John Alford, Carolyn Funk, and John Hibbing, in their American Political Science Review article (hereafter referred to as “AFH”), employ twin studies to argue that political ideologies (i.e., conservative or liberal political attitudes) are highly heritable and thus strongly influenced by genetics.Footnote 1 AFH state that genetic explanations of behavior are “conspicuously absent” from the political science dialogue and charge that “political scientists do not take seriously the possibility of nonenvironmental influences.”Footnote 2 They urge “political scientists to incorporate genetic influences” into models of political behavior and to “join forces” with behavioral geneticists.Footnote 3 Given the significant criticisms of studies in which twins are used to understand the origins of complex human behaviors and problems with the scientific data supposedly supporting the validity of this method, such a move would be, at best, premature.Footnote 4
AFH argue that the classical twin approach for studying complex human social behaviors is well supported by the evidence. Here, we examine the empirical studies cited by AFH in support of a critical assumption underlying the validity of these twin studies, the equal environments assumption (EEA).Footnote 5 We do not claim that genetic factors play no role in human behavioral traits, but rather we show that the empirical evidence used in support of the underlying premise of these twin studies is weak, far less certain than AFH would have readers believe. In fact, many of the studies cited as supporting the validity of the twin method include data that violate predictions of the EEA. As a result, the conclusions drawn on the basis of classical twin studies, as those presented by AFH, are of dubious scientific value.
The Heritability Concept
Heritability is a statistical term that should represent, for a trait of interest, the fraction of variation in a population that is due to genetic contributions.Footnote 6 Expressed as a number ranging from 0.0 (no heritability) to 1.0Footnote 7 (complete heritability), heritability is measured for a specific population, at a specific time, interacting with a specific environment.Footnote 8 Thus, as pointed out by Schaffner, “if the environment changes, the heritability will almost certainly change.”Footnote 9 This property of heritability is highlighted by a study showing that IQ differences exhibit high heritability in families of higher socioeconomic status, but near-zero heritability in families of lower socioeconomic status.Footnote 10 In other words, in an advantaged population, a large fraction of IQ variability was attributed to genetic influences, while in a disadvantaged population, IQ variability was attributed almost entirely to environmental influences. In effect, incorporating measures of environmental influences into such studies can yield dramatic variations in heritability estimates. Relatively few such twin studies have treated environment very deeply, usually attempting to rule it out as having a confounding effect on genetic conclusions.Footnote 11 As leading behavioral geneticist Eric Turkheimer has put it, this is “hardly the stuff of good environmental analysis.”Footnote 12
The “Crucial” Equal Environments Assumption
AFH acknowledge that the equal environments assumption is “crucial to everything that follows from twin research.”Footnote 13 They conclude that the “caveats” raised have “been subject to sustained and varied investigation and … [have] been found to hold up under empirical scrutiny.”Footnote 14 In support of this claim, they cite studies that look at the effects of measures of similarity of appearance, treatment, and environment (including so-called “misperceived zygosity studies”) on shared traits between twins.Footnote 15 We review the evidence that they argue supports their conclusions.
Studies in which heritability of behavioral traits is based on comparisons of pairs of monozygotic (MZ) and dizygotic (DZ)Footnote 16 twins rely on the validity of the EEA. The EEA assumes that MZ and DZ twins each share environments to the same extent. In particular, it is assumed that the two types of twins do not differ in those environmental influences that might affect the trait under study (trait-relevant environments). However, critics argue that, because of their identical appearance, MZ twins are treated and interacted with by parents and the outside world much more similarly than are DZ twins or ordinary siblings. Consider the potential influence of shared physical features of MZ twins on social interactions. Children who are obese, for example, might suffer indignities that others do not. Children thought of as “attractive” might experience a very different world than those deemed “ugly.” MZ twins on average share closer bonds than DZ twins by a number of measures. Both this suggested closer bond between MZ twins and the potentially similar responses of family and society to the identically-featured MZ twins could well influence the behavioral development of the children in the direction of greater similarity. If this were to be the case, the incorrect acceptance of the EEA could lead to heritability measurements in twin studies that would be highly inflated, despite the fact that the trait under study was strongly environmentally influenced.
Thus, in order for heritability estimates to be valid for a particular trait, MZ and DZ twins should experience equal trait-relevant environments. This is a primary problem in the studies cited by AFH, since determining which environmental factors to assess may be difficult to achieve. After all, such studies are meant to sort out the factors that influence a trait, and should not assume to know a priori all of the relevant environmental influences.Footnote 17 Even those researchers who do attempt to measure environmental influences, as the authors of one study point out, choose out of “a virtually endless array of possible environmental characteristics.”Footnote 18 Indeed, none of the studies we reviewed here sought specific “trait-relevant” environmental effectors, but only attempted to rule out effects of general measures of environmental similarity.
“Misperceived Zygosity” StudiesFootnote 19
Misperceived zygosity studies, considered to be tests of the EEA, take advantage of twin pairs for which their twin status (i.e., MZ or DZ) has been wrongly assigned. These studies are made possible by instances where biologically identical twins call themselves or are called by their parents “fraternal,” or where fraternal twins call themselves or are called by their parents “identical.” Trait correlations for these misidentified twins are then compared to those for correctly identified twins in order to test the effects of “true zygosity” (presumed to measure a genetic effect) versus “perceived zygosity” (presumed to measure an environmental effect). Thus, the argument goes, if MZ twins who perceive themselves to be DZ are just as similar in trait measures as MZ twins who correctly identify themselves (or DZ twins who misidentify as MZ just as dissimilar as DZ twins who correctly identify themselves), then the environmental influence of more similar treatment for MZ twins can be concluded “to be at best irrelevant.”Footnote 20
However, in order for “perceived zygosity” to be a meaningful test of the EEA, it must be a true surrogate for the environmental similarity a pair of twins' experiences. For example, a genetically identical twin pair who misidentify themselves as “fraternal” must believe themselves as well as be perceived by others to be non-identical fraternal twins. Otherwise, this may merely be a case of genetically MZ twins, who are likely to appear remarkably similar, but confused or were incorrectly told their twin status. Unless self-identification of zygosity is shown to correlate with beliefs about degree of similarity and actual physical similarity, it is questionable whether merely mis-categorization is a meaningful measure of how a twin pair will be treated, when MZ twins will still appear physically similar and DZ twins dissimilar. Nevertheless, the authors of these studies, without evidence, assume that DZ twin pairs misidentified as MZ pairs “view themselves as more similar [and] will be treated more alike by parents, family, and society.”Footnote 21 This is an untested assumption. We propose the term “misidentified zygosity” as more appropriate for such twin pairs, as “misperception” has not been demonstrated.
In fact, data in two “misperceived zygosity” studies appear to contradict the notion that physical similarity is the basis for the perception of zygosity status. One study found that, compared to MZ or DZ twins who were correct about their zygosity, MZ twins who incorrectly believed they were DZ “were much less likely to base their opinion on physical appearance, but more likely to base their opinion on what their parents were told at their birth.”Footnote 22 Another study found that “when mothers were mistaken about the zygosity of their twins, they usually relied on information given to them from a doctor at the time of the twins' birth,” as opposed to similarity of physical appearance.Footnote 23 These findings raise questions about the utility of using “misperceived zygosity” as a test of the EEA. As the authors of one study using this method acknowledge, “a single measure of self-perceived zygosity may not be the most accurate representation of what critics mean by biasing perception of the two kinds of twins.”Footnote 24
The limitation of using “perceived zygosity” as a surrogate for environmental similarity is revealed by the data presented in a 2002 study.Footnote 25 MZ twins sharing classes were significantly more likely to share symptoms of separation anxiety disorder, attention deficit-hyperactivity disorder (ADHD), and oppositional defiant disorder. However, “perceived zygosity” was not associated with shared symptomology, indicating that misidentified zygosity did not capture environmental effects that were seen when accounting for the likelihood of sharing classes. Furthermore, sharing friends by DZ twins predicted similar ADHD scores, although a “perceived zygosity” score for shared ADHD did not have a statistically significant effect. This suggests that “perceived zygosity,” as measured by these studies, is unlikely to be a good indicator of environmental similarity, and thus will fail to truly test the EEA.
Violations of the EEA
One of the clearest and most consistent results of the studies cited by AFH was that MZ twins are indeed more similar on measures of environmental and physical similarity. MZ twins were “markedly more confusable in appearance than DZ twins;”Footnote 26 significantly more likely than DZ twins to report being treated “as two of a kind” by parents, to have dressed identically, to have been in the same class at school, to have played together “almost always,” and to have spent time together “almost always”;Footnote 27 and share a much more similar environment based on mothers' accounts of childhood environmental similarity (including sharing friends, sharing classes, and dressing alike).Footnote 28 As the authors of one study state, “we must conclude that the MZ twins in our sample were treated more similarly than were the DZ twins and that therefore the equal environments assumption was violated.”Footnote 29
Ultimately, the authors of all of these studies conclude that despite apparent violations of the EEA generally, their measures of environmental similarity did not influence the traits under study, and therefore “trait-relevant” EEA was not violated. However, a closer look at the studies in support of the EEA finds that many actually contain significant data arguing that “trait-relevant” EEA was indeed violated. For example, one study found that DZ twins who were more likely to be confused in appearance were more likely to have similar scores for activity and impulsivity traits. MZ and DZ twins as a whole who were more similar in appearance were more likely to share all four measures of personality traits assessed.Footnote 30
In another study, the authors report that “for personality measures MZ twins were significantly more similar than DZs, and DZ twins who believed they were monozygotic were more similar than those … who correctly believed they were fraternal pairs.” In other words, “both true and perceived zygosity were related to cotwin similarity on personality measures.”Footnote 31 Furthermore, they found that physical similarity (as assessed by eight raters) predicted more similar Raven IQ scores for monozygotic twins. The authors admit their data is evidence of “bias in perceived and physical similarity creating greater cognitive similarity among MZ twins,” although they downplay the effect. Nevertheless, they do conclude, “for personality variables, perceived zygosity may have some effect on fraternal pairs who believe themselves to be monozygotic.”Footnote 32
Another of these studies found significant associations between measures of DZ twins' preferences for similar treatment, similarity of environment, and preference for similar experience with anxiety and depression scores.Footnote 33 For MZ twins, an association was found between measures of environment and depression. The authors conclude that the “equal environments assumption appears to be invalid.”Footnote 34 Likewise, a different study found that measures of environmental similarity predicted greater similarity in symptom scores for nearly half of the tested traits.Footnote 35 For example, MZ twin pairs who were more frequently in the same classes were significantly more likely to share similar symptom scores for all of the measured disorders than those who were in the same classes less often. Dressing alike predicted greater symptom correlation among MZ twins for separation anxiety disorder, ADHD, or conduct disorder.
Further indicating that environmental similarity of MZ twins confounds twin studies is the fact that the data also show that DZ twins who more often share friends, dress alike, or share classes, were significantly more similar on scores for ADHD, separation anxiety disorder, and conduct disorder, respectively.Footnote 36 In addition, they found a significant effect of “perceived zygosity” for MZ twins on shared conduct disorder, and for DZ twins on shared separation anxiety and conduct disorders. These statistically significant findings call into question the EEA. This data is all but ignored by the authors of this 2002 study, and where the effect of environmental similarity is discussed, it is downplayed.Footnote 37
If environmental similarity does influence MZ twin concordance for a trait in question, then this would be expected to inflate estimates of heritability. Accounting for the unequal environments would decrease estimated heritability. In a study measuring the “contact frequency” of twin pairs, the authors found that heritability estimates dropped dramatically when the samples were stratified by level of contact between twins.Footnote 38 On average, their heritability measures for twins with infrequent contact dropped 31 percent below twins with frequent contact. Compared with the total sample of twins, controlling for the frequency of contact indicates that their measures of heritability were overestimated on average by 27 percent. The authors note that “the heritability estimates were consistently higher among twin pairs with frequent contact, suggesting a potential violation of EEA.”Footnote 39
Given that several of these studies report findings that appear to violate both general and trait-relevant EEA, how do all of these studies conclude that the EEA is valid?Footnote 40 Despite small sample sizes and weak statistical power, several studies do so solely on the basis of negative results, or a failure to find statistically significant correlations between their measures of similarity and the trait in question.Footnote 41 Not all of the studies, however, based their EEA-validating conclusions on negative results. One study presenting significant evidence that the EEA did inflate heritability estimates nevertheless emphatically concluded that they “found no support for violation of EEA.”Footnote 42 In order to draw this conclusion, they used 40,000 simulated twin pairs and the new assumption that MZ twins select their environment on the basis of their genetics.Footnote 43 As if to highlight their conclusions in spite of their actual empirical data, they state that “the simulation analyses thus provide an interpretation of our empirical results without violation of the EEA.”Footnote 44
Retrospective Nature of the Studies
Another problem with studies evaluating the validity of the EEA is that important data is often obtained from retrospective information. One study asked adult twins with a mean age of 34 how they were treated as children. As the authors put it, the subjects were “reporting on events and their feelings about those events some 20 years after their occurrence.”Footnote 45 Another study measured emotional and behavioral problems via mother-reported retrospective telephone interviews, recording mothers' reports of twins' zygosity, as well as recounts of childhood environmental similarity.Footnote 46 This retrospective format is suspect, given that even in a study where parents gave contemporary reports of their children, mothers and fathers independently agreed on the answers only about one-third of the time. When asked the same questions only three months later, their answers changed over 25 percent of the time.Footnote 47 This calls into question the reliability of retrospective data over a matter of weeks, let alone decades.
In addition to methodological flaws, all of the studies cited by AFH in support of the EEA used small samples, ranging in size from 41 twin pairsFootnote 48 to 201 twin pairsFootnote 49 having “misperceived zygosity.” One of the largest studies (with 3,155 total twin pairs, 185 of which had “misperceived zygosity”) included a power analysis indicating that even that sample size offered weak to no statistical power to detect an influence of misidentified zygosity.Footnote 50 A study cited by AFH summarizes the statistical weakness of these earlier reports, noting that studies supporting the EEA may have reached their conclusions “because of limited statistical power” and “because the validity of the assumption is usually established by failing to find a significant effect.”Footnote 51
Conclusion
We agree with the authors AFH when they state that the EEA is essential for conclusions drawn from twin studies concerning human behavioral traits. However, we conclude that the EEA has not been well tested nor validated. We point out numerous problems with the studies cited by AFH. These studies report small sample sizes with low statistical power, problematic retrospective interviews, and assumptions about misidentified zygosity that are often not supported by data. The attempts to control for trait-relevant environments in these efforts are quite limited, admittedly because it is difficult to know exactly which environmental factors to control for. Furthermore, several of the studies cited in support of the EEA contain evidence that could be interpreted as arguing against the assumption. It is difficult not to get the sense that in the face of inconvenient data, a fair amount of intellectual acrobatics was necessary to arrive at some of the conclusions drawn.
AFH have made a plea for the political science research community to take more seriously twin studies that suggest a strong genetic component affecting political beliefs. While, to many, twin studies seem to present an ideal opportunity to study the relative contributions of genetics and environment to human behavioral traits, these studies face many more problems than has been acknowledged. In a research area that is often misrepresented to the public, we urge political scientists to take a more critical look at the studies that supposedly provide the foundation for this field.
Finally, we point out that in recent years, some researchers in the field of human behavioral genetics have begun to take more seriously the role of environment in influencing human behavioral traits. We have already referred to the studies of Eric Turkheimer and colleagues. Other examples include studies that have incorporated measurements of child abuse into a study of genetic effects on anti-social behavior.Footnote 52 A study genuinely accounting for social factors found that certain measures of greater similarity among MZ twins compared to DZ twins diminished or disappeared.Footnote 53 The findings of James Flynn and his colleaguesFootnote 54 that the mean IQ of populations in many countries (including the U.S.) has been steadily increasing over the last 50 years raise questions about the more deterministic views of genetic influences on human cognition. A recent article looking at concentrated disadvantage found neighborhood-level effects on verbal ability among African-American childrenFootnote 55 and indicates more sophisticated measures of environment may be necessary to truly detect significant environmental influences on complex human traits.
