No Two Alike

by Judith Rich Harris

  But the developing brain isn’t fussy. As long as the visual stimulation, or the mother, or the language, fall within the range normally encountered by the species, the fine-tuning proceeds according to plan. Only highly abnormal conditions—conditions of severe deprivation—cause permanent deficits. The environment needs only to provide the bare minimum; beyond that minimum, there is no evidence that variations in quality or quantity make a difference.29 Babies who get to look at pretty pictures and intricate mobiles do not develop better visual systems than those reared in bare huts.

  Nor is there evidence that these early experiences cause the brain to be permanently disposed to expect more of the same. Toddlers whose first language is Polish or Korean or sign language are not handicapped if they later discover that the language of their community is spoken English. The hearing children of deaf parents generally learn sign language first, but they soon become competent speakers of the language spoken outside the home. Which language is learned first doesn’t matter in the long run, just as long as the child learns some language. Children are open to the possibility that the language used outside the home may differ from the language they learned from their parents.30

  Attachment appears to work the same way. As long as the infant manages to form some attachment in the first few years of life, development proceeds according to plan. If the first relationship doesn’t work out well, the child is open to the possibility that the next one will be better; he doesn’t expect all people to behave like his mother. Isn’t this the way you would design the baby, if it were up to you? If a child happens to have an abusive or neglectful mother but nonetheless manages to survive, why should the poor kid be burdened with additional handicaps? The goal, from an evolutionary viewpoint, is to minimize the damage caused by early adversity.

  I said at the beginning of this chapter that all the currently popular theories of personality development are based on the assumption that children learn something in one environmental setting or with one social partner, and that this learning subsequently affects the way they behave in other places and with other people. What makes it hard to test this assumption is that genetic influences, too, can produce similarities of behavior across contexts. Thus, a convincing test would have to provide some way of eliminating or controlling for the misleading effects of genes.

  I’ve already described two ways to do this. The first is to use behavioral genetic methods. The behavioral genetic evidence provides no support for the assumption that patterns of behavior learned at home affect behavior or personality outside the home. Nor does it confirm the predictions of attachment theory. Two people—even identical twins—who were cared for from birth by the same mother are no more alike in personality than two who had different mothers.

  The second way is to do a birth order study. In the previous chapter I described the strengths of this method: the fact that there are no systematic genetic differences between firstborns and laterborns but there are systematic differences in their experiences at home. The evidence from birth order studies does not support the notion that patterns of behavior learned at home are generalized to other contexts: firstborn children are more aggressive at home but not on the playground. In adulthood there are no reliable differences between firstborns and laterborns in their scores on personality tests.

  A third way is to look at the effects of context on a social behavior that is not influenced by genes. My favorite example is language. Though its acquisition is the product of a specialized mechanism in the human brain, once language is acquired it becomes a social behavior, sensitive to social context. Unlike most other social behaviors, however, it is free of genetic influence: differences in dialect are not due to differences in genes. Children are not born with a predisposition to speak English or Polish or Korean, or to use the accent of a particular region or social class. These things are entirely learned.

  When we eliminate genetic influences by looking at language and accent, the apparent carryover of behavior from one context to another vanishes. A child who spoke Polish for the first few years of his life, and who continues to speak Polish at home with his parents, will speak unaccented English outside his home if he grows up in a neighborhood of native-born English-speakers.31 The children of immigrants do not blend their two language contexts; they do not end up with some sort of a compromise. No trace of their parents’ foreign accent leaks out into the language they use with their peers.32

  That this phenomenon has something to do with the regulation of social behavior is shown by what happens if something goes wrong with the ability to respond normally to social inputs. Simon Baron-Cohen, the British cognitive scientist who studies autism, found that autistic children reared by immigrant parents do not lose their parents’ foreign accents.33

  The example of language also shows that consistencies in behavior across contexts are not invariably due to genetic influences. Most children who grow up in English-speaking countries speak English both at home and at school—not because they inherited a predisposition to speak that language but because experience has taught them that it works in a variety of contexts. Their behavior (speaking English) is similar in different contexts because their experiences (hearing people speak English, being understood when they speak English) have been similar in different contexts.

  Some behaviors, such as speaking English, work in a variety of contexts. Others, such as being aggressive, work in some but not in others. The need to be aware of social context is as pressing in childhood as it is in adulthood. Nowadays many parents teach their children that it’s good to express emotions. If you feel like crying, go ahead and cry. If you need a little loving, throw your arms around me. All very well and good at home, but if you take these behaviors outside with you, they could get you into trouble. The child who cries too much in school is likely to be teased or avoided. To say nothing of the one who throws his arms around people willy-nilly.

  A fourth way to test the assumption that behaviors learned at home generalize to other contexts is to do an experiment.

  Most research in developmental psychology is not experimental: it is observational. Typically, developmentalists make use of preexisting, spontaneous differences in child-rearing practices among the parents in their study. The problem is that a parent’s child-rearing practices are not chosen at random. Parenting behavior is, in part, a response to the child’s behavior, which is influenced by the child’s genes. In part it is a function of the parent’s own personality, which is influenced by the parent’s genes.34 Thus, families in which the parents use a harsh child-rearing style may differ genetically from those in which the parents use a more permissive style. They may also differ in other ways—in socioeconomic class, for instance.

  A similar problem exists in medical research. Medical researchers know that studies in which subjects make their own decisions about treatments may produce misleading results. That’s why the “gold standard” of medical research is the randomized control trial, in which subjects are randomly assigned to the experimental group or the control group.35 Random assignment ensures that the only way the two groups of subjects differ systematically is in whether or not they are getting the treatment being tested. All the other relevant ways in which medical subjects can differ—how healthy they are to begin with, how cautious or impulsive in making health-related decisions, and so on—will average out in the two groups, if the N is sufficiently large. So the two groups will be approximately equal at the start of the experiment.

  In developmental psychology, the equivalent of the randomized control trial is an intervention study. In a well-done study of this type, parents are randomly assigned to an intervention group or a control group. Those in the intervention group are taught better ways of dealing with their children—that is, ways that the researchers believe are better. Then the researchers look for effects of the changed parental behavior on the children’s behavior.

  If the intervention is successful in changing the way the parents behave at
home, then the researchers are likely to find changes in the way the children behave at home. The question is, Will these changes in the children’s behavior transfer to other settings? Will a child who is less troublesome at home also be less troublesome at school?

  In my talk at the NICHD conference—the conference I told you about in chapter 3, the one in which the primatologist Stephen Suomi accused me of ignoring interactions—I went out on a limb. I said that my theory predicts that interventions designed to improve parents’ child-rearing methods might change children’s behavior at home but will not affect their behavior at school.36

  No one stood up in the question period and told me I was wrong. Instead, three very nice interventionists came up to me afterwards and told me politely that there was lots of evidence that interventions aimed at parents can improve children’s behavior in school.

  In fact, one such study was presented at the conference that very afternoon, by Philip Cowan and Carolyn Pape Cowan, a husband-and-wife team at the University of California at Berkeley. Philip is the director of Berkeley’s Institute of Human Development. The message the Cowans delivered to a receptive audience was that an intervention study can prove that parenting matters.37

  To my regret, I was unable to attend the Cowans’ presentation—I had to skip some of the sessions and spend the time resting in my hotel room—but afterwards I wrote to Philip Cowan in e-mail. I requested a copy of his talk and he promptly sent it to me. He also proved willing to answer my questions in e-mail. He was motivated, I believe, by a sincere desire to convince me of the error of my ways. Interventionists, I found, are a lot like the people who want to convert you to their religious beliefs so that you won’t go to hell.

  The paper Philip Cowan sent me began with an excellent critique of conventional methodology in developmental psychology, explaining why correlational data—even from studies that follow children over periods of years—are ambiguous and cannot be used as proof of parental influence. The best way to clear up the cause-or-effect ambiguities of correlational data, Cowan and Cowan advised, is to do an intervention study. Their own intervention—a four-month program of group counseling for the parents, designed to improve the marital relationship as well as the parent-child relationship—clearly demonstrated that parental behaviors play a role in child outcomes, they said. Their results showed that if you improve the parents’ relationship with each other and with their child, you can improve the child’s performance in school. The children had higher academic achievement and fewer behavior problems in kindergarten, and better adaptation to first grade.38

  The words were impressive but I had trouble understanding their data. Cowan and Cowan used complex statistics—“latent variable path models”—to analyze their results. I looked for a straightforward comparison between the children in the experimental group and those in the control group and couldn’t find one. So I asked Philip Cowan for a clarification:

  You said…“When parents shifted in these positive ways, their children did better in adjusting to the challenges of kindergarten.” I’m not clear on what that means. Does it mean that positive effects of the intervention were found only for those children whose parents responded favorably to the intervention—only in those whose parents showed a measurable improvement in marital relations and/or parenting skills?39

  Evidently I had interpreted the results correctly. In his reply, Cowan told me that this outcome is exactly what one should expect from an intervention aimed at improving parental behavior:

  Would we expect that all kids whose parents were in the intervention to benefit? No. Just the ones whose parents improved. That’s what we expected and that’s what we found.40

  Hmm. So the parents who improved as a result of the intervention had kids who did better in kindergarten. Wait, hasn’t this analysis turned your experiment into a correlational study? What you’re doing is looking for parent-child correlations within the intervention group. In order to prove that the intervention had an effect on the children’s behavior in school, don’t you have to show that there’s a difference between the children in the intervention and control groups? Isn’t that the whole point of randomly dividing the subjects into two groups?

  When I expressed these doubts to Philip Cowan, he understandably got a bit testy. “There are dozens of intervention studies that show effects of interventions with parents on children, with measures occurring outside the home,” he said, naming several researchers. “None of them is cited in your work. This begins to raise skepticism about your openness to data that might not support your position.”41

  Ah. I am often accused of being unwilling to look at data that might contradict my position, but usually the accusers make vague reference to “dozens of studies” and then fail to name them. Philip Cowan had done me the favor of naming some of them.

  To follow up on this I needed the kind of assistance that Brent Carradine provided to the detective in The Daughter of Time, so I got in touch with Joan Friebely. Joan did a thorough search of the intervention literature and sent me articles by all the researchers Cowan had named and by many others as well. Thanks to her efforts I have several thick folders full of intervention studies.

  What a mess. In many of the studies there was no objective measure of the children’s behavior outside the home; the success of the intervention was assessed by asking the parents who had received the counseling or training to judge their children’s behavior. In some studies there was no control group. In others, the intervention took place at school as well as at home—teachers as well as parents were given training in how to deal with unruly behavior—so if the children behaved better in school you couldn’t tell if it was due to a change in the parents’ behavior or in the teachers’. I have made detailed criticisms of these studies elsewhere42 and I’m not going to do that here, but there’s one study in particular I’d like to tell you about, because the design was exemplary and the results were crystal clear.

  The report, published in a peer-reviewed journal, is titled “Parenting through Change: An Effective Prevention Program for Single Mothers.” The authors are Marion Forgatch—one of the researchers whose work Philip Cowan had advised me to look at—and David DeGarmo. The goal of the intervention was to reduce “coercive parenting” in the divorced mothers of young school-age sons; these mothers received group training sessions in “positive parenting,” followed up by reminder phone calls from the interventionists. The children’s behavior was assessed both at home and at school. There was a control group. Assessments at school were made by teachers who received no training.

  To any casual reader of Forgatch and DeGarmo’s article, the intervention appears to have been a resounding success. According to the authors’ summary, it led to an improvement in parenting behavior: afterwards, the mothers in the intervention group engaged in less “coercive parenting” than those in the control group. Furthermore, “improved parenting practices correlated significantly with improvements in teacher-reported school adjustment,” a result that led the researchers to conclude that “the intervention indirectly benefitted child outcomes through improved parenting practices.”43 This was exactly the result Philip Cowan had told me was to be expected from an intervention aimed at the parents, and it was based on the same statistical method he used: a path analysis.

  But Forgatch and DeGarmo also provided a simpler statistic, the one I had been unable to find in the Cowans’ paper: the results of a direct comparison between the children in the intervention and control groups. “Contrary to expectations,” admitted Forgatch and DeGarmo, “there were no direct effects of the intervention on child adjustment.” Based on assessments by the teachers, there were no significant differences in school behavior between children in the two groups. Those in the control group ended up as well-behaved and well-adjusted, on average, as those in the intervention group.44 In other words, an intervention focused on improving the parents’ behavior at home had failed to improve the children’s behavior at school.

>   The strength of an intervention study is the random assignment of subjects to treatment and control groups. A path analysis can produce misleading results because it circumvents the rigorous control of randomization and looks within the groups of parents to identify which ones followed the advice given by the interventionists and which ones did not. The analysis showed that the children of the parents who complied with the interventionists’ advice did better in school. But the parents who complied with the interventionists’ advice were not selected randomly. Compliant parents may differ systematically from those who do not comply with an intervention.

  I didn’t know it at the time I was corresponding with Philip Cowan, but a similar error in data analysis also crops up in medical research. The path analysis introduces what medical epidemiologists call a “compliance-determined susceptibility bias.” Here’s an actual example. A randomized trial of a drug designed to reduce blood lipids showed that patients who faithfully took the drug had lower fatality rates than those who stopped taking it or took it irregularly. Sounds like a good drug, huh? The catch was that patients who faithfully took the placebo also had lower fatality rates. So it wasn’t the drug: it was some other difference between the compliers and the noncompliers. People who comply and people who don’t comply with doctors’ instructions differ in personality and perhaps in intelligence as well. These differences cause them to do other things—other than taking or not taking the drug—that affect their life expectancy.45