Human Diversity
Page 26
A Simple Model
Now for the nuts and bolts of calculating the heritability ratio for narrow heritability, h2. Let’s return to the example of height. For the denominator, the calculation is trivially simple: Measure the height of the people in your sample and do the arithmetic shown in Appendix 1 to calculate variance. The problem comes when you try to estimate the numerator.
By definition, all the variance in a given trait in a population must come from some combination of genes and environment. The simplest model used to operationalize that combination (I will get to the complications later) assumes that total population variance (VP) consists of the additive effects of three components:
VP = A + C + E
In the technical literature, it is commonly referred to as the ACE model. The letter A refers to additive genetic variance. The letter C originally referred to common environment, now customarily called shared environment. Plausible candidates for the shared environment are such things as parental income, occupation, education, age, parenting practices, family structure, the quality of the neighborhood, and the quality of the schools.[8] The letter E originally stood for error but now refers to a combination of two things: measurement error and environmental influences that twins do not share (and that make them different from one another), known as nonshared environment.
Both aspects of the nonshared environment need some explanation. Measurement error bedevils all the sciences but is especially troublesome in the social sciences. The error can consist not only of inaccurate measurement (e.g., family income is misreported), but also of the gap that separates the construct (what the researcher wants to measure) and the indicator (what the researcher is actually measuring). For example, the number of books in the home is an indicator used to measure the construct “environment for stimulating intellectual development.” But such an environment is far more complicated than can be represented by a count of books. Even if the count of books is technically accurate, the indicator will have a lot of error as a representation of the construct. Because of this, the common practice of referring to E as the nonshared environment without mentioning measurement error is seriously misleading.
But some portion of the E component of the ACE model is likely to reflect environmental influences that siblings do not share. In 1987, Robert Plomin, one of the leading students of the nonshared environment, suggested five unshared sources of such phenotypic differences.9 One is a catch-all nonsystematic category (e.g., accidents, illnesses, trauma). Four others are systematic: family composition (e.g., birth order, sex differences), sibling interaction (e.g., differential sibling jealousy), parental treatment (e.g., differential maternal affection), and extrafamilial networks (e.g., local peer groups, teachers, social media).10 Actually demonstrating an important role for these potential components of the nonshared environment has proved to be frustrating, as I will discuss in chapter 13. For now, it is enough to recognize that the reality of the nonshared environment is not surprising. If you have siblings, think of all the ways in which you and they had different experiences growing up that were unrelated to your parents’ socioeconomic status or to their common parenting practices. You are probably thinking about differences generated by the nonshared environment.
The Analytic Power of Comparing MZ and DZ Twins
This brings us to the unique advantages of studying MZ twins and DZ twins:
MZ twins share virtually 100 percent of their genes.11
DZ twins share approximately 50 percent of their genes.12
Skipping the algebra that lies between, the first two of those advantages mean that you can solve for A, C, and E in the ACE model. Geneticist Douglas Falconer developed the equations for doing so. They look like this:
A = 2(RMZ − RDZ)
C = 2RDZ − RMZ
E = 1 – (A + C)
Put in words, narrow heritability equals twice the difference between the correlations of the samples of MZ and DZ twins. The shared environmental contribution equals twice the correlation among DZ twins minus the correlation among the MZ twins. The nonshared environment equals 1 (the total variance) minus the sum of heritability and the shared environment. The note gives an explanation of how this bottom line is reached.[13] In the case of the height example, suppose that the correlation for height of our sample of DZ twins is +.55 and the correlation for our sample of MZ twins is +.95.
A = 2 × (.95 – .55) = .80
C = 2 × .55 – .95 = .15
E = 1 – (.80 + .15) = .05
All full siblings share about 50 percent of their genes. The advantage of using DZ twins instead of full siblings who are not twins is that twins are born at the same time. Siblings born several years apart can be born into radically different environments depending on what’s happened to the parents’ marital relationship, jobs, income, or the location of their home in the meantime. None of those or many other important environmental forces are likely to vary objectively for twins, even though the twins may react to them differently.
TWINS RAISED APART
For many people, “twin studies” brings to mind the famous Minnesota Study of Twins Reared Apart.14 It got so much publicity because it produced so many dramatic examples of similarities in adults who had never met each other. When the separated twin brothers who inspired the study were reunited in their 30s, it was discovered that as children both had had a dog named Toy. As adults, both had been married twice, first to wives named Linda and then to wives named Betty. They had independently taken family vacations to the same three-block strip of Florida beach, both driving light blue Chevrolets from their homes in the Midwest. Both smoked Salems, both had worked part-time in law enforcement as sheriffs, and both had a habit of scattering love notes to their wives around the house.15
Studies of twins raised apart—the Minnesota study was one of several—produced valuable information, but the method’s potential was limited. Separation of identical twins at birth happens so seldom that large sample sizes are impossible. The range of environments in which separated twins are raised is narrow—adoption agencies don’t knowingly place infants with impoverished or dysfunctional parents.16 In contrast, it is not difficult to assemble large samples of twins who have been raised together. Their home environments span the range.
The Validity of Twin Studies
The ACE model makes a strong claim: It can disentangle the roles of nature and nurture. You will not be surprised to learn that many challenges to the validity of that model have been mounted.
The logic I have just presented entails five primary assumptions. Three of them, discussed in the note, involve fewer problematic issues.[17] Two of the assumptions are at center stage in the debate over the validity of twin studies:
Humans mate randomly (no assortative mating).
DZ and MZ twins experience their common environments equally, known in the literature as the equal environments assumption (EEA).
The Random Mating Assumption
The Falconer equations assume that DZ twins share on average 50 percent of their genes, which in turn depends on their parents having mated randomly for any given trait. When this assumption is violated, the statistical estimate of heritability will be too low.18
To see why, suppose that height is 100 percent heritable but that people mate randomly relative to height. In the Falconer equations, the role of shared environment equals twice the correlation among DZ twins minus the correlation among the MZ twins. If true heritability is 100 percent, the MZ twin correlation for height is expected to be +1.0 and the DZ twin correlation is expected to be +.5. Falconer’s calculation of the shared environment (C) would be 2 × .5 – 1.0 = 0, which is the correct answer.
What happens if the assumption is wrong? Let’s say that people tend to marry others who are in the same part of the distribution of height (e.g., shorter-than-average men tend to marry shorter-than-average women). The higher that correlation between the heights of the parents, the more that DZ twins resemble each other over and above the d
egree that would be predicted by their shared genes, but for a reason that has nothing to do with the environment. Suppose that the assortative mating increased the observed DZ correlation to +.6. In that case, the MZ twin correlation is unchanged at 1.0, but the Falconer formula would determine that the value of the shared environment, C, is 2 × .6 – .6 = .6, or 60 percent, which is inflated.
In the real world, assortative mating is routine. At least when it comes to marriage, people tend to marry others who are similar on a wide variety of traits. The empirical reality of that statement has been established for a long time, beginning with Steven Vandenberg’s review of the early literature in 1972.19 Since then, extensive additional research has documented assortative mating for education, intelligence, political affiliation, mental illness, substance abuse, aggressive behavior, and criminal behavior. Often these correlations are substantial, in the region of +.4 to +.5.20 The expectation must be that assortative mating leads to consistent though modest underestimates of A in the ACE model.
The Equal Environments Assumption
On average, parents of MZ twins treat them more similarly than do parents of DZ twins. As John Loehlin put it, “identical twins are indeed treated more alike—they are dressed alike more often, are more often together at school, play together more, and so forth.”21 But the same scholars who found such differences in treatment of MZ and DZ twins also found, in the words of Devon LoParo and Irwin Waldman, that “the presence of higher levels of physical and environmental similarity in MZ twins than in DZ twins is a violation of the EEA only if these aspects of the environment are etiologically relevant to the phenotype of interest.”22 LoParo and Waldman found instead that it doesn’t seem to make any causal difference that the environments of MZ twins are more similar in some respects than the environments of DZ twins. A few exceptions exist,23 but such is the finding of the bulk of the literature in twin studies of intelligence, personality, schizophrenia, eating attitudes and behaviors, major depressive disorder, generalized anxiety disorder, phobias, parent-child interactions, hyperactivity, ADHD, PTSD, social attitudes, aggression, alcoholism, and externalizing behaviors.[24] A comprehensive 2014 evaluation of the EEA by Jacob Felson found that controlling for environmental similarity reduced heritability significantly for just one out of 32 outcomes. He concluded, “All things considered, it seems unlikely that the EEA is strictly valid, but it also seems likely that violations of the EEA are relatively modest.”25
To sum up: Twin studies have come under criticism for overstating the role of genes. The reality is that violations of the random mating assumption are common and lead to modest understatement of the role of genes, whereas violations of the equal environments assumption have even more modest effects in the other direction and are uncommon. Overall, heritability as estimated by twin studies appears to be accurate, with errors tending on net to slightly underestimate heritability rather than overestimate it.[26]
Recapitulation
To understand the causal dynamics that lead to financial and professional success for some people and not for others, and thereby ultimately determine socioeconomic classes, it is essential to understand the comparative roles of environment and genes. Disentangling those roles with ordinary samples of people is extremely difficult and subject to endless disputes. Even before the genome was sequenced, twins offered a uniquely powerful solution. By comparing MZ twins and DZ twins, it is possible to determine the proportional roles of genes and of the environmental conditions shared by children in the same family. Armed with this method, scholars have conducted thousands of twin studies. It’s time to look at what they have found.
11
The Ubiquity of Heritability and the Small Role of the Shared Environment
Proposition #8: The shared environment usually plays a minor role in explaining personality, abilities, and social behavior.
This chapter has two purposes. The first is to establish that all cognitive repertoires are heritable to some degree and that the ones most likely to affect success are substantially heritable. You are free to apply your own definition of substantial to the many numbers I present. My own interpretation is that substantially heritable doesn’t have to mean more than half but it should be at least a third. My second purpose is to ask you to think about environmental factors from a new perspective. The conventional perspective is that the way parents raise a child makes a big difference for all sorts of important qualities. Send children to really good schools, and it will make a big difference in how well they perform intellectually. The right kind of parenting can foster self-discipline and grit, or, for parents with different priorities, foster their children’s creativity and free-spiritedness. Proposition #8 says “Not really.” The environment in which a child grows up usually makes a difference, but seldom because of the things that parents can control.
The Unexpected Story of the Shared Environment
In the decades since the ACE model was developed, thousands of twin studies have been published. As early as 2000, the pattern of results had been so consistent and so striking that behavior geneticist Eric Turkheimer was led to set out the three laws of behavior genetics:
First Law. All human behavioral traits are heritable.
Second Law. The effect of being raised in the same family is smaller than the effect of genes.
Third Law. A substantial portion of the variation in complex human behavioral traits is not accounted for by the effects of genes or families.1
In the two decades since Turkheimer stated them, no one who accepts the validity of twin studies has found reason to dispute them.
Turkheimer’s first law is the least interesting of the three. How many people are really surprised to learn that all human traits are heritable to some degree? How can anyone who has been a parent be surprised?2 The second law is the most provocative. Turkheimer stated it circumspectly, but the typical finding is starker than his wording indicates. It’s not just that the role of the shared environment is less than that of genes; that role is usually small, especially with regard to the child’s eventual cognitive repertoires as an adult.
This story has been known in broad outline within the behavior genetics community since the 1980s. It was first exposed to a general audience in 1998 when Judith Harris published The Nurture Assumption.3 It got wider attention in 2002 when Steven Pinker recounted it in his bestseller The Blank Slate.4 It has subsequently been referenced in many magazine and book-length discussions of parenting.5
At first glance, the claim that parental socialization doesn’t make much difference is counterintuitive. The family environment, including socioeconomic status (SES), must surely have a major influence on children’s outcomes. That’s why psychologists John Loehlin and Robert Nichols of the University of Texas were bemused by the data on 850 twin pairs that they analyzed in the early 1970s. They unveiled their unexpected finding in a book titled Heredity, Environment, and Personality, published in 1976.
Thus, a consistent—though perplexing—pattern is emerging from the data (and it is not purely idiosyncratic to our study). Environment carries substantial weight in determining personality—it appears to account for at least half the variance—but that environment is one for which twin pairs are correlated close to zero.… In short, in the personality domain we seem to see environmental effects that operate almost randomly with respect to the sorts of variables that psychologists (and other people) have traditionally deemed important in personality development.6
In 1981, psychologists David Rowe and Robert Plomin named this mysterious source of variation the nonshared environment, introduced in chapter 10.7 A year later, psychologists Sandra Scarr and Susan Grajek put the expanding evidence into language no one could fail to understand:
Lest the reader slip over these results, let us make explicit the implications of these findings: Upper-middle-class brothers who attend the same school and whose parents take them to the same plays, sporting events, music lessons, and therapists, and use similar chil
d rearing practices on them are little more similar in personality measures than they are to working class or farm boys whose lives are totally different. Now, perhaps this is an exaggeration of the known facts, but not by much.8
The findings that Scarr and Grajek described in 1982 have subsequently been confirmed and reconfirmed. The establishment of the truth—for truth it seems to be—that the childhood family environment explains little about the cognitive repertoires of the adult is one of the more important achievements of the social sciences in the last four decades.9 Do not understand this truth too quickly, however. Three clarifications need to be kept in mind.
First, I am using the word “explain” in its statistical sense, not an explicitly causal sense. Proposition #8 says that even though you may think that your parenting style and your family’s resources make a big difference in how your children turn out as human beings, using a straightforward model for identifying that effect fails to turn up evidence for it.