It was that—the negative part of The Nurture Assumption, not my theory of group socialization—that got all the publicity. DO PARENTS MATTER? was the typical headline. Professors in psychology departments all over the country were asked for their opinion and most of them disagreed with me. The developmentalists were especially vehement in their opposition. According to Newsweek, “Many of the nation’s leading scholars of child development accuse Harris of screwy logic, of misunderstanding behavioral genetics, and of ignoring studies that do not fit her thesis.”13 Actually, I had spent many pages taking apart those studies and showing exactly what was wrong with them. As for misunderstanding behavioral genetics, the behavioral geneticists didn’t think so.
The charge that I misunderstood behavioral genetics was replaced, as time went on, by the charge that I relied on it too heavily, too trustingly. My conclusions, however, were not based solely on behavioral genetic evidence; most of the research I discussed in that book (and most of the research I will tell you about in this one) came from other areas of psychology. Nevertheless, understanding the behavioral genetic evidence is essential because it provides some important clues. Clues about the environment.
In order to understand how the environment produces differences in personality, we first have to allow for the differences due to genes. We have to know what the genes contribute before we can figure out what the environment does. Those who don’t take genes into account are likely to jump to the wrong conclusions about how the environment affects personality. Psychologists from Freud on down have made that mistake. Freud noticed that finicky, uptight patients tended to have finicky, uptight mothers and attributed their plight to the way they were toilet trained.
Behavioral geneticists have been likened to Nazis and denounced as racists. They have been accused—are still being accused14—of being genetic “determinists,” though no one seems to worry much about environmental determinists. I’m not going to go into the politics of the nature-nurture debate; see Steven Pinker’s The Blank Slate. But one persistent—and misguided—criticism of behavioral genetics really gets my goat: that damn rectangle. Here, for example, is Paul Ehrlich, a professor of biology and population studies at Stanford, writing in The Chronicle of Higher Education in the year 2000:
We can’t partition the responsibility for aggression, altruism, or charisma between DNA and upbringing. In many such cases, trying to separate the contributions of nature and nurture to an attribute is rather like trying to separate the contributions of length and width to the area of a rectangle, which at first glance also seems easy. When you think about it carefully, though, it proves impossible.15
Ehrlich is saying that it’s impossible to figure out how much heredity and environment contribute to personality. They’re both important!
He is right in a way: it is nonsense to say, about a particular individual, that a certain percentage of his or her personality is due to heredity and a certain percentage to environmental influences. But no behavioral geneticist has ever made such a statement. That’s not what behavioral genetics is about. It’s about human variation, human differences. Not variation within a single individual, but differences from one individual to another within a population or group of people.
For a single rectangle, it indeed makes no sense to say that it owes its area more to one dimension than the other. But it makes perfect sense to make that statement about the variation in a group of rectangles—the ones in the top diagram, for example. These rectangles vary considerably in vertical length but are about the same in width, so most of the variation in area is due to the differences in length. In contrast, the rectangles in the bottom diagram vary considerably in width and are about the same in vertical length, so most of the variation in area is due to the differences in width.
Let’s say you have a bunch of rectangles that vary in area. You want to figure out how much of the variation—the differences in area within this set of rectangles—can be attributed to differences in width and how much to differences in length, but you can’t measure either dimension directly. How would you go about it? Right, you’d try to hold one dimension constant and see what happens to area when you vary the other. That is the principle behind behavioral genetic methods.
The innovative feature of these methods is that subjects (children or adults) are measured in pairs, usually pairs of twins or siblings. By choosing pairs that vary in genetic similarity, the researchers can study the effects of being more or less alike genetically. By choosing pairs that vary in environmental similarity, they can study the effects of having more or less similar environments. So, to vary the genetic dimension, the researchers might compare identical twins with fraternal twins, or ordinary siblings with adoptive or step-siblings. (Genetic similarity, defined as the proportion of genes shared by common descent, is 1.00 for identical twins, .50 for fraternal twins and ordinary siblings, and zero for adoptive siblings.) To vary the environmental dimension, they compare pairs reared in the same family with pairs of the same genetic similarity reared in different families.
As expected, genetic similarity makes identical twins more alike in personality than fraternal twins, and ordinary siblings more alike than adoptive siblings. Environmental similarity was expected to have a corresponding effect, but it did not. Twins who grew up together are no more alike in personality than those reared in different families, and adoptive siblings reared in the same family aren’t alike at all.
Admittedly, the environmental measure is not as well calibrated as the genetic one. Only two levels of environmental similarity are considered: either a pair grew up in the same home or they didn’t. However, it is not necessary to assume that the two who grew up in the same home had identical environments, or that the two reared separately had altogether different ones. What this method does is to assess the effects of any aspects of the environment that are the same for siblings reared together but that might differ for siblings reared apart. These include most of the things that developmentalists have been talking about for years: whether the home is headed by one parent or two, whether the parents (if there are two of them) get along well or poorly, whether they are college graduates or high school dropouts, whether their philosophies of child-rearing are autocratic or laissez-faire. The mother might stay at home or go out to work every day, as John Cheever’s mother did; she might be in robust mental health or subject to depressions like Alice James. The home itself might be urban or rural, crowded or spacious, well organized or chaotic, stocked with art supplies or strewn with used auto parts. All these things are alike for two siblings reared together—particularly if they are twins—but may differ for two reared apart.
And that’s not all. The things that might differ for the reared-apart pair also include features of the environment found outside the home. The reared-together pair not only share a home: they also grow up in the same neighborhood, attend the same schools, and—particularly if they are twins—may belong to the same peer group.
In effect, the same-home/different-home contrast skips to the bottom line. It captures most of the environmental variables that the developmentalists have been studying for fifty years and puts them all together. This is good in one way and bad in another. It’s a powerful tool, because even if each little difference between two homes has only a small effect, something should show up when they’re added together. On the other hand, it’s a blunt instrument: it can’t discriminate these effects from one another. In particular, it can’t separate the effects of family life from the effects of what happens to children outside the home. Everything is lumped together.
And yet this lumped-together environment has little or no measurable effect on the siblings who share it. The most common outcome in behavioral genetic studies, especially in the newer, larger ones, is that the estimated effects of the “shared” environment—the environment shared by two siblings who grew up together—do not differ significantly from zero. Growing up in the same home does not make twins, ordinary siblings, or ado
ptive siblings more alike in personality.16
In the clarity that hindsight brings, some developmentalists denied that they ever expected the home environment to make siblings more alike.17 This expectation, however, didn’t come out of the blue; it was based on the same reasoning as research based on traditional methods. Consider an article, written by three developmentalists, that is typical of their field: a survey of research on emotional expressiveness in children. The studies reviewed in this article showed that parents who express emotions freely tend to have emotionally expressive children, and those who keep their emotions buttoned up tend to have emotionally unexpressive children—a correlation the authors attributed to “parental socialization of emotion.”18
Let me rephrase those results. There was a statistical trend for the children of expressive parents to be above average in expressiveness and for the children of inexpressive parents to be below average in expressiveness. This means that two children, growing up in two different homes, whose parents happen to be equally expressive are, on average, more alike in expressiveness than two whose parents happen to differ in that respect. Which implies, in turn, that two children raised by the same parents should be more alike in expressiveness than two raised by different parents.
So the prediction that two children raised in the same home should be more alike in behavior than two raised in different homes follows logically—indeed, mathematically—from the results of studies using traditional methods. Prevailing theories of development, such as the theory that children are socialized by their parents to express emotions or to bottle them up, had led researchers to expect (if they thought about it at all) that being reared by the same parents would make siblings more alike.
The developmentalists found that the children’s behavior was correlated with the parents’ behavior and attributed the correlation to the effects of the home environment. Though they realized that heredity might account for some of the correlation, they never considered the possibility that heredity might account for all of it. But that is exactly how it turned out. Once the effects of genetic similarities were estimated and skimmed off, the correlation declined to zero. The putative effects of the home environment disappeared.
In other fields too, new methods of research sometimes produce results that conflict with prevailing theories. In medical science, for instance, according to epidemiologist Alvan Feinstein, “almost every plausible concept that has been held throughout the centuries about the causes, mechanisms, and treatment of diseases has been either wholly wrong or so deficient that it was later overthrown and supplanted by other concepts.”19 But plausible concepts don’t give up the ghost without a struggle. For some reason humans don’t like to admit—even to themselves—that they were wrong. Nevertheless, physicians eventually gave up their faith in purging and bloodletting, which gives me hope for psychology.
In science, the struggle between old concepts and new ones often focuses on research methods. Quoting Feinstein again,
When the conclusion suggested by the research is compared with the belief held by the reader or by the scientific community, all further aspects of rational analysis may vanish. If the results confirm what we believe, the customary human tendency is to assume that they must be right. The research methods need not be examined closely because there is no need to do so. Having produced the right answer, the methods must also be correct. Conversely, if the results are contrary to what we believe, the research methods must be wrong, no matter how good they seem.20
If the behavioral geneticists’ research had produced the expected results—that sharing genes makes people more similar in personality, and that growing up in the same family also makes people more similar—no one would have complained about their methods. Since the research didn’t produce the expected results, many members of the scientific community concluded that there must be something wrong with the methods. The jurors in the O. J. Simpson trial rejected the DNA evidence rather than give up their belief in his innocence.
But the strength of behavioral genetics is that it isn’t a single method: it’s a set of methods, designed to home in on an answer from two or more directions at once. Depending on which kinds of subject pairs take part in the study, different formulas are used to calculate the contributions of genes and environment. There is no mathematical necessity for these formulas to produce the same results—no a priori reason why estimates of environmental effects based on identical twins reared apart should agree with those based on adoptive siblings reared together—but they do. Estimates of genetic effects vary somewhat from one study to another, which is what we’d expect. But estimates of the effects of the shared home environment seldom stray far from zero, regardless of the method that is used.21
In the past three decades, the behavioral geneticists have examined a wide variety of human behaviors and psychological characteristics, measured in a variety of ways: standard personality tests; judgments by parents, teachers, spouses, or peers; observations by researchers. Objective measures—whether or not subjects got divorced or had run-ins with the law—have also been used. The remarkable thing about these results is their consistency. Occasionally an aberrant result does turn up—somebody finds a significant effect of the home environment on something or other—but aberrant results, like the glove that didn’t fit O. J., turn up occasionally in every field. One has to look at the overall pattern of the evidence, not at just a single finding. The behavioral genetic results obtained in different studies, using different subjects and measuring different things in different ways, back each other up to a degree that is rarely found in psychology.
The rectangle criticism is generally made by people who know nothing about behavioral genetics. Those with a little knowledge are more likely to base their criticism on the pie. The claim is that there’s something dodgy about the notion that variation in behavior or personality can be cut up, like a pie, into genetic and environmental sectors.22 That’s not exactly what behavioral geneticists do, however. In fact, what they do is quite conventional: like researchers in other areas of psychology, they look for correlations in their data and use them to account for some of the variation in the measured outcome.
Traditional developmentalists count their research a success if they find a correlation between the outcome variable they are interested in and one of the environmental variables they’ve measured. If they find a correlation of .30 between children’s aggressiveness and how often they were spanked, it would mean that the researchers have, in a technical sense, “accounted for” some of the variation in aggressiveness (though, as I said, the correlation doesn’t prove that spanking caused the aggressiveness). A correlation of .30 accounts for 9 percent of the variance—you square the correlation to get the proportion of variance accounted for.
Correlations are used in a somewhat different way by behavioral geneticists but the principle is the same. They look at pairs of subjects who vary in genetic or environmental similarity and use that variation to account for the variation in the outcome. To the degree that subjects who are more similar genetically are more similar in aggressiveness, they’ve accounted for some of the variation in aggressiveness. To the degree that subjects who have more similar environments are more similar in aggressiveness, they’ve accounted for some additional variance. Instead of using an environmental variable—spanking—to account for variance, they’re using similarity in genes and similarity in environment.
But what the behavioral geneticists are doing with the pie is no different from what the developmentalists have been doing for so long: they’re slicing off whatever portions of the variance they can explain by the correlations in their data. The main difference is that the behavioral geneticists can take a much bigger slice, which naturally has aroused some resentment. The developmentalists have to content themselves with morsels; to them, 9 percent is a feast. But the behavioral geneticists help themselves to a whopping wedge—the 45 percent of the variance that can be attributed to the effects of being mo
re or less similar genetically.23 If the effects of sharing or not sharing a home are detectable in their data (in most cases they are not), they can take another little sliver.
But they can’t have the whole pie. A little more than half still remains after the behavioral geneticists have taken all the slices their data entitle them to. What they’ve done in their published reports is to attribute this portion to a mysterious perpetrator they’ve dubbed “the nonshared environment.” I suspect they would have had fewer public relations problems if they had simply admitted their ignorance and left the uneaten portion on the plate.
The 55 percent of the variance in behavior or personality that the behavioral geneticists attribute to the nonshared environment is variance they can’t account for, variance they don’t know how to explain. It is not genetic, so by default they call it environmental. A more accurate name for it is “unexplained variance,” and that is what I will call it. The unexplained variance is the variance that remains after all the correlations—all the measured similarities between siblings—are taken out. It includes something called “measurement error”: imprecisions in measuring the outcome variable. But mainly it represents differences in personality—real differences—among the subjects who took part in the study. Not just differences between siblings but differences among all the subjects. All the variation that can’t be explained by the differences in the subjects’ genes or in the homes in which they grew up.
I don’t want you to get the idea that the unexplained variance is a mathematical abstraction. On the contrary, it’s plainly visible to the naked eye. The poster children of the unexplained variance are reared-together identical twins: two people with identical genes who grew up in the same home at the same time, and yet they are different. Even conjoined twins who live their entire lives together, like Laleh and Ladan Bijani and the famous “Siamese twins,” Chang and Eng, have different personalities.24
No Two Alike Page 6
