HERITABILITY STUDIES: METHODOLOGICAL FLAWS, INVALIDATED DOGMAS, AND CHANGING
CALLIE H. BURT
University of Washington
Citation: Burt, Callie H. Heritability studies: Methodological flaws, Invalidated dogmas, and changing
paradigms. Advances in Medical Sociology: Health, Genetics, & Society 16, forthcoming 2015.
Running Head: HERITABILITY STUDIES IN THE POSTGENOMIC ERA
* Direct all correspondence to Callie Burt, Department of Sociology, University of Washington, Box 353340,
Seattle, WA 98195; email: email@example.com.
HERITABILITY STUDIES: METHODOLOGICAL FLAWS, INVALIDATED DOGMAS, AND CHANGING
Purpose: Heritability studies attempt to estimate the contribution of genes (vs. environments) to variation in
phenotypes (or outcomes of interest) in a given population at a given time. The current chapter scrutinizes
heritability studies of adverse health phenotypes, emphasizing flaws that have become more glaring in light of
recent advances in the life sciences and manifest most visibly in epigenetics.
Design/methodology/approach: Drawing on a diverse body of research and critical scholarship, this
chapter examines the veracity of methodological and conceptual assumptions of heritability studies.
Findings: The chapter argues that heritability studies are futile for two reasons: (1) heritability studies suffer
from serious methodological flaws with the overall effect of making estimates inaccurate and likely biased
toward inflated heritability, and, more importantly, (2) the conceptual (biological) model on which heritability
studies depend—that of identifiably separate effects of genes vs. the environment on phenotype variance—is
unsound. As discussed, contemporary bioscientific work indicates that genes and environments are enmeshed in
a complex (bidirectional, interactional), dynamic relationship that defies any attempt to demarcate separate
contributions to phenotype variance. Thus, heritability studies attempt the biologically impossible. The
emerging research on the importance of microbiota is also discussed, including how the commensal relationship
between microbial and human cells further stymies heritability studies.
Originality/value: Understandably, few sociologists have the time or interest to be informed about the
methodological and theoretical underpinnings of heritability studies or to keep pace with the incredible advances
in genetics and epigenetics over the past several years. The present study aims to provide interested scholars
with information about heritability and heritability estimates of adverse health outcomes in light of recent
advances in the biosciences.
Keywords: heritability study, twin study, epigenetics, plasticity, postgenomics, microbiome
HERITABILITY STUDIES: METHODOLOGICAL FLAWS, INVALIDATED DOGMAS, AND CHANGING PARADIGMS
“In the real world of humans, in a given context everything is heritable to some extent and environmental to some
other extent, but the magnitudes of the proportions are variable from situation to situation, and have nothing
whatsoever to do with the causal properties of genes and environment for the trait in question, unless one is
interested in the pointless null hypothesis that one of the components is zero. This message has taken 100 years to
soak into genetic social science, and it has not fully soaked in yet: not a month goes by without another outbreak
of credulous surprise that one trait or another has turned out to be 50% heritable.” (Turkheimer 2011: 598).
The nature vs. nurture debate continues in the social and behavioral sciences in the form of the heritability study.
These studies aim to partition the variation in an outcome of interest (phenotype) into a proportion caused by
genes (heritability) and a proportion caused by the environment. The general conclusion emerging from these
studies is that variance in every examined characteristic—including preferences, social behaviors, and diseases—is
significantly shaped (~50%) by genetic influences (Turkheimer 2000). For example, recent studies show
substantial heritability of everything from breastfeeding (Colodro-Condre et al. 2013), breakfast eating patterns
(Keski-Rahkonen et al. 2004), “treatment responses to tooth whitening” (Corby et al 2014), and sickness absences
(Svedberg et al. 2012) to ADHD, autism, depression, hypertension, high cholesterol, and diabetes (e.g., Plomin,
DeFries, Knopik, & Neiderhiser 2012).
While heritability studies continue to abound, the past decade has witnessed significant advances in our
understanding of the function of “genes,”1 including the nature of genetic influences on adverse health phenotypes.
These advances in knowledge have not answered questions about how much variation is genetic versus
environmental (or nature vs. nurture), but rendered such questions manifestly obsolete (Keller 2010). Recent
advances in molecular genomics reveal that the causal interactions among genes, proteins, cells, and physical and
social environments on health and development are entangled, dynamic, and context dependent. Such advances
have exposed the conceptual framework of heritability studies—that of identifiably separate effects of genes vs.
environments—as unsound (Burt & Simons 2014; Charney 2012; Charney & English 2013) and have fostered a
1Indeed, “several recent discoveries have cast serious doubt on the idea that there are coherent entities [clearly identifiable
particulate units of inheritance] in our DNA that can unambiguously be called ‘genes’” (Moore 2014: 46). Furthermore, although
the gene was once considered the sole unit of inheritance, there is growing evidence to suggest that there are other mechanisms of
biological inheritance, such as the epigenetic inheritance system (e.g., Charney 2012; Jablonka & Lamb 2006). Despite this, to
maintain consistency with discussed studies and avoid confusion, I will continue to use the terms “genetic influences” and “genes” to
refer to inherited DNA sequences.
growing scholarly awareness, including among former proponents of heritability studies, that heritability studies
are superannuated (e.g., Rutter 2006; Turkheimer 2011). Yet, this misguided enterprise continues, primarily in
the social and behavioral sciences, despite the fact that the biological model undergirding heritability studies is
incompatible with contemporary bioscientific knowledge (Charney & English 2012; Duster 2003).
In the present chapter, I discuss heritability studies with two goals. First, I aim to assist readers in
understanding heritability and making more judicious assessments of heritability estimates. Given the prevalence
of heritability studies, the failure of some studies to clearly describe the shaky foundations of the model, and the
potential misuse of these estimates, as well as the fact that these studies will continue (at least in the short term), it
is vital that scholars are cognizant of the meaning and limitations of heritability estimates (e.g., “The heritability of
ADHD in children is an estimated 75%”; Rietveld et al. 2004). Thus, I first discuss the concept of heritability and
the technical/statistical limitations of the model, focusing on the most common method at present, the “twin
study.” Second, I critique the conceptual (biological) model of heritability studies, emphasizing fatal flaws that
have become more glaring in light of recent advances in the life sciences, focusing on their relevance for studies of
health. Echoing the arguments of prominent scholars (e.g., Charney 2012; Crusio 2012; Joseph 2004; 2006), I
argue for an end to heritability studies as well as recognition of the problematic nature of existing heritability
In this chapter I argue against the continuation of heritability studies in the social sciences. I maintain that
heritability studies are futile for two reasons. First, heritability studies suffer from serious methodological flaws
with the overall effect of making estimates (potentially highly) inaccurate and biased toward inflated heritability
and deflated shared environmental influences. Second, and more importantly, the conceptual (foundational)
model on which heritability models depend—that of identifiably separate effects of genes vs. the environment on
phenotype variance—is misguided. Although the conceptual model of heritability studies, like the technical
model, has been criticized for more than half a century, profound scientific advances over the past decade, perhaps
best exemplified in epigenetic research, provide clear evidence that the dichotomous view of “genetic” vs.
“environmental” effects on phenotype variance assumed by heritability studies is untenable (Burt & Simons 2014;
2015; Charney 2012).“[D]evelopment is not merely a process involving a battle between nature (genes) and
nurture (experience) but the interweaving of dynamic processes within a system that is inseparably both the
organism and its environment” (Kaplan & Rodgers 2003: 5). Thus, regardless of their technical merits, the very
goal of heritability studies is biologically impossible.
After laying out my argument that heritability studies should be discontinued, I conclude with a brief
discussion of the need to move beyond heritability to focus on the plasticity of human development and the
flexibility of genetic expression in response to environmental influences. As I hope is clear, I attempt to push the
field forward, not by denying the role of genes or biological factors, but by recognizing the complexity of the
biopsychosocial system. Many nuances and details cannot be covered in this chapter; fortunately, there is an
extensive body of work that I point the reader to for more detail.2 Before moving into the limitations and flaws in
heritability studies, I provide a brief introduction to heritability, a concept that is frequently misunderstood.
The heritability construct was originally developed in agricultural research to assist animal and plant breeders to
predict the outcomes of controlled breeding programs. Since the 1960s, this concept has been promoted as the
“nature-nurture ratio,” or the relative influence of heredity (through genes) and environmental experiences on
trait variance within human populations (Joseph 2006; Lewontin, Rose, & Kamin 1984). Heritability is defined
narrowly by Wahlsten (1994: 224) as “the proportion of variance in a measure of behavior or other phenotype in a
breeding population that is attributable to genetic variation,” and more loosely by Plomin and colleagues
(2012:87) as “the proportion of phenotypic variance that is accounted for by genetic differences among
individuals.” The heritability coefficient is the numerical index of heritability that ranges from 0 (no genetic
contribution) to 1.0. This estimate is derived from a model based on Mendelian principles as well as a number of
assumptions, discussed below. Given that the construct of heritability is frequently misunderstood, it is useful to
clarify some misunderstandings about heritability estimates.
2 Notably, most of the arguments in this chapter are not original but are those of prominent scientists, many of whom I cite, whose
criticisms accumulating over the past 80 years come together, in my view, to demonstrate the futility of the heritability study.
Perhaps most importantly, heritability estimates do not indicate the effect of genes on a particular trait.
Instead, they index the genetic contribution to trait variability in a population. Heritability estimates are often
wrongly interpreted as indicating the former by laypersons as well as some scholars (Hirsch 2004; Joseph 2006).3
The methods of Mendelian genetics are responsive only to the tiny portion of genes that are polymorphic and
make us different; they do not allow for conclusions about the role of heredity in general (Wahlsten 1990). “It is
an illegitimate inference to assume that an assumption that variance for a trait is due largely to ‘genes’ as opposed
to ‘environment’ entails that the trait itself is inherited (i.e., genetically transmitted from parents to offspring).
Causes of a trait, and causes of trait variance are not the same thing” (Charney 2012:61). As Joseph (2004: 138-9)
notes “This leads to a paradoxical situation in which a trait could be 100% inherited, yet have a heritability of
zero—human beings having two eyes for example.” Genetics, of course, explains humans’ “eyedness”; we have
two eyes because of our genetic endowment. Because everyone (or nearly everyone) who has fewer than two eyes
is that way as a result of life experiences, the heritability of “eyedness” is zero (see also Lewontin et al. 1984). As
such, heritability estimates do not indicate, for example, what proportion of a population’s or an individual’s risk
for a disease is due to genes. Moreover, heritability estimates signify nothing about the number of genes involved
nor do they contribute to the identification of specific candidate genes (e.g., Schielzeth & Husby 2014).
Notably, heritability is not a fixed estimation of the overall effects of genetics on trait variance but rather
is time and population specific. As such, findings of genetic influences on differences within populations do not
extrapolate into explaining differences observed in traits across populations (e.g., adults vs. children; rural vs.
urban populations; low SES vs. high SES populations; Lewontin 1986). Indeed, even for trait variance that is
entirely heritable within a population, the cross-population variance may be due completely to environmental
factors (see Lewontin 1986; Joseph 2004). This aspect of heritability is wonderfully illustrated by Lewontin and
colleagues (1984). Suppose one takes two handfuls of heterogeneous corn seed and plants one handful on a field of
3 Much of the misunderstanding over the meaning of heritability has to do with the confusion of the words “heritable” and
“inherited” or “heredity” (Joseph 2004; Stoltenberg 1997). The two concepts, according to Hirsch (1997:220) “have been
hopelessly conflated…[b]ecause of their assonance.” Indeed, this assonance and confusion has led Stoltenberg (1997) to propose
that the word “heritability” be replaced by the word “selectability” so as not to confuse the scientific and folk definitions of the
nutrient-rich soil, and the other in a field of nutrient-poor soil. When the seeds have grown, we can see that there
is variation within fields in plant height, as well as variation between fields, specifically with lower plant height for
the poor quality soil field. Not allowing any environmental variation within the respective fields (equal light,
water, etc.), differences in the resulting plant height within the fields is totally due to genetic factors (heritability =
1.0). Differences in plant height between the two fields, however, are completely due to environmental factors,
specifically soil quality (heritability = 0).
Importantly, heritability estimates do not speak to the responsiveness of a phenotype to environmental
interventions. Traits can be highly heritable and yet be drastically altered or eliminated by changes in the
environment (e.g., Lewontin 1974; Plomin et al. 2012; see Joseph 2004, p. 140, for a discussion of the classic
example of PKU). Thus, high heritability does not mean inevitability of a phenotypic outcome. In addition, the
heritability of a trait can change as a result of changes in the environment. Returning to the cornfield example, if
one were to plant trees around the corn field, thereby producing unequal sun exposure among the corn plants
within a field, heritability would decrease as the resulting variation in sunlight exposure would account in part for
differences in plant height. Thus, there is not and cannot be any static absolute value for the strength of genetic
influences on trait variance.
Grounded in the classical gene-centric paradigm (e.g., the central dogma of the genomic era), especially the
assumption that genes and environments make identifiably separate contributions to phenotypes, heritability
studies attempt to estimate the relative influence of genes on trait variance in a population. Under the prevailing
methodology, the heritability of a given phenotype is usually estimated by comparing concordances and
discordances between subjects relative to their presumed degree of genetic similarity. Various models have been
used in traditional studies of heritability, including family, adoption, twins-reared-together, and twins-reared-
apart studies. Although these different models are grounded in different assumptions, they are united by the fact
that all compare phenotypes across varying degrees of genetic relationships and use these comparisons to estimate
genetic and environmental influences without actually measuring either. (A newer method, GCTA, which does
not rely on family comparisons and measures genetic variants, will also be briefly discussed later.) At present the
study of reared-together twins (referred to as the “twin study”) is the primary workhorse of behavioral genetic
studies of heritability; thus, I focus on twin studies.4 Below I briefly discuss the set of technical/statistical
assumptions and problems inherent in twin studies (for more detail on these technical flaws see, e.g., Jackson
1960; Joseph 2004; 2006; Lewontin et al. 1984; Wahlsten 1990).
THE TWIN STUDY (TWINS REARED TOGETHER)
INTRODUCTION TO AND FUNDAMENTALS OF TWIN STUDIES
Over the past several decades the “twin study” has become the primary method for estimating heritability.5 As we
know, identical (monozygotic or MZ) twins come from one fertilized egg and are assumed to share 100% of their
genes (genetic clones), whereas fraternal (dizygotic or DZ) twins come from two fertilized eggs and are presumed
to share on average half of their genes (the same amount as other non-twin biological siblings). Given this and
along with several assumptions, researchers have used MZ and DZ twins as a natural experiment to estimate the
relative influence of nature and nurture (Joseph 2004; Plomin et al. 2012).
Twin studies separate phenotypic variation into three components: genetics (h), shared environment (c),
and unshared environment (e). Notably, the shared and unshared environments are misnomers (Burt & Simons
2014). The shared environment consists of non-genetic influences that make twins similar to each other, such as
parenting, SES, and community factors. Although it is frequently assumed that the influences that cause trait
similarity are in fact shared, technically, this component captures any factors (shared or not) that cause twins to be
more similar to one another (Plomin 2011; Suhay & Kalmoe 2010). The unshared environment, on the other
hand, consists of all non-genetic factors that create differences in traits among twins, such as different peers,
classrooms, interactional experiences and the like, as well as model error. As with the shared environment, while
it is often assumed that environmental influences that cause differences are “unshared” in the technical sense (e.g.,
4 Space constraints do not permit an extensive examination of the statistical features and assumptions of the two other classic
heritability models, adoption and twins-reared apart studies, which also rest on dubious empirical and conceptual foundations.
Fortunately, excellent critical discussions of these models in relation to adverse health outcomes have appeared elsewhere (see,
e.g., Joseph 2004; 2006; see Burt & Simons 2014 for a more concise overview).
5 Some have misguidedly overgeneralized arguments against twin studies of heritability as arguments against the use of twin samples
in research generally (e.g., Moffitt & Beckley 2015). To be clear, my critique is focused on twin studies of heritability, and it is
silent on the use of twin-based research for other purposes (e.g., studies of MZ twin discordance).
being treated differently by parents), all factors that produce differences among twins are incorporated into the
unshared environment, whether actually shared or not. Scholars not infrequently fail to describe what is meant by
these terms, and some have made inappropriate and widely publicized conclusions about the irrelevance of
parenting or community factors based on shared environmental estimates (e.g., Harris 1998; Rowe 1994).
The basic logic of the twin study model is to compare twin concordances for phenotypes and, based on
several assumptions mentioned below, assign the greater similarity of MZ co-twins relative to DZ co-twins to
their greater genetic similarity. The key assumption in this natural experiment is that everything is the same
between the two types of twins except for the fact that MZ co-twins are genetic clones while DZ co-twins share on
average 50% of their genes. In the simplest approach, heritability is usually estimated from twice the MZ-DZ
differences in correlations. More recently, complex latent variable models have been utilized; however, the basic
logic underlying these more sophisticated models is the same as that for the more naïve models.
The assumptions of the twin study have remained largely unchanged since the 1920s and remain crucial to
the model given that neither genetic nor environmental influences are measured.6 Some of these assumptions are
relatively unproblematic, but others are quite dubious. Prominent questionable assumptions include the
following: (Charney 2012; Joseph 2006; Plomin et al. 2012).7
1. The environments of MZ co-twins are no more similar than that of DZ co-twins for trait-relevant
environments (trait-relevant equal environments assumption).
2. The relevant genes exert effects additively.
3. The risk of receiving a diagnosis for a phenotype (e.g., ADHD, depression, diabetes) is the same
between MZ and DZ co-twins.
4. Phenotypic variation can be demarcated into genetic (h), shared environmental (c), or unshared
environmental (e) components.
6 The exception is the modification of the equal environments assumption to its trait-relevant form given clear evidence that the
former was invalid (see Joseph 2004; 2006, for an overview of this shift.)
7 As noted this is an incomplete list of assumptions, and excludes one, no assortative mating, whose violation has the effect of
biasing heritability downward (e.g., Rutter 2006). Given space constraints combined with the fact that the no assortative mating
assumption is more often noted in heritability studies (often implying or stating that the h2 estimate might be an underestimate), I
focus on these four assumptions.
In general, few studies mention these assumptions, and even fewer describe them or the likely impact of their
violation, which is generally inflated heritability estimates and deflated shared environmental influences. Below, I
discuss (and question the validity of) these key assumptions from the classical genetic perspective that undergirds
the heritability model, leaving aside the conceptual deficiencies that are glaring in light of recent advances in
molecular genomics. I then turn to a critique of the model from a postgenomic perspective.
SOME METHODOLOGICAL LIMITATIONS OF TWIN STUDIES
Equal Environments Assumption (#1)
The trait-relevant equal environments assumption (hereafter EEA) is the assumption that the covariance between
genetics and the environment is zero for environments that influence variance in the outcome under study. In
other words, the model assumes that MZ co-twins experience the same degree of (trait-relevant) environment
similarity as DZ co-twins (assuming, for example, that MZ co-twins are treated no more similarly than DZ co-
twins). Without the EEA, the greater similarity of MZ co-twins could be due to genetics or more similar
environments (see Suhay & Kalmoe 2010, for an algebraic demonstration of the importance of the EEA). The
consequence of the EEA is that all greater phenotype similarity between MZ than DZ twin pairs is credited to
genetics. As a result, EEA violations (greater trait-relevant environmental similarity among MZ than DZ twin
pairs) serve to artificially inflate heritability and deflate shared environmental estimates.
Although the extent of EEA violations varies across different traits, in general, research evinces that MZ
co-twins experience much more similar environments that DZ co-twins (see Joseph 2006 for an excellent
review). Indeed, for years many scholars have argued that the more similar environments for MZ than DZ co-
twins bias heritability upwards to an unknown degree for many, if not most, phenotypes (e.g., Horwitz et al.
2003; Joseph 2004; Lewontin et al. 1984). For example, research shows that MZ co-twins are more likely to be
treated similarly by parents and others (Evans and Martin 2000), to have the same friends (Cronk et al. 2002;
Horwitz et al. 2003), to share the same classroom (Cronk et al. 2002; Morris-Yates et al. 1990), and to spend
time together (and therefore experience the same environments more frequently; Horwitz et al. 2003; Rende et
al. 2004). Thus, MZ co-twins are more likely to experience the same physical environments (including exposure
to pollution and chemicals, food/nutrition, even traumatic experiences) and engage in the same activities (e.g.,
exercise; Carlsson et al. 2006), many of which can have profound and diverse effects on adverse health outcomes.
Evidence that EEA violations are consequential (i.e., inflate heritability) is found in studies that examine
the association of greater environmental similarity with trait concordance (see e.g., Beckwith & Morris 2008;
Joseph 2004). For example, Cronk and colleagues (2002) found that DZ co-twins who more often share classes,
share friends, and dress similarly were more similar on health outcomes, such as ADHD and anxiety disorder. The
authors also reported a significant influence of “perceived zygosity”8 on conduct disorder and anxiety. Similarly,
another study found that measures of environmental similarity predicted greater phenotype similarity in symptom
scores for almost half of the tested health outcomes (Morris-Yates, Andrews, & Henderson 1990). Thus, evidence
suggests that the more similar environments of MZ co-twins are trait-relevant (influence concordance) for adverse
health outcomes and have the effect of upwardly biasing heritability estimates.
The EEA is even less reasonable for studies that include opposite-sex DZ twin pairs in their models. In
these cases, the model assumes that MZ co-twins (all of whom are same sex) are treated no more similarly than
opposite-sex DZ twins. Given the voluminous research on sex/gender differences in experiences, this assumption
seems questionable. Evidence for this interpretation is found in Meier et al. (2010) who compared the correlation
for childhood conduct disorder among opposite-sex and same-sex DZ twins and found that the opposite-sex
correlation was significantly smaller (approximately half) than that of the same-sex DZ twins (see also Saudino,
Ronald, and Plomin, 2005).
An even more questionable practice is the use of sibling data, which includes MZ twins, DZ twins, full
siblings, half siblings, and sometimes cousins. In these models, kinship pairs are compared with their average level
of genetic relationship, ranging from 1 for MZ twins to .25 for half siblings and .125 for cousins. These models
assume that the environments of opposite-sex, different age cousins who reside in different homes are no less
similar than that for MZ co-twins. The EEA seems absurd for studies of kinship pairs.
8 See Beckwith and Morris (2008) for a discussion of “perceived zygosity” tests of the validity of the EEA.
Moreover, while the above-mentioned evidence focuses on the greater postnatal environmental similarity
between MZ than DZ co-twins (and in some cases siblings), the greater environmental similarity for most MZ co-
twins actually begins in the womb (see Charney 2008; Joseph 2006). Although this greater prenatal similarity may
not be relevant for some traits, many adverse health phenotypes could conceivably be influenced by intrauterine
exposure to chemical toxins, viruses, or the like.9
In sum, for many health outcomes, the greater prenatal and postnatal environmental similarity for MZ
versus DZ co-twins (and especially when pairs of non-twin siblings or cousins are utilized) likely violates the trait-
relevant EEA (e.g., Charney 2012; Suhay & Kalmoe 2010). To be sure, for most health-related phenotypes, there
is much more work to be done in identifying pre- and postnatal environmental factors that contribute to the
development of adverse health outcomes. As such, the twin study model rests on the heroic assumption of the
equality of trait-relevant environments (many unknown) for MZ and DZ co-twins. To be sure, some studies
recognize and make a laudable effort to control for the more similar environments for MZ co-twins (e.g.,
Boardman et al. 2011); however, capturing the manifold of ways in which MZ co-twins environments are more
similar with survey items (especially for influences that are as yet unknown) is impracticable.10 Importantly, even
minor violations of the EEA can substantially inflate heritability estimates (thereby underestimating the influence
of the shared environment; e.g., Suhay & Kalmoe 2010). Thus, it is a credible interpretation of twin-study
findings that heritability estimates are upwardly biased due in no small part to MZ co-twins having more similar
shared environments than DZ co-twins (e.g., Horwitz et al. 2003; Jackson 1960; Joseph 2006). Surprisingly,
many recent twin studies of adverse health outcomes fail to even mention the EEA, and even fewer discuss the
implications of EEA violations.
9 While out of the scope of the present paper, scholars have noted that MZ and DZ co-twins experience prenatal environments that
are more stressful than that of singletons, thus calling into question the generalizability of twin studies (Charney 2012a: 19).
10 Some twin-study proponents have attempted to test the validity of EEA (at least postnatally) and/or the implications of EEA
violations on heritability estimates using survey items about environmental similarity as well as cases of misidentified zygosity (e.g.,
Boardman et al. 2011). Certainly, contradictory evidence has emerged from these studies (discussed in the text); however, some
studies report findings that are interpreted as bolstering the validity of the EEA. As discussed elsewhere, these tests are both limited
and problematic (e.g., Beckwith & Morris 2008; Charney 2008; Joseph 2004; Suhay & Kalmoe 2010). For example, these tests
frequently employ only a few broad (not necessarily trait-relevant) retrospective survey items about environmental similarity in
attempt to capture similarity in environments. As discussed, there are numerous (and some unknown) trait-relevant environments,
and these evade measurement in broad surveys especially with a limited number of items.
Genetic Additivity Assumption (#2)
Genetic influences can be separated into additive and nonadditive components. Nearly all twin studies assume an
additive model of gene combinative effects such that genetic variants each contribute small individual effects that
add up to shape variation in a phenotype. Nonadditive genetic variance is that which occurs as a consequence of
interactions between genes such that the phenotype is different from the sum of the individual genetic effects
(Stoolmiller 1999). Nonadditive variance can be one of two types. Dominance is that which arises when the alleles
at a given locus (one from each parent) interact to produce a phenotype and occurs among genes that operate with
a strict dominance-recessive mode of inheritance. Epistasis occurs when several genes (alleles at different loci)
interact to produce a behavior (Stoolmiller 1999). Although the extent of nonadditive genetic variance for various
health phenotypes is not known, there is good reason to believe that it is operative for all complex traits (Plomin
et al. 2012).
As with the EEA, violations of the additivity assumption inflate heritability and deflate shared
environmental estimates (see Grayson 1989; Stoolmiller 1999 for more detail). Basically, because MZ co-twins
share all of their genes, genetic nonadditivity will reduce the genetic correlation for DZ co-twins but not MZ co-
twins (because MZ co-twins have identical interacting alleles). Although dominance interactions can be modeled
in twin studies, this is not often done and comes with problems of its own (e.g., shared environmental effects are
ignored because they cannot be modeled simultaneously; Neale & Cardon 1992). Moreover, modeling epistatic
interactions with human kinship data is generally considered impracticable (Eaves 1988). Regarding nonadditivity,
renowned behavioral geneticists Plomin and colleagues (2012: 401) stated: “These types of effects complicate
model fitting because there are many forms in which they could occur. Normal twin study designs do not offer
much hope for identifying them.” In short, dominance and epistasis almost certainly upwardly bias heritability
estimates to an unknown degree (e.g., Burt & Simons 2014). Moreover, as with the EEA assumption, many twin
studies of adverse health outcomes fail to explain the implications of assuming genetic additivity and the
consequences of violation.
Nonblind Diagnosis (#3)
As noted, twin studies assume that MZ co-twins are not more likely than DZ co-twins to receive a diagnosis or
label in response to their twin’s diagnosis (or recognize their symptoms in the case of symptom reports). While
space constraints and a lack of empirical research on this issue precludes a detailed discussion, I believe it is worth
noting the possibility that nonblind diagnoses of co-twins with adverse health outcomes might bias heritability
It seems reasonable to expect that, in general, MZ co-twins are more likely to receive adverse health
diagnoses in response to a diagnosis of their co-twin than are DZ co-twins, all else equal, given the belief in the
heritability of many disease/disorder states, the recognition that MZ co-twins probably share more similar
environments, and the tendency to treat them more similarly. Moreover, given their greater attachment,
closeness, and feelings of similarity—as well as their recognition that they are genetic clones—MZ co-twins may
be more likely than DZ co-twins to identify shared symptoms as problematic or seek medical attention for such
symptoms in response to their co-twins diagnoses or symptom identification. Inasmuch as these practices are
operative, heritability would be inflated and shared environmental effects would be deflated.
Separate, Identifiable Contributions of Genes and Environments (#4)
As noted, the goal of heritability studies is to parse the effects of nature and nurture on phenotypic variance. Of
course, then, the model assumes that genes (G) and environments (E) have identifiably separate effects on
phenotypes. However, evidence suggests this is a fallacious assumption.
From a classical genetic perspective, there are two issues that confound the partitioning of G and E
effects: genetic-environmental covariance and gene environment interactions (G x E). Genetic-environmental
covariance is the association of certain genotypes with particular environments. An example in the health realm
could be individuals’ genetically-influenced tastes (Breen, Plomin, & Wardle 2006) shaping preferences for highly
processed, high sugar foods, which, in turn, could increase likelihood of diabetes, in combination with a host of
environmental influences (SES, physical activity level, parental monitoring/regulation of food consumption). It is
an ongoing matter of debate as to how to classify this covariance in the calculation of heritability. Perhaps most of
us would agree that this process fits neatly in neither the G nor the E category. Several behavioral geneticists have
argued that such gene-environment correlations should be classified as genetic effects (e.g., Fowler, Baker, &
Dawes 2008; Segal & Johnson 2009); however, as Rutter (2002: 4) noted, “it is misleading to suppose that just
because genetic factors influence the occurrence of an environmental risk factor, this must mean that the risk
process is genetically mediated. This assumption does not follow because there is no necessary connection
between the causes of the origin of a risk factor [taste preferences] and its mode of risk mediation [metabolic
effects].” I illustrate this point with the use of skin pigmentation.
In the U.S.A., a person genetically coded to have darker skin pigmentation will experience a different
social environment, on average, than one with lighter skin. There is a wealth of evidence that health is influenced
by environmental factors that are pervasively and systematically patterned along racial lines in the U.S. (Krieger
2000). Darker skin pigmentation is associated with exposure to environmental risk factors for adverse health
outcomes, such as exposure to noxious chemicals, due in part to its association with socioeconomic status,
community disadvantages, and environmental racism. Darker skin pigmentation is also associated with
interpersonal racial discrimination, which has been linked to a variety of health outcomes, including depression,
anxiety, hypertension, and the like (e.g., Burt, Simons, & Gibbons 2012; Brody et al. 2014; Williams 1999).
Surely, we can all agree that classifying adverse health outcomes that result from the interaction of skin
pigmentation with societal racism as due to genetic endowment is preposterous. However, even less manifestly
spurious cases are problematic, as the twin-study model has to classify such complex, interactional relationships as
either genetic or environmental. Such dynamic biopsychosocial associations defy such neat classification (Spencer &
Harpalani 2004; Wahlsten 1990).
G x E interactions, situations in which the effect of genotype on phenotype is conditioned by environmental
input (or environmental influences on phenotypes are conditioned by genotype), also thwart attempts to partition
genetic from environmental influences. The traditional twin study requires that one assumes G x E interactions
are nonexistent or that their effects are trivial (Wahlsten 1990). There is mounting evidence that this is not the
case, and that G x E interactions are the rule rather than the exception (e.g., Bagot & Meaney 2010; Rutter
2007). For example, a wealth of research indicates that much genetic variation among individuals influences their
sensitivity to environmental influences, rather than genotype having an unconditional effect on phenotypes (e.g.,
Caspi et al. 2003; Belsky & Pluess 2009, see Boardman & Fletcher 2014). Although evidence of G x E interactions
are not new (e.g., Hogben 1933), recent evidence documenting their prevalence is mounting with the advent of
new technologies that make genotyping easier and more cost effective, especially for complex socially-mediated
phenotypes (e.g., Belsky & Pluess 2013).
In the traditional way of partitioning, the effects of gene-environment correlations and interactions will
be included in the heritability estimate if the environmental effect is shared (such as skin pigmentation) and in the
nonshared environmental estimate if the gene-environment interplay operates in a twin-specific fashion. The
rationale is that because origins of the environmental risk factor (e.g., taste preferences or racial discrimination)
derive from genes (involved in taste buds or skin pigmentation), it is reasonable to attribute the whole of the
environmental effect to genetics (Rutter 2002). Clearly, this argument is unsound. Such ubiquitous interactional
relationships almost certainly render heritability estimates wildly inaccurate and deflate estimates of the shared
environment (Burt & Simons 2014).
Additional Technical Issues
Confidence Intervals and Model Fitting. An additional reason to view twin-study heritability estimates with caution is
that they tend to have large confidence intervals (Burt & Simons 2014; Rutter 2006). Frequently the point
estimate is the focus of attention, a practice that serves to reify an inherently imprecise estimate. Perhaps more
troublesome is the potential for bias as a result of these larger confidence intervals in the more sophisticated
structural models that have been more widely adopted. These models operate on the principle of parsimony, and
the usual approach involves a systematic comparison of different models with the aim of finding the simplest
model that fits the data. When the elimination of a parameter does not worsen the fit of the model, it is standard
practice to drop the parameter.11 Importantly, the decision to drop a parameter is based on the lower end of the
95% confidence limits without reference to the upper end, and given the large confidence intervals could result in
a parameter being dropped that in fact has a significant effect (Rutter 2006). Although many studies do not report
11 This is customary practice in much structural equation modeling; however, it is important to remember that the lack of a
significant effect of the shared environment is not tantamount to evidence of a zero effect (Burt & Simons 2014).
the estimates from the ‘inferior’ models, when these estimates are presented, it is not uncommon to see effects
dropped that have confidence intervals that range from 0 to more than 30% (e.g., Boardman, Alexander, &
Stallings 2011). In practice, this approach has often led to the elimination of shared environmental effects with the
consequence of exaggerating genetic and nonshared environmental estimates (Rutter 2002).
Phenotype Specification. Heritability studies are, of course, dependent on the accuracy of the measure of the
phenotype. Although phenotype ambiguity is perhaps more obvious in some domains, such as “liberal” and
“conservative” political phenotypes and criminal phenotypes, phenotype ambiguity in adverse health outcomes is
also an issue. Several scholars have written on this issue in relation to other phenotypes (e.g., Charney 2008;
Duster 2003; Press 2006) so I will not belabor the issue, but a few deserve brief mention. Here I focus on two
points that relate to the inherent imprecision and social construction involved in the creation of (most) health-
related phenotypes: treating quantitative variables as dichotomous and treating ongoing processes as fixed.
Perhaps most importantly, none of the phenotypes that fall under the heading of health or illness are
unequivocally biologically-given classifications. The classification of health and sickness is not a unique division of
biological reality, but rather there is an inherent degree of social construction and vagueness in these classifications
(e.g., Hubbard & Wald 1999). To be sure, some phenotypes clearly involve more social influence in
categorization than others. ADHD, for example, is extremely elastic and relies on the presence of symptoms such
as “Often talks excessively” and “Is often easily distracted.” Not to be captious, but “often” “excessively” and
“easily” are rather indefinite terms, and these broad statements conflate domain-specific attributes. However,
other classification systems that may seem more given and unproblematic involve a degree of indeterminacy and
arbitrariness, such as the classification of obesity (BMI >=30.0) and high blood pressure (systolic greater than 140
or diastolic greater than 90). Biology does not provide a neat cutoff for obesity at 30.0 and hypertension at greater
than 140/90. This tendency to treat continuous variables as dichotomous or categorical and biologically given is
thus unavoidable in health research and inevitably shapes results (Joseph 2004; Press 2006).
Social forces influence not only the creation of classification systems, but also their application and
meaning. Where along the line that we (or others) decide we are “sick” or “healthy” depends on a number of
individual and social characteristics (Hubbard & Wald 1999; Rose 2006). Focusing on mental illnesses, for
example, Kessler and colleagues (2005) concluded that “about half of Americans will meet the criteria for a DSM-
IV disorder” in their lifetimes, and roughly 26% in the past year. Yet, less than half of individuals meeting
diagnostic criteria for these disorders receive treatment and thus are (potentially) diagnosed (Kessler et al. 2001).
For many adverse health outcomes, then, the focus of attention is on the perception by the individual or his/her
medical team that symptoms are problematic and/or qualify as a disease or disorder.
Additionally, it is often the case that in measurement we treat categories as fixed variables (e.g., a
“smoker”), neglecting the reality that these are often ephemeral states in a long developmental process. Individuals
may often, at different points in their lives, experience episodic states of depression, anxiety, high cholesterol,
hypothyroidism, alcoholism, and other conditions. Classifying such processes into static phenotypes ultimately
myopically focuses on a tiny snapshot of an ongoing developmental reality.
To be sure, I am not arguing that because most health phenotypes are not given by biology and thus are in
some degree social constructions, this implies that these categories are not based on something real, that they
should not be used, or that they are not scientific.12 The fuzziness and plasticity of classification does not invalidate
them on their face. For example, the fact that the precise boundary between day and night cannot be demarcated
does not mean that night and day cannot be meaningfully differentiated (Leahey & Leahey 1983). Rather, this
discussion is intended to remind that phenotypes in heritability studies are not clear, biologically-given categories
that are fixed within the individual or across time and space. Instead, phenotype classification necessarily simplifies
a much more complicated biological and social process (Dupre 2012).
Although a thorough discussion of an additional point is not possible given space limitations, some
phenotypes, in my view, are ill-suited for genetic investigation (e.g., “smoking desistance” or “age of first
cigarette”), as they are manifestly the result of a multitude of individual traits that combine with numerous social
influences in an ongoing process of development (much like “age of first game of golf” or “bacon-eating
12 Recognizing that science is a social enterprise that is produced by human agents implies that science is to some degree a social
construction. This does not imply an anti-realist position or that science is not about truths. Science is about truths, but schemes of
classification are not inherent in the nature of most phenomena of interest to social and biological scientists (see Dupre 2012 for an
excellent discussion of this point).
desistance”). Smoking studies often utilize samples of nonsmokers, former smokers, and current smokers thereby
reifying these as “actual categories of human beings,” and “neglect[ing] the profound and unique role of human
cognition and self-reflexivity in regard to behavior” (see Press 2006: 143-44).
Summarizing Technical/Statistical Limitations
For at least 80 years scholars have criticized the technical limitations of heritability studies from a classical genetic
perspective. Scholars have argued that given these flaws, which altogether have the potential to wildly bias
heritability, render the estimates of heritability studies too problematic to be of scientific value (Charney 2008;
Joseph 2004; 2006; Lewontin 1974). I concur with these scientists. Regardless of one’s views on this issue,
however, understanding the assumptions and inherent technical limitations of heritability studies is important
given the prevalence of these studies and estimates.
Although I think understanding these methodological limitations is useful, as I discuss below, in my view
the fatal flaw with heritability studies is conceptual. Although critics have denounced the conceptual model of
heritability studies as flawed for decades (Hogben 1933; Lewontin 1974; Wahlsten 1979), recent evidence
emerging from molecular genomics evinces the inseparability of genetic from environmental influences and has
begun to identify mechanisms through which environmental factors influence genetic activity and thereby shape
phenotype development. These recent advances have uprooted nearly all of the assumptions of the classic gene-
centric paradigm undergirding heritability studies, including the notions that DNA is the sole biological agent of
heritability and that genes are fixed entities that are encapsulated within the cell, produce proteins in a
straightforward manner, and serve as the driver/executive of genetic activity within the cell (Charney 2012a;
Griffiths & Stolz 2013; Keller 2010). Indeed, this new knowledge has transformed the scientific understanding of
the gene and the relationship between genotype and phenotype and marks a shift from the classic genetic paradigm
to a postgenomic view (Charney 2012a; Jablonka & Lamb 2006; Keller 2010). The postgenomic paradigm “is
characterized by extreme complexity, variability, multilevel reciprocal interactionism, and stochasticity as an
inherent property of biological systems, all of which contribute to what might be called the blurring of
boundaries, in particular, the boundary between genes and the environment” (Charney 2012a:332). As I discuss,
heritability studies have no place in the postgenomic era.
POSTGENOMIC CHALLENGES TO HERITABILITY STUDIES
“In the emerging postgenomic paradigm, we are confronted with a biological world that is in many ways the opposite of that
which has thus far enabled the methodologies of behavioral genetics” (Charney 2012a: 61).
When the Human Genome Project (HGP) was launched in 1990, many scholars anticipated finding a
straightforward relationship between the complexity of an organism and its number of genes. With over 100,000
identified proteins in the human body, assuming each gene codes for a different protein, researchers projected that
the human genome consisted of at least 100,000 genes. These projections were quite inaccurate. Completion of
the HGP in 2001 revealed that the human genome has only roughly 23,000 genes (this number is continuously
being revised), slightly more than the fruit fly and the roundworm and less than corn (Charney & English 2012;
Claverie 2001; Venter et al. 2001). This finding was clear evidence that the predominant one gene—one protein
assumption was incorrect (Silverman 2004). We now know that a single gene can code for multiple proteins,
something that is estimated to occur in as much as 90% of all human genes (Charney 2008). These findings
combined with the failure of genome-wide association studies (GWAS) to find strong main effects of genes on
phenotypes, led to a rethinking about genetic functions, including questioning the utility and meaning of a “gene”
(Jablonka & Lamb 2006; Keller 2010).
In response to the flaws inherent in the classic “gene-centric” paradigm, the new postgenomic paradigm
has begun to emerge (Charney 2012a). Key to the postgenomic paradigm is the recognition of an interactional,
bidirectional relationship among genes, cells, organisms, and environments as well as adaptive developmental
plasticity—the capacity of organisms to modify physiological, morphological, or behavioral phenotypes in
response to environmental conditions (Charney & English 2013). Rather than being a code, program, or blueprint
for development, genes are now understood to be cellular resources (Gottlieb 1997; Keller 2010; Charney
2012a). The activity of the genome is regulated as part of the individual’s general developmental-physiological
adaptation to environmental signals—a “constant interplay between biology and experience” (Lickliter &
Honeycutt 2003: 820). Although these recent advances have become dominant/influential in biological science,
regretfully they have not yet fully penetrated the social and behavioral sciences. Below I briefly discuss advances in
molecular genetic knowledge that debunk the genetic model that undergirds heritability studies.
“…[I]t is important to remember that by itself, DNA is an inert, sticky glop.” (Hubbard 2013:22)
Genes, while certainly integral to life and variation, are now understood to be inert molecules incapable of doing
anything on their own. For many years, genes were thought to be a self-activating code for the production of
protein products (the “central dogma” of molecular biology); however, we now know this is not the case. Instead,
we now understand that cells are the directors of genetic and other cellular activity, and genes are merely a
resource upon which cells draw to respond to environmental signals. In thinking of this revised role of DNA and
genetic function, Hubbard and Wald (1999) used an excellent analogy of DNA as a cookbook. Here, genes are
recipes, and cells are the cook. In order to make a complex dish, a cook needs a recipe, which is contained in a
cookbook (genome). Of course, the cookbook does not determine the recipe, and the recipe does not make the
dish or determine how it will turn out. The dish depends on a number of external factors, including those that
have nothing to do with the recipe itself or the cook. Such is the relationship among genes, cells, DNA, and the
environment. Cookery is also an apt metaphor because it allows for a degree of flexibility that is also inherent in
DNA action, as genes are utilized by the cell to respond to environmental changes (Hubbard & Wald 1999).
Although molecular epigenetic research is highly specialized and technical, understanding some basics is
useful given that this provides evidence of the effects of environmental and behavioral influences on genetic
activity. As I have noted, genes are not self-activating. Instead, DNA has to be transcribed to produce RNA and
proteins,13 but in order to be transcribed it has to be “turned on”. The process of gene activation, often called gene
expression, is regulated by the epigenome—the biochemical regulatory system that can turn on, silence (leave off),
or change the transcriptional availability/activity of genes (Charney 2012a; Martiensen, Riggs, & Russo 1996). In
a multitude of ways (see e.g., Charney 2012a; Jablonka & Lamb 2006), the epigenome regulates gene expression.
13 In a process known as transcription, the DNA molecule is used by a cell to produce messenger RNA (mRNA), which in turn
serves as a template for the synthesis of polypeptides, which form proteins (a process called translation).
Most gene regulation occurs in response to the immediate demands of the environment, takes place in
time spans ranging from seconds to weeks, and differs between specialized cells (Francis 2011; Jablonka & Lamb
2006). For example, when eating a hamburger, our tongue, stomach, and pancreatic cells will react differently
and many cells types will not react at all. With very few exceptions, these different cell reactions are due to
epigenetic variation across the cells (Jablonka & Lamb 2006). In recent years, scientists have focused on gene
regulation that takes place over much longer intervals. This rapidly advancing field, known as epigenetics,14
focuses on the mechanisms of gene regulation implicated in changes in gene expression and phenotype, which can
last for weeks, months, years, or across the life span and can even be transmitted onto future generations
(epigenetic inheritance) (Bagot & Meaney 2010; Bollati & Baccarelli 2010; Charney 2012a; Francis 2011).
Importantly for our purposes, epigenetic research reveals that the epigenome is responsive to environmental input
(both internal and external to the cell); thus, the environment influences gene activation through the epigenome
(Charney 2012a; Jablonka & Lamb 2006).
Although new epigenetic mechanisms are still being uncovered, the most well understood fall into a class
known as chromatin markers, often referred to as “epigenetic markers.” Two of the most well researched
epigenetic markers are histone acetylation (Grunstein 1997) and DNA methylation (Bird 2002). These markers
(posttranslational modifications) affect cellular activity and phenotypes by influencing the accessibility of DNA to
transcription factors, thereby shaping whether (and to what extent) a gene is activated. Research documenting the
effects of environmental factors on epigenetic markers is accumulating, thereby illustrating ways in which gene
expression is influenced by the environment (Bagot & Meaney 2010; Charney 2012a; Francis 2011).
Environmental epigenetics focuses on the effects of social-environmental factors such as pollution, nutrition,
parental care, and stress on gene regulation (Landecker & Panofsky 2013). Studies in this field (primarily on non-
human animals) identify lasting effects of environmental factors on phenotypes through the epigenome. Research
14 The term epigenetics has undergone many transformations in meaning since the term was coined, and even now continues to be
employed loosely and interchangeably with other related concepts (see Jablonka et al. 2009). Conrad Waddington coined the term
epigenetics in 1942 to refer to the developmental processes involved the development of a phenotype, recognizing that such a
process is not a simple genotype !phenotype relationship. However, the modern use of the term epigenetics is usually narrower,
and can be thought of as the study of the processes of gene regulation (e.g., Meloni 2014).
in the emerging field of neuroepigenetics reveals that adverse early environments and injurious social experiences
generate epigenetic changes that shape brain architecture and consequently responses to later social experiences
and adversities throughout the lifespan (e.g., Hoffmann & Spengler 2014; Sweatt 2013). Recent evidence from
animal models suggests that such “experience-dependent DNA memories,” inscribed epigenetically, can play key
roles in development and stress responsivity (e.g., Hoffman & Spengler 2014; Murgatroyd et al. 2009) and can
have transgenerational effects (i.e., environmentally-induced epigenetic changes can be transmitted to offspring;
Franklin et al. 2010).
For illustration, in their now famous studies, Weaver and colleagues (2004; 2005; 2006) connected stress
phenotypes (HPA responses to stress) in adult rats to early maternal care. Female rats who engaged in more
frequent licking and grooming of their pups during the first days of life raised offspring that were less reactive to
stress across their lifespan and, among the females, went on to become high-licking and high-grooming mothers
themselves. Weaver and colleagues (2005; 2006) linked this early maternal care to epigenetic changes (DNA
methylation) in the glucocorticoid receptor promoter, and further work has identified similar epigenetic changes
linking early maternal care to hundreds of other genes in the rat pups, demonstrating the manifold biological
effects of early care on the developing brain (at least in the rat). Importantly, while these results showed that
early-life parenting experiences have a stable and broad effect on several health-related phenotypes, they also
revealed that these epigenetic markers are reversible in adulthood (Weaver et al. 2005; 2006) consistent with the
notion of organisms’ adaptive phenotype plasticity in response to changes in social conditions (e.g., Ellis et al.
In addition to maternal care, studies of rodents have linked exposure to a number of substances, such as
pesticides, fungicides, BPA, cigarette smoke, vehicle exhaust, and cocaine, to methylation or other epigenetic
markers with concomitant changes in gene expression and phenotypes, including physiological function,
behavioral outcomes, and propensity to cancers and metabolic disorders (e.g., Feil & Fraga 2011; Landecker &
Panofsky 2013). For example, pesticide exposure has been linked to sperm defects and subfertility as well as
decreases in fearful behaviors (see Skinner 2008). These changes occur without changes in DNA sequences, but
rather through changes in the epigenome (e.g., DNA methylation) associated with genes linked to these
phenotypes (Charney 2012a).
While there are many more examples, these studies are sufficient to illustrate the fact that epigenetic
research evinces that genetic and environmental effects are interactional and dynamic, and thus differences in
phenotypes cannot be separated into genetic versus environmental influences. Epigenetic markers embody the
blurred boundaries between genes and environments and the difficulty of classifying molecular mechanisms in
gene expression in the G versus E dichotomy. For example, as Charney (2012b) questioned, how would
heritability study proponents have us classify experience-dependent epigenetic marks (e.g., chromatin
architecture) in the G vs. E dichotomy: As G because chromatin attaches to the genome and plays a role in
silencing or turning genes on? Or as E given that chromatin structure is shaped by environmental input? What
about the inheritance of environmentally-induced epigenetic reprograming in the absence of the original inducing
environment (e.g., pesticide exposure, maternal care): As G because it is inherited? Or as E because it originated
with an environmental stimulus? Or is E transmuted to G? These instances of the blurring of boundaries between
genes and environments are a reoccurring theme in modern genomics research and point to the utter impossibility
of separating “genetic” from “environmental” in development (Charney 2012a).
In short, the model of a distinct, particulate gene that directs cellular activity and performs one job
(produces one protein product) independent of environmental input has been discredited in the past decade. The
mere presence of a gene as part of a genotype does not ensure that it will be activated. Moreover, although
evidence suggest that environmental events that occur early in life (e.g., prenatal care; early maternal care;
exposure to toxic substances) tend to produce more pronounced epigenetic effects than those that occur later,
methylation and other epigenetic processes continue throughout the lifespan (Francis 2011; Weaver, Meaney, &
Szyf 2006) and are involved in (and necessary for) both healthy development as well as disease progression.
Epigenetic research demonstrates that the genome is flexible in its expression, and this expression is responsive to
context, experience, and developmental history (Griffiths & Stoltz 2013). As Dupré (2012: 3) noted: “One way of
beginning to think about epigenetics is to realize that the genome, as much as the organism, is a process rather
than a static thing.”
Several assumptions are necessary for heritability studies to be meaningful, but the most crucial of these is
the (biological) assumption that genes and environments have identifiably separate effects on phenotypes (Charney
2012a). Biological evidence is clear that this is not how genes work.
ADDITIONAL GERMS FOR THOUGHT: WHAT ABOUT MICROBES?
As I noted, the finding from the HGP that the human genome consists of only ~23,000 genes was a surprise to
many, as researchers projected roughly 100,000 genes. However, if the view of what constitutes the “human
genome” is expanded, then 100,000 genes is a drastic underestimate (Turnbaugh et al. 2007). The human body is
a complex, symbiotic whole comprised of many different microbial organisms (e.g., bacteria and fungi) of
different kinds and genomes that live on or inside human tissues. Astonishingly, evidence suggests that within our
bodies 90% of the cells are actually microbes, most of which inhabit the gastro-intestinal tract, and 99% of the
genes on or under our skin are actually bacterial (Xu & Gordon 2003). Thus, the microbiome contains at least 100
times as many genes as our own genome (Gill et al. 2006).
Over the past decade along with (or in the wake of) new technologies for studying microbiota (a
microbial population), interest in the role of the microbiome (the set of microbial genes) in human health has
surged (Cho & Blaser 2012). Ambitious large-scale projects, including the U.S. Human Microbiome Project
(Turnbaugh et al. 2007) and the European project MetaHIT (Erlich 2011), have begun to advance knowledge on
the biological properties and medical significance of the human microbiome and its collective genes (metagenome;
Cho & Blaser 2012). These projects in combination with smaller scale efforts have begun to elucidate the
composition and development of individuals’ microbiomes as well as their influential role in health and
development. Several findings have emerged, including the density of the microbiome at various sites in the
human body, the extent of inter- and intra-individual variation, and the dynamic and significant interplay between
microbial and human cells. Acknowledging the significant role of the microbiome, Gill et al. (2006: 1354) noted
that the microbiome “endows us with physiological properties that we have not had to evolve on our own.”
Over the course of a lifetime, each individual develops a densely populated microbiome. Individuals,
even from the same region and similarly aged, have quite different microbiomes, and differences between
individuals appear to be much greater than intra-individual variation (at least over short periods of time; Cho &
Blaser 2012).15 The colonization of the human gut begins at birth, as infants pick up billions of microbes as they
move from a sterile womb through a microbe-laden vagina (if a vaginal birth, which of course was the only form
of birth throughout most of human history). Thus, an individuals’ microbiome appears to be a consequence of this
important initial composition from the mother as well as a dynamic interrelationship between environmental
(e.g., diet, stress, exposure to antibiotics and probiotics; David et al. 2013; Jakobsson et al. 2010) and biological
(likely including genetic) factors throughout development.
Much recent research has focused on the role of intestinal microbes in maintaining health and
homeostasis. Scholars have explained that our gut microbiota represent a “virtual inner organ” (Forsythe et al.
2010: 9). Although we are only beginning to understand how microbiome composition and function influences
human health, a spate of recent studies suggests that microbes have significant influences (in interaction with
genomes and environments) in shaping health outcomes (e.g., Cho & Blaser 2012). In particular, we have clear
evidence of the role of gut microbes in processing indigestible foods, such as plant polysaccharides, synthesizing
essential amino acids, and therefore obtaining necessary key nutrients from our food (e.g., Gill et al. 2006). More
recently, evidence of the influence of gut microbes through gut-brain signaling in homeostasis and nervous system
functioning is beginning to accumulate (Mayer 2011; Forsythe et al. 2010). An expanding body of literature
explores pathways of influence and links individuals’ microbiota to adverse physical health outcomes through
complex channels involving the immune system and host metabolism (Forsythe et al. 2010; Grice & Segre 2012).
Although this work is still in its infancy, these nascent findings suggest “that there is much communication and
interaction between microbial and human cells, and there is growing evidence that (gut) microbes have significant
distal influences on health and disease through their effects, for example, on the immune system and on
15 This is complicated by the fact that research suggests that bacteria from a specific body part have more in common than those
from a specific person. Your armpit microbes have more in common with my armpit microbes than they do with your oral
microbia (Ding & Schloss 2014).
neurotransmission” (Dupre 2012: 13). At present, gut microbes have been implicated in such diverse outcomes as
chronic inflammation, anxiety, mood, aggressiveness, eating disorders, cognitive functioning, diabetes, psoriasis,
autism, allergies, and atherosclerosis (Forsythe et al. 2010; Grice & Segre 2012; Mayer 2011; Turker et al. 2014).
This discussion of the role of microbes, which is necessarily abbreviated but hopefully stimulating, is
intended to further cloud the reductionist, false genes versus environment dichotomy in shaping human
phenotypes. Microbes exist within us and greatly outnumber our cellular and genetic material, and our
microbiomes are a consequence of a complex interaction of biological and environmental factors, which
consequently mediate and condition the effects of genetic and environmental influences on health. Indeed,
emerging research indicates that microbiota influence host gene expression (for example, gut microbiota influence
the expression of genes involved in host digestive functions and metabolism; Combe et al. 2014).
How would heritability study proponents have us classify the effects of microbes? As G because the
microbiome, although not strictly human genes, fulfills functions that we have not had to evolve on our own? As
G because microbiota can influence the epigenome and gene expression? Or as E because they are not simply
human DNA sequences, even though many are intergenerationally transmitted (through the birth canal)?
Moreover, although evidence suggests that individuals’ microbiomes become increasingly stable over time (e.g.,
Forsythe et al. 2010), changes in one’s lifestyle (diet, activity level) and medications (e.g., antibiotics) can alter
the composition of one’s microbiome, along with ontogenetic development (e.g., Cho & Blaser 2012; David et al.
2014). Humans are living systems that exist and develop in a synergetic, commensal relationship with other
organisms (and their genomes), and such relationships defy classification into (human) genetic versus
EVEN IF WE COULD: HERITABILITY ESTIMATES LACK UTILITY
Quite aside from their many flaws, heritability estimates have little practical utility, and this reality is recognized
at present even by prominent behavioral geneticists. For illustration, Eric Turkheimer, one of the most well-
known behavioral geneticists, who developed the “three laws of behavioral genetics,” provided a summary of why
he thinks heritability studies lack utility: “…it is not about cause. Practitioners of the art wanted it to be about
cause, in the sense that the relative magnitudes of the various components were supposed to tell us something
about the importance of genetic and environmental causes underlying a trait, but they do not” (Turkheimer 2011:
598; emphasis added).
Heritability estimates do not forecast the developmental endpoints for individuals or groups, the
consequences of interventions, or the causal processes or mechanisms involved in phenotype variation (Burt &
Simons 2014: 246; see also Rutter 2002). A high heritability coefficient says nothing about the amount of genetic
influence on a trait (recall the ‘eyedness’ example), quantifying heritability does not contribute to the
identification of specific candidate genes16 (e.g., Schielzeth & Husby 2014), and heritability implies nothing about
the potential for change through environmental alterations (see Joseph’s (2004) PKU example).
Moreover, it bears repeating that heritability estimates are time, context, and population-specific
averages. A heritability estimate could be analogous to an estimation of the average temperature of North America
based on taking the average temperature from randomly selected cities on a given day (Kruger et al. 2008). The
usefulness of such an estimate is scanty to naught. Importantly, and in contrast to the average temperature of
North America, which has a true value, “[t]he proportion of variance attributable to A [additive genetic variance],
C [“shared” environments] and E [“unshared” environments] depend on the variances of A, C and E, and the
variances of A, C and E have no ‘true’ values to be estimated” (Turkheimer 2011: 598).
To be sure, this is not an argument that nothing positive has come out of heritability studies, because
whatever their flaws (and there are many) they have likely significantly contributed to the acceptance of genetic
influences on various phenotypes among social scientists. However, at the present state of knowledge, the numerical
estimates of heritability, even if they could be accurately estimated, are neither of practical use nor do they
advance scientific understanding of cause (see also, Turkheimer, Pettersson, & Horn 2014).17 As Chaufan
16 As Schielzeth and Husby (2014:45-6) note, “there is in general no reason to expect that traits with a high proportion of additive
genetic variance should harbor large-effect variants rather than many loci with small effects...From a statistical viewpoint, what
matters is a low amount of residual variance, not a high heritability per se...a direct link between heritability and the number and
effect size of trait loci is not compelling.”
17 Indeed, Turkheimer (2011:598) noted that “Plomin and Daniels  marked the beginning of the end” of heritability studies as
a useful scientific endeavor.
(2008:24) noted, heritability estimates “provide no information relevant to … medical practices and public health
policies that we do not already have.”
Finally, given their problems, heritability estimates can (and likely have) misinformed and misguided
efforts to understand the etiology of and therefore prevent or reduce adverse health outcomes (e.g., Lewontin
1974). For example, the absorption of environmentally-induced, intergenerationally-transmitted epigenetic
markings into the G in heritability studies could lead scholars to conclude that the cause is inherited genetic
defects, and hence environmental culprits (e.g., pesticides, maternal care) may not be considered (see Charney
CUTTING-EDGE TECHNIQUES MEET OUTDATED NATURE VS. NURTURE DICHOTOMY
Biometrical Moderation Models
Advances in statistical modeling over the past decade have enabled heritability researchers to incorporate
environmental contingencies into their variance partitioning analyses, “allowing the quantification of phenomena
that have traditionally been characterized as gene-environment interaction and correlation” (Kruger et al. 2008:
1485). Several recent studies of adverse health outcomes have incorporated environmental contingencies (e.g.,
Boardman et al. 2012; Johnson et al. 2010; Kruger et al. 2008; Tuvblad, Grann, & Lechtenstein 2006). For
example, Boardman and colleagues (2011: 1517; 2008) utilize biometrical moderation models with various
smoking phenotypes, exploring, for example, “the effect of social policy on the extent to which genes influence
[variance in] smoking desistance.”
Although this might seem to be a useful strategy to overcome the G x E and rGE conundrum besetting
heritability studies, here is why I think it is not. First, these models are still based on the flawed twin study model
(with its various assumptions aforementioned). Second, these studies seek to identify one (or a select few)
environmental contingency out of an infinite number of possibilities thereby having a significant risk of omitted
variable (environmental contingency) bias. For example, in their study of body mass (BMI), Boardman and
colleagues (2012: 370) compared the heritability of BMI across schools (using a twin-study of kinship pairs nested
within schools) and examined the relevance of “school differences in both health-related policies and social norms
regarding body size” on heritability estimates using the Add Health data. The authors examined whether various
school characteristics (school administrators’ views of the severity of deviant behavior problems; punitiveness of
school punishments, prevalence of weight loss resources), as well as school norms about body size (average school
BMI for students who classify themselves as ‘normal weight’) and norm enforcement (school variance around this
average for ‘normal weight’ self-classified students) influence the degree to which BMI is heritable. Boardman et
al. (2012: 383) found that heritability estimates varied across several of these environmental characteristics, and
they concluded that school social norms and institutional policies influence genetic contributions to weight gain
and thus “provide new insights into social and institutional factors, while also helping to better implement social
Certainly it is possible, indeed likely, that school-level factors have some effect on BMI and therefore the
degree of genetic influences on variation in BMI. However, it is also plausible that wider contextual factors (e.g.,
community disadvantage, population SES, prevalence of healthy food options in the community) shape school
norms. Moreover, the reverse causal pathway may be equally or more significant. In other words, students’ BMIs
may be driving the observed “school effect.” For example, factors exogenous to the school (SES, geographic
location, school funding (monies devoted to more expensive, healthier food options)) may be shaping average
school BMI, and school policies may be endogenous to these factors. Concomitantly, a community (and its
residents’) history of disadvantage (or advantage) may have fostered unhealthy (healthy) eating patterns among a
population, which, in turn, may engender epigenetic changes transmitted to future offspring that increased the
likelihood of metabolic problems and a higher BMI. Analogously, school foodstuffs (e.g., cafeteria meals, snacks,
soda availability) influence the food consumption of students, which over time likely influences the microbiomes
of the population of students, and, in turn, interacts with genetic and environmental factors to shape metabolism
and food preferences, and hence BMI. In short, there are many alternative explanations that are not examined
(and some that cannot be examined at present), many of which are not neatly classified as genetic or
Notably, such estimates are not about mechanisms or cause. In contrast to what some studies argue or
imply, identifying variation in heritability across various environments does not shed light on the genetic
mechanisms underlying phenotypes (or phenotype variance) nor the potential efficacy of environmental
interventions. Genetic mechanisms are not the focus of such studies. [Thus, to use the BMI example, such
biometrical moderation models do not indicate “the extent to which the genetic influences on body mass are
different,” but rather differences in the extent to which such observed variation is credited to additive genetics
(Boardman et al. 2012: 370)]. For elucidation, identifying the school factors that shape heritability of BMI can be
analogized with the earlier cornfield example. Showing that cornfields with trees around them have lower
heritability than those without tells us nothing about the causes of plant height, genetic or environmental. In the
same way, identifying school factors tells us nothing about the causes of BMI, genetic or environmental.
Furthermore, as noted above, a high heritability signifies nothing about the potential success of environmental
interventions. Thus, identifying differences in heritability does not inform “debates about the most effective means
to improve the health of the public” nor does a high heritability indicate that “policies may be ineffective for those
who are more likely to gain weight because of very small differences across their genome” (Boardman et al. 2012:
Moreover, and as should now be clear, my view is that this attempt to separate G from E is impossible. It
is not merely that separate G and E effects interact to influence a phenotype in ways that are potentially separable
at the population level; instead they do not have separate influences on phenotypes in the first place. As
Turkheimer (2011: 600) notes, “individual differences in complex human characteristics do not, in general, have
causes, neither genetic nor environmental. Complex human behaviour emerges out of a hyper-complex
developmental network into which individual genes and individual environmental events are inputs. The
systematic causal effects of any of those inputs are lost in the developmental complexity of the network.”
Genome-Wide Complex Trait Analysis
Recently, scientists have developed a new methodology for identifying gene variants in human populations called
genome-wide complex trait analysis (GCTA; Yang et al. 2010; 2011), and over the past four years its use for
health phenotypes has burgeoned. In contrast to standard heritability estimates, which do not measure genetic
factors and instead rely on various assumptions around genetic relatedness, GCTA studies require that individuals
are unrelated and rely on a genetic relationship matrix comprised of common single-nucleotide polymorphisms
(SNPs) (Yang et al. 2011). GCTA studies avoid many of the abovementioned technical limitations that result from
the fact that standard heritability models measure neither genes nor environments and thus rely on a host of
assumptions. Basically, GCTA studies involve scanning hundreds of thousands of SNPs of thousands of (ostensibly)
unrelated persons in a sample with the goal of determining whether phenotype similarity can be linked to various
SNPs (Yang et al. 2010). The goal is an estimate of the heritability of a phenotype that is accounted for by the
additive effects of shared SNPs (Charney 2013).
At present, GCTA results appear to support the arguments of twin and adoption study critics as they have
yielded significantly lower and in some cases nonsignificant heritability estimates even when substantial twin study
heritability estimates are found in the same sample (Charney 2013). For example, in their study of five anxiety-
related traits (after a GWAS finding of “no common genetic variants of large effects that contribute to the
heritability of these traits”), Trzaskowski and colleagues (2013) reported an average GCTA heritability estimate of
10%, which they noted is less than one fifth of the average twin-study heritability estimate of 55%. Similarly,
comparing GCTA and twin study estimates, Trzaskowski, Dale, and Plomin (2013: 1048) concluded: “Behavioral
problems in childhood [which included health-related phenotypes such as depression]—whether rated by parents,
teachers, or children themselves—show no significant genetic influence using GCTA, even though twin study
estimates of heritability are substantial in the same sample…” Thus, GCTA studies provide further evidence that
something is awry in the high behavioral genetic heritability estimates of adverse health phenotypes (Burt &
To be sure, the lower estimates from GCTA studies may be due in part to the fact that they only capture
the additive effects of genetic variants that are in linkage disequilibrium with the common SNPs analyzed (rare
SNPs are excluded) in genome wide association research (Yang et al. 2010). However, like standard heritability
studies, GCTA studies rely on problematic assumptions whose violations can increase the likelihood of spurious
associations and hence flawed estimates, such as population stratification (PS) (Charney 2013).18 PS raises the
possibility of false associations due to gene-environment correlations (i.e., more genetically similar individuals are
more likely to experience certain environments, discussed above with the example of skin pigmentation).
Notably, research has demonstrated that GCTA studies are particularly vulnerable to PS confounding (e.g.,
Browning & Browning 2011; Janss et al. 2012). While the developers of GCTA are cognizant of the problems
posed by PS and do make attempts to correct for it (Yang et al. 2011; in practice see Boardman et al. 2014), these
corrections have been shown to be insufficient (for more information, see Charney 2013).
In sum, like other heritability studies, GCTA studies suffer from methodological limitations that
undermine the legitimacy of their findings and augment the likelihood of spurious genetic associations. Ultimately,
however, GCTA studies, like other heritability models, rest on an outdated, false gene vs. environment
dichotomy. To repeat, the lack of scientific merit in heritability studies is not limited to technical/statistical
problems that can be improved over time or can be addressed with new cutting-edge techniques that do not rely
on the same dubious methodological assumptions. The problem is conceptual (biological): genes and
environments do not have identifiably separate effects on complex phenotypes.
“One of the most striking features of the nature-nurture debate is the frequency with which it leads to two apparently
contradictory results: the claim that the debate has finally been resolved (i.e., we now know that the answer is neither nature nor
nurture, but both), and the debate’s refusal to die.” (Keller 2010: 1).
In this chapter, I have argued that there is compelling evidence that heritability studies are methodologically
flawed, especially for complex adverse health phenotypes. I have also argued, drawing on recent advances in
molecular genomics and epigenetics, that heritability studies are grounded on a specious conceptual foundation.
Recent advances in molecular genomics have debunked nearly every assumption that underlies heritability studies.
This new evidence manifestly supports, indeed proves, the arguments that critics of heritability studies have been
18 Population stratification refers to the nonrandom distribution of polymorphisms in different populations (ethnic or geographical)
due to unique ancestral patterns of migration, mating practices, and reproductive patterns (Charney & English 2013; Hunley,
Healy, & Long 2009). Most nonfamilial populations exhibit PS, including those considered relatively homogenous (e.g., among
Icelanders; Charney 2013).
making for decades that the goal of partitioning genetic and environmental effects on variance in phenotypes is
unsound (e.g., Lewontin et al. 1984). In short, heritability studies attempt the impossible. Furthermore, that
these methods are fatally flawed is no great loss to science; heritability estimates lack utility and are not about
cause (Rutter 2002; Turkheimer 2011). Joining the arguments of others (e.g., Charney 2012; Lewontin 1974;
Turkheimer 2011), I contend that further attempts to estimate heritability for complex phenotypes, given what
we now know, are a manifest misuse of scholarly energy and attention (e.g., Chaufan 2008; Joseph 2004;
Lewontin 1974). Moreover, given their many flaws, I urge scholars to recognize and acknowledge the
problematic nature of existing heritability estimates and end the frequent use of the phrase: “We know from a
wealth of behavioral genetic studies that the heritability of [insert health-related phenotype] is roughly xx
percent.” (Notably, this statement is misguided because heritability estimates are time, space, and population
specific.) As I hope is now clear, these estimates are highly dubious at best and should not be presented as facts,
given the flawed methodology and misguided conceptual model. No amount of quantitative genetic research can
establish the validity of such heritability estimates. “Technically flawed and conceptually unsound models—no
matter how often published or repeated—do not by virtue of their numbers make for sound evidence” (Burt &
Simons 2014: 252).
Heritability studies rely on an outdated gene-centric biological model. Although such a deterministic
model of genetic function may be ideally suited for heritability studies, it is chimerical (Charney 2012a). The
remarkable advances in our understanding of genetic function that are most visible in molecular epigenetics
debunk the oversimplified model of the genome and genotype-phenotype relationship on which such work relies
(Charney & English 2013). Indeed, the more we learn about development, the less meaningful seems any attempt
to estimate genetic (vs. environmental) contributions to phenotype variance—and the less important (Burt &
Notwithstanding my strong objections to heritability studies, I am enthusiastic about biopsychosocial
health research that recognizes the interactional, bidirectional relationship between genes, cells, organisms, and
environments. The challenge is harnessing these rapid advances in molecular genomics to enhance our
understanding of the etiology of adverse health outcomes. In so doing, we need to leave the unproductive nature
vs. nurture debate behind to focus on understanding the mechanisms and developmental processes of adverse
health outcomes, “[starting] from the premise that biological and environmental systems are indivisible” (Braun
2004: 143). To understand how these factors dynamically interact requires study of the processes of development,
rather than simply trying to link genomic variation to some sort of fixed end point (adverse health outcome).
Attention should shift from heritability to humans’ remarkable degree of adaptive phenotype plasticity (the
capacity of a single genotype to support a range of phenotypes) and the biological pathways (e.g., epigenetic
effects) that underlie such developmental adaptations.
Notably, in highlighting postgenomic findings and lines of research that recognize the “porousness of the
biological with the social,” it was not my intent to oversell the evidential strength of fields as nascent as epigenetics
and medical microbiology (Meloni 2014: 6). Much in these burgeoning fields remains controversial and debated,
and scientists face new challenges as a consequence of the epistemological shift from a dichotomous separation of
“genetic” versus “environmental” causes to the “inextricable mixture of social and biological factors typical of the
epigenetics and postgenomic landscape” (Meloni 2014: 6). However, this recognition of the difficult
methodological and epistemic questions facing molecular genetics is no reason to deny that epigenetic research
invalidates the nature/nurture dichotomy or to downplay the great potential of epigenetics in elucidating “the
pathways through which the social shapes and is literally inscribed into the body” (Meloni 2014: 6). Indeed, that
epigenetics and medical microbiology remain open, debated fields underscores the need for social scientists to
engage in this debate from the beginning to shape biosocial research as it “[hesitates] at a crossroads between
reductionism and holism” (Morange 2006: 356). As a recent editorial in Nature (2012: 143) opined: “It is time for
sociologists and biologists to bury the hatchet and cooperate to study the effects of environmental stress on how
This matters for medical sociology. The field is at a crossroads where the intersection between biology
and the social is ripe for research and theorizing (e.g., Landecker & Panofsky 2013; Meloni 2014). Rejecting
heritability studies and the false nature-nurture dichotomy and gene-centric model on which they are grounded
will pave the way for a reconceptualization of the link between the biological and the social in shaping health
outcomes, one which is consistent with contemporary bioscientific knowledge. Achieving a comprehensive
understanding of these processes will not be easy and will take time. It will also require an integrative approach;
social scientists, geneticists, microbiologists, and neuroscientists need to form interdisciplinary alliances to
facilitate the holistic study of the development of adverse health outcomes from cell chemistry to whole body
physiology, recognizing environmental influences from the cellular to the macrostructual levels.
Heritability studies and estimates of adverse health outcomes have gained a lot of traction and attention in
recent years based in part on the belief that they are undergirded by rigorous, state-of-the art science.19 There is
value in showing how this work does not measure up to that billing both methodologically and theoretically (Burt
& Simons 2015). Although I, among others, find fault in the heritability model, I hope that those who read this
critique take it in the spirit of constructive criticism in which it was offered. I hope this critique advances science.
19 See Panofsky 2014 for an enlightening discussion of the historical development of the field of behavioral genetics and strategies
used to maintain its footing in various scientific fields.
The author would like to thank Brea Perry and two anonymous reviewers for valuable comments on earlier drafts
of the manuscript, and a special thanks to Kara Hannula for extensive feedback on multiple drafts. The arguments
presented herein are entirely those of the author and do not reflect the views of those who provided feedback.
Bagot, R. C., & Meaney, M. J. (2010). Epigenetics and the biological basis of gene x environment interactions.
Journal of the American Academy of Child and Adolescent Psychiatry, 49, 752–771.
Beckwith, J., & Morris, C. A. (2008). Twin studies of political behavior: Untenable assumptions? Perspectives
on Politics, 6, 785-791.
Belsky, J., & Pluess, M. (2009). Beyond diathesis stress: Differential susceptibility to environmental influences.
Psychological Bulletin, 135, 885-908.
Belsky, J., & Pluess, M. (2013). Beyond risk, resilience, and dysregulation: Phenotypic plasticity and human
development. Development and Psychopathology, 25, 1243-1261.
Bird, A. (2002). DNA methylation patterns and epigenetic memory. Genes & Development, 16, 6-21.
Boardman, J. D., Alexander, K. B., & Stallings, M. C. (2011). Stressful life events and depression among
adolescent twin pairs. Biodemography and Social Biology, 57,53-66.
Boardman, J. D., Blalock, C. L., Pampel, F. C., Hatemi, P. K., Health, A. C. & Eaves, L. J. (2011). Population
composition, public policy, and the genetics of smoking. Demography, 48, 1517-1533.
Boardman, J. D., Domingue, B. W., & Daw, J. (2014). What can genes tell us about the relationship between
education and health? Social Science & Medicine, Advance online publication
Boardman, J. D., & Fletcher, J. M. (2014). To cause or not to cause? That is the question, but identical twins
may not have the answer. Social Science & Medicine, Advance online publication.
Boardman, J. D., Roettger, M. E., Domingue, B. W., McQueen, M. B., Haberstick, B. C., & Harris, K. M.
(2012). Gene-environment interactions related to body mass: School policies and social context as
environmental moderators. Journal of Theoretical Politics, 24, 370-388.
Boardman, J. D., Saint Onge, J. M., Haberstick, B. C., Timberlake, D. S., & Hewitt, J. K. (2008). Do schools
moderate the genetic determinants of smoking? Behavior Genetics, 38, 234-246.
Braun, L. (2004). Commentary. In C. Garcia Coll, E. L. Bearer, & R. M. Lerner (Eds.), Nature and Nurture: The
Complex Interplay of Genetic and Environmental Influences on Human Behavior and Development (139-143).
New York: Psychology Press.
Breen, F. M., Plomin, R., & Wardle, J. (2006). Heritability of food preferences in young children. Physiology
and Behavior, 88, 443-447.
Brody, G. H., Lei, M., Chae, D. H., Yu, T., Kogan, S. M., & Beach, S. R. H. (2014). Perceived discrimination
among African American adolescents and allostatic load: A longitudinal analysis with buffering effects. Child
Development, 85, 989-1002.
Browning, S. R., & Browning, B. L. (2011). Letter to the editor: Population structure can inflate SNP-Based
heritability estimates. American Journal of Human Genetics, 132, 191-193.
Burt, C. H., & Simons, R. L. (2014). Pulling back the curtain on heritability studies: Biosocial criminology in
the postgenomic era. Criminology, 52, 223-262.
Burt , C. H., & Simons, R. L. (2015). Heritability studies in the postgenomic era: The ‘fatal flaw’ is conceptual.
Criminology. Advance online publication..
Burt, C. H., Simons, R. L., & Gibbons, F. X. (2012). Racial discrimination, ethnic-racial socialization, and
crime: A micro-sociological model of risk and resilience. American Sociological Review, 77, 648-677.
Carlsson, S., Andersson, T., Lichtenstein, P., Michaelsson, K., & Ahlbom, A. (2006). Genetic effects on
physical activity: Results from the Swedish Twin Registry. Medicine and Science in Sports and Exercise,
Caspi, A., Sugden, K., Moffitt, T. E., Taylor, A., Craig, I. W., Harrington, H. L., McClay, J., Mill, J., Martin, J.,
Braithwaite, A., & Poulton, R. (2003). Influence of life stress on depression: moderation by a polymorphism in
the 5-HTT Gene. Science, 301, 386-389.
Charney, E. (2008). Genes and ideologies. Perspectives on Politics, 6, 299-319.
Charney, E. (2012a). Behavior genetics and postgenomics. Behavioral and Brain Sciences, 35, 331–410.
Charney, E. (2012b). Humans, fruit flies, and automatons. Behavioral and Brain Sciences, 35, 381-392.
Charney, E. (2013). Still Chasing Ghosts: A new genetic methodology will not find the “missing heritability”.
Independent Science News. URL to article: http://www.independentsciencenews.org/health/still-chasing-ghosts-
Charney, E., & English, W. (2012). Candidate genes and political behavior. American Political Science Review,
Charney, E., & English, W. (2013). Genopolitics and the science of genetics. American Political Science
Review, 107, 382–95.
Chaufan, C. (2008). Unpacking the heritability of diabetes: The problem of attempting to quantify nature and
nurture. International Journal of Critical Statistics, 2, 23-38.
Cho, I., & Blaser, M. J. (2012). The human microbiome: At the interface of health and disease. Nature Reviews
Genetics, 13, 260-270.
Colodro-Conde, L., Sanchez-Romera, J. F., & Ordonana, J. R. (2013). Heritability of initiation and duration of
breastfeeding behavior. Twin Research and Human Genetics, 16, 575-580.
Combe, B. E., Defaye, A., Bozonnet, N., Putheir, D., Royet, J., Leulier, F. (2014). Drosophila microbiota
modulates host metabolic gene expression via IMD/NF-κB Signaling. PLoS ONE, 9, e104120.
Corby, P. M. A., Biesbrock, A., Gerlach, R., Corby, A. L., Moreira, A., Schork, N. J., & Bretz, W.A. (2014).
Treatment responses to tooth whitening in twins. Twin Research in Human Genetics, 17, 23-26.
Cronk, N. J., Slutske, W. S., Madden, P. A. F., Bucholz, K. K., Reich, W., & Heath, A. (2002). Emotional and
behavioral problems among female twins: An evaluation of the equal environments assumption. Journal of the
American Academy of Child and Adolescent Psychology, 41, 829–37.
Crusio, W. E. (2012). Heritability estimates in behavior genetics: Wasn’t that station passed long ago?
Behavioral and Brain Sciences, 35, 361–62.
David, L. A., Materna, A. C., Friedman, J., Campos-Baptista, M. I., Blackburn, M. C., Perrotta, A., Erdman, S.
E. & Alm, E. J. (2014). Host lifestyle affects human microbiota on daily timescales. Genome Biology, 15, R89.
David, L. A., Maurice, C. F., Carmody, R. N., Gootenberg, D. B., Button, J. E., Wolfe, B. E., Ling, A. V.,
Devlin, A. S., Varma, Y., Fischbach, M. A., Biddinger, S. B., Dutton, R. J., & Turnbaugh, P. J. (2013). Diet
rapidly and reproducibly alters the human gut microbiome. Nature, 505, 559-563.
Ding, T., & P.D. Schloss. (2014). Dynamics and associations of microbial community types across the human
body. Nature, 509, 357-360.
Duster, T. (2003). Backdoor to Eugenics. New York: Routledge.
Eaves, L. J. (1988). Dominance alone is not enough. Behavior Genetics, 18, 27–33.
Ellis, B. J., Del Guidice, M., Dishion, T.J.. Figueredo, A. J., Gray, P. et al. 2012. The evolutionary basis of risky
adolescent behavior: Implications for science, policy, and practice. Developmental Psychology, 48,598-623.
Erlich, S. D., & The MetaHIT Consortium. (2011). MetaHIT: The European Union Project on Metagenomics of
the Human Intestinal Tract. In K.E. Nelson (Ed.), Metagenomics of the Human Body. (Pp.307-316). New York:
Feil, R., & Fraga, M. F. (2011). Epigenetics and the environment: Emerging patterns and implications. Nature
Reviews Genetics, 13, 97-109.
Franklin, T. B., Russig, H., Weiss, I. C., Graff, J., Linder, N., Michalon, A., Vizi, S., & Mansuy, I. M. (2010).
Epigenetic transmission of early stress across generations. Biological Psychiatry, 68, 408-415.
Forsythe, P., Sudo, N., Dinan, T., Taylor, V. H., & Bienenstock, J. (2010). Mood and gut feelings. Brain,
Behavior, and Immunity, 24, 9-16.
Fowler, J. H., Baker, L. A., & Dawes, C. T. (2008). Genetic variation in political participation. American
Political Science Review 102: 233-248.
Francis, R. C. (2011). Epigenetics: The Ultimate Mystery of Inheritance. New York: Norton.
Gill, S. R., Pop, M., DeBoy, R. T., Eckburg P. B., Turnbaugh, P. J., Samuel, B. S., Gordon, J. I., Relman, D. A.,
Fraser-Liggett, C. M., & Nelson, K. E. (2006). Metagenomic analysis of the human distal gut microbiome.
Science, 312, 1355-1359.
Gottlieb, G. (1997). Synthesizing Nature-Nurture: Prenatal Roots of Instinctive Behavior. Mahway, NJ.
Gottlieb, G. (1998). Normally occurring environmental and behavioral influences on gene activity: From central
dogma to probabilistic epigenesist. Psychological Review, 105, 792-802.
Grice, E. A., & Segre, J. A. (2012). The human microbiome: Our second genome. Annual Review of Genomics
and Human Genetics, 13, 151-170.
Grunstein, M. (1997). Histone acetylation in chromatin structure and transcription. Nature, 389, 349-352.
Harris, J. R. (1998). The Nurture Assumption: Why Children Turn Out the Way They Do. New York: Free Press.
Hirsch J. (1997). Some history of heredity-vs-environment, genetic inferiority at Harvard(?), and The
(incredible) Bell Curve. Genetica, 99, 207-224.
Hirsch, J. (2004). Uniqueness, diversity, similarity, repeatability, and heritability. In C. Garcia Coll, E. L.
Bearer, and R. M. Lerner (Eds.) Nature and nurture (127-138). New York: Psychology Press.
Hogben, L. (1933). Nature and Nurture. London: Williams and Norgate.
Hoffmann, A., & Spengler, D. (2014). DNA memories of early social life. Neuroscience, 264, 64-75.
Horwitz, A. V., Videon, T. V., Schmitz, M. F., & Davis, D. (2003). Rethinking twins and environments:
Possible social sources for assumed genetic influences in twin research. Journal of Health and Social Behavior,
Hubbard, R. (2013). The mismeasure of the gene. In S. Krimsky & J. Gruber (Eds.), Genetic Explanation: Sense
and Nonsense. Cambridge, MA: Harvard University Press.
Hubbard, R., & Wald, E. (1999). Exploding the Gene Myth: Boston, MA: Beacon Press.
Hunley, K. L., Healy, M. E., & Long, J. C. (2009). The global pattern of gene identity variation reveals a history
of long-range migrations, bottlenecks, and local mate exchange: Implications for biological race. American
Journal of Physical Anthropology, 139, 35-46.
Jablonka, E., & Lamb, M. J. (2006). Evolution in Four Dimensions. Genetic, Epigenetic, Behavioral, and
Symbolic Variation in the History of Life. Cambridge, MA: MIT Press.
Jackson, D. D. (1960). A critique of the literature on the genetics of schizophrenia. In D. D. Jackson (Ed.), The
Etiology of Schizophrenia. New York: Basic Books.
Jakobsson, H. E., Jernberg, C., Andersson, A. F., Sjolund-Karlsson, M., Jansson, J. K., & Engstrand, L. (2010).
Short-Term Antibiotic Treatment Has Differing Long-Term Impacts on the Human Throat and Gut Microbiome.
PLoS One, 5, e9836.
Janss, L., de los Campos, G., Sheehan, N., & Sorensen, D. (2012). Inferences from genomic models in stratified
populations. Genetics, 192, 693-704.
Johnson, W., Kyvik, K. O., Mortensen, E. L., Skytthe, A., Batty, G. D., & Deary, I. J. (2010). Education reduces
the effects of genetic susceptibilities to poor physical health. International Journal of Epidemiology, 39, 406-
Joseph, J. (2004). The Gene Illusion: Genetic Research in Psychiatry and Psychology Under the Microscope.
New York: Algora.
Joseph, J. (2006). The Missing Gene: Psychiatry, Heredity, and the Fruitless Search for Genes. New York:
Kaplan, G., & Rodgers, L. J. (2003). Gene Worship: Moving Beyond the Nature/Nurture Debate over Genes,
Brain, and Gender. New York: Other Press.
Keller, E. F. (2010). The Mirage of a Space between Nature and Nurture. Durham, NC: Duke University Press.
Keski-Rahkonen, A., Viken, R. J., Kaprio, J., Rissanen, A., & Rose, R. J. (2004). Genetic and environmental
factors in breakfast eating patterns. Behavior Genetics, 34, 503–514.
Kessler, R. C., Berglund, P. A., Bruce, M. L., et al. (2001). The prevalence and correlates of untreated serious
mental illness. Health Services Research, 36, 987-1007.
Kessler, R. C., Chiu, W. T., Demler, O., & Walters, E. E. (2005). Prevalence, severity, and comorbidity of 12-
month DSM-IV disorders in the National Comorbidity Survey Replication. Archives of General Psychiatry, 62,
Krieger, N. (2000). Discrimination and health. In L. F. Berkman & I. Kawchi (Eds.), Social Epidemiology (Pp.
36-75). Oxford University Press.
Kruger, R. F., South, S., Johnson, W., & Iacono. W. (2008). The heritability of personality is not always 50%:
Gene-environment interactions and correlations between personality and parenting. Journal of Personality, 76,
Landecker, H., & A. Panofksy. (2013). From social structure to gene regulation and back: A critical introduction
to environmental epigenetics and society. Annual Review of Sociology, 39, 333-357.
Leahey, T. H., & Leahey, G. E. (1983). Psychology’s Occult Doubles: Psychology and the Problem of
Pseudoscience. Chicago: Nelson-Hall.
Lewontin, R. C. (1974). The analysis of variance and the analysis of causes. American Journal of Human
Genetics, 26, 400–411.
Lewontin, R. C. (1986). Race and intelligence, In N. J. Block and G. Dworkin (Eds.), The IQ Controversy (pp.
107-112). New York: Pantheon Books.
Lewontin, R. C., Rose, S., & Kamin, L. J. (1984). Not in our Genes: Biology, Ideology, and Human Nature.
New York: Pantheon Books.
Mayer, E. A. (2011). Gut feelings: The emerging biology of gut-brain communication. Nature Reviews
Neuroscience, 12, 453-466.
Meier, M. H., Slutske, W. S., Heath, A. C., & Martin, N. G. (2010). Sex differences in the genetic and
environmental influences on childhood conduct disorder and adult antisocial behavior. Journal of Abnormal
Psychology, 120, 377-388.
Meloni, M. (2014). The social brain meets the reactive genome: Neuroscience, epigenetics, and the new social
biology. Frontiers in Human Neuroscience, 8, 309.
Moffitt, T. E., & Beckley, A. (2015). Abandon twin research? Embrace epigenetic research? Premature advice
for criminologists. Criminology, advance online publication.
Moore, D.S. (2014). Big B, little b: Myth #1 is that Mendelian genes actually exist. In S. Krimsky & J. Gruber
(Eds.), Genetic Explanations: Sense and Nonsense (pp.43-50). Cambridge MA: Harvard University Press.
Morris-Yates, A., Andrews, G., Howie, P., & Henderson, S. (1990). Twins: A test of the equal environments
assumption. Acta Psychiatrica Scandinavica, 81, 322–326.
Murgatroyd, C., Patchev, A. V.,Wu, Y., Micale, V., Bockmühl, Y., Fischer, D., Holsboer, F., Wotjak, C. T.,
Almeida, O. F. X., & Spengler, D. (2009). Dynamic DNA methylation programs persistent adverse effects of
early-life stress. Nature Neuroscience, 12, 1559-1568.
Neale, M. C., & Cardon, L. R. (1992) Methodology for Genetic Studies of Twins and Families. Boston, MA:
Plomin, R. (2011). Commentary: Why are children in the same family so different? Non-shared environment
three decades later. International Journal of Epidemiology, 40, 582–92.
Plomin, R., & Daniels, D. (1987). Why are children in the same family so different from each other. Behavioral
and Brain Sciences, 10: 1-16.
Plomin, R., DeFries, J. C., Knopik, V. S.,& Neiderhiser, J. M. (2012). Behavioral Genetics, 6th ed. New York:
Press, N. (2006). Social construction and medicalization: Behavioral genetics in context. In E. Parens, A.R.
Chapman, & N. Press (Eds.), Wrestling with Behavioral Genetics: Science, Ethics, and Public Conversation
(pp.131-149). Baltimore: John Hopkins University Press.
Reitveld, M. J. H., Hudziak, J. J., Bartels, M., van Beijstervelt, C. E. M., & Boomsma, D. I. (2004). Heritability
of attention problems in children: Longitudinal results form a study of twins, age 3 to 12. Journal of Child
Psychology and Psychiatry, 45, 577-588.
Rende, R., Slomkowski, C., Lloyd-Richardson, E., & Niaura, R. (2005). Sibling Effects on Substance Use in
Adolescence: Social Contagion and Genetic Relatedness, Journal of Family Psychology, 19, 611-618.
Rowe, David C. (1994). On the Limits of Family Influence: Genes, Experience, and Behavior. New York:
Rutter, M. (2002). Nature, nurture, and development: From evangelism through science toward policy and
practice. Child Development, 73, 1-21.
Rutter, M. (2006). Genes and Behavior: Nature-Nurture Interplay Explained. Malden, MA: Blackwell
Rutter, M. (2007). Gene-environment interdependence. Developmental Science, 10, 12–18.
Rutter, M., & Plomin, R. (1998). Child development, molecular genetics, and what to do with genes once they
are found. Child Development, 69, 1223-1242.
Saudino, K. J., Ronald, A., & Plomin, R. (2005). The etiology of behavior problems in 7-year-old twins:
Substantial genetic influence and negligible shared environmental influence for parent ratings and ratings by
same and different teachers. Journal of Abnormal Child Psychology, 33,113–130.
Scheilzeth, H., & Husby, A. (2014). Challenges and prospects in genome-wide quantitative trait loci mapping of
standing genetic variation in natural populations. Annals of the New York Academy of Sciences, 1320, 35-57.
Segal, N. L., & Johnson. W. (2009). Twin studies of general mental ability. In Y. Kim (Ed.) Handbook of
Behavior Genetics (pp. 81-100). London: Springer.
Skinner, Michael K. (2008). What is an epigenetic transgenerational phenotype: F3 or F2? Reproductive
Toxicology, 25, 2-6.
Spencer, M. B., & Harpalani, V. (2004). In C. Garcia Coll, E.L. Bearer, & R.M. Lerner (Eds.), Nature and
Nurture (pp.53-77). New York: Psychology Press.
Stoltenberg, S. F. (1997). Coming to terms with heritability. Genetica ,99, 89-96.
Stoolmiller, M. (1999). Implications of the restricted range of family environments for estimates of heritability
and nonshared environment in behavior-genetic adoption studies. Psychological Bulletin 125:392–409.
Suhay, E., Kalmoe, N. P. (2010). The equal environment assumption in twin studies of political traits: Social
confounds and suggested remedies. Unpublished manuscript:
Sweatt, J. D. (2013). The emerging field of neuroepigenetics. Neuron, 80, 624-632.
Svedberg, P., Ropponen, A., Alexanderson, K., Lichtenstein, P., & Narusyte, J. (2012). Genetic susceptibility to
sickness absence is similar among women and men: Findings from a Swedish birth cohort. Twin Research and
Human Genetics 5: 642-648.
Turker, N., Toh, Z. Q. Karagiannis, T. C., & Liccidardi, P. V. (2014). Anti-inflammatory effects of probiotics
and their metabolites: Possible role for epigenetic effects. In N. Maulik and T. Karagiannis (Eds.), Molecular
Mechanisms and Physiology of Disease: Implications for Epigenetics and Health (pp. 127-150) New York:
Turkheimer, E. (2000). Three laws of behavior genetics and what they mean. Current Directions in
Psychological Science, 9, 160-169.
Turkheimer, E. (2004). Spinach and ice cream: Why social science is so difficult. In L.F. DiLalla (Ed.) Behavior
Genetics Principles: Perspectives in Development, Personality, and Psychopathology (pp.161-189).
Washington, DC: American Psychological Association.
Turkheimer, E. (2011). Commentary: Variation and causation in the environment and genome. International
Journal of Epidemiology, 40, 598-601.
Turkheimer, E., Pettersson, E., & Horn, E. E. (2014). A Phenotypic Null Hypothesis for the Genetics of
Personality. Annual Review of Psychology, 65, 515-540.
Turnbaugh, P. J., Ley, R. E., Hamady, M., Fraser-Liggett, C. M., Knight, R., & Gordon, J. I. (2007). The Human
Microbiome Project. Nature, 449, 804-810.
Tuvblad, C., Grann, M., & Lichtenstein, P. (2006). Heritability for adolescent antisocial behavior differs with
socieoeconomic status: Gene-environment interaction. Journal of Child Psychology and Psychiatry, 47, 734-
Wahlsten, Douglas 1979. A critique of the concepts of heritability and heredity in behavioral genetics. In eds.
J.R. Roye & L.P. Mos (Eds.), Theoretical Advances in Behavioral Genetics. Alphen aan den Rijn, the
Netherlands: Sijthoff & Noordhoff.
Wahlsten, Douglas. (1990). Insensitivity of the analysis of variance to heredity-environment interaction.
Behavioral and Brain Sciences 13: 109-120.
Wahlsten, D. (1994). The intelligence of heritability. Canadian Psychology, 35, 244-258.
Weaver, I. C.G., Cervoni, N., Champagne, F. A., et al. (2004). Epigenetic programming by maternal behavior.
Nature Neuroscience, 7, 847-854.
Weaver, I. C. G., Champagne, F. A., Brown, S. E., et al. (2005). Reversal of maternal programming of stress
responses in adult offspring through methyl supplementation: Altering epigenetic marking later in life. The
Journal of Neuroscience, 25, 11045-11054.
Weaver, I. C. G., Meaney, M. J., & Szyf, M. (2006). Maternal care effects on the hippocampal transcriptome
and anxiety-mediated behaviors in the offspring that are reversible in adulthood. PNAS, 103, 3480-3485.
Williams, D. R. (1999). Race, socioeconomic status, and health: The added effects of racism and discrimination.
Annals of the New York Academy of Sciences, 896, 173-188.
Yang, Jian, Beben Benyamin, Brian P. McEvoy, et al. (2010). Common SNPs explain a large proportion of the
heritability for human height. Nature Genetics, 42, 565-569.
Yatsunenko, T., Rey, F. E., Manary, M. J., Trehan, I., Dominguez-Bello, M. G., Contreras, M., Magris, M.,
Hidalgo, G., Baldassano, R. N., Anokhin, A. P., Heath, A. C., Warner, B., Reeder, J., Kuczynski, J., Caporaso, J.
G., Lozupone, C. A., Lauber, C., Clemente, J. C., Knights, D., Knight, R., & Gordon, J. I.. (2012). Human gut
microbiome viewed across age and geography. Nature, 486 222-228.