No Two Alike

by Judith Rich Harris

  Joan and I never found any evidence, in the published works of Suomi and his colleagues, that the cross-fostering experiment had ever progressed beyond the small “first cohort” of baby monkeys. If thirty-six monkeys were in fact cross-fostered, I suspect that the results from the larger N didn’t support the findings described in the 1987 chapter. Anyone who has done research knows how risky it is to get excited about preliminary results. The drug that seemed so promising when given to fifty patients may fail to live up to its promise when given to five hundred. It’s the law of small numbers and it means you shouldn’t bet the ranch on early results.

  In July 2000, I informed Eleanor Maccoby and Andrew Collins of the results of the investigation that Joan and I had carried out. I told them that the claim they made in their American Psychologist paper, that high-reactive baby monkeys reared by calm foster mothers “develop normally and indeed rise to the top of their dominance hierarchy,” was based on evidence that appeared to involve fewer than eight cross-fostered monkeys and that had been published only in chapters in edited books. I also pointed out a discrepancy between what they reported in their article and what Suomi had said in his chapters. Collins and Maccoby had claimed that high-reactive babies reared by calm foster mothers did well after they were separated from their mothers, but what Suomi had been saying since 1987 was that high-reactive babies reared by nurturant mothers did well.

  I never heard back from Collins, but Eleanor Maccoby replied. She was apologetic about having confused calm foster mothers with nurturant ones. For the other things she said about Suomi’s monkeys, she told me, she had relied mainly on one of Suomi’s book chapters, one of the ones I had read. “True,” Maccoby admitted, “the chapter gave us only a summary, not the numbers we are all eager to see when the new monograph becomes available.”29

  Are you familiar with the term “vaporware”? It’s a word used to describe a product announced and promoted by a software company, but that hasn’t yet been delivered and probably never will be. Suomi’s monograph proved to be vaporware. He never sent me a copy of it and, as far as Joan and I have been able to determine, it was never published. We have searched repeatedly and never found an account of the cross-fostering experiment containing the details that would be required for acceptance by a journal.30

  Two things of interest did appear in print, however. The first, in 2002, was a book based on the NICHD conference, Parenting and the Child’s World. It contains a chapter by everyone who participated in the conference—a written version of their talk—and naturally I was curious to see what Suomi would say in his chapter. By the time he was preparing the final draft he must have known, not only what I had discovered about his research, but also that I had been informing others of my discoveries. What would he do? Would he repeat what he said at the conference? Nope, he backed down. Those high-reactive baby monkeys who had played such a prominent role in his talk—the ones who rose to the top of their dominance hierarchy because they were reared by the right kind of foster mothers—do not appear in Suomi’s chapter.31 They had been left on the cutting room floor.

  The second thing was an article by another primatologist, Dario Maestripieri of the University of Chicago, which appeared in 2003 in the journal Developmental Psychobiology. Maestripieri reported the results of a cross-fostering experiment with baby rhesus monkeys, the same species Suomi had studied and pretty much the same method. But Maestripieri had followed his subjects for three years (Suomi’s monkeys were only fifteen months old when he summarized his findings in the 1987 chapter), and Maestripieri’s results were quite different from Suomi’s—in fact, opposite to them. He made quantitative observations of the young monkeys’ social behavior and reported that “no clear behavioral similarities between offspring and foster mothers were observed at any age” in fact, the offsprings’ social behaviors “resembled those of their biological mothers.”32 Their biological mothers, not their foster mothers.

  But what really clinched it was what Maestripieri said about his N. He had to report it, of course, since this was a journal article. Only ten baby monkeys took part in the study, and Maestripieri explained why there weren’t more. “Cross-fostering experiments are difficult to perform because [monkey] mothers adopt unrelated infants only under very restricted circumstances,” he said. “As a result, very few studies have used this procedure to investigate the intergenerational transmission of behavior, and these studies typically involved very small sample sizes.”33 Here he cited Suomi’s 1987 chapter—the only work by Suomi he referred to.

  Primatologists studying rhesus monkeys keep informed about one another’s work. If Suomi had really run thirty-six baby monkeys in a cross-fostering experiment, I’m sure Maestripieri would have known about it. What Maestripieri was suggesting was that his ten baby monkeys might not be much, but they were more than Suomi’s “very small sample.”

  Perhaps you’re wondering why I’m making such a fuss about a bunch of monkeys. After all, they’re not humans; they’re not even apes. Why should I care whether Suomi’s evidence on cross-fostering in monkeys was or wasn’t genuine? Why should I care if the monkey mothers whose foster children allegedly turned out well were nurturant or calm?

  I’ll tell you why. The nurturant monkey mothers, as Suomi described them in his 1987 chapter, sound very much like the kind of human mothers sometimes described as “overprotective.” The nurturant monkeys and the overprotective human mothers both spend a lot of time holding their babies; both are relatively tolerant of disobedience. This kind of indulgent mothering has been reported (by Suomi) to have good effects on high-reactive infant monkeys, but it has been reported (by the developmentalist Jerome Kagan) to have bad effects on high-reactive infant humans.34 And these conflicting results have both been used as evidence that I’m wrong: that mothering style does have important effects on how the offspring turn out, but that the effects come in the form of an interaction—only high-reactive infants show these effects.

  High-reactive infant humans (also called timid, fearful, inhibited, or shy) featured prominently in a Newsweek cover story about The Nurture Assumption. In that article, I was accused of “ignoring studies that do not fit [my] thesis,” and a study by Kagan was presented as “Exhibit A” among the studies I had allegedly ignored. Here’s how it was described:

  Exhibit A: the work of Harvard’s Kagan. He has shown how different parenting styles can shape a timid, shy child who perceives the world as a threat. Kagan measured babies at 4 months and at school age. The fearful children whose parents (over)protected them were still timid. Those whose parents pushed them to try new things—“get into that sandbox and play with the other kids, dammit!”—lost their shyness. A genetic legacy of timidity was shaped by parental behavior, says Kagan, “and these kids became far less fearful.”35

  The evidence does sound impressive. These were not short-term results; according to Newsweek, the children in this study had been followed to school age. The fearful babies whose parents had refrained from overprotectiveness had lost their timidity by school age.

  Your question now should be: Have these results been published in a peer-reviewed journal? Funny you should ask; the answer is no. Joan and I found no evidence that the study described in Newsweek was ever done or, if it was done, that it produced the reported results. It wasn’t just that it hadn’t been published in a peer-reviewed journal—as far as we could determine, it hadn’t been published anywhere.

  What had been published was a preliminary report. In his book Galen’s Prophecy, which came out in 1994, Jerome Kagan described a study carried out by his student Doreen Arcus. It was the research she did for her 1991 doctoral dissertation at Harvard. Arcus studied twenty-four babies who were judged high-reactive on the basis of a test given when they were four months old, and twenty-six babies who were judged low-reactive. She found that the high-reactive babies whose mothers were relatively indulgent—who picked them up and held them when they cried and who were relatively wimpy about setting lim
its when they began to crawl—were more likely to be fearful at fourteen months. At twenty-one months, the high-reactive babies who were still fearful tended to be those whose mothers said that they didn’t particularly value obedience. The low-reactive babies were unlikely to be fearful at fourteen or twenty-one months regardless of their mothers’ child-rearing practices or philosophies.36

  Galen’s Prophecy summarizes fifteen years of research on fearful children by Kagan and his students. Doreen Arcus’s unpublished dissertation research—twenty-four high-reactive babies followed to the age of twenty-one months—is the only evidence Kagan gives in the book to support his conviction that the way parents raise their children plays a role in whether or not the children are fearful.

  The problem here is not the N: it’s that twenty-one months is not school age, and the Newsweek article had said that the babies were measured “at 4 months and at school age.”

  In 2001, three years after the Newsweek report, a chapter by Arcus appeared in an edited book. The chapter is about biological and social influences on temperament, and the stability of temperament during development. In it, Arcus refers to a “longitudinal study of 98 infants” and reports data on stability of temperament for these subjects from the age of four months to four and half years. (At four and a half years, the children who had been judged high-reactive at age four months were, on average, less chatty and outgoing than those who had been judged low-reactive.) Clearly, these children had been followed at least to the age of four and a half years. And yet when Arcus describes her research on the relationship between mothers’ child-rearing style and children’s fearfulness, she reports results only up to the age of fourteen months.37

  Promising preliminary results often fail to hold up when more subjects are added to a study. The same thing can happen when subjects are followed for a longer period of time.

  Assume, for a moment, that Newsweek’s description of the research by Kagan and his students was correct, and that correlations between mothers’ child-rearing style and children’s fearfulness could still be detected at school age. What would that show? The findings for fourteen and twenty-one months are presented in Galen’s Prophecy as evidence of a gene-environment interaction: an interaction between a particular kind of infant temperament, assumed to be genetic, and a particular mothering style, part of the infant’s environment. Babies who are born with a genetic predisposition to be fearful can, according to Kagan, avoid this fate if they are reared in the right kind of environment.

  Kagan and his students used a test given at four months to classify babies as high-reactive or low-reactive. This is a method often used by developmentalists: instead of using twin or adoption studies as a way of controlling for genetic predispositions, they use some kind of pretest. They test or observe a bunch of children at an early age, look at them again at a later age, and assume that any changes between Time 1 and Time 2 must be due to environmental factors. They assume that genetic effects show up early, like inherited money, and that environmental effects show up later, like compound interest.

  But genetic effects can show up at any age. There are genes that kick in early and genes that bide their time; this is one of the reasons why you become more like your parents as you get older. Genes that determine whether a female will grow large or small breasts at puberty don’t begin to reveal their intentions until the biological clock strikes twelve. Genes that determine male baldness work on an even more leisurely schedule. You can’t tell by looking at a baby how much hair he’s going to have in middle age. Counting the hairs on the baby’s head is not a good control for genetic influences on hairiness at age forty.

  The environmental side of the alleged interaction is equally problematic. First of all, the way a mother takes care of a baby is in part a reaction to the baby’s behavior. A baby destined to be fearful may be sending out signals that Kagan’s four-month test misses but that get through to its mother. The mother might respond by being more protective.

  Second, a mother’s style of caring for her baby is a function, not only of the baby’s characteristics, but also of her own.38 Mothers who hold their infants a lot, especially when they cry, and who are wimpy about enforcing discipline sound to me like anxious mothers. Anxious mothers may have fearful children (and fearful children may have anxious mothers) for genetic reasons. The heritability of personality traits is about .45. A child who is anxious or fearful may have inherited this characteristic from one or both of his parents.

  It’s an odd thing. Though many developmentalists are now willing to admit that babies differ from one another at birth and that these differences are largely genetic, they still haven’t come to terms with the fact that babies get their genes from their parents. They still haven’t figured out how to deal with the fact that babies and their parents are likely to resemble each other in personality for genetic reasons alone. One sign of this unwillingness to face facts is their avoidance of the word “heredity.” They tend to use euphemisms such as “genetic” and “nature” (as in “nature and nurture”), which acknowledge that children have genes without acknowledging where they get them from.

  Behavioral geneticists, who are keenly aware of where children get their genes, study adopted children for this very reason. Adoptees get their genes from one set of parents and their home environment from a different set. Suomi’s cross-fostering study of monkeys is an adoption study; the idea is that you can do things with monkeys that you cannot do with humans. But even though high-reactive human babies aren’t systematically placed with calm or nervous adoptive mothers, or with nurturant or punitive ones, we have enough data from human adoptions to know that the adoptive mother’s personality and her philosophy of child-rearing have no long-term effects on the personality of her adopted child. The near-zero correlation between adoptees’ personalities and those of their adoptive parents means that adoptees reared by nervous adoptive mothers are, on average, no more anxious or fearful than those reared by calm ones.39

  Reports of Kagan’s timid babies and Suomi’s nervous monkeys have traveled beyond the academic world into publications aimed at a general audience. Suomi’s cross-fostering experiment was recently described, for example, in a book titled Liars, Lovers, and Heroes, by neuroscientists Steven Quartz and Terrence Sejnowski. Here’s the description these authors gave of Suomi’s experiment:

  Suomi, a primatologist at the National Institute of Child Health and Human Development, selectively bred monkeys for both inhibited and bold temperaments. Then, he rearranged their rearing environment to explore how environmental factors and temperament interact, all while tracking their neurotransmitter levels. In Suomi’s experiments, a fearful infant monkey was put in the care of an uninhibited, nurturing foster mother…. The young monkey became less fearful, and its levels of noradrenaline dropped. Regarding temperament, Suomi notes, “Our work shows that you can modify these tendencies quite dramatically with certain types of early experiences.”40

  No reference was given in the endnotes for the experiment or for the quote from Suomi, so I wrote to the authors and asked for their source. Terry Sejnowski replied. He said the description of the procedure and results came from a talk Suomi had given at his institute a few years earlier, and that the quote came from a 1997 article in Discover magazine.41

  I looked up the article. What Suomi was referring to when he told the Discover journalist “You can modify these tendencies quite dramatically with certain kinds of early experiences” was not the cross-fostering experiment Quartz and Sejnowski described in the passage I quoted from their book: Suomi was talking about peer-reared versus mother-reared monkeys. But later on in the same article the Discover journalist did refer to the cross-fostering experiment:

  Suomi…has shown that even monkeys who are born anxious and inhibited can overcome their temperamental handicap—and even rise to the top of the dominance hierarchy in their troop—if they are raised by ultra-nurturing super-moms. Kagan’s work confirms that mothering can alter the cours
e of an inhibited child’s development.42

  The Discover journalist did not point out that “ultra-nurturing super-moms” are alleged to alter the inhibited monkey’s development one way and the inhibited child’s development in the opposite way. Nor did she mention that both results were evidently based on small samples followed for an insufficient length of time and that neither has been published in a peer-reviewed journal, despite the fact that years have gone by since the preliminary reports.

  A good story—especially one that appears to confirm what people already believe—acquires a life of its own. Once it becomes folklore it is highly resistant to disproof.43 I am showing you how these myths get started. Suomi gives talks about his cross-fostered monkeys; Kagan tells journalists about his overprotected kids. The stories, perhaps improved a bit by the journalists, get into the popular press—the Discover article, the cover story in Newsweek—and then are repeated, with further improvements, by other writers. When you can no longer remember where you heard or read something, you are unlikely to question its accuracy. It’s just something you know.

  In 1951, when Josephine Tey’s The Daughter of Time was published, every schoolchild in Britain could tell you that Richard III was a heartless villain. Tey spent more than two-thirds of her book trying to persuade her readers that what everyone knew wasn’t true. The case against Richard III, she showed, had been trumped up. The histories of Richard’s brief reign—one of which was the basis for Shakespeare’s play—had been written by people who had something to gain from the posthumous trashing of Richard’s reputation. They were allies of his enemies.