In her 2010 paper on diverging family structure, Amy Wax names the traits and stops. Discount rate, executive function, self-control, future orientation, conscientiousness, cognitive ability. She correlates them with education, occupation, and demographic group. She observes that they determine who thrives under normative deregulation. Then she declines to say where they come from. She knows what the next question is. Everyone reading the paper knows what the next question is. She writes that the source of the observed differences, whether grounded in culture or individual characteristics, warrants further investigation, and leaves the sentence to do the work of a chapter.
Prudence explains part of it. A working paper for an Elgar volume on family law cannot carry a full treatment of behavior genetics and survive as a paper about family law. Any paper that takes on the heritability of self-regulation across populations becomes a paper about that, and the local-versus-global model gets read as a vehicle rather than as a claim. Wax had already written enough to guarantee hostility. Adding the origins question would have guaranteed that nothing else in the paper got discussed.
Legal scholarship has no procedure for adjudicating this kind of empirical dispute. A law review cannot referee a quarrel among statistical geneticists. What it can do is cite, and citation in a contested field is taken as endorsement. Wax’s choices were to cite the hereditarian literature and be read as endorsing it, cite the environmentalist literature and be read as endorsing that, or cite neither. She cited neither, with two exceptions buried in footnotes: a 1981 Sandra Scarr volume on race, social class, and individual differences in IQ, and a reference to average racial and ethnic differences in personality profiles on employment screening tests.
Her argument runs: a latent distribution of self-regulatory capacity existed before 1960; strong norms masked it; deregulation revealed it; the revealed differential now compounds across generations. A theory that explains how a latent difference becomes manifest after a regime change is incomplete without a theory of where the latent difference came from. Without one, the framework risks circularity. Populations differ in outcomes because they differ in decision style; we infer the decision style from the outcomes.
Start with what the three literatures say.
The first literature, on within-population heritability of cognitive ability and personality, is about as settled as anything in social science. Twin, adoption, and sibling designs across many Western samples converge on adult heritability estimates for measured cognitive ability in the range of fifty to eighty percent, with the estimate rising through childhood and adolescence and stabilizing in adulthood. Personality traits, including conscientiousness, run lower, in the range of forty to fifty percent. These are variance-partitioning results within the populations sampled, and they do not say what any individual’s score is caused by. They are also not seriously contested by anyone who works with the data. Robert Plomin (b. 1948) has spent a career establishing them and Eric Turkheimer, who dislikes almost every use to which they are put, does not dispute them.
The second literature, on between-population differences, is where the confidence collapses. The classic point from Richard Lewontin (1929-2021) holds: within-group heritability licenses no inference about between-group causes. Two batches of seed with identical genetic variance, grown in different soils, produce mean differences that are entirely environmental while heritability within each batch stays high.
The obvious response is that molecular genetics should settle what quantitative genetics cannot. It has not. Polygenic scores for cognitive ability and educational attainment, built from genome-wide association studies conducted almost entirely in European-ancestry samples, lose most of their predictive power when applied to other ancestry groups. The loss comes from differences in linkage disequilibrium patterns, differences in allele frequencies, and confounding by population structure that within-sample controls do not remove. Comparing mean polygenic scores across ancestry groups, which several authors have attempted, is a procedure that statistical geneticists reject on the grounds that the tools currently cannot do the job.
Against the persistence of measured gaps stands the Flynn effect. James Flynn (1934-2020) documented gains in measured cognitive ability across the twentieth century of roughly three points per decade in many countries, which is to say that the environmental headroom on population means exceeds the size of the differences under dispute. Whatever produced those gains was not genetic change. Some portion of the American Black-White gap has narrowed over the same period, by amounts that vary by test and cohort. Transracial adoption studies exist on both sides of the question and all of them are small, old, and confounded.
The between-population question is open, the confidence expressed on both sides exceeds what the data support, and anyone who tells you it is settled is telling you about his coalition.
The behavior-genetic literature does contain a finding that wounds Wax’s argument. Wax’s fallback, and the fallback her sympathetic readers reach for, is the developmental loop. Family structures produce the developmental environments of children; those children acquire the traits; the traits reproduce the family structures. She writes that restraint, moderation, emotional control, trustworthiness, persistence, and sobriety develop best when children live with both biological parents in stable marital homes. The loop generates path-dependent equilibria without invoking any heritable substrate difference, which is exactly why it appeals.
The trouble is that the loop specifies a shared-environment effect, and shared environment is the component that behavior genetics has had the hardest time finding. In standard designs, the shared environment contribution to adult cognitive ability in Western samples shrinks toward zero. For personality traits including conscientiousness it is near zero at most ages. Siblings raised in the same home resemble each other on these traits at about the rate predicted by their genetic relatedness and little more. Adoptive siblings, who share the home and not the genes, correlate near zero as adults. If growing up with a married father transmitted self-command through the home environment, the designs that measure home environment should find it. They mostly do not.
The dissolution is available. Judith Rich Harris (1938-2018), in The Nurture Assumption, argued that the near-zero shared environment finding does not mean environment fails to socialize children. It means the socializing environment is the peer group. Children calibrate conduct to the group they expect to live among, and the calibration happens outside the family. From the standpoint of a twin design conducted within neighborhoods, peer influence loads onto the non-shared component, or vanishes into the residual, because siblings in the same home often occupy different peer positions and because the design cannot see variance that a whole community shares.
Harris’s account fits Wax’s framework better than Wax’s own account does. A normative regime operates at the level of a community. The pre-1960s architecture Wax describes worked because everyone in the neighborhood was following the same rule, which is what made following it cheap. That is a community-level variance component, and a twin study conducted inside a single normative environment measures nothing about it. Ask a twin design what fraction of variance in age at first intercourse is attributable to shared environment in Utah, and the answer tells you about differences among Utah families.
Shared environment estimates differ sharply by outcome. For measured intelligence and for personality traits, shared environment washes out. For behaviors like religiosity, political attitudes, smoking initiation, age at first intercourse, and fertility timing, shared environment persists into adulthood at levels that are not trivial. This is the structure Wax’s theory requires. Norms govern conduct, and conduct is the thing on which community-level environment continues to register.
Trait distribution may be substantially heritable, and the question of its cross-population structure may remain open, and neither fact touches the claim that a normative regime converts a given trait distribution into a given behavioral distribution. Wax’s argument is about a conversion. The origins question determines the ceiling on what environmental intervention might achieve.
Behavior geneticists have a name for what a strong normative regime does. Michael Shanahan and Scott Hofer call it social control, one of four types in the typology they set out in “Social Context in Gene-Environment Interactions: Retrospect and Prospect,” alongside triggering, compensation, and enhancement. Norms canalize genetically diverse people, they write, and as norms grow more effective and choices narrow, genetic differences carry less consequence. When a rule pushes everyone toward the same conduct regardless of disposition, dispositional variance stops appearing in behavior and the heritability of the outcome falls. Remove the rule and the outcome tracks the disposition. The same logic runs through the Scarr-Rowe finding, where heritability of cognitive ability comes out lower in low-socioeconomic-status American samples and shows up less consistently in European samples with stronger institutional floors.
The demographer J. Richard Udry (1928-2012) reached the same position from the other direction in “Biosocial Models of Low-Fertility Societies,” fourteen years before Wax wrote. The more freedom individuals hold over reproductive behavior, he argued, the more the variation in that behavior reflects biological differences between them.
Wax cites neither Udry nor Shanahan and Hofer. Her account of normative architecture arrives at a claim that demographers and behavior geneticists had already stated in their own idiom, and the two literatures have never met.
Where the origins question bites is on policy. If the trait distribution responds to environment, environmental intervention can compress it, and the pessimism about the post-deregulation equilibrium is overstated. If the distribution has a substantial heritable component that intervention cannot reach, the only arrangement that compensates for the differential is one that restores external rules, and the argument for restoration is arithmetic.
A paper that names the traits, correlates them with racial groups, and refuses to say where they come from leaves the reader to supply the missing term. Readers tend to supply the one that is convenient. The hostile reader concludes she means heredity and is hiding it. The friendly reader concludes she means heredity and is being brave about it. Both conclusions are inferences from a gap, and both are unfalsifiable, since she wrote nothing to check them against. Silence in a combustible field transfers authorship of the missing paragraph to whoever wants to write it.
The alternative she had available was to state the disjunction and show that her argument survives both branches. That paragraph runs about two hundred words. It says the origins of the trait distribution are contested. It says molecular methods have not settled the contest and may not for some time. It says the descriptive claim holds under either resolution, and that the policy implications diverge, with the environmental branch supporting intervention and the heritable branch supporting institutional restoration. Writing it would have cost her the accusation of hiding something and gained her the accusation of raising it. Whether that trade made sense in 2010 is a judgment about the field she was writing in rather than about the evidence, and the field has moved enough since that the calculation might come out differently now.
Nobody knows the cross-population structure of the traits Wax describes. The tools that could answer the question do not work across ancestry groups yet, and the inferential shortcut from within-group heritability is invalid. The within-population heritability of the traits is high and not in dispute. The shared-environment finding rules out the simplest family-transmission story and leaves community-level normative transmission standing, which is the story Wax needs and the one she should have told.
Her framework, stripped of the origins question, makes a prediction about cohort heritability. The reasoning is ordinary behavior genetics. When a rule carries everyone to the same conduct regardless of temperament, temperament stops appearing in the outcome. A man with a high discount rate and a man with a low one both marry at twenty-three, both stay, and both have their children inside the marriage, because the alternatives carry costs neither will pay. The dispositional variance sits there with nowhere to express. Heritability of the outcome runs low, and the community rule the siblings grew up under does the work. Remove the rule and the same two men diverge, because the outcome now tracks the disposition and nothing else is steering. Heritability rises.
The prediction has a literature, and it runs older than Wax’s paper.
Hans-Peter Kohler, Joseph Rodgers, and Kaare Christensen asked in “Is Fertility Behavior in Our Genes? Findings from a Danish Twin Study” whether genetic influence on fertility shifts across demographic regimes, and found it high for Danish women born around 1880, low through the middle of the century, and high again after 1955. They followed it with “Between Nurture and Nature,” and by 2003 Kohler and Rodgers were stating the argument outright: reduced social constraint and greater egalitarianism raise the relevance of genetically mediated differences. Felix Tropf and colleagues found the same pattern in British female twins born between 1919 and 1968, naming war and economic crisis as constraints that suppress genetic expression and “normative freedom” as what releases it.
The American version exists too. Daniel Briley, Paige Harden, and Elliot Tucker-Drob ran it on 933 twin pairs from the Midlife in the United States study, framed by the second demographic transition, in “Genotype by Cohort Interaction on Completed Fertility and Age at First Birth.” Genetic influence on both outcomes was trivial for the 1920 to 1935 cohort and rose substantially for 1936 to 1955. That is the prediction, confirmed on American data, published five years after Wax wrote.
The third cohort is where the result stops being a confirmation and starts being an instrument. For twins born between 1956 and 1970, genetic influence on completed fertility persists, and genetic influence on age at first birth falls back, with shared environment returning. The model behind this is a diffusion curve rather than a release curve. Briley and colleagues cite Rogers on the diffusion of innovations in their first footnote, and the logic follows: when a new behavior becomes available, people differ in willingness to adopt it, those differences are partly heritable, and so genetic influence peaks during the transition and falls once the new behavior settles into a norm of its own. Old norm, low dispositional expression. Norm breaking, maximum differentiation. New norm, convergence.
A hump has consequences for anyone reading the number. Any given value of the additive component is consistent with two positions on the curve, early or late, and reading it requires knowing where the cohort sits in the transition. Deregulation does not raise heritability indefinitely. It raises it while the rule is contested and lowers it once a replacement rule is in place, which means a low estimate is evidence of a settled regime without telling you which regime.
Smoking supplies the same curve from a second direction. Jason Boardman, Casey Blalock, and Fred Pampel used the same MIDUS twins in “Trends in the Genetic Influences on Smoking,” estimating heritability of regular smoking year by year from 1922 to 1968. It runs high for those born in the early 1930s, collapses to zero for those born around 1942, climbs to a maximum for those born in 1954, and falls to zero again for those born in the mid-1960s. The 1942 minimum belongs to the cohort that was in its early twenties when the Surgeon General reported in 1964. The final decline belongs to the cohorts who came of age under clean-indoor-air laws. Boardman and colleagues distinguish three accounts of what moves the number, and only two of them are causal. Social control suppresses expression. Social trigger permits it. Social push, borrowed from Adrian Raine, changes how visible genetic differences are to the observer without changing how they operate, and they assign the entire 1935 to 1954 stretch to it. The estimate rose because social smokers quit after the Surgeon General and left the physiologically dependent behind, which is a change in who is being measured.
There is a further complication, and it is arithmetic. Heritability is a ratio. Constraint that closes the behavioral margin cuts the numerator. Constraint that homogenizes circumstances cuts the denominator, and the ratio rises. Gülüm Kosova, Mark Abney, and Carole Ober built a study on exactly that in “Heritability of Reproductive Fitness Traits in a Human Population,” choosing the Hutterites because communal life assures uniform and equal access to resources, and recovering significant heritabilities for reproductive traits in both sexes. A norm that dictates family size can suppress genetically influenced variation in desired family size. A norm that forbids contraception can expose genetically influenced variation in fecundity. Both are norms, and they move the estimate in opposite directions. Constraint compresses heritability only when it closes the margin through which the dispositional variance would otherwise reach the outcome.
None of which leaves the structural rivals where a first reading places them. It looks as though Wilson should predict stable heritability or rising shared environment, since wages and sex ratios fall on neighborhoods and families and siblings share both. That does not follow. A shock reaching both twins in a pair need not register as shared environment. Reliable contraception hands individuals control over reproduction, and if people differ heritably in discount rates, sexuality, conscientiousness, or desired family size, the technology raises the additive component for the same reason a moral collapse does. Udry’s model depends on that possibility. Kohler and Rodgers give female labor-force entry as another case. The decomposition registers that a constraint fell. It does not name the constraint.
Timing, geography, and race do separate the accounts. Akerlof, Yellen, and Katz put the break between 1965 and 1973 and predict a common break everywhere, because the technology arrived everywhere at once. Wilson puts it in deindustrialization, a decade later, concentrated in the labor markets that lost manufacturing and in the populations living in them. Those come apart on a date and a map. Geronimus comes apart the same way, since weathering ties early childbearing to local mortality and health schedules. Wax and Akerlof do not come apart at all here, since a moral collapse and a contraceptive revolution ran on the same clock and reached the same people.
One comparison does separate them. Populations that kept their own norm-producing institutions while the surrounding society shed its own hold the technology and the economy roughly constant and vary the norm. Boardman’s own sources show the pattern in another domain, with Judith Koopmans and colleagues and David Timberlake and colleagues finding genetic influence on tobacco muted or absent among people raised under strict religious prohibition. Any registry carrying religious affiliation can run the family-formation version, and that comparison, rather than the raw cohort trend, is what would discriminate a normative account from a technological one.
Two design constraints govern anyone who runs it. Family structure is not one phenotype. Matt McGue and David Lykken established genetic influence on divorce risk in 1992, Victor Jockin and colleagues traced part of it to heritable personality, and Beth Jerskey and colleagues then found in Vietnam-era twins that the genetic influences on marrying and the genetic influences on divorcing are largely distinct. Marriage timing, completed fertility, nonmarital birth, and dissolution need separate decompositions, and a result on one carries nothing to another. Hans Fredrik Sunde and Philipp Dierker have since argued from 353,210 Norwegian marriages that divorce is a couple-level outcome and that conventional twin models inflate what they attribute to individuals.
The shared environment component also behaves less obligingly than the framework suggests. A national norm experienced by every twin pair does not become shared environment because everyone shares it. If it holds constant within a cohort it contributes no variance at all. It shifts the mean, or suppresses the additive component, or interacts with it. Briley and colleagues report raw variance components rather than standardized proportions for this reason, and warn that the components need not sum to one.
Which leaves the gap that matters most. Every study named here runs on white, Nordic, or British samples. Briley is MIDUS. Boardman is MIDUS. Kohler is Danish, Tropf is British, Kosova is Hutterite, Jerskey is Vietnam-era veterans. The American argument among Wax, Wilson, Akerlof, Yellen, Katz, and Geronimus is centrally about a divergence between Black and white family structure, and a heritability estimate is a within-population number that says nothing about differences between groups in either direction. Minnesota, Virginia, Denmark, Sweden, and Australia cannot answer an American question about two populations. The sibling samples in the National Longitudinal Survey of Youth can, across two cohorts, which is the structure Roland Fryer used for the wage decomposition. The precedent for the population half exists: a 2009 study of cohort differences in educational attainment among African American twins read Jim Crow as an external ceiling suppressing individual differences and predicted that removing the ceiling would show up as declining similarity within pairs. Nobody has run the family-formation version.
The Minnesota, Virginia, Swedish, Danish, and Australian registries hold cohorts on both sides of the break with marital and fertility histories attached, and they will answer the European version of the question. The American version requires the sibling samples, split by race, on marriage timing and nonmarital first birth rather than on completed fertility.
Udry and Shanahan and Hofer built the theory. Kohler and Briley and Tropf calibrated it on fertility, Boardman calibrated it on smoking, Kosova established what the ratio does when the environment is homogenized, and four decades of results show what the instrument can and cannot see. It has never been pointed at the outcomes the American argument turns on, in the populations that argument is about. The experiment does not need to be invented. It needs to be imported, and the importing is the harder half.
Notes
On Wax’s own record. The paper under discussion is Amy L. Wax, “Diverging Family Structure and ‘Rational’ Behavior: The Decline in Marriage as a Disorder of Choice,” University of Pennsylvania Public Law and Legal Theory Research Paper No. 10-17, available at https://ssrn.com/abstract=1592424. The two footnote citations described above appear at notes 76 and 99 of that draft. The earlier statement of the demographic argument is “Engines of Inequality: Class, Race, and Family Structure,” 41 Family Law Quarterly 567 (2007). Race, Wrongs, and Remedies: Group Justice in the 21st Century (Hoover Institution Press, 2009) is the book where she walks through the exogenous explanations for the test score gap and declines the hereditarian step, at https://archive.org/details/racewrongsremedi0000waxa. The paper that shows the agnosticism doing positive work is “The Poverty of the Neuroscience of Poverty: Policy Payoff or False Promise?“, 57 Jurimetrics 239 (2017), at https://scholarship.law.upenn.edu/faculty_scholarship/1711/ and https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2888600. Her argument there turns on the point that imaging research cannot separate innate from environmental causes of brain characteristics, which she uses against interventionist optimism while asserting nothing about origins. “Educating the Disadvantaged,” National Affairs (Summer 2017), sits at https://www.nationalaffairs.com/publications/detail/educating-the-disadvantaged. Her faculty listing, which still carries rational choice and family structure among works in progress, is at https://www.law.upenn.edu/faculty/awax/.
On the within-population findings. Eric Turkheimer, “Three Laws of Behavior Genetics and What They Mean,” 9 Current Directions in Psychological Science 160 (2000), states the two propositions the essay leans on: traits are heritable, and the effect of being raised in the same family runs smaller than the effect of genes. Journal version at https://journals.sagepub.com/doi/10.1111/1467-8721.00084, author copy at https://uva.theopenscholar.com/files/eric-turkheimer/files/three_laws_of_behavior_8.pdf. Turkheimer’s own reading of the laws differs sharply from the use hereditarian writers make of them, and his Substack is the place to see him say so. Christopher Chabris and colleagues added a fourth law in 2015, at https://journals.sagepub.com/doi/full/10.1177/0963721415580430. For the heritability estimates and their rise with age, Robert Plomin and Ian Deary, “Genetics and Intelligence Differences: Five Special Findings,” Molecular Psychiatry (2014), at https://www.nature.com/articles/mp2014105, and Thomas Bouchard’s account of the age gradient at https://www.cambridge.org/core/journals/twin-research-and-human-genetics/article/wilson-effect-the-increase-in-heritability-of-iq-with-age/FF406CC4CF286D78AF72C9E7EF9B5E3F.
On the shared-environment problem and its resolution. Judith Rich Harris, “Where Is the Child’s Environment? A Group Socialization Theory of Development,” 102 Psychological Review 458 (1995), and The Nurture Assumption (Free Press, 1998), carry the argument that the socializing environment is the peer group. Harris is the least-cited figure in this dispute relative to her usefulness, largely because both camps read her as conceding to the other. On the interaction between socioeconomic position and heritability, see the Scarr-Rowe literature and the failure to replicate it outside the United States: Ken Hanscombe and colleagues, “Socioeconomic Status and Children’s Intelligence,” PLoS ONE (2012), at https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3270016/. The Dickens and Flynn model of environmental multipliers, which explains how high heritability coexists with large environmental movement in population means, sits here.
On why the cross-population question stays open. The logical point is Richard Lewontin’s, most accessibly in his 1970 reply to Arthur Jensen in the Bulletin of the Atomic Scientists. For the technical reason molecular methods cannot currently answer it, see Alicia Martin and colleagues, “Clinical Use of Current Polygenic Risk Scores May Exacerbate Health Disparities,” Nature Genetics (2019), at https://www.nature.com/articles/s41588-019-0379-x, with the open preprint at https://www.biorxiv.org/content/10.1101/441261v3, and the companion survey of the field’s ancestry composition at https://www.nature.com/articles/s41467-019-11112-0. More recent work shows prediction accuracy falling continuously along the ancestry continuum rather than in discrete population steps, at https://www.nature.com/articles/s41586-023-06079-4, which further undercuts any procedure that compares group mean scores. A recent review of the open problems in portability is at https://www.nature.com/articles/s41467-026-68565-3. For the skeptical case stated at full strength, Jonathan Kaplan and colleagues, “Genes, Heritability, ‘Race’, and Intelligence,” Genes (2022), at https://www.mdpi.com/2073-4425/13/2/346.
On the theory of normative constraint and genetic expression. Michael J. Shanahan and Scott M. Hofer, “Social Context in Gene-Environment Interactions: Retrospect and Prospect,” 60 Journals of Gerontology: Series B 65 (2005), at https://academic.oup.com/psychsocgerontology/article/60/Special_Issue_1/65/2965138, sets out the four ideal types, of which social control is the one Wax needs. J. Richard Udry, “Biosocial Models of Low-Fertility Societies,” 22 Population and Development Review 325 (1996), at https://doi.org/10.2307/2808017, states the demographic version and is the ancestor of everything in this literature that followed.
On the cohort-heritability test, which requires the most care. Hans-Peter Kohler, Joseph Lee Rodgers, and Kaare Christensen, “Is Fertility Behavior in Our Genes? Findings from a Danish Twin Study,” 25 Population and Development Review 253 (1999), and the same authors’ “Between Nurture and Nature: The Shifting Determinants of Female Fertility in Danish Twin Cohorts, 1870-1968,” 49 Social Biology 218 (2002), at https://pubmed.ncbi.nlm.nih.gov/14652919/, report that genetic influence on female fertility runs high in cohorts born around 1880, falls through the middle of the twentieth century, and rises again for cohorts born after 1955. They tie the movement to changing socioeconomic and normative conditions rather than to normative deregulation as such. Daniel Briley and colleagues, “Genotype by Cohort Interaction on Completed Fertility and Age at First Birth,” Behavior Genetics (2015), at https://link.springer.com/article/10.1007/s10519-014-9693-3, is the American version and the source of the diffusion model that turns the prediction into a hump. Felix Tropf and colleagues, “Genetic Influence on Age at First Birth of Female Twins Born in the UK, 1919-68,” 69 Population Studies 129 (2015), at https://doi.org/10.1080/00324728.2015.1056823, names normative freedom as the moderator. The Danish twin fertility design is at https://read.dukeupress.edu/demography/article/38/1/29/170416/Behavior-genetic-modeling-of-human-fertility, and a two-registry treatment of completed fertility at https://link.springer.com/article/10.1007/s10519-016-9805-3.
On the analogy and its warnings. Jason D. Boardman, Casey L. Blalock, and Fred C. Pampel, “Trends in the Genetic Influences on Smoking,” 51 Journal of Health and Social Behavior 108 (2010), at https://pmc.ncbi.nlm.nih.gov/articles/PMC3158572/, is free in full and is the methodological template for anything anyone runs on family structure. Their social push category, borrowed from Adrian Raine, is the one that should worry anyone reading a heritability trend as evidence of normative change, since it describes an estimate moving because the composition of the measured population moved. Their religious-upbringing sources are Judith Koopmans and colleagues at https://pubmed.ncbi.nlm.nih.gov/10857244/ and David Timberlake and colleagues at https://pubmed.ncbi.nlm.nih.gov/16497606/. For the arithmetic point about the ratio, Gülüm Kosova, Mark Abney, and Carole Ober, “Heritability of Reproductive Fitness Traits in a Human Population,” 107 PNAS Suppl. 1, 1772 (2010), at https://www.pnas.org/doi/10.1073/pnas.0906196106, which uses Hutterite uniformity as an instrument for detecting genetic variance rather than as a constraint suppressing it.
On family structure as several phenotypes. Matt McGue and David T. Lykken, “Genetic Influence on Risk of Divorce,” 3 Psychological Science 368 (1992); Victor Jockin, Matt McGue, and David T. Lykken, “Personality and Divorce: A Genetic Analysis,” 71 Journal of Personality and Social Psychology 288 (1996); Beth A. Jerskey and colleagues, “Marriage and Divorce: A Genetic Perspective,” 49 Personality and Individual Differences 473 (2010), which finds the genetic influences on marrying and on divorcing largely distinct. Hans Fredrik Sunde and Philipp Dierker, “Are Spouses Similar in Divorce? Investigating Spousal Similarity in Couple-Shared Outcomes,” Twin Research and Human Genetics (2026), argues that couple-level outcomes and assortative mating inflate what conventional twin models attribute to individuals.
Anyone running the test might want to read Michael Rutter’s skeptical chapter in Offspring: Human Fertility Behavior in Biodemographic Perspective (National Academies Press, 2003), at https://www.ncbi.nlm.nih.gov/books/NBK97281/, which argues that secular trends in heritability estimates have proved inconsistent across traits and countries and that the interpretive temptation runs ahead of the evidence. His example is the rise in heritability of educational attainment in Norway for males and not females over a period when opportunity widened for both. That objection applies to the argument advanced here with full force. Note also that the existing work covers fertility outcomes rather than marriage timing, marital dissolution, and childbearing outside marriage, and covers Denmark, Britain, and a largely white American sample rather than the two populations whose divergence the American argument is about. For the precedent on the population half, see “Education in Time: Cohort Differences in Educational Attainment in African-American Twins” (2009), at https://pmc.ncbi.nlm.nih.gov/articles/PMC2765648/, which runs the same design on Jim Crow.
Further reading. On the older quantitative-genetic critique, David Layzer, “Heritability Analyses of IQ Scores: Science or Numerology?“, Science (1974), at https://www.science.org/doi/10.1126/science.183.4131.1259, states the case against inference from heritability estimates in a form that most later criticism restates. On the demographic side of the family argument, Andrew Cherlin, The Marriage-Go-Round (2009), Charles Murray, Coming Apart (2012), Robert Putnam, Our Kids (2015), and Melissa Kearney, The Two-Parent Privilege (2023), carry the descriptive picture forward from where Wax leaves it. Orlando Patterson’s writing on culture and agency, and William Julius Wilson’s account of the professional silence that followed the Moynihan report in the opening chapter of The Truly Disadvantaged (1987), supply the disciplinary context for the omission this essay examines.