Criticism #2: MZ twins experience life more similarly starting as fetuses. DZ twins are “dichorionic,” meaning that they have separate placentas. In contrast, 75 percent of MZ twins share one placenta (i.e., are “monochorionic”).* Thus most MZ twin fetuses share maternal blood flow more than do DZ twins, and thus are exposed to more similar levels of maternal hormones and nutrients. If that isn’t recognized, greater similarity in MZs will be misattributed to genes.
Various studies have determined what the chorionic status was in different MZ pairs and then examined end points related to cognition, personality, and psychiatric disease. By a small margin, most studies show that chorionic status does make a difference, leading to overestimates of genetic influence. How big of an overestimation? As stated in one review, “small but not negligible.”24
Criticism #3: Recall that adoption studies assume that if a child is adopted soon after birth, she shares genes but no environment with her biological parents. But what about prenatal environmental effects? A newborn just spent nine months sharing the circulatory environment with Mom. Moreover, eggs and sperm can carry epigenetic changes into the next generation. If these various effects are ignored, an environmentally based similarity between mother and child would be misattributed to genes.
Epigenetic transmission via sperm seems of small significance. But prenatal and epigenetic effects from the mother can be huge—for example, the Dutch Hunger Winter phenomenon showed that third-trimester malnutrition increased the risk of some adult diseases more than tenfold.
This confound can be controlled for. Roughly half your genes come from each parent, but prenatal environment comes from Mom. Thus, traits shared more with biological mothers than with fathers argue against a genetic influence.* The few tests of this, concerning the genetic influence on schizophrenia demonstrated in twin studies, suggest that prenatal effects aren’t big.
Criticism #4: Adoption studies assume that a child and adoptive parents share environment but not genes.25 That might approach being true if adoption involved choosing adoptive parents randomly among everyone on earth. Instead, adoption agencies prefer to place children with families of similar racial or ethnic background as the biological parents (a policy advocated by the National Association of Black Social Workers and the Child Welfare League).* Thus, kids and adoptive parents typically share genes at a higher-than-chance level; if this isn’t recognized, a similarity between them will be misattributed to environment.
Researchers admit there is selective placement but argue over whether it’s consequential. This remains unsettled. Bouchard, with his twins separated at birth, controlled for cultural, material, and technological similarities between the separate homes of twin pairs, concluding that shared similarity of home environments due to selective placement was a negligible factor. A similar conclusion was reached in a larger study carried out by both Kendler and another dean of the field, Robert Plomin of King’s College London.
These conclusions have been challenged. The most fire-breathing critic has been Princeton psychologist Leon Kamin, who argues that concluding that selective placement isn’t important is wrong because of misinterpretation of results, use of wimpy analytical tests, and overreliance on questionable retrospective data. He wrote: “We suggest that no scientific purpose is served by the flood of heritability estimates generated by these studies.”26
Here’s where I give up—if super smart people who think about this issue all the time can’t agree, I sure don’t know how seriously selective placement distorts the literature.
Criticism #5: Adoptive parents tend to be more educated, wealthier, and more psychiatrically healthy than biological parents.27 Thus, adoptive households show “range restriction,” being more homogeneous than biological ones, which decreases the ability to detect environmental effects on behavior. Predictably, attempts to control for this satisfy only some critics.
—
So what do we know after this slog through the criticisms and countercriticisms about adoption and twin studies?
Everyone agrees that confounds from prenatal environment, epigenetics, selective placement, range restriction, and assumptions about equal environment are unavoidable.
Most of these confounds inflate the perceived importance of genes.
Efforts have been made to control for these confounds and generally have shown that they are of less magnitude than charged by many critics.
Crucially, these studies have mostly been about psychiatric disorders, which, while plenty interesting, aren’t terribly relevant to the concerns of this book. In other words, no one has studied whether these confounds matter when considering genetic influences on, say, people’s tendency to endorse their culture’s moral rules yet rationalize why those rules don’t apply to them today, because they’re stressed and it’s their birthday. Lots more work to be done.
The Fragile Nature of Heritability Estimates
Now starts a bruising, difficult, immensely important subject. I review its logic every time I teach it, because it’s so unintuitive, and I’m still always just words away from getting it wrong when I open my mouth in class.
Behavior genetics studies usually produce a number called a heritability score.28 For example, studies have reported heritability scores in the 40 to 60 percent range for traits related to prosocial behavior, resilience after psychosocial stress, social responsiveness, political attitudes, aggression, and leadership potential.
What’s a heritability score? “What does a gene do?” is at least two questions. How does a gene influence average levels of a trait? How does a gene influence variation among people in levels of that trait?
These are crucially different. For example, how much do genes have to do with people’s scores averaging 100 on this thing called an IQ test? Then how much do genes have to do with one person scoring higher than another?
Or how much do genes help in explaining why humans usually enjoy ice cream? How much in explaining why people like different flavors?
These issues utilize two terms with similar sounds but different meanings. If genes strongly influence average levels of a trait, that trait is strongly inherited. If genes strongly influence the extent of variability around that average level, that trait has high heritability.* It is a population measure, where a heritability score indicates the percentage of total variation attributable to genetics.
The difference between an inherited trait and heritability generates at least two problems that inflate the putative influence of genes. First, people confuse the two terms (things would be easier if heritability were called something like “gene tendency”), and in a consistent direction. People often mistakenly believe that if a trait is strongly inherited, it’s thus highly heritable. And it’s particularly bad that confusion is typically in that direction, because people are usually more interested in variability of traits among humans than in average levels of traits. For example, it’s more interesting to consider why some people are smarter than others than why humans are smarter than turnips.
The second problem is that research consistently inflates heritability measures, leading people to conclude that genes influence individual differences more than they do.
Let’s slowly work through this, because it’s really important.
The Difference Between a Trait Being Inherited and Having a High Degree of Heritability
You can appreciate the difference by considering cases where they dissociate.
First, an example of a trait that is highly inherited but has low heritability, offered by the philosopher Ned Block:29 What do genes have to do with humans averaging five fingers per hand? Tons; it’s an inherited trait. What do genes have to do with variation around that average? Not much—cases of other than five fingers on a hand are mostly due to accidents. While average finger number is an inherited trait, the heritability of finger number is low—genes don’t explain individual differences much. Or s
tated differently: Say you want to guess whether some organism’s limb has five fingers or a hoof. Knowing their genetic makeup will help by identifying their species. Alternatively, you’re trying to guess whether a particular person is likely to have five or four fingers on his hand. Knowing whether he uses buzz saws while blindfolded is more useful than knowing the sequence of his genome.
Next consider the opposite—a trait that is not highly inherited but which has high heritability. What do genes directly have to do with humans being more likely than chimps to wear earrings? Not much. Now consider individual differences among humans—how much do genes help predict which individuals are wearing earrings at a high school dance in 1958? Tons. Basically, if you had two X chromosomes, you probably wore earrings, but if you had a Y chromosome, you wouldn’t have been caught dead doing so. Thus, while genes had little to do with the prevalence of earrings among Americans in 1958 being around 50 percent, they had lots to do with determining which Americans wore them. Thus, in that time and place, wearing earrings, while not a strongly inherited trait, had high heritability.
The Reliability of Heritability Measures
We’re now clear on the difference between inherited traits and their degree of heritability and recognize that people are usually more interested in the latter—you versus your neighbor—than the former—you versus a wildebeest. As we saw, scads of behavioral and personality traits have heritability scores of 40 to 60 percent, meaning that genetics explains about half the variability in the trait. The point of this section is that the nature of research typically inflates such scores.*30
Say a plant geneticist sits in the desert, studying a particular species of plant. In this imaginary scenario a single gene, gene 3127, regulates the plant’s growth. Gene 3127 comes in versions, A, B, and C. Plants with version A always grow to be one inch tall; version B, two inches; C, three inches.* What single fact gives you the most power in predicting a plant’s height? Obviously, whether it has version A, B, or C—that explains all the variation in height between plants, meaning 100 percent heritability.
Meanwhile, twelve thousand miles away in a rain forest, a second plant geneticist is studying a clone of that same plant. And in that environment plants with version A, B, or C are 101, 102, or 103 inches tall, respectively. This geneticist also concludes that plant height in this case shows 100 percent heritability.
Then, as required by the plot line, the two stand side by side at a conference, one brandishing 1/2/3 inch data, the other 101/102/103. They combine data sets. Now you want to predict the height of one example of that plant, taken from anywhere on the planet. You can either know which version of gene 3127 it possesses or what environment it is growing in. Which is more useful? Knowing which environment. When you study this plant species in two environments, you discover that heritability of height is miniscule.
Neon lights! This is crucial: Study a gene in only one environment and, by definition, you’ve eliminated the ability to see if it works differently in other environments (in other words, if other environments regulate the gene differently). And thus you’ve artificially inflated the importance of the genetic contribution. The more environments in which you study a genetic trait, the more novel environmental effects will be revealed, decreasing the heritability score.
Scientists study things in controlled settings to minimize variation in extraneous factors and thus get cleaner, more interpretable results—for example, making sure that the plants all have their height measured around the same time of year. This inflates heritability scores, because you’ve prevented yourself from ever discovering that some extraneous environmental factor isn’t actually extraneous.* Thus a heritability score tells how much variation in a trait is explained by genes in the environment(s) in which it’s been studied. As you study the trait in more environments, the heritability score will decrease. This is recognized by Bouchard: “These conclusions [derived from a behavior genetics study] can be generalized, of course only to new populations exposed to a range of environments similar to those studied.”31
Okay, that was slick on my part, inventing a plant that grows in both desert and rain forest, just to trash heritability scores. Real plants rarely occur in both of those environments. Instead, in one rain forest the three gene versions might produce plants of heights 1, 2, and 3 inches, while in another they are 1.1, 2.1, and 3.1, producing a heritability score that, while less than 100 percent, is still extremely high.
Genes typically still play hefty roles in explaining individual variability, given that any given species lives in a limited range of environments—capybaras stick to the tropics, polar bears to the Arctic. This business about heterogeneous environments driving down heritability scores is important only in considering some hypothetical species that, say, lives in both tundra and desert, in various population densities, in nomadic bands, sedentary farming communities, and urban apartment buildings.
Oh, that’s right, humans. Of all species, heritability scores in humans plummet the most when shifting from a controlled experimental setting to considering the species’ full range of habitats. Just consider how much the heritability score for wearing earrings, with its gender split, has declined since 1958.
—
Now to consider an extremely important complication.
Gene/Environment Interactions
Back to our plant. Imagine a growth pattern in environment A of 1, 1, and 1 for the three gene variants, while in environment B it’s 10, 10, and 10. When considering the combined data from both environments, heritability is zero—variation is entirely explained by which environment the plant grew in.
Now, instead, in environment A it’s 1, 2, and 3, while in environment B it’s also 1, 2, and 3. Heritability is 100 percent, with all variability in height explained by genetic variation.
Now say environment A is 1, 2, and 3, and environment B is 1.5, 2.5, 3.5. Heritability is somewhere between 0 percent and 100 percent.
Now for something different: Environment A: 1, 2, 3. Environment B: 3, 2, 1. In this case even talking about a heritability score is problematic, because different gene variants have diametrically opposite effects in different environments. We have an example of a central concept in genetics, a gene/environment interaction, where qualitative, rather than just quantitative, effects of a gene differ by environment. Here’s a rule of thumb for recognizing gene/environment interactions, translated into English: You are studying the behavioral effects of a gene in two environments. Someone asks, “What are the effects of the gene on some behavior?” You answer, “It depends on the environment.” Then they ask, “What are the effects of environment on this behavior?” And you answer, “It depends on the version of the gene.” “It depends” = a gene/environment interaction.
Here are some classic examples concerning behavior:32
The disease phenylketonuria arises from a single gene mutation; skipping over details, the mutation disables an enzyme that converts a potentially neurotoxic dietary constituent, phenylalanine, into something safe. Thus, if you eat a normal diet, phenylalanine accumulates, damaging the brain. But eat a phenylalanine-free diet from birth, and there is no damage. What are the effects of this mutation on brain development? It depends on your diet. What’s the effect of diet on brain development? It depends on whether you have this (rare) mutation.
Another gene/environment interaction pertains to depression, a disease involving serotonin abnormalities.33 A gene called 5HTT codes for a transporter that removes serotonin from the synapse; having a particular 5HTT variant increases the risk of depression . . . but only when coupled with childhood trauma.* What’s the effect of 5HTT variant on depression risk? It depends on childhood trauma exposure. What’s the effect of childhood trauma exposure on depression risk? It depends on 5HTT variant (plus loads of other genes, but you get the point).
Another example concerns FADS2, a gene involved in fat metabolism.34 One variant is associated with
higher IQ, but only in breast-fed children. Same pair of “what’s the effect” questions, same “it depends” answers.
One final gene/environment interaction was revealed in an important 1999 Science paper. The study was a collaboration among three behavioral geneticists—one at Oregon Health Sciences University, one at the University of Alberta, and one at the State University of New York in Albany.35 They studied mouse strains known to have genetic variants relevant to particular behaviors (e.g., addiction or anxiety). First they ensured that the mice from a particular strain were essentially genetically identical in all three labs. Then the scientists did cartwheels to test the animals in identical conditions in the labs.
They standardized everything. Because some mice were born in the lab but others came from breeders, homegrowns were given bouncy van rides to simulate the jostling that commercially bred mice undergo during shipping, just in case that was important. Animals were tested at the same day of age on the same date at the same local time. Animals had been weaned at the same age and lived in the same brand of cage with the same brand and thickness of sawdust bedding, changed on the same day of the week. They were handled the same number of times by humans wearing the same brand of surgical gloves. They were fed the same food and kept in the same lighting environment at the same temperature. The environments of these animals could hardly have been more similar if the three scientists had been identical triplets separated at birth.
Behave: The Biology of Humans at Our Best and Worst Page 25