In response to:
Parallel Lives from the November 30, 2006 issue
To the Editors:
Frank Sulloway writes [“Parallel Lives,” NYR, November 30, 2006] that “genes are responsible for…about 60 percent of individual differences in general intelligence.” Although this claim is frequently made by researchers in the field of behavioral genetics, I do not believe it stands up to careful scrutiny.
Estimates of the proportion of variability in general intelligence that is explained by genetics are based on three types of research—studies of identical twins raised apart, studies of biologically unrelated siblings raised in the same home, and studies comparing identical to fraternal twins. Each of these types of studies is based on one or more highly questionable assumptions.
Studies of identical twins raised apart assume that the twins were separated soon enough after birth that their common upbringing prior to separation had no appreciable effect on their IQ scores when measured later in life. But many of the subjects of these studies either lived together for a substantial period of time prior to separation or were reunited long before adulthood. In the largest study, published by J. Shields in 1962, 13 of 44 pairs lived together during at least the first year of life. Of these 13, 8 lived together until at least the age of two and 5 until at least the age of four. One pair lived together until the age of eight. Of the 31 pairs separated prior to their first birthday, 9 were reunited by the age of twelve or earlier, including 2 pairs who were reunited at the age of five.
Just as importantly, these studies assume that there is no correlation between the post-separation environments of pairs of twins—that their environments differ just as much on average as they would if the twins had been randomly assigned to households in the general population. But in reality many, if not most, pairs of twins in these studies were raised by different members of their families. A common arrangement in the Shields study was that one child would be raised by the mother or father and the other by either a grandparent, an aunt, or an uncle.
Adoption studies are equally problematic. These studies assume that adoptive families do not differ from conventional ones in any way that has a significant effect on measured intelligence. But this is by no means self-evident, since adoptive parents are both self-selected and screened by agencies. It would seem plausible that the environments provided by adoptive families are better on average and less variable than those provided by families generally. It is hard to imagine, for example, a child being adopted by an unmarried teenage girl, or by a seriously dysfunctional or impoverished married couple.
Furthermore, there is statistical evidence supporting this hypothesis. Most studies of adopted children do show much less variability in IQ scores than in the general population, indicating that adoptive children have either less genetic or less environmental variability than biological children.
That leaves studies comparing identical to fraternal twins. The main problem here is that identical twins may, on average, experience environments that are more similar than do fraternal twins. They may, for example, be more likely to share the same friends. Or they may on average spend more time together. Or they may be treated in a more similar way by parents, teachers, and others. It is known unambiguously that fraternal twins experience more similar environments than do ordinary siblings. Many studies have shown a higher correlation for fraternal twins than for ordinary siblings on intelligence and other variables, and the only possible explanation for this is environmental, since the genetic resemblance between fraternal twins is the same as that between ordinary siblings. If fraternal twins experience environments that are substantially more similar than do ordinary siblings, isn’t it plausible that the same thing is true for identical twins compared to fraternal ones?
Professor of Statistics
Frank J. Sulloway replies:
Professor Kaplan has criticized several key assumptions that underlie heritability estimates for IQ. He seems unaware, however, that these issues have all been addressed empirically during the last two decades and are no longer regarded as “highly questionable.” For example, Thomas Bouchard and colleagues, in 1990, analyzed 56 sets of identical twins reared apart and later reunited in adulthood.1 What is particularly instructive about Bouchard et al.’s study is their extensive analysis of data on the rearing environments of the twins included in their sample. Bouchard’s team assessed nine different aspects of the twins’ rearing environments, including the adoptive parents’ levels of education, the adoptive father’s socioeconomic status, parental treatment, and various cultural and intellectual resources available within the home.
Bouchard et al. found that the rearing environments of the twins were indeed moderately correlated (r = .22), confirming a potential bias in the estimation of heritability. Bouchard’s team then assessed the direct influence of each of these environmental factors on IQ, from which they were able to determine the total contribution to heritability. The maximum contribution for any one of these nine individual environmental measures turned out to be a minuscule .03, and the mean contribution for the nine measures was just .006. In other words, the heritability estimate for this set of reared-apart twins, which was .69, would be reduced to about .66 based on environmental similarities among the adoptive homes of the twins.
Bouchard and his colleagues also tested the possible contribution made to IQ by pre-separation and post-reunion contact between the twins, another of the key issues raised by Professor Kaplan. Total contact time between the twins averaged 5.1 months prior to separation and 20.3 months following first reunion, when the twins were mostly in their late twenties and early thirties. Degree of contact accounted for almost none of the similarity in the twins’ IQs (r = .14, which was not significantly different from zero).
The last of the questionable assumptions about heritability estimates involves the issue of whether identical twins experience environments that are more similar than do fraternal twins. This issue is important because heritability is sometimes calculated as twice the difference between the correlations for IQ scores among identical as opposed to fraternal twins. Behavioral geneticists themselves have long acknowledged the potentially problematic nature of this method, although some research suggests that fraternal twins are not substantially different from identical twins in this regard. In any event, an obvious way around this potential confounding influence is to compare the IQ scores of identical twins reared apart with those of fraternal twins reared apart. With this approach, a shared “twin environment” is no longer a confounding factor for either class of twins.
Five previous studies of identical twins reared apart have yielded a mean-weighted heritability estimate of .75 (based on 158 twin pairs)—that is, 75 percent of individual differences in IQ scores are accounted for by heredity.2 Although conducted in different countries and time periods, and with different measurement instruments, these results have been remarkably consistent (.64 to .78). Two previous studies of fraternal twins reared apart have yielded a mean-weighted heritability estimate of .38 (based on 73 twin pairs). These collective data yield a heritability estimate—unconfounded by shared twin environments—of .74 (.75 minus .38, multiplied by 2). It is worth noting that these 7 studies of twins reared apart actually yield a higher heritability estimate than do the 75 studies involving identical and fraternal twins reared together, which is only .52 (based on 4,672 identical and 5,546 fraternal twin pairs).3 The difference between these two heritability estimates is likely due to the relatively younger age of the participants in studies where twins have been raised together, as heritability estimates increase with age.
In sum, for twins, heritability estimates for IQ appear to be between .50 and .70, depending on the particular method by which IQ is calculated, the age of the study participants, and measurement error. The possible confounding influences mentioned by Professor Kaplan appear to make almost no difference in any of these findings. This general conclusion does not mean that environmental influences on IQ are unimportant. On the contrary, abundant evidence has shown that family environments make a substantial contribution to intelligence, especially before children reach adulthood and especially in impoverished environments that do not allow for the full development of genetic predispositions.4 The increasingly evident portrait of human development that has emerged from these twin studies is one of nature via nurture, as people growing up are drawn to environments that provide the best outlets for their inborn dispositions and abilities.
March 15, 2007
Thomas J. Bouchard, Jr., David T. Lykken, Matthew McGue, Nancy L. Segal, and Auke Tellegen, “Sources of Human Psychological Differences: The Minnesota Study of Twins Reared Apart,” Science, Vol. 250 (1990), pp. 223–228. ↩
Nancy L. Segal, Entwined Lives: Twins and What They Tell Us about Human Behavior (Dutton, 1999), pp. 135–136. ↩
Thomas J. Bouchard, Jr., and Matthew McGue, “Familial Studies of Intelligence: A Review,” Science, Vol. 212 (1981), pp. 1055–1059. ↩
Kathleen McCartney, Monica J. Harris, and Frank Bernieri, “Growing Up and Growing Apart: A Developmental Meta-Analysis of Twin Studies,” Psychological Bulletin, Vol. 107 (1990), pp. 226–237; Eric Turkheimer, Andreana Haley, Mary Waldron, Brian D’Onofrio, and Irving I. Gottesman, “Socioeconomic Status Modifies Heritability of IQ in Young Children,” Psychological Science, Vol. 14 (2003), pp. 623–628. ↩