I love writing reviews of bad books. There’s nothing like the feeling when you’re halfway through a book and it dawns on you that this book is really bad, and that it deserves a pummeling in print for misleading the reader, spreading dodgy ideas, or getting tangled up in the barbed wire of its own prose.
Sometimes, though, a bad book review comes out that shames us bad book reviewers. A review that has all the snark of a good bad book review, but none of the substance. A review that displays not a sharp critical take, but a predictable argument based on its author’s political prejudices.
It happened earlier this month. The review in question, in the New York Review of Books, is by two Stanford University professors: Marcus Feldman, a biologist, and Jessica Riskin, a historian of science. It’s called “Why Biology is Not Destiny”, and it’s about Kathryn Paige Harden’s book The Genetic Lottery. In this post, I’m going to review the review.
My conflict of interest statement here is threefold: Paige Harden is a friend of mine; we have the same literary agent; and I wrote one of the blurbs on the back of The Genetic Lottery, so if the book is as bad as this review implies, I’ll have egg on my face.
Prior to my review-review, I’ll briefly sketch out what Harden says in the book:
There’s a lot of evidence that many of the important differences between people in psychological attributes—attributes that help them do well in our current society—are in part influenced by their genetic differences;
There’s a debate around this topic that’s often taken to the extremes - on one side people act as if these genetic differences are the most important thing ever; on the other they act as if they don’t exist at all;
There’s a moderate position where you can both accept the science and have an interesting debate about what our genetic predispositions mean for social equality.
You can read more about her ideas in a long New Yorker profile that came out just before the book’s release. In my view The Genetic Lottery has the clearest, best-written popular explanation of old studies and new developments in behaviour genetics that I’ve seen. And although I slightly wish there was more of a firewall between the sections on the basic science and Harden’s own interpretation, I’m still very happy to recommend it to anyone who wants to know more about this field.
To Feldman and Riskin, though, there’s nothing new about the evidence Harden describes.
“Biological essentialism, aimed at demonstrating an innate hierarchy of intelligence, is going strong after more than two centuries of empirical failure.”
Ouch! It’s all like this: the review is bad-tempered, confusing, and tendentious. It shows little acquaintance with the current research on behavioural genetics. We’ll go through all that. But to me the most interesting thing about the review is that, if you follow its logic, you end up with pure nihilism. In their rush to debunk Harden’s arguments about behavioural genetics, Feldman and Riskin end up in a position where it’s “impossible”—their term—to do any research in genetics at all. It might as well have been called “Why Biology—or At Least Genetics—is Not Science”.
“I refute it thus!”
A big part of Feldman and Riskin’s argument is that social scientists should stay out of the field of genetics. There are things that it’s sensible to look at using a genetic lens, things where the use of genetics is ambiguous, and things where it’s completely absurd. Measures from social science are in the latter category, and yet the incentives are pushing social scientists towards genetics:
If you find a magical hammer that, whenever you swing it, rewards you with funding and professional advancement, you look at your research area and see nothing but nails. Genome-wide association studies are the social sciences’ new magical hammer.
Feldman and Riskin’s example of a real, objective trait is macular degeneration, an eye disease that often comes with age and which causes blurred vision and sometimes blindness. It makes sense to analyse this in terms of its genetic basis, Feldman and Riskin write; and indeed, in one of the earliest examples of a genome-wide association study (GWAS), scientists:
…found two sets of single-nucleotide polymorphisms [common genetic variants] where the groups [macular degeneration sufferers and controls] differed significantly.
But then we come to the intermediate, ambiguous traits:
For complex diseases such as schizophrenia or bipolar disorder, however, genome studies haven’t revealed any spots showing statistically important differences between the focus and control groups; but there are thousands of spots showing statistically tiny differences.
This is already dreadful science writing. For macular degeneration, there were genes that “differed significantly” - does this mean “differed statistically significantly”? That is, are they using the word “significantly” in the technical sense? One would assume so. But then they say studies haven’t found “statistically important” differences for schizophrenia and bipolar disorder. Since in fact the studies have found hundreds of statistically significant differences for schizophrenia and dozens for bipolar, they must mean something different. But then, what does “statistically tiny” mean? If they mean that the hundreds or dozens of differences found in those studies were all tiny, why say that “thousands of spots” had tiny differences? It’s a confusing mess.
With regard to the “magical hammer” of GWAS, Feldman and Riskin argue that macular degeneration “seems plausibly to be a nail”. Schizophrenia, on the other hand “appears not [their italics] to be a nail, though it might have some structural features a hammer could help with.” (I have no idea what this means. Recent studies have found both lots of small-effect genetic variants and a small number of rarer variants that also predispose you to the disease).
But then we get to social-science traits, which are the real objects of Feldman and Riskin’s ire:
The things social scientists have been swinging at aren’t just non-nails. They are to nails as ships to sealing wax, as cabbages to kings. To suggest that macular degeneration has genetic causes is to make an empirically testable proposal; to suggest that “grit” or “openness to experience” has genetic causes is to make a category mistake. These are interpretive descriptions, made of ideas, opinions, and practices, not molecules.
Note that somehow they forgot to mention schizophrenia or bipolar here. Do they think that it’s similarly ridiculous to argue that, say, schizophrenia has (at least partial) genetic causes? They never say, instead preferring the softer targets of grit and openness to experience. I’m pretty sure Feldman and Riskin aren’t dualists, proposing that we have “ideas, opinions, and practices” that float free from our brains or the molecules that make them, so it’s very difficult to know what bright philosophical line they’re drawing between eyes and brains that means that genes can merrily affect the first but can never touch the second.
One possibility is that it might be due to measurement: as they note in their review, it’s harder to measure grit than it is to measure macular degeneration. They ridicule Harden for talking about psychological variables because “there must be disagreement among researchers” about them - “more so… than about what constitutes macular degeneration”. Yeah! Macular degeneration is totally objective, unlike those silly psychological variables! Except… take five minutes to look at the macular degeneration literature, and you find that there’s so much disagreement about diagnosis that researchers in that field have to put together conferences to try and come up with consensus definitions of, for example, what counts as late-stage atrophy in macular degeneration, and how to measure it. So, again: where is the philosophical line between psychologists debating how to measure grit, and physicians debating how to measure macular atrophy?
At base, all these psychological traits—whether “openness to experience”, “grit”, or “schizophrenia”—are descriptions of behavioural tendencies. For openness, the specific behaviour is “tends to say on an openness questionnaire that they find it pleasurable when people talk about ideas (among other things)”. For grit, it’s “tends to say on a grit questionnaire that they work hard and get tasks done (among other things)”. For schizophrenia, it’s “has acted in such a way that made a psychiatrist give them a diagnosis of schizophrenia, presumably because they appear to suffer from delusions, hallucinations, thought disorder, and so on”.
To see how these behavioural tendencies might potentially have genetic causes, you just need to walk back through the hypothetical causal chain. The questions on the grit questionnaire are in part about focusing on tasks. So maybe people who have trouble focusing report lower levels of grit. Maybe they have trouble focusing because they have higher levels of impulsivity. Maybe they have higher impulsivity because of a difference in the structure of their prefrontal cortex. Maybe that difference is there in part because of differences in particular proteins that played a role in their brain’s development. Maybe those proteins were different because of different genetic variants in their DNA.
I’m not saying this is the actual causal chain for grit. This is just a toy example which could be entirely wrong; the real story will be vastly more complicated. It’ll involve the environment too, as well as lots of random luck. But that’s the kind of thing people mean when they say a psychological trait might have genetic causes. And it’s the kind of thing they design studies, including both genetic and environmental data, to test. This is doing science: coming up with testable hypotheses about a difficult, complex problem, and then linking together evidence from studies on various different parts of this chain about what causes what.
What Feldman and Riskin are doing is no better than the classic “appeal to the stone” fallacy, where Samuel Johnson, in a fit of pique at being unable to think of an actual argument to refute Bishop Berkeley’s philosophical argument about how objects don’t truly exist in the world, kicked a stone along the road and exclaimed “I refute it thus!”. Sure, there’s no actual line to draw between a “physical” and a “psychological” measurement when it comes to genetic analysis. But look, some people have macular degeneration! Eye refute it thus.
Cause I said so
Harden argues in the book that the kind of causal chain I just sketched is, for whatever reason, much more difficult for people to believe if it’s about genetics than about the environment. She says that people hold a double-standard: they’re happy to agree that an environmental intervention was causal even if they don’t know the specific mechanisms for how it had its effect.
To illustrate it, she uses an example from when children were rescued from orphanages in Romania, and showed increases in their IQ. She says that, although this study clearly shows that being fostered caused the IQ increase, we don’t know the specific mechanism for how it happened: which parts of the brain, which neurons, which synapses were impacted by the foster care. And that’s fine - people are still perfectly happy to say the fostering improved the IQ.
But mention genetics as a cause and people suddenly raise their hackles: they want much more information before they’ll even consider that genetics might in any way be involved.
Here’s how Feldman and Riskin respond to this point:
Well, of course we know why it was better to be in foster care. Any or all the mechanisms Harden lists may have been involved, but the essential explanation is simple: care causes children to thrive; neglect causes them to languish.
They then make perhaps their most bizarre argument so far:
Harden’s insistence that no one really knows how it worked reproduces some important steps in the old eugenic circular logic…
I’m not making this up or misquoting or anything like that: this is what they say in their review. If you point out that we don’t yet know the specific brain-level mechanisms of how foster care causes children to develop more healthily - well, you must be a eugenicist.
Then, in their next paragraphs, they go on to demand very specific causal mechanisms for how a genetic variant might cause differences in a psychological trait. In other words, they beautifully illustrate the exact double-standard Harden described. If it’s about things they like: “well, of course we know” - no research required. If it’s about things they don’t like: tell us the precise mechanism right now or give up. Scott Alexander once warned his readers to beware isolated demands for rigour. This one is a classic. Importantly, if we held our genetic investigations of macular degeneration to these same standards, we’d never get any of them done either.
This dismissive, incurious streak is also in evidence later in the review. Harden had made the point that huge swathes of research on developmental psychology are compromised because of “genetic confounding”. The logic is the following. Imagine you find a correlation between a parental behaviour and a characteristic of their child: a classic example is that parents who read to their children more often have children who are, on average, better at reading. You might assume that such a correlation would imply that reading to kids causes them to learn to read better.
And it might! But there’s a big reason to take extra care: parents don’t just give their kids parenting - they also give them genes. It’s possible that kids who are better at reading are better in part for genetic reasons, and those same genes also made the parents more interested in reading. If you just do a study with no genetic component, you’ll completely miss this, and overestimate the effect of the environment.
In this way (and others), including genetics in your social-science research is one of the best ways to reduce potential confounding factors, and truly understand what’s happening in the environment. The “natural experiment” of twins helps us to hold genetics constant while we look at the other stuff. It’s not a perfect experiment, by any means - all study designs have their weaknesses. But doing something like this is necessary if you want a full picture of what contributes to psychological—or any—differences between children.
Now, one of my own criticisms of Harden’s book is that she doesn’t give enough examples of this kind of thing: it really is the case that, for example, developmental psychology is full of studies that ignore genetics, but I think the reader deserved a few more specifics to really convince them. But Feldman and Riskin don’t even bother trying to engage. They claim that Harden is saying this with a “poker face” (she’s secrectly interested in only the genetics due to her eugenic views, or something!), and then blithely state that to disentangle genetics and environment is “impossible, even as an ideal”. Indeed:
We can no more unbraid genetics and environment than we can unbraid history and culture, or climate and landscape, or language and thought.
It’s not only geneticists: historians, climate scientists, and linguists might as well give up too. They can’t even aspire to do their work: it’s impossible. The dreariest thing about this nihilism is that it’s not in the slightest bit new. Steven Pinker wrote an essay responding to these exact points as long ago as 2004. His conclusion was the following:
“The human brain has been called the most complex object in the known universe. No doubt hypotheses that pit nature against nurture as a dichotomy or that correlate genes or environment with behavior without looking at the intervening brain will turn out to be simplistic or wrong. But that complexity does not mean we should fuzz up the issues by saying that it’s all just too complicated to think about, or that some hypotheses should be treated a priori as obviously true, obviously false, or too dangerous to mention. As with inflation, cancer, and global warming, we have no choice but to try to disentangle the multiple causes.”
Over at the Normielisation Substack, my friend Jeremy Driver coined the term “Cheems Mindset”. It’s defined as:
“the reflexive belief that barriers to policy outcomes are natural laws that we should not waste our time considering how to overcome.”
Maybe there’s a Scientific Cheems Mindset too: one where people reflexively orient towards—and sometimes revel in—problems and complexities rather than try to think of ways to overcome them. And that’s particularly frustrating when overcoming such issues is among the most important parts of doing science.
The nice thing is that behaviour geneticists and their colleagues have, particularly in recent years, been opening all sorts of new windows on this complexity. Not only have they begun to understand, quantify, and interrogate all the specific genetic variants that are linked to behavioural and psychiatric outcomes, but they’re combining them with family data, understanding how factors like environmental influences, assortative mating, population stratification, and all kinds of gene-environment correlation might affect their results.
And although this field is hardly immune to hype, it also has the self-critical streak you’d hope for, with new papers regularly popping up to criticise previous work and point to biases, confounders, and other problems, while researchers develop new statistical methods to account for them. No cheems here. But for Feldman and Riskin, all this is highly suspect. No scientific progress on this issue has been made since Francis Galton - and nor could it ever be. Shut it all down.
Weasels and giraffes
Feldman and Riskin argue that at every turn, Harden is desperate to argue for genetic determinism.
“When merely implying causation, she uses weasel words: genes are ‘relevant’ for educational attainment; they are ‘associated with’ first having sex at an earlier age; they ‘matter’ for aggression and violence; social and economic inequalities ‘stem from’ genetics.”
I could just about agree with them on “relevant” and “matter” and “stem from”. Those are a bit vague, if forgivable in a popular work. But “associated with”? This is totally normal, well-understood scientific language that means “there is a correlation between these two things”. You can even put a number on it, between -1 and 1. And “associated with” doesn’t “imply causation” - it’s explicitly saying that there’s just, well, an association.
The problem here is that Feldman and Riskin aren’t above using this kind of rhetorical trick themselves. Two examples from elsewhere in the review:
“This is not to say that genes are inessential to social life.”
“True, genes help shape people…”
“Not inessential”? “Help shape”? Don’t those sound a bit, well, weaselly to you? They do to me, and I think they serve a very specific purpose. These authors know that it’s perverse to say “genes don’t have any influence on behaviour” - but they want to talk out of both sides of their mouth: they want to say that anyone looking at genetics and social outcomes is hopelessly misguided (and maybe a bit evil), while also having a couple of sentences to refer back to, to cover themselves if they’re accused of denying the obvious genetic influences on behaviour.
The reason they need to do this is that they end up at an extreme—and again, nihilistic—position. If behaviours aren’t to some extent related to genes, they can’t be selected for. If selection doesn’t happen on what could clearly be fitness-promoting or fitness-reducing behaviours, we’d have to throw out lot of what we know about natural selection - and certainly we couldn’t do any more research on it. Surely Feldman and Riskin aren’t out to rewrite the theory of evolution - are they?
Well… about that. There’s an amusing coda to this episode. In a bonus Q&A that Feldman and Riskin did with the NY Review of Books, we learn is that Riskin is a fan of the Chevalier de Lamarck, who in the 19th Century proposed an early and incorrect theory of evolution that involved the “inheritance of acquired characteristics”. The famous example, which Riskin mentions, is a giraffe who strives to reach leaves on higher branches, and in doing so extends his neck, then somehow passes longer-neckness down to his descendants. Riskin states:
“Lamarck’s theory languished in exile for over a century, from roughly the 1880s until the 1990s, when the possibility that organisms might transform themselves heritably began re-entering mainstream biology in areas such as epigenetics.”
Riskin is talking about “transgenerational epigenetic inheritance” - the idea that the environment might change people’s genomes and that these changes pass on to future generations (the way she characterises it—organisms “transform[ing] themselves”—is a strangely teleological way to put it, implying that organisms, like the giraffe, are doing this intentionally).
This is a highly controversial area of modern biology. Although we know the required mechanism is certainly possible, as it applies to humans and their traits we have little more than a junkyard of small-sample, low-quality, over-hyped studies. That’s not to dismiss it entirely: let’s see the evidence. But our level of uncertainty about transgenerational epigenetic inheritance is a million miles higher than that for basic behavioural genetics - yet it’s the latter that Riskin spent a 5,000-word review trashing.
At the end of their review, Feldman and Riskin implore their readers to do their best to escape the “essentialist tradition” Harden and her behavioural-genetics research apparently follow. Their advice is essentially the following: Stop asking questions. Stop reading scientific research. Stop thinking about stuff. And indeed, if you’re a scientist: stop doing research that tries to understand how genetics might influence people.
It’s a pretty strange attitude for two academics to hold - but then again, their review is a very strange and unseemly thing altogether. I haven’t even gone into their disgraceful attempts to smear Harden as a crypto-racist, despite her book explicitly arguing against racist interpretations of behaviour genetics at every turn. That kind of thing is so beneath professors at Stanford, so bone-headed in its guilt-by-association logic (“bad people in the past did stuff to do with genetics, therefore only bad people in the present must be doing it”) that it deserves only contempt.
And that’s the thing: Harden couldn’t be clearer about her progressive politics (indeed, this has garnered her criticism from a very different angle). Before her book came out, I predicted that this might be the thing that helped scientists of that same political persuasion realise that behaviour genetics isn’t inherently an evil, right-wing, eugenic exercise done by racist biological determinists. To be sure, some of those people exist - and Harden takes aim at them in her book, showing that you can accept the science and hold a non-extreme position. It’s often inspiring and mind-changing to see someone on your own “side” of the aisle taking a position you previously disagreed with. If this review is anything to go by, my prediction was wrong.
But, as the sociologist Jeremy Freese (also at Stanford) put it in a great Twitter thread, this kind of review is of a piece with criticism of behaviour genetics that’s been around since at least the 1980s - it’s full of the same thought-terminating clichés that are rolled out again and again, regardless of the kind of evidence that’s at issue.
Even the title of the review, “Why Biology is Not Destiny” is a clichéd criticism of a strawman argument that nobody made - and which Harden couldn’t be clearer in denying.
It’s dispiriting to think that our public debate on genetics is going to be forever hobbled by this kind of thing, and from ostensibly authoritative sources. Other fields like evolutionary biology, climate science, and immunology have their creationists, climate change deniers, and anti-vaxxers, but they’re not usually Stanford professors. In this area, the nuisance call—the nihilist call, encouraging us to cut off our entire scientific nose to spite our behaviour-genetic face—is coming from inside the house.
Thanks for reading Science Fictions. If you’d like to receive new posts and support my writing, consider becoming a free or paid subscriber.
Image Credit: Getty