4.1 A Controversial Implication of Turkheimer’s Nonshared Environment Equals Free Will Thesis

In the second paragraph of ‘Genetics and Human Agency’, Turkheimer poses the question ‘What is the right way to think about genetic influences on sexual orientation?’ (Turkheimer, Reference Turkheimer2011, p. 825). Two pages later he answers that question: ‘The nonshared environment captures what is at stake when we are concerned about whether people are able to control their own weight or choose their own sexual orientation or whether they should be held responsible for their criminal behavior’ (Turkheimer, Reference Turkheimer2011, p. 827 [emphasis added]). Note first the framing of sexual preference as a ‘choice’ in this sentence. Note also the juxtaposition of that ‘choice’ with criminal responsibility. Now let us revisit Turkheimer’s table of nonshared environmental influences, which he suggests are ‘ordered in terms of their controllability and moral relevance: weight more than height, personality more than intelligence, and depression more than schizophrenia’ (Turkheimer, Reference Turkheimer2011, p. 827 [emphasis added]).

In Table 1 (see Part I) we have reproduced Turkheimer’s table of traits in exactly the same order that he listed them. Specifically, note that the table does not order the traits from lowest to highest nonshared environmental variance, but instead arranges them into the pairs that Turkheimer contrasts in the sentence quoted above: weight versus height, personality versus intelligence, depression versus schizophrenia. We have added gridlines to emphasize these pairwise groupings. Only one of these trait comparisons in Table 1 is left conspicuously unstated: sexual preference versus criminality. The clear but unspoken conclusion that Turkheimer invites his readers to draw is that sexual orientation (1-rMZ = 0.5) displays more ‘controllability and moral relevance’ than criminality (1-rMZ = 0.4); that is, that people have more choice over their sexuality than whether or not they commit a crime. This is a remarkable thing to suggest.

Of course, this does not imply that Turkheimer takes personal issue with people’s sexual orientation one way or another — indeed his statement that ‘the term “genetics of sexual orientation” is to be preferred over the term “genetics of homosexuality’’’ (Turkheimer, Reference Turkheimer2011, p. 825) in order to avoid pathologizing homosexuality is clear evidence that he does not.

Nevertheless, it should be evident that Turkheimer’s interpretation of the nonshared environment as ‘the free-will coefficient’ (Harden, Reference Harden2021a, p. 12.5), his inferences about what this means for the ‘controllability and moral relevance’ of sexual preference, and his ill-considered juxtaposition of sexual preference with criminal behavior risk buttressing homophobic viewpoints that see homosexuality as a morally charged lifestyle choice.

The incendiary implication that we have more control over our sexual orientation than whether we commit a crime should not prevent us from dispassionately assessing whether Turkheimer’s theory that the nonshared environment is a measure of human agency is valid (though we will presently argue that this theory is not valid). It does, however, show that Turkheimer’s theory of human agency poses downstream political risks remarkably similar to those that are sometimes attributed to ‘hereditarian’ interpretations of behavioral genetic research; namely, of potentially providing ideological ammunition to hate groups or helping to sustain reactionary political ideologies. This example highlights the dangers of using indirect risks as a principle to determine which ideas are discussed and which findings are published. Indirect risks are not always immediately obvious, and the flight from facts or ideas that are perceived to be dangerous can sometimes drive us towards greater dangers.

4.2 Sexual Orientation Is Not a Choice

An important reason it troubles us to hear sexual orientation characterized as more controllable than criminal behavior is that it suggests we can change our sexual orientation at will. But it should be self-evident that this is not the case. Few of us could simply decide to be sexually attracted to a different sex (or sexes) from the one (or two) that we are currently attracted to, any more than we can decide which individuals we are or are not attracted to. We can choose not to act on our desires, and we can even choose to act against our desires, but we cannot normally choose what our desires and preferences are.

One of the things that made the heritability of sexual preference so appealing to gay rights campaigners was that it seemed to imply that same-sex-attracted individuals were ‘born this way’ (Bean, Reference Bean2010; Gaga, Reference Gaga2011). However, this turns out to be a mischaracterization. Yes, there is a genetic component to sexual preference that makes it more likely that monozygotic twins share the same sexual preferences than nonidentical twins, but it is nevertheless quite common for identical twins to have different sexual orientations to each other. Genetics and shared family background both play a smaller role in who we are sexually attracted to than the general public has been led to believe.

Nevertheless, the peculiar circumstances that lead us to be attracted to men versus women, blondes versus brunettes, or whatever other idiosyncratic sexual preferences we find ourselves having, are hardly under our control. When one member of an identical twin pair turns out to be gay and the other turns out to be straight, it is not because they made different choices about who they find attractive. Our sexual orientation, whatever it may be, is not a choice but a brute fact about who we are, shaped by our genetic history, our family history, and our unique personal history. Likewise, just because personality shows nonshared environmental influence of 60%, this does not mean we chose to be outgoing, people-pleasing, neurotic, or conformist, or that we can choose to have a different personality tomorrow to the one we have today.

As Plomin writes in Blueprint, ‘A … crucial discovery … is the unexpected way in which the environment makes us who we are … the environment makes siblings reared in the same family as different as siblings reared in separate families. What makes us different environmentally are random experiences, not systematic forces like families’ (Plomin, Reference Plomin2018, pp. xiii–ix). The nonshared environment often defines who we are as much as our genetic endowment does, substantially explaining our differences from other people. But these random experiences still cause our subsequent behavior as surely as our genetic influences or shared environmental influences do.

4.3 The Nonshared Environment Is Not Experienced as Free Will

This gets us to a fundamental misconception at the heart of Turkheimer’s theory that the nonshared environment measures human agency. Turkheimer writes that, ‘We have a special concern for the genetics of behavior because behavior genetics is experienced genetics’ (Turkheimer, Reference Turkheimer2011, p. 827), and for him, the nonshared environment represents the most important domain of human experience because ‘From the point of view of the individual, this represents the degree to which [e.g.,] intelligence is under that individual’s control, as opposed to the degree to which it is limited by either genes or upbringing’ (Turkheimer, Reference Turkheimer2011, p. 827). This assertion is the very core of Turkheimer’s argument, and it is false. The nonshared environment is not experienced as the domain of personal control and free will.

If one monozygotic twin suffers a lifelong cognitive impairment because he was struck by lightning, this is an environmental influence he is unlikely to share with his co-twin, but this event can hardly be said to be under his control. Likewise, if a monozygotic twin sustains a life-altering brain injury because a drunk driver collided with her bicycle, her reduced IQ score relative to her co-twin will hardly be a matter of choice. Perversely, Turkheimer’s theory of free will seems to hold the less fortunate twin in these examples responsible for the accident that reduced their cognitive abilities. Even more perversely, it holds the drunk driver only semi-responsible for his crime, because the nonshared environmental estimate for criminal behavior is 0.5.

It might be countered that these kinds of events constitute ‘exogenous within-family environmental differences’ (Turkheimer, Reference Turkheimer2011, p. 826) of the kind that Turkheimer lists as further discounts to our agency after accounting for heritable and shared environmental variation, but Turkheimer seems to insist that it is only measured environmental differences that should be discounted in this way. If being struck by lightning or being hit by a drunk driver is not on the checklist of variables that researchers tend to collect as potential predictors of twin differences in adult intelligence, it seems unjust that twins should be held responsible for these differences. Once we recognize that the nonshared environment largely consists of unmeasured environmental differences that twins cannot be held responsible for — illnesses, accidents, and countless other serendipitous advantages and disadvantages that are outside of people’s control — the nonshared environment is entirely invalidated as a meaningful index of human agency.Footnote ²¹

These examples also throw up another serious problem with Turkheimer’s theory that ‘the nonshared environment … is free will’ (Turkheimer, Reference Turkheimer2011, p. 826), namely, that this population variance component tells us nothing about the level of agency any particular individual has over a given trait. The fact that the nonshared environment accounts for ∼20% of the variation in adult intelligence in developed Western countries (see Table 1) reveals nothing about the extent to which genetic, shared environmental, or nonshared environmental influences explain why a particular individual scored well or badly on an IQ test relative to the rest of the population.Footnote ²² In the examples just given, a nonshared environmental insult explains virtually all of the IQ differences between the identical twins and also explains why one twin from each pair performs substantially worse on adult IQ tests than members of the general population. In summary, not only is the magnitude of the nonshared environmental variance uninformative about the extent to which people are responsible for their trait differences across a population, it is also uninformative about the extent to which any individual is responsible for a given behavior — and it is nearly always the extent to which individuals are responsible for their behavior that is the question that concerns us in discussions of personal responsibility.

4.4 ‘Sliding Doors’ as a Parable About Nonshared Environmental Luck

The fact that the attributes of one monozygotic twin cannot be perfectly predicted from the attributes of their co-twin does not necessarily mean the differences between them are under their control, or that the nonshared environmental variance component is a measure of human agency. We can vividly illustrate how the nonshared environment is not a domain of personal choice by enlisting the 1990s Gwyneth Paltrow movie Sliding Doors (Howitt, Reference Howitt1998) which Harden frequently enlists as a teaching device in her own writings and talks. The conceit of Sliding Doors is that the life of Paltrow’s character, Helen, is set on two different courses based on the accident of whether or not she manages to board a particular train. In the reality in which she catches her train, she arrives home in time to find her long-term boyfriend in bed with another woman. Conversely, in an alternate reality where she narrowly misses her train, she never discovers that her boyfriend is having an affair. This discovery or nondiscovery sets her two lives on very different trajectories, one happy, the other tragic.

In her opening remarks for an Intelligence Squared debate in 2019 (Open to Debate, 2019), Harden used Sliding Doors to illustrate that our lives might have turned out very differently if we had have been born into a different family in a different society, saying, ‘What if you had the same genes but you were raised … by a Hazda family in Tanzania?’Footnote ²³ This anecdote enlists Sliding Doors as a parable about the luck of the shared environment, the between-family differences that have preoccupied psychologists and other behavioral scientists since Freud. But fundamentally Sliding Doors is a parable about the luck of the nonshared environment, the serendipitous events or unhappy accidents that set our lives on different courses.

Recall that, in twin studies, the nonshared environment is estimated as the fraction of the variation in a particular trait that is not shared by identical twins (1-rMZ). It is an omnibus measure of all the things that lead identical twins to differ from each other after accounting for their identical genes and the environmental experiences they hold in common. In this respect, the two Helens in Sliding Doors are effectively super-identical twins who not only share the same genes, and the same family environments, but even share the same nonshared environmental histories up until a moment in Helen’s mid-20s when her life forks along two different paths in two alternative realities. Moreover, the environmental difference that sets Helen’s life on two different paths is a nonshared environmental experience par excellence. In one reality a little girl blocks Helen’s path down the train station’s stairway. In the other reality the little girl’s mother pulls her out of Helen’s way (see Figure 3). Helen had no control or responsibility over the decisive environmental factor that sets her life on two different courses.

Figure 3. A trivial environmental difference sets Helen’s life on two separate paths in the film Sliding Doors (Howitt, 1998).

In a 1987 paper that first drew wide attention to the major role of the nonshared environment across human behavioral traits, Plomin and Daniels (Reference Plomin and Daniels1987) floated the possibility that the nonshared environment might largely consist of ‘unsystematic, idiosyncratic, or serendipitous events’ like these,Footnote ²⁴ and give the example of a Sliding Doors moment that changed the course of Charles Darwin’s life and thereby altered the course of science and history. Darwin writes: ‘The voyage of the Beagle has been by far the most important event in my life, and has determined my whole career; yet it depended on so small a circumstance as my uncle offering to drive me thirty miles to Shrewsbury’ (Plomin & Daniels, Reference Plomin and Daniels1987, p. 8). However, Plomin and Daniels do not make the mistake of assuming that Darwin was somehow responsible for the ‘serendipitous events’ that secured him a place on the Beagle. They recognize them as matters of pure luck.

As with Darwin, the two Helens are no more responsible for the random incident that set their lives on two different paths than identical twins are for the stochastic developmental variation that makes them different from each other. Harden rightly recoils from the proposition that we might be held responsible for random developmental differences in our brain wiring, which is why she feels compelled to subtract that unknown quantity from the nonshared environmental variation that quantifies human agency (albeit only in a footnote). But she fails to grasp the nettle and take the important final step, which is to recognize that there is no variation in behavior that luck does not explain at some level. All human behavior is ultimately due to antecedent causes of behavior that are outside of the individual’s control. There is no gap in Harden’s (and Turkheimer’s) God-of-the-gaps theory of free will for human agency to get a foothold.Footnote ²⁵ A solid foundation for human agency must instead be sought elsewhere.

4.5 The Nonshared Environment as the Archetypal Domain of Luck

The throughline of Harden’s book, like Turkheimer’s (Reference Turkheimer2011) article, is that ‘Your genotype, like the social class of your family, is an accident of birth over which you had no control’ and ‘a source of luck in your life’ (Harden, Reference Harden2021b, p. 204). On this basis, Harden argues that one’s genetic endowment and one’s family background are both fortuitous sources of lifelong (dis)advantage for which one can take no blame or credit. But the glaring omission in both Turkheimer and Harden’s account is the more obvious role that luck plays in the nonshared environment. Indeed, nonshared environmental luck epitomizes what we ordinarily think of as ‘luck’.

Luck, as traditionally defined, concerns the ordinary fortunes and misfortunes that befall us over the full course of our lives. It does not normally concern the circumstances of our birth. We ordinarily invoke luck when random events work unexpectedly in our favor (or work against us) as compared with how we expected events to unfold based on our efforts and abilities. This has been the traditional conception of ‘Lady Luck’ since ancient times, personified in the Roman goddess Fortuna and the Greek goddess Tyche, who represents the capricious role of Chance in our lives. She is the fickle goddess who can propel a mediocre athlete to victory by making stronger competitors stumble or who can turn a battle in a smaller army’s favor through a serendipitous weather event. These lucky and unlucky breaks don’t account for all of the nonshared environmental variance. For instance, one identical twin in each pair can be reliably faster than the other because they train harder or have more fast twitch muscle fibres. But unlike differences in abilities, which produce predictable and reliable differences in people’s performance, luck in this conventional sense, is unpredictable and unreliable by definition. Buying a winning lottery ticket is the textbook example of this kind of luck, where overnight a person’s financial prospects are improved beyond what a lifetime of ordinary strivings could accomplish.

The Genetic Lottery invites us to take this quintessential example of nonshared environmental luck and extend it to the genetic and the shared environmental domains — and, of course, there is a certain sense in which our genetic endowments and the type of family we were born into are also matters of what we might call ‘cosmic’ luck.Footnote ²⁶ But Harden has neglected the role of luck in the nonshared environment, which is its archetypal domain. If we follow Harden’s exculpatory logic that we are not responsible for any behavior that can ultimately be traced to fortuitous antecedent causes, then we are forced to accept that it’s luck all the way down — to accept, as the philosopher Galen Strawson famously put it, that ‘luck swallows everything’ (Strawson, Reference Strawson2016).Footnote ²⁷ This is precisely the conclusion reached by free will skeptics like the philosopher Greg Caruso who, in the passage below, describes three kinds of luck that we encounter in our lives:

These three categories of luck essentially describe shared environmental luck, genetic luck, and nonshared environmental luck respectively, and for Caruso these three categories are exhaustive. Under this view, there is no corner of human behavior that luck does not explain.

5.1 The Three Major Positions on Free Will in Academic Philosophy

In academic philosophy, there are three main positions staked out in the free will debate: compatibilism, free will skepticism, and libertarianism:

Compatibilists (sometimes called ‘soft determinists’) maintain that human agency, personal responsibility, and moral desert (that is, praiseworthiness or blameworthiness for our actions) are fully compatible with a deterministic universe in which causation reigns and in which history was always going to unfold as it did. They argue that the key question for deciding whether an agent is responsible for his actions is not whether his behavior was determined and/or has antecedent causes, but rather whether he could have acted differently if he had wanted to do so badly enough (M. S. Moore, Reference Moore2016).Footnote ²⁸ The philosopher Daniel Dennet is one of the best-known contemporary evangelists for this position (Dennett, Reference Dennett1984, Reference Dennett2004), but notable historical compatibilists include Thomas Hobbes and David Hume (Hobbes, Reference Hobbes1656; Hume, Reference Hume1739).

Free will skeptics (sometimes called ‘hard determinists’) maintain that we cannot be held responsible for any of our actions because the ultimate causes of those actions are outside of our control. Some of the best known free will skeptics in contemporary academic philosophy are Galen Strawson and Derk Pereboom (Pereboom, Reference Pereboom2001; Strawson, Reference Strawson2016). The neuroscientists Sam Harris and Robert Sapolsky are both prominent advocates for this position (Harris, Reference Harris2012; Sapolsky, Reference Sapolsky2023).

Libertarians maintain that human behavior is not entirely determined — either because indeterministic properties of the universe (e.g., at the quantum level) make human decisions inherently unpredictable, or because human beings are endowed with metaphysical powers that free them from the laws of causation that govern the rest of the natural world, allowing them to become uncaused causes of subsequent events. A dualistic, supernatural model of agent causation is the implicit position on free will maintained in many religious traditions, but ‘folk’ conceptions of free will often uncritically inherit the libertarian conception of human beings as self-causing agents even in secular communities (Sarkissian et al., Reference Sarkissian, Chatterjee, De Brigard, Knobe, Nichols and Sirker2010). Prominent agent-causation libertarians include Roderick Chisholm, Richard Taylor, and Timothy O’Connor (Chisholm, Reference Chisholm1964; Taylor, Reference Taylor1966; O’Connor, Reference O’Connor2000). The neurogeneticist Kevin Mitchell supports a kind of quantum libertarianism (Mitchell, Reference Mitchell2020a, Reference Mitchell2020b).

Libertarianism and free will skepticism are both classified as ‘incompatibilist’ positions because they hold that human agency and moral responsibility are incompatible with conditions under which our behavior is determined by antecedent causes. The main difference between these two positions is that free will skeptics hold that all of our behavior is ultimately dictated by antecedent causes, while libertarians claim that human behavior is incompletely determined by preceding events and that this affords us wiggle room to shape our own behavior to some extent.

It should be noted, however, that neither compatibilism nor free will skepticism hinge on determinism being strictly true. Proponents of both positions acknowledge that there are properties of the universe that appear to be indeterministic at the quantum scale and are prepared to accept the possibility that these indeterministic phenomena could have repercussions at the macroscopic scale in which humans operate. Neither position would be invalidated, then, if quantum indeterminacy somehow rendered human behavior inherently unpredictable to some extent and if more than one future were strictly possible at any given moment.

The key debate between compatibilists and free will skeptics is whether human beings can be held responsible when the ultimate causes of their actions are outside of their control. Both positions are indifferent to the question of whether those causes are deterministic or indeterministic. For this reason, some free will skeptics prefer to describe their position as ‘hard incompatibilism’ rather than ‘hard determinism’ (Pereboom, Reference Pereboom2001).Footnote ²⁹ Alternatively, they sometimes refer to their position as ‘hard luck’ (Levy, Reference Levy2011), highlighting the notion that we cannot be held responsible for our behavior when the particular set of causes that drove our behavior are ultimately matters of chance.Footnote ³⁰

5.2 Turkheimer Is a Libertarian, Not a Compatibilist

Confusingly, Turkheimer and Harden both present Turkheimer’s nonshared environmental theory of free will as a compatibilist theory when it is not. For example, in ‘Heritability and Biological Explanation’, Turkheimer writes: ‘Philosophers are less likely to agree than even social scientists, but a broad area of consensus has emerged about how these questions might be answered. In the free-will literature, this point of view is known as compatibilism’ (Turkheimer, Reference Turkheimer1998, p. 790). Turkheimer then represents his own position on free will as belonging to this philosophical consensus and repeatedly namedrops the compatibilist philosopher Daniel Dennett as though they were fellow travelers in the same philosophical tradition.

Harden follows suit, and uses the title of one of Dennett’s best known books to describe Turkheimer’s theory of free will, writing: ‘To borrow a phrase from the philosopher Daniel Dennett, [the nonshared environment] lets you know how much “elbow room” you have to choose who are you [sic] going to be’ (Harden, Reference Harden2021b, p. 202). But the compatibilist theory of free will outlined in Dennett’s (Reference Dennett1984) book Elbow Room and defended across his philosophical career diametrically opposes the position staked out by Turkheimer and Harden.Footnote ³¹ Neither does Turkheimer’s theory correspond to the position defended by any other notable compatibilist philosopher.

Far from being a compatibilist, Turkheimer is, in fact, a libertarian incompatibilist. He is an incompatibilist in the sense that he holds that antecedent genetic and environmental causes of behavior infringe upon human agency, and a libertarian in the sense that he treats antecedent causes of human behavior as nonexhaustive, leaving some uncaused fraction of behavioral variation that human agency explains.

The unusual sense in which Turkheimer describes himself as a ‘compatibilist’ is merely to signal that he is he is a secular materialist who supports the existence of free will in contrast to those who would invoke supernatural explanations for human agency. We see this, for example, when he writes, ‘The nonshared environment, in a phrase, is free will. Not the kind of metaphysical free will that no one believes in anymore, according to which human souls float free above the mechanistic constraints of the physical world, but an embodied free will, tethered to biology’ (Turkheimer, Reference Turkheimer2011, p. 826 [emphasis added]). However, despite his peremptory dismissal of metaphysical libertarianism in this passage, Turkheimer’s conception of free will bears several fundamental characteristics in common with it.

First, note that under many, if not most, libertarian conceptions of free will, human behavior remains partly subject to physical laws rather than being completely independent of them. Second, consider that under a metaphysical conception of free will in which human beings are partially self-causing agents, we would expect the miraculous, contra-causal workings of the unique soul of each identical twin to manifest precisely as nonshared environmental variance in behavior genetic analyses (i.e., as differences in behavior that could not be predicted from their shared genetic and environmental backgrounds). This would be a special subcategory of nonshared environmental variation that was different from the prosaic kind produced by the ‘exogenous within-family environmental differences’ (Turkheimer, Reference Turkheimer2011, p. 826) that behavioral scientists try to identify — just like the subcategory of nonshared environmental variance that Turkheimer says is a measure of free will. It would also be different from the nonshared environmental variance contributed by measurement error, or by stochastic developmental variation.

By suggesting that nonshared environmental variance is a measure of human agency, Turkheimer is effectively proposing that twin differences in behavior are self-caused rather than externally caused. He explicitly writes that the nonshared environment captures ‘the self-determinative ability of humans to chart a course for their own lives’ (Turkheimer, Reference Turkheimer2011, p. 826). Far from offering a sophisticated materialist conception of human agency that offers an alternative to supernatural conceptions of free will, Turkheimer’s theory simply rehearses the dualistic notion that there are special domains of human behavior which ‘float free above the mechanistic constraints of the physical world’ (Turkheimer, Reference Turkheimer2011, p. 826) without presenting a nonsupernatural explanation for how human beings can possibly accomplish that (see Figure 4).

Figure 4. Turkheimer suggests the nonshared environmental variation is an index of agent self-causation without explaining how non-miraculous self-causation can possibly occur. Cartoon by Sidney Harris (image reproduced with permission of ScienceCartoonsPlus.com©).

5.3 Indeterministic and Chaotic Causes of Behavior Cannot Support Free will

The fundamental challenge facing secularized libertarian perspectives on free will is that they beg the question of precisely how human beings are supposed to acquire the power of self-causation (i.e., of being an ‘uncaused cause’ or an ‘unmoved mover’). The possibility of quantum indeterminism at the human scale does not supply this positive account. It is unclear why people should be considered any more responsible for actions dictated by a cosmic coin toss than for actions dictated by the inexorable march of history.Footnote ³² These indeterministic causes would still present the same problem as deterministic causes, that is, of being antecedent sources of our behavior that are outside of our control.Footnote ³³ As Sam Harris writes:

Similar challenges confront attempts to base free will on ‘deterministic chaos’ (i.e., the unpredictability of complex systems which are highly sensitive to infinitesimal differences in starting conditions). Turkheimer repeatedly appeals to the ‘complexity’ and ‘interactivity’ of genetic and environmental causes as though these characteristics could somehow generate human agency, but a cause is a cause, irrespective of whether it is simple or complex. If human agency requires us to be the ultimate cause of our own actions, it is disqualified by antecedent causes of any kind, regardless of whether they are systematic or unsystematic, deterministic or indeterministic.Footnote ³⁴

Harden notes that Turkheimer takes ‘Unpredictability [as] a sign of freedom’ (Harden, Reference Harden2021b, p. 202), but while complex or indeterministic causes pose serious challenges for the scientific prediction of human behavior, they provide no firmer basis for personal responsibility than simple or deterministic causes do. Under the logic of incompatibilism, the only kind of unpredictability that could potentially signal human agency would be the unpredictability arising from self-causing agents whose actions were thereby genuinely independent of antecedent events. However, human behavior can be unpredictable for many, far less exotic reasons, meaning unpredictability cannot be safely interpreted as an index of human agency. Besides, without a positive, secular account of how agent self-causation might be possible, there is no compelling reason to believe that it actually exists (Pereboom, Reference Pereboom1995). Belief in self-causing agents seems to demand the same leap of faith as the religious belief in an immaterial soul.

5.4 Harden’s Libertarian Incompatibilism: Less Libertarian and More Incompatibilist

In The Genetic Lottery, Harden provides a sympathetic summary of Turkheimer’s nonshared environmental theory of free will that is even longer than the article in which Turkheimer originally presented it (Harden, Reference Harden2021b). She also dedicates several pages of her review in the Annual Review of Psychology to his theory (Harden, Reference Harden2021a). But despite giving Turkheimer’s theory substantial real estate in two publications intended to convey her overarching philosophy of behavioral genetics, Harden never gives an unequivocal endorsement of Turkheimer’s position, leaving some uncertainty about whether his and her views are entirely the same.Footnote ³⁵ We will show that both of their positions are philosophically very close, but with some important practical differences.

When we look at Harden’s explicit statements about the nature of free will, we find her initially reserving judgment about whether human behavior is fully determined by antecedent causes or whether we have a metaphysical capacity for self-causation. In The Genetic Lottery she writes: ‘Whether the universe is deterministic, whether such a thing as free will actually exists — these questions are beyond the scope of this book’ (Harden, Reference Harden2021b, p. 200).

Likewise, in her Annual Review article she writes:

Note that while Harden reserves judgment about whether human behavior is completely determined here, she has already tacitly accepted the incompatibilist premise that we are not free to the extent that we can ultimately trace the causes of our behavior to events outside of our control. She also appears to have accepted the incompatibilist premise that ‘real’ free will consists of the metaphysical ability to act differently from how antecedent conditions lead us to behave. Harden then proceeds to give several indications that she does not believe the world is completely deterministic. A first indication is in her Annual Review article, where she writes:

This passage implicitly rejects determinism. In a deterministic universe, if someone ‘rewound the tape’, your life (and everyone else’s) would unfold in exactly the same way as before. You would still have gotten divorced, made precisely the same amount of money, and been just as shy and disorganized. More generally, under determinism, every individual was always going to follow the path they did, because the history of the universe was always inevitably going to unfold the way it has from the moment of the big bang. It is only in an indeterministic universe that we could expect things to turn out any differently if we ‘rewound the tape’ and let it play out again — either because of random quantum-style indeterminacy or because human beings have special contra-causal powers of self-determination.

Another indication that Harden does not take the prospect of determinism seriously is the brevity with which she discusses hard determinism, dedicating just a short, two-sentence aside to this position. She writes, ‘If you think that the universe is deterministic, and the existence of free will is incompatible with a deterministic universe, and free will is an illusion, then genetics doesn’t have anything to add to the conversation. Genetics is just a tiny corner of the universe where we have worked out a little bit of the larger deterministic chain’ (Harden, Reference Harden2021b, p. 200).Footnote ³⁶ However, if hard determinism gets short shrift, compatibilism is overlooked completely, with Harden providing no discussion about how or why personal responsibility might be compatible with antecedent causation of behavior. Instead, we are treated to an extended disquisition on how genetic and environmental causes of behavior diminish people’s responsibility for any crimes they commit and reduces the credit they can take for their achievements in life.Footnote ³⁷

What we are left with is a de facto libertarian incompatibilism in which some unspecified sliver of human behavioral variation mysteriously survives the general annihilation of personal responsibility that antecedent causes otherwise produce. Harden’s libertarianism is manifest in the fact that she still clearly believes that some forms of responsibility exist. Despite dedicating an entire chapter of The Genetic Lottery to the topic of how genetic and environmental causes of our behavior diminish our responsibility, the book is filled with moral and political exhortations about what we ‘must’ or ‘should not’ do, and repeatedly discusses ‘our social responsibilities’ to reduce inequality.Footnote ³⁸ In addition, Harden’s belief in responsibility is displayed in her scathing moral condemnation of scientists like Francis Galton, Karl Pearson, Ronald Fisher, and Henry Goddard, at whose feet she lays the ultimate blame for Nazi genocide and for forced sterilization programs carried out elsewhere.Footnote ³⁹

Harden fails to provide a coherent positive account of how human agency can arise under conditions in which our behavior has antecedent causes, but like Turkheimer she sometimes appeals to deterministic chaos to explain the nonshared environmental variance. She references a scene from the film Jurassic Park in which Jeff Goldblum’s character, a mathematician specializing in chaos theory, illustrates the inherent unpredictability of complex systems by showing how two droplets of water fail to follow the same path down his interlocutor’s hand. ‘It changed. Why?’, he explains, ‘Because tiny variations — the orientation of the hair on your hands, the amount of blood distending your vessels, imperfections in the skin … — never repeat and vastly affect the outcome’ (Spielberg, Reference Spielberg1993). From this, Harden concludes: ‘[The nonshared environment] represents differences between people that are not due to differences … in their DNA or in the social circumstances into which they were born. [The nonshared environment] reflects the degree to which two identical drops of water, beginning in the same place, fall off in different directions’ (Harden, Reference Harden2021b, p. 201).

However, if infinitesimal differences in initial conditions are the ultimate wellspring of the twin differences captured by the nonshared environment, twins can hardly be held responsible for those differences in starting conditions. When two water droplets follow different paths down a surface, we do not say the droplets have agency. As discussed above, attempts to base human agency on deterministic chaos founder on the problem that these antecedent causes of human behavior are also outside of the individual’s control — just like people’s genetic endowment or the family they were born into.

Unlike Turkheimer, Harden does not suggest that people should be held responsible for actions caused by the kind of indeterministic chaos that makes identical twins different from each other, but this makes it unclear how personal responsibility is supposed to arise at all in Harden’s philosophical framework. Her remarks about the contribution of deterministic chaos to the nonshared environmental variation mirror her previously mentioned comments about the contribution of stochastic developmental variation. We are again left with the implication that the nonshared environment represents at best, ‘an upper bound of the extent to which people have agency’, after stochastic developmental variation and other sources of deterministic chaos have been removed (alongside measurement error, and ‘exogenous within-family environmental differences’). As with Turkheimer, we are left with the conception that some portion of human behavioral variation is effectively attributed to agent self-causation. However, once again no positive account is provided as to how this self-causation can possibly arise.

The main difference between Harden’s libertarian incompatibilism and Turkheimer’s is that Turkheimer reserves almost the entirety of the nonshared environmental variation as a refuge for free will, while Harden is prepared to concede that nearly all of it might be attributable to antecedent causes of our behavior that are outside of our control (whether we have identified the specific causes of that nonshared environmental variation or not). Harden is prepared to accommodate this anemic conception of human agency because it is easy to reconcile with the Rawlsian equity ethics that she champions in her book. Rawls likewise endorses a conception of free will and personal responsibility that is minimal at best, as we explore briefly in section 8 (see Part III). This makes Harden considerably more sanguine about genetic and shared environmental causation of human behavior than Turkheimer.

Turkheimer, by contrast, is keen to preserve a much larger role for human agency. However, as a libertarian incompatibilist, this goal obliges him to carve out a large fraction of human behavioral variation that is not explained by antecedent causes. In the next section, we will explore an alternative formulation of Turkheimer’s libertarian theory of free will that allows him to substantially expand the scope of human agency beyond the confines of the nonshared environmental variance to encompass parts of the heritable and shared environmental variation as well. Turkheimer adopts two questionable strategies to achieve this goal: first, he seeks to reduce all causes of human behavioral variation down to the causes that we know about; and second, he actively seeks to downplay our knowledge of those causes of variation.

6.1 The Moral Inversion of the ‘Gloomy Prospect’

Turkheimer’s investment of the nonshared environment with the special property of being able to measure human agency has its roots in a longstanding scientific debate about the nature of that variance component. The main contours of that debate were initially outlined by Plomin and Daniels (Reference Plomin and Daniels1987) who wrote: ‘What is happening environmentally to make children in the same family so different from one another? One gloomy prospect is that the salient environment might be unsystematic, idiosyncratic, or serendipitous events such as accidents, illnesses, and other traumas, as biographies often attest.’ (p. 8 [emphasis added]).

While they were the first to raise this ‘gloomy prospect’, Plomin and Daniels suggested that we should first investigate whether specific nonshared environments could be identified that accounted for this large source of individual differences in human behavior before throwing in the towel, writing: ‘[U]ntil more systematic research on nonshared environmental variance sources is conducted it is too early to conclude that the large environmental component of variance … is brought about by idiosyncratic experiences’ (Plomin & Daniels, Reference Plomin and Daniels1987, p. 6). However, after embarking on a multi-decade search for specific nonshared environments that explain this large fraction of behavioral variation, Plomin has become convinced of the reality of the gloomy prospect, writing: ‘After thirty years of searching unsuccessfully for systematic non-shared environmental influences, it’s time to accept the gloomy prospect. Non-shared environmental influences are unsystematic, idiosyncratic, serendipitous events without lasting effects’ (Plomin, Reference Plomin2018, p. 80).

By contrast, Turkheimer and his co-author Mary Waldron had already declared science’s defeat at the hands of the gloomy prospect in the year 2000, barely a decade after the search had begun. After conducting a review of existing studies that found that less than 3% of the variance in behavioral traits could typically be explained by measured nonshared environmental influences, Turkheimer and Waldron (Reference Turkheimer and Waldron2000) declared that the search should be called off. ‘The gloomy prospect is true’, Turkheimer (Reference Turkheimer2000) separately proclaimed (p. 163).

Turkheimer and Waldron partly present this verdict as a lesson in scientific humility, for example, writing ‘The limitations of our existing social scientific methodologies ought not provoke us to wish that human behavior were simpler than we know it to be’ (Turkheimer & Waldron, Reference Turkheimer and Waldron2000, p. 93). But crucially, they also went much further than this, actively celebrating the fact science may have hit an epistemological barrier. They wrote: ‘a world in which human behavior could be understood all the way down in terms of correlations between difference scores … would present its own gloomy prospects in the ethical evaluation of human agency’ (Turkheimer & Waldron, Reference Turkheimer and Waldron2000, p. 93 [emphasis added]).

Turkheimer (Reference Turkheimer2000) hit the same celebratory note, writing, ‘Finally, it must be remembered that the gloomy prospect is gloomy only from the point of view of the working social scientist. … In the long run, the gloomy prospect always wins, and no one would want to live in a world where it did not. … The gloomy prospect isn’t [gloomy].’ (Turkheimer, Reference Turkheimer2000, p. 164 [emphasis added]). Here we see a complete inversion of Plomin and Daniels’ gloomy prospect, in which scientific pessimism about the prospects of explaining human behavior becomes a source of moral and metaphysical optimism.

Note that the sentiment expressed here is no longer simply concern about scientific hubris and overreach. It is a particularly candid expression of what Daniel Dennett has called the fear of creeping exculpation: ‘Is science going to show us that nobody ever deserves punishment? Or praise, for that matter?’ (Dennett, Reference Dennett2004, p. 21). It is a libertarian incompatibilist concern that human beings will lack agency to the extent that their behavior is shown to have antecedent causes. Readers are invited to hope that some fraction of the nonshared environmental variation will never be explained by science, as full explanation would mean the extinction of human agency. It therefore becomes ‘methodologically vexing [but] humanistically pleasing’ (Turkheimer, Reference Turkheimer2000, p. 161) when science fails to isolate the specific causes of differences in human behavior.

In Turkheimer and Waldron (Reference Turkheimer and Waldron2000), then, we already see the skeleton of the nonshared environmental theory of free will that Turkheimer would flesh out 11 years later in ‘Genetics and Human Agency’ (Turkheimer, Reference Turkheimer2011). Moreover, in his later writings, Turkheimer would make the relationship between his inverted gloomy prospect and his theory of free will even clearer, writing: ‘The gloomy prospect … is gloomy for working scientists, but for individual people it is freedom; it is human agency’ (Turkheimer, Reference Turkheimer2019a, p. 125 [emphasis added]).

6.2 Turkheimer Extends the ‘Gloomy Prospect’ to Heritability

The foundations for Turkheimer’s nonshared environmental theory of free will may have been laid in Turkheimer and Waldron (Reference Turkheimer and Waldron2000), but in ‘The three laws of behavior genetics and what they mean’, Turkheimer (Reference Turkheimer2000) extended that theory in an important new direction. After reiterating Turkheimer and Waldron’s conclusion that the gloomy prospect prevails for the nonshared environment, and embellishing their fledgling idea that the gloomy prospect might represent a refuge for human agency, Turkheimer (Reference Turkheimer2000) generalized the gloomy prospect beyond the nonshared environment to other variance components, with a particular emphasis on the heritable variation. Turkheimer writes:

While Turkheimer (Reference Turkheimer2011) would later argue that any variance in the nonshared environment left unexplained by specific environmental measures quantified human agency, the extension of the gloomy prospect in Turkheimer (Reference Turkheimer2000) had already laid the grounds for a more radical position: namely, that the variance unexplained by specific variables under any variance component quantifies human freedom.

Both positions are libertarian, free-will-by-subtraction theories of human agency in which the total causes of human behavioral variation are treated as being limited to and co-extensive with the known causes of behavioral variation. However, the generalized genetic-plus-environmental gloomy prospect in Turkheimer (Reference Turkheimer2000) accommodates an even larger potential role for agent self-causation, mainly because any heritable variation which is not explained by specific genetic variants is also deemed a potential refuge for human agency. While the genetic dimension of this extended gloomy prospect was neglected in Turkheimer’s theory of free will as it was presented in Turkheimer (Reference Turkheimer2011), championing the genetic gloomy prospect has been a persistent focus of Turkheimer’s public commentary as the molecular behavioral genetics revolution has gained momentum.

6.3 Turkheimer’s Epistemic Libertarianism: Defining Ignorance as Freedom

It is worth taking a moment to stress what an extraordinary position Turkheimer has staked out by making himself the flagbearer of the gloomy prospect. The gloomy prospect is not merely the claim that science has already hit (or will someday hit) a limit regarding how much behavioral variation it can explain. It is an active celebration of the prospect that science will hit that limit — and the sooner the better! It is nothing less than the hope that behavioral science will stop advancing. This is a remarkably antiscientific position for a practicing behavioral scientist to take.

Unfortunately, Turkheimer’s theory of human agency occasionally lurches in an even more antiscientific direction. Instead of implicitly defending metaphysical libertarianism as discussed above, he sometimes defends a position that we might call epistemic libertarianism, that is, rather than arguing that the nonexhaustive scientific explanation of human behavior leaves room for the possibility of self-causation, Turkheimer sometimes defends the eccentric position that it is ignorance of the causes of human behavior that generates human agency.Footnote ⁴²

We see Turkheimer’s epistemic libertarianism in action when he argues that unexplained behavioral variation captures human agency (‘the gloomy prospect … is freedom’, Turkheimer, Reference Turkheimer2019a, p. 125) while simultaneously acknowledging that unexplained variation is attributable to unknown causes that might one day yield to scientific investigation. This implies that it is our knowledge of the causes of human behavior that infringes on our freedom rather than causation as such.

Turkheimer and Waldron (Reference Turkheimer and Waldron2000) acknowledge that the unexplained variation consists of unidentified causes when they write, ‘something about the environment must be causing differences among genetically related siblings reared together’ (Turkheimer & Waldron, Reference Turkheimer and Waldron2000, p. 90 [emphasis added]).Footnote ⁴³ They likewise acknowledge that those unidentified causes of behavior might eventually yield to scientific explanation when they write:

Turkheimer (Reference Turkheimer2000) also acknowledges that the nonshared environment consists of antecedent causes of human behavior but proposes that these causes are currently too complex for scientists to understand. He writes, ‘Nonshared environmental variability predominates … because of the unsystematic effects of all environmental events, compounded by the equally unsystematic processes that expose us to environmental events in the first place’ (Turkheimer, Reference Turkheimer2000, p. 163). He then proceeds to say that ‘Psychology is at least one good paradigm shift away from an empirical answer to the gloomy prospect’, conceding that much of the variation in human behavior that remains unexplained by specific genetic and environmental measures might someday yield to scientific investigation.

However, perhaps the starkest example of Turkheimer defending the idea that ignorance is itself freedom is in a Twitter thread he published in September 2022 where he wrote:

Here Turkheimer openly acknowledges that actual genetic constraints on our behavior exist but then suggests that whether we have an identical twin or not affects how free (vs. determined) our behavior is simply by conferring knowledge about how much of our behavior is predictable from our co-twin’s behavior.

To consider what an extraordinary statement this is, consider the documentary film Three Identical Strangers about three monozygotic triplets reared apart who discover at age 19 that they have genetically identical siblings (Wardle, Reference Wardle2018). The lives of these three men were not suddenly more determined after they discovered they were triplets. The discovery simply revealed how determined they already were by their genetic influences. This error in Turkheimer’s logic was pointed out almost immediately by a respondent on Twitter who replied: ‘Comprehension isn’t a prerequisite for determinism. A deterministic system is still deterministic even in the absence of observers with knowledge of its workings’ (KuremaSan, 2022). Turkheimer never replied to this astute remark.Footnote ⁴⁴

As the philosopher and legal scholar Michael Moore has noted, epistemic libertarianism ‘is a non-starter on its face’, because ‘[t]he degree of our knowledge of the causes of someone else’s choice can obviously have no bearing on the degree of their responsibility’ (M. S. Moore, Reference Moore2016, p. 59). It entails the ‘crude mistake’ of ‘slip[ping] from one’s own partial ignorance of causation to … partial absence of causation’ (p. 59 [emphasis added]).Footnote ⁴⁵ Moore concludes that ‘to [clearly] state this view is pretty much to refute it’ (p. 59).Footnote ⁴⁶

How is it, then, that Turkheimer sometimes finds himself advancing such an indefensible position? We submit that he is backed into this awkward intellectual corner because he is caught between his libertarian conception of free will and his professed secular materialism. As a secular materialist he cannot concede the existence of miraculously uncaused human behavior, yet his libertarian conception of human agency simultaneously demands it. Epistemic libertarianism offers a convenient way to hold these two contradictory ideas at the same time. But ignorance provides no escape hatch from the iron logic of incompatibilism. If one accepts the notion that human agency is incompatible with antecedent causation and also accepts that all human behavior has antecedent causes, one is obliged to accept hard incompatibilism and the nonexistence of free will (see Figure 5). Appeals to the irreducible complexity or scientific intractability of those causes cannot avert this conclusion.

Figure 5. The free will trilemma. The three main positions adopted in the academic free will debate represent three possible solutions to a trilemma in which only two of the three circled propositions can be logically subscribed to simultaneously: Libertarianism (A+B), Compatibilism (A+C), or Free will skepticism (B+C). Turkheimer champions a libertarian theory of free will which emphasizes the existence of personal responsibility (A), and the incompatibility of antecedent genetic and environmental causes of behavioral variation with personal responsibility (B); however, Turkheimer’s secular materialism compels him to simultaneously support the proposition that all human behavior has antecedent causes (C). But proposition C cannot be logically supported without rejecting proposition A or B. Turkheimer appeals to the scientific intractability of complex causes of behavior as a refuge for personal responsibility, but it is left unclear why personal responsibility would only be infringed by the causes known to science. Note that Turkheimer sometimes claims to be a compatibilist on the basis that he nominally supports propositions A and C, but this is untenable without rejecting proposition B.

Turkheimer attempts to paper over this problem by labelling himself a ‘compatibilist’, but compatibilism is not the proposition that human agency resides in the causes of human behavioral variation that remain unexplained. It is the position that free will and personal responsibility are compatible with the comprehensive causation of human behavior (encompassing known and unknown causes of human behavioral differences). By insisting we can ‘quantif[y] human agency … as the variability that remains after genetic and environmental … constraints have been taken into account’ (Turkheimer, Reference Turkheimer2000, p. 827), Turkheimer has forfeited the argument that human agency is compatible with causation. There is no coherent basis for arguing that people are more responsible when the causes of their behavior are unknown rather than known. This is selective incompatibilism, not compatibilism. Mislabeling epistemic libertarianism as a ‘compatibilist’ position only serves to cloak a philosophically indefensible position in the authority of a respectable and widely held position in academic philosophy.

An unfortunate upshot of the eccentric position that knowledge itself infringes on human agency is that it implicitly creates a moral imperative to stymie and even reverse the advances of behavioral science in order to protect and expand the scope of human agency. Thankfully, Turkheimer has never followed this strand of his thinking through to its logical conclusion and embarked on an explicitly antiscientific campaign. He only enlists epistemic libertarianism to buttress his de facto metaphysical libertarianism, which it achieves by sustaining an ambiguity over whether unexplained human behavioral variation has antecedent causes or is self-caused.

As noted earlier, however, this de facto metaphysical libertarianism is sufficient in itself to motivate profound moral resistance to advances in the behavioral sciences, without tipping into the outright antiscientific stance that epistemic libertarianism would imply if that position were embraced wholeheartedly. We opened Part I of this series with a quote from Daniel Dennett that nicely captures the anxieties metaphysical libertarianism can produce:

When ‘causal explanation’ is seen to ‘nibble away at the soul’ and narrow the scope of human agency, we can expect both genetic and environmental explanations for human behavioral differences to be met with considerable prejudice. Pessimism about scientific progress in the behavioral sciences becomes metaphysical optimism (‘the gloomy prospect isn’t [gloomy]’, Turkheimer, Reference Turkheimer2000, p. 164) and each failure of science to advance becomes something to be celebrated as a triumph for human agency and the human spirit (‘the gloomy prospect … is freedom’, Turkheimer, Reference Turkheimer2019a, p. 125).

7.1 Moving Goalposts on Genetic Associations and Genetic Prediction

Two of Turkheimer’s signature scientific predictions were that the intractably complex causes of human behavior would mean the associations between genetic variants and particular behaviors ‘would fail to replicate and accumulate’ (Turkheimer, Reference Turkheimer2000, p. 163) and that the likelihood of finding behavioral domains where they do would be ‘inversely related to the complexity of the behavior under study’ (Turkheimer, Reference Turkheimer2000, p. 164). Both predictions have since been disproven by subsequent genomewide association studies (GWAS).

Turkheimer has himself subsequently acknowledged that genetic associations with behavior replicate, arguing in a public exchange with another GWAS critic, ‘Isn’t the success of a polygenic score in a new sample an indication that the effect sizes of SNPs are “replicating” in a general way? Otherwise, the scores wouldn’t predict anything’ (Turkheimer, Reference Turkheimer2019b). In a follow-up exchange with Turkheimer a few months later, the statistical geneticist Michel Nivard listed several notable examples of statistically significant genetic associations (‘hits’) replicating out of sample, writing: ‘Howard [et] al replicate 97 out of 102 depression hits out of sample, Lee et al replicate a substantial number of 160 EA2 hits in the new data with no overlap. Okbay does the same for the 3 EA1 hits. Saying we don’t replicate out of sample is misleading’ (Nivard, Reference Nivard2019).Footnote ⁴⁷

As of the latest, GWAS of educational attainment (EA4), Okbay et al. (Reference Okbay, Wu, Wang, Jayashankar, Bennett, Nehzati, Sidorenko, Kweon, Goldman, Gjorgjieva, Jiang, Hicks, Tian, Hinds, Ahlskog, Magnusson, Oskarsson, Hayward, Campbell and Young2022) were able to predict 13−17% of the variance in EA in independent samples. This shows that, against Turkheimer’s prediction, genetic associations identified in the GWAS do, in fact, accumulate into powerful genetic predictors of individual differences. Moreover, contrary to Turkheimer’s prediction that success in finding genetic associations that replicate and accumulate would be ‘inversely related to the complexity of the behavior under study’, some of the greatest successes so far have been for educational attainment, a highly complex and socially contingent behavioral trait. This is primarily because it has been easier to obtain data on educational qualifications for large numbers of genotyped subjects. The EA4 GWAS involved more than 3 million genotyped individuals.

The latest EA4 polygenic scores predict educational attainment about as accurately as parental education levels, one of the best phenotypic predictors we have for this outcome (see e.g., Figure 4b of Lee et al., Reference Lee, Wedow, Okbay, Kong, Maghzian, Zacher, Nguyen-Viet, Bowers, Sidorenko, Linnér, Fontana, Kundu, Lee, Li, Li, Royer, Timshel, Walters, Willoughby and Cesarini2018). That we can now guess someone’s educational attainment more accurately from their DNA than from their parents’ income strikes many laypeople as something from the realms of science fiction, and as recently as 2014 this prospect appeared to strike Turkheimer the same way. Reacting to an early GWAS of children’s intelligence (Benyamin et al., Reference Benyamin, St Pourcain, Davis, Davies, Hansell, Brion, Kirkpatrick, Cents, Franić, Miller, Haworth, Meaburn, Price, Evans, Timpson, Kemp, Ring, McArdle, Medland and Visscher2014), Turkheimer wrote:

Note that, in the blogpost cited above, the ‘bottom line’ was about genetic prediction of behavior. Indeed, Turkheimer concluded that blogpost by writing ‘The real question is whether, short of growing everyone an identical twin, … we can build useful … prediction models’ (Turkheimer, Reference Turkheimer2013 [emphasis added]).Footnote ⁴⁸ Nevertheless, eight years later, Turkheimer unironically greeted the results of EA4 with the dismal question: ‘Has GWAS of complex human behavior turned out to be a disappointment?’ (Turkheimer, Reference Turkheimer2022a). Moreover, despite the above remarks about genetic prediction being a matter of public record, Turkheimer has insisted that his predictions ‘have been consistently right’ (Turkheimer, Reference Turkheimer2022d) and that ‘it is GWAS-world that has moved the goalpost, not the critics’ (Turkheimer, Reference Turkheimer2022c).

7.2 The Conspicuous Alignment of Turkheimer’s Views About Genetic Influence on Behavior

What we seek to highlight here is that the prior assumptions Turkheimer brings to his interpretation of many scientific results in behavioral genetics (and his pessimistic predictions about future scientific developments) are not based simply on empirical priors informed by the weight of previous evidence, but on philosophical priors based on the existential threat to free will that substantial genetic causation of human behavior is feared to imply. Without recognizing this it would be difficult to explain the improbable alignment of Turkheimer’s positions across a broad range of scientific issues.

If substantial genetic causation of individual differences in behavior is felt to represent an intolerable infringement on human agency, a particular constellation of scientific beliefs can predictably be expected to follow:

• • The interpretation of heritability as the proportion of behavioral variation that is caused by genetic differences between individuals will be rejected (Turkheimer, Reference Turkheimer1998, Reference Turkheimer2000, Reference Turkheimer2011).

• • The distinction between genetic and environmental influences will be blurred (Turkheimer, Reference Turkheimer1998, Reference Turkheimer2000).

• • Heritability estimates will be characterized as barely informative and of little scientific interest (Turkheimer, Reference Turkheimer1998, Reference Turkheimer2011).

• • Despite the purported irrelevance of heritability estimates, potential sources of upward bias in those estimates will be emphasized (e.g., nonadditive genetic influences, violation of equal environments assumption) while potential sources of downward bias will be neglected (e.g., assortative mating, rare variants) (Turkheimer, Reference Turkheimer2022b, Reference Turkheimer2022e).

• • The methodologies that produce the lowest heritability estimates (e.g., RDR h², within-family SNP h²) will be systematically preferred over other methods (Turkheimer, Reference Turkheimer2021a, Reference Turkheimer2021b).

• • Heritability estimates will be characterized, not merely as population-specific by definition, but as highly labile and changeable across social contexts (Turkheimer et al., Reference Turkheimer, Haley, Waldron, D’Onofrio and Gottesman2003).

• • Any unresolved confounding in (e.g., SNP) heritability estimators and polygenic scores will be assumed to be mostly environmental/non-genetic (Turkheimer, Reference Turkheimer2021d, Reference Turkheimer2022d, Reference Turkheimer2023).

• • The (within-family) effects of polygenic scores that only capture the effects of genotyped SNPs in current GWASs will be represented as though they capture the total genetic influence on a trait (Turkheimer, Reference Turkheimer2021d, Reference Turkheimer2022d, Reference Turkheimer2023).

• • Finally, insofar as any quantifiable genetic influences on behavioral differences are grudgingly conceded, it will be assumed that these effects largely result from the arbitrary prejudices and prepositions that society harbors towards people based on incidental, heritable characteristics such as their height, weight, or physical attractiveness (Turkheimer, Reference Turkheimer2018, Reference Turkheimer2021a, Reference Turkheimer2021c).

Turkheimer has publicly subscribed to all of the positions just described.

We stress that some of these positions are legitimate scientific hypotheses that may ultimately be borne out. What is conspicuous is how they are all aligned in a common direction that uniformly resists the prospect of substantial genetic causation of individual differences in human behavior. Such a uniformly one-sided skepticism is not scientifically impartial or disinterested. After all, there is no scientific reason why the default hypothesis should be that the genetic contribution to differences in human behavior should be statistically minute, and there are many empirical findings that suggest exactly the opposite.

While it is true that extraordinary claims demand extraordinary evidence, scientific claims should only be considered extraordinary to the extent that they contradict previous empirical findings, not on the basis that we dislike their potential moral or political implications. Otherwise, scientific skepticism risks becoming an instrument of wish-fulfilment. A classic dictum in philosophy called Hume’s guillotine is that one cannot derive an ‘ought’ from an ‘is’, that is, that one cannot derive a moral injunction from a set of facts. In behavioral genetics, we are at risk of making a more egregious philosophical error: the moralistic fallacy of deriving an ‘is’ from an ‘ought’ — of deciding that some things cannot be true because they must not be true.Footnote ⁴⁹

Persistent skepticism about genetic influences on human behavior has had the welcome side-effect of driving rigor and technical innovation in behavioral genetics (Lee & McGue, Reference Lee and McGue2016), but selective and insatiable demands for scientific rigor from motivated skeptics risks placing behavior genetic research on a scientific treadmill that indefinitely postpones a public reckoning with potentially inconvenient truths (Alexander, Reference Alexander2014). Some of Turkheimer’s pessimistic predictions for behavioral genetics may well turn out to be correct but, of course, a blanket scientific pessimism will always turn out to be justified in some instances. The danger is that Turkheimer’s (explicitly) motivated scientific pessimism risks becoming a self-fulfilling prophecy if it is allowed to derail productive lines of scientific enquiry.