduction

Claims about the genetic determination of psychological traits, like intelligence, continue to be talked about in the academic literature using twin and adoption studies and more recently GWAS and PGSs. High heritability is frequently taken to indicate that a trait is “largely genetic”, that genetic differences are explanatorily primary, or that environmental interventions are comparatively limited in their effects. These inferences capture both soft and strong hereditarianism, ranging from explicit genetic determinism to probablistic genetic influence.

But at the same time, critics like Lewontin (1974), Block and Dworkin, Oyama (2000), Rose (2006), and Moore and Shenk (2016)—among others—have emphasized that heritability is a population- and environment-relevant statistic describing the proportion of phenotypic variance associated with genetic variance under specific background conditions. It does not, by itself, identify causal mechanisms, degrees of genetic control, or the developmental sources of individual traits. Moore and Shenk have labeled the inference from high heritability to genetic determination the “heritability fallacy.” Despite widespread agreement that heritability is often misinterpreted, genetic conclusions continue to be drawn from heritability estimates, suggesting that this underlying inferential error has yet to be identified.

I will argue that the persistence of hereditarian interpretations of heritability reflect a deeper logical one than a mere empirical one. I will show that any attempt to infer genetic determination from heritability must take one of two forms. Either high heritability is treated as evidence for genetic determination, in which case the inference is logically invalid committing the fallacy of affirming the consequent; or high heritability is treated as entailing genetic determination, in which case the inference rests on a false case of what heritability measures. These two options exhaust the plausible inferential routes from heritability to hereditarian conclusions, and in neither case does the inference succeed.

So by formalizing the dilemma, I will show the logical structure of the heritability fallacy, and show that it is not confined to outdated arguments. This dilemma applies equally to contemporary and cautious forms of hereditarian causation that emphasize probablistic causation, polygenicity, or gene-environment interaction. Because heritability does not track causal contribution, developmental control, or counterfactual stability, no refinement of statistical technique can bridge the inferential gap that hereditarianism requires.

This argument doesn’t deny that genes are necessary for development, that genetic variation exists or that genetic differences may correlate with phenotypic differences under certain conditions. Rather, it shows that heritability cannot bear the explanatory weight that hereditarianism requires of it. Thus, hereditarianism fails not because of missing data or insufficiently sophisticated models, but because the central inference on which it relies cannot be made in principle.

The argument

This argument has two horns: (a) high heritable entails genetic determination or (b) high heritability correlates with genetic determination.

P1: If heritability licenses inferences about genetic determination, then either: (a) high heritability entails genetic determination or (b) high heritability merely correlates with genetic determination.

P2: If (a), the inference from high heritability to genetic determination is valid only if heritability measures genetic determination, which it does not.

P3: If (b), the inference commits the fallacy of affirming the consequent.

P4: Heritability measures only population-level variance partitioning, not individual-level causal genetic determination.

C: Therefore, heritability cannot license inferences about genetic determination.

C1: Therefore, hereditarianism fails insofar it relies on heritability.

When it comes to (a), the argument is valid (if H -> G; H; ∴ G), but heritability doesn’t measure or capture genetic determination. For it to do that it would have to measure causal genetic control, track developmental mechanisms, generalize across environments, and distinguish genes from correlated environments. So horn (a) is a category mistake, a category error, and a misinterpretation of heritability.

When it comes to (b), it’s an invalid inference since it affirms the consequent. The form is (If G -> H; H; ∴ G). Basically, if genetic determination produces high heritability, then high heritability must imply genetic determination. This doesn’t follow because there are other causes of high heritability other than G: environmental uniformity, social stratification, canalized developmental pathways, institutional sorting, and shared rearing conditions, among other things. Thus, the inference is invalid since H does not uniquely imply G. The failure of horn (b) is structural, not empirical. Ultimately, the consequent does not uniquely imply the antecedent, therefore the argument affirms the consequent.

The hereditarian group differences argument is basically: (1) Within-group heritability -> individual genetic causation; (2) individual causation -> group mean difference; (3) group differences -> biological explanation. This is something that Herrnstein and Murray argued for, but Joseph and Richardson (2025) show that there is no evidence for the claim that they made.

Possible responses

“We don’t claim determination, just strong genetic influence.” “Strong genetic influence” can mean a multitude of things: large causal contribution, constraint on developmental outcomes, stability across environments, or explanatory primacy. But heritability doesn’t entail any of those options. High heritability is compatible with large environmental causal effects, massive developmental plasticity, complete reverseability under intervention, and zero mechanistic understanding. The dilemma still applies.

“Genes set limits, environments only operate within them.” This is similar to the “genes load the gun and environment pulls the trigger” claim. Unfortunately, for this claim to be meaningful, hereditarians must provide: a non-circular account of “genetic limits”, counterfactual stability across environments, and a developmental mechanism mapping genes -> phenotypes. However, no such limits are specified, reaction norms are unknown, and developmental systems theory shows that genes are necessary enabling conditions, not boundary-setting sufficient causes (eg Oyama, 2000; Noble, 2012). This claim just assumes what needs to be proven.

“Polygenic traits are different; heritability is about many small causes.” Unfortunately, polygenicity increases context-sensitivity, developmental underdeterminination, and environmental mediation. GWAS hits lack mechanistic understanding and are environment- and population-specific. Therefore, polygenicity undermines causal genetic determination.

“Heritability predicts outcomes so it must be causal.” This is just confusing prediction with explanation. Weather predicts umbrella use, and barometer predict storms. But does that say anything about prediction and causation? Obviously, prediction doesn’t entail causation. Heritability predicts variance under strict, fixed background conditions. It says nothing about why the trait exists, how the trait develops, and what would happen infer intervention.

Population stratification and the misattribution of genetic effects

GWAS is heritability operationalized without theory. It assumes additive genetic effects, stable phenotype measurement, linear genotype-phenotype mapping, and environment as noise. But association studies don’t identify causal mechanisms and unfortunately, more and more data won’t be able to ultimately save the claim, since correlations are inevitable (Richardson, 2017; Noble, 2018) But for traits like IQ (since that seems to be the favorite “trait” of the hereditarian GWAS proponent), measurement is rule-governed and non-ratio, development is socially scaffolded, and gene effects covary (non-causally, Richardson, 2017) with institutions, practices, and learning. So GWAS findings are statistically real, explanatorily inert and overinterpreted. PGS reifies population stratification, encodes social stratification, and fall under environmental change. Therefore, GWAS falls infer the same argument that is mounted here. Richardson (2017) has a masterful argument on how these differences arise, which I have reconstructed below.

High heritability within stratified societies does not indicate that observed group differences are genetically caused. Rather, social stratification simultaneously produces (i) persistent genetic stratification across social groups and (ii) systematic environmental differentiation that directly shapes cognitive development. As a result, between-group differences in cognitive test performance will necessarily mirror the structure of stratification, even when no trait differences are genetically determined.

In such contexts, heritability estimates capture covariwbce within an already-structured population, not causal genetic contribution. The same stratification processes that maintain patterned mating, residential segregation, and ancestry clustering also produce differential schooling quality, resource exposure, discrimination, and developmental opportunities. These environmental processes directly affect cognitive performance while remaining statistically aligned with ancestry. So the resulting group differences reflect the pattern of stratification rather than the action of genes.

P1: In socially stratified populations, social stratification produces both (a) persistent genetic stratification, and (b) systematic environmental differentiation relevant to cognitive development.

P2: Persistent environmental differentiation directly contrived to cognitive performance outcomes.

P3: Because genetic stratification and environmental differentiation are jointly generated by the same stratification processes, they are necessarily aligned across groups.

C1: Therefore, between-group differences in cognitive test performance in stratified populations necessarily reflect the pattern of ongoing social stratification.

C2: Therefore, the alignment of cognitive differences with genetic population structure does not imply genetic causation, but follows from the way stratification co-produces genetic clustering and environmental differentiation.

A discerning eye can see what I did here. The heritability pincer argument leads hereditarians into one of two inferences. A hereditarian can now say “Even if heritability isn’t causal, it still tracks genetic causation.” But the Richardson stratification argument blocks this by showing that social stratification sorts both genomes and environments, so cognitive differences can arise without genes being a causal mechanism of them.

Conclusion

The heritability fallacy consists in inferring genetic determination from high heritability, an inference that is either invalid (affirming the consequent) or valid only by presupposing the false premise that heritability measures causal genetic determination rather than population-relative variance.

The argument I have developed shows that hereditarian interpretations of heritability cannot license claims of genetic determinism. If heritability is to entail generic causation, then the inference presupposes that heritability measures causal genetic influence, which it doesn’t. If heritability is instead feared as merely correlates with genetic causation, then the inference from high heritability to genetic causation commits the fallacy of affirming the consequent. In either case, heritability is a statistic of variance partitioning at the level of populations, not a measure of individual-level or developmental causation. It does not identify what produces, sustains or structures phenotypic differences.

The population stratification argument clarifies why this logical failure is also an empirical one. In socially stratified societies, historical and ongoing processes of segregation, assortative mating, migration sorting, institutional selection, and unequal exposure to developmental environments generate joint covariwbce between socially structured genetic ancestry and socially structured environmental conditions. Persistent group differences in cognitive test performance in these contexts are therefore not independent of genetic causation. Rather, they necessarily reflect the underlying pattern of stratification that simultaneously produces environmental differentiation. Apparent “genetic signals” in behavioral or cognitive outcomes can therefore emerge even when the operative causal mechanisms are environmental and developmental.

Together, these argument converge on a single conclusion. Heritability estimates do not—and cannot—discriminate between causal pathways that arise from genes and those that arise from socially reproduced environments they covary with genetic structure. To infer genetic determination from heritability is therefore doubly unwarranted: it is a fallacy in form and is unreliable in stratified populations where social structure organizes both gene frequencies and developmental conditions along the same lines. Hereditarianism fails insofar as it treats heritability, or stratification-aligned outcome differences, as evidence of intrinsic generic causation rather than as reflections of the socially constructed systems that produce and maintain them.

The way that heritability is interpreted, along with how genes are viewed as causal for phenotypes, leads to the conclusion that DNA is a blueprint—Plomin claimed in his 2018 book that DNA is a “fortune-teller”, while genetic tools like GWAS/PGSs are “fortune-telling devices.” Genes don’t carry context-independent information, so genes aren’t a blueprint or recipe (Schneider, 2007; Noble, 2024).

This pincer argument along with the social stratification argument shows one thing: hereditarianism is a failure of an explanation and that environment can explain these observations, because ultimately, genes don’t do anything on their own and done act in the way hereditarians need them to for their conclusions to be true and for their research programme to be valid.

The hereditarian has no out here. Due to the assumptions they hold about how traits are fixated and how traits are caused, they are married to certain assumptions and beliefs. And if the assumptions and belief they hold are false, then their beliefs don’t hold under scrutiny. Heritability is a way to give an air of scientific-ness to their arguments by claiming measurement and putting numbers to things, but this argument I’ve mounted here just shows yet again that hereditarianism isn’t a valid science. Hereditarian beliefs about the causation of traits belong in an age where we don’t have the current theoretical, methodological and empirical knowledge on the physiology of genes and developmental systems. Hereditarians need to join the rest of us in the new millennium; their beliefs and arguments belong in the 1900s.