This Article Abstract Full Text (PDF) Services Download to citation manager Citing Articles Citing Articles via HighWire PubMed PubMed Citation

Social Context in Gene–Environment Interactions: Retrospect and Prospect

¹ University of North Carolina at Chapel Hill.
² Pennsylvania State University, State College.

Address correspondence to Michael J. Shanahan, Department of Sociology, University of North Carolina, CB 3210, 212 Hamilton Hall, Chapel Hill, NC 27599-3210. E-mail: mjshan{at}unc.edu

	Abstract

TOP Abstract A Typology of Genotype-Social... Critique of Studies of... Implications for Future Research References

 
While many behavioral scientists believe that gene–environment (GE) interactions play an important and perhaps pervasive role in human development and aging, little attention has been devoted to a fundamental conceptual issue: What is it about social context that could alter gene expression? We draw on existing examples of GE interactions to formulate a typology that identifies a set of generic mechanisms by which E moderates G. Empirical studies suggest four ideal types: Social context can trigger a genetic diathesis, compensate for a genetic diathesis, act as a control to prevent behaviors for which there is a genetic predisposition, and enhance adaptation through proximal processes. This typology highlights several problems, however, with prior empirical research, which may explain, in part, why so few GE interactions have actually been observed. These problems include inattention to the dynamic nature of social experience, the manifold, often-intercorrelated dimensions of social context ("EE interactions"), mediators that link social context and the genotype, and analytic models that examine GE interactions as processes that characterize individual development. In turn, these insights call for the integration of life course sociology and behavioral genetics to foster ways of studying genes, context, and aging.

STUDIES based on human and nonhuman populations and drawing on a range of research designs support the conclusion that there is a genetic basis for most complex behavior (broadly defined to include, e.g., facets of personality, cognition, motivation, psychopathology, various forms of deviance and problem behaviors, motor skills, health-related behaviors, and physical well-being) (e.g., Boomsma et al., 1999; Crabbe, 2002; Plomin, DeFries, McClearn, & McGuffin, 2001; Rose, 1995). Yet the link between genotype and phenotype is apparently complex, characterized by developmental processes involving constellations of variables at multiple levels of analysis (Johnston and Edwards, 2002; Li, 2003). One source of this complexity is gene–environment (GE) interactions, which occur when genes alter the organism's sensitivity to specific environmental features or environmental features exert differential control over genetic effects (Kendler & Eaves, 1986).

Many students of human behavior, development, and aging believe that the study of GE interactions will promote a better understanding of complex behaviors (e.g., McClearn et al., 2001; McGue, 1999; Rowe, 2001; Rutter & Silberg, 2002; Sawa & Snyder, 2002; van Os & Marcelis, 1998; Wahlsten, 1999). Yet surprisingly few empirical examples of GE interactions have been identified in the study of behavioral phenotypes in humans, and failures to find such interactions have been noted (e.g., Heath et al., 2002; McGue & Bouchard, 1998, p. 12). There are, unfortunately, many plausible reasons for this discrepancy between the presumed commonality of GE interactions and their relative infrequency of detection. Some of these reasons are methodologic, including power issues associated with inadequate sample size, strong and naturally occurring GE correlations, restricted variability in measurement and sampling, inappropriate measurement levels, differences between objective and effective contextual factors, large and multivariate reaction ranges, and the improper linear modeling of nonlinear relationships (McCall, 1991; Rutter & Pickles, 1991; Stoolmiller, 1999). Further, most studies of GE interaction are cross-sectional in design, failing to use the stronger within-person longitudinal designs for evaluating developmental patterns (Hofer & Sliwinski, 2002; Molenaar, 1999; Wohlwill, 1973).

Beyond methodologic challenges, little attention has been devoted to a basic conceptual problem: What is it about social context that would matter for gene expression? As Rutter and Pickles (1991) observe, no single theory of context is applicable to all GE interactions, and research should start with hypotheses on the nature of processes that link social context and gene expression. In this article, we review select studies that report GE interactions involving social context. In organizing this material, we suggest a typology that identifies four generic mechanisms by which social context moderates gene expression. The typology is provisional, given its reliance on extant research, and heuristic, given that its categories are not always clearly delineated. Nevertheless, it organizes existing evidence for GE interactions in a parsimonious framework, and, when viewed from a life course developmental perspective, it highlights several limitations concerning the conceptualization and measurement of social context. After discussing the typology, we review these limitations and then consider their implications for genetically informed research on aging.

	A TYPOLOGY OF GENOTYPE–SOCIAL CONTEXT INTERACTIONS

TOP Abstract A Typology of Genotype-Social... Critique of Studies of... Implications for Future Research References

 
At least four lines of evidence suggest that GE interactions may explain diverse phenotypes in humans. First, there is impressive evidence for the pervasiveness of GE interactions from animal studies, the designs of which often allow for the selection, control, and manipulation of both genotypes and environments (Festing, 1979; McClearn et al., 2001). To date, GE interactions have been observed among nonhuman animals on a wide range of behavioral phenotypes, including stress reactivity, longevity, and manifold dimensions of health, psychosocial adjustment, and cognitive ability (e.g., Mackay, 2001; McClearn, 2004; Suomi, 2004). For example, Viera and colleagues (2000) mapped quantitative trait loci (QTLs) for the adult life span of Drosophila melanogaster in five different environments (standard, high, and low temperature, heat shock, and starvation); 17 QTLs were identified, although none of them was expressed in all environments, suggesting pervasive and important GE interactions in the aging of the fruit fly. There is no principled reason to suppose that GE interactions help explain the diverse phenotypes of nonhuman animals but not humans.

Second, the hunt for specific genes that are linked to specific phenotypes appears largely unsuccessful to date (Plomin & McGuffin, 2003). For example, Levinson and colleagues (2002) report that despite substantial evidence for genetic linkage of schizophrenia to chromosome 1q, a detailed analysis of pedigrees among eight individual samples observed no evidence for such a linkage or for heterogeneity in allele sharing among the different samples. As is likely to be the case in the hunt for genes that underlie complex behaviors (Hamer, 2002), the authors conclude that if 1q is a risk factor for schizophrenia, its population-wide, additive effect is likely to be quite small.

Third, interest in GE interactions acknowledges that behavior is seldom randomly distributed according to indicators of social experiences. Just as virtually all behaviors are heritable, virtually all behaviors are distributed in the population according to such characteristics as socioeconomic status, race/ethnicity, gender, and present and past experiences with the family, peer group, school, neighborhood, and workplace. For example, Sampson, Morenoff, and Gannon-Rowley (2002) observed that "the range of childhood and adolescent outcomes associated with concentrated [neighborhood] disadvantage is quite wide." Apart from concentrated disadvantage, differences among neighborhoods (e.g., unemployment rates) are associated with many forms of problem behaviors. Most behaviors can be predicted, to varying degrees, by social circumstance, and GE correlations are unlikely to explain these behaviors fully.

Finally, the promise of GE interactions in explaining behavior is suggested by an increasing number of human studies showing that GE interactions play an important role in explaining individual differences in well-being and disease. Pharmacogenomic and nutrigenomic studies reveal that the effects of many environmental toxins, ingested substances, and dietary patterns are known to interact with specific alleles to significantly change the likelihood of health outcomes (Evans & McLeod, 2003; Grigorenko, 2005; van Ommen & Stierum, 2002). For example, the apoliprotein 4 allele may influence the efficacy of the commonly prescribed antidepressants mirtazapine and paroxetine (Murphy, Kremer, Rodrigues, & Schatzberg, 2003a), and the number of c alleles of HTR2A may be associated with increased side effects when taking paroxetine (Murphy, Kremer, Rodrigues, & Schatzberg, 2003b). In instances such as these, the "environment" refers to exposure to chemical compounds or physical assaults that have effects on human physiology.

Aside from "E as exposure," however, what is meant by "E as social context" and what is known about G x "E as social context" interactions? Social context typically refers to patterns of interpersonal relationships, roles, and behavioral routines that tend to endure because they are embedded in a relatively stable social frame (see Giddens, 1984; Sewell, 1992). This frame comprises agreed-upon and tacitly approved rules and norms but also social forms that include, for example, the family, the workplace, the neighborhood, and the community. Social roles, relationships, and internalized rules and habits are proximal in that they define the rights, responsibilities, opportunities, and limitations that immediately confront the person. The distal social context includes comparatively macrosociological units of analysis such as formal organizations and societal institutions (Collins, 1981). Ideally, ecologically informed accounts of human behavior examine features of both the distal and the proximal context as found, for example, in studies that link social change—which coincides with changes in institutions and organizations—with relatively proximal settings, including communities, neighborhoods, and families (Elder & Shanahan, in press).

A review of empirical studies suggests a typology of ideal types that identifies generic processes by which genes and social context interact. Given our focus on contextual mechanisms, we simplify our discussion with respect to the genotype, assuming that a person either does or does not have a genetic predisposition for a phenotype. We also use the term "GE interaction" in its broadest sense to refer to any situation in which the genetic variance associated with a phenotype depends on variability of the environment, or vice versa. In most cases, variations in both genotype and environments have an influence on particular phenotypes, each with independent main effects (sometimes called "joint GE effects"). GE interactions may further be distinguished from GE correlations, which refer to the developmental processes (i.e., occurring across generations and within persons over time) (Shanahan, Sulloway, & Hofer, 2000) whereby certain genotypes are systematically distributed across levels of environmental variables. We focus, however, on the moderating effect of genotype and environment on behavioral outcomes.

Contextual Triggering
Pharmacogenomic studies typically rely on a triggering mechanism whereby the effect of exposure to a socially based toxin or other biological assault moderates the expression of a genotype. That is, the phenotype is triggered when the contextual feature and specific genotype are combined. In the strong variant of the triggering interaction, neither the genotype nor the context influences the likelihood of the phenotype: Their combination is necessary. In the weak variant, the genotype and/or context have an additive effect on the likelihood or intensity of the phenotype, but their combination significantly, additionally influences the manifestation of the phenotype. Apart from the "biological events" that serve as triggers in pharmacogenomic research, however, social context may also trigger specific genetic expressions through stress processes.

According to the stress–diathesis model, environmental stressors interact with personal predispositions to produce disease states, illness, and decrements in well-being. Although diatheses encompass a wide range of personal characteristics (e.g., cognitive patterns, mood, history of emotional disturbance), the model is believed to apply to the activation of genetic predispositions for at least some outcomes (Rende & Plomin, 1992). For example, Kendler and Kessler (1995) examined whether stressful life events could trigger a predisposition for onset of major depression among female–female twins. Genetic risk was assigned based on lifetime incidence of major depression before the study began: both monozygotic twins unaffected (1, or lowest-risk group), both dizygotic twins unaffected (2), both dizygotic twins affected (3), and both monozygotic twins affected (4, or highest-risk group). In the absence of a severe life event, the likelihood of onset of major depression is roughly equal across these risk groups. The impact of a severe life event on major depression, however, is substantially greater for individuals at greater genetic risk. Thus, in the presence of a severe life event, the risk of onset of major depression in the lowest-risk group (i.e., lifetime unaffected monozygotic twins) is 6.2%, whereas the risk is 14.6% in the highest-risk group. In other words, given a severe life event, membership in the high-genetic-risk group is associated with 2.4 times greater likelihood of major depression than membership in the lowest-genetic-risk group (see also Silberg et al., 2001).

Many studies have relied on adoption designs to study GE triggering interactions. These designs typically compare adoptees with and without a family history of an illness or disorder and examine whether conditions in the adoptive home moderate the genetic risk represented by the two groups. For example, Tienari and colleagues (1994) draw on data from the Finnish Adoptive Study to compare the prevalence of serious mental health disorders (including personality disorder, borderline disorder, and psychosis) among the adopted children of schizophrenic mothers (the index group) and a control group. Findings showed that index adoptees run a significantly higher risk of serious disorder if both adoptive parents had ever had a serious disorder than if only one had ever had a serious disorder. Further analyses revealed a GE interaction: Reality testing in the family—including congruence between the interviewer's and family's view of the family, sense of family values and traditions, and sense of humor—strongly moderated genetic risk. Children with an apparent genetic risk and low levels of reality testing in the adopted family ran a significantly higher risk of mental disorder than children with just a genetic risk or just a low level of reality testing. Similar GE triggering interactions have been observed with adoption designs for alcoholism (e.g., Cutrona et al., 1994; Sigvardsson et al., 1996) and adolescent aggression and conduct disorder in adulthood (e.g., Cadoret et al., 1995; Mednick, Gabrielli, & Hutchings, 1984).

All of these studies provide circumstantial evidence of GE triggering because genetic risk is inferred, either from the subjects themselves or, more commonly, from the exhibited phenotype of the biological parents. Caspi and colleagues' (2002) study of adult antisocial behavior provides what is currently a rare example of a gene–social context interaction tested with measures of both the social context and the genotype itself. Animal studies show that early maltreatment leads to alterations in neurotransmitter systems (norepinephrine [NE], serotonin [5-HT], and dopamine [DA]) that, in both mice and humans, continue into adulthood and are positively related to aggressive behavior. The monoamine oxidase-A (MAOA) gene encodes for the MAOA enzyme, which metabolizes neurotransmitters such as NE, 5-HT, and DA. The authors reasoned that low MAOA activity, indicated by a variable-number tandem repeat polymorphism of the MAOA gene, would be insufficient to constrain maltreatment-induced changes in these neurotransmitters, which would increase the likelihood of antisocial behavior in later life. In effect, childhood maltreatment without high levels of MAOA activity alters these neurotransmitters, which, in turn, predisposes persons to adult antisocial behavior.

Drawing on males (the MAOA gene is on the X chromosome) in the Dunedin Multidisciplinary Health and Development Study, Caspi and colleagues (2002) examined how this genotyped polymorphism and childhood maltreatment (characterized as none, probable, and severe) predicted four indicators of antisocial behavior (conduct disorder, convictions for violent offenses, disposition toward violence, and antisocial personality disorder symptoms, according to the Diagnostic and Statistical Manual of Mental Disorders [4th ed.]). For all four dimensions, a significant interaction was observed such that low levels of MAOA activity coupled with severe maltreatment led to significantly elevated levels of antisocial behavior. Roughly 80% (10 of 13 boys) of the severely maltreated, low-MAOA boys exhibited adolescent conduct disorder compared with roughly 40% (8 of 20 boys) of the severely maltreated high-MAOA boys. Referring to the composite of the four dimensions, the authors conclude that 85% of the severely maltreated boys with low MAOA developed some form of antisocial behavior. (For an additional triggering example based on the measured genotype and context, see Caspi et al. [2003], which shows the moderation of life events on depression by 5-HTT, a functional polymorphism in the promoter region of the serotonin transporter.) Although the pattern of results depends on a small group of boys, the study is consistent with a long line of research that suggests the validity of the triggering mechanism.

Thus, a wide range of studies illustrates the idea that social context can act as a stressor that activates a genetic diathesis. Further, whereas studies supporting the triggering mechanism typically view outcomes dichotomously (e.g., psychiatric disorders), triggers could also operate so as to significantly aggravate a condition that is continuous in nature.

Social Context as Compensation
GE triggering refers to a detrimental context combining with a genetic diathesis; a GE compensation interaction refers to a positive, possibly enriched setting that prevents the expression of a genetic diathesis. In some instances, compensation and triggering are ends of a continuum: Absent significant stressors, people with a diathesis do not exhibit distress (compensation), but as the level of stressors increases, the likelihood of distress increases (triggering). For example, the 1995 study by Kendler and colleagues showed that in the absence of life events, all groups defined by differing levels of genetic susceptibility exhibited the same propensity for major depression; only when exposed to a severe life event (assault, serious marital problem, divorce or separation, death of close relative) did people's likelihood of major depression differentiate by genetic susceptibility. The results of Caspi et al. (2002) show a similar pattern: Regardless of MAOA activity level, boys were not different in their antisocial behavior (composite index) unless exposed to severe childhood maltreatment.

In some cases, however, compensation may refer to situations in which only pronouncedly enriched settings can neutralize a genetic diathesis. In these cases, compensation refers not to an absence of a detrimental context (e.g., life events or childhood maltreatment) but rather to the presence of markedly positive features in the environment. For example, lines of rats bred for poor and good maze performance vary in predictable ways in their error rates in a Hebbs–Williams maze when reared in their usual environments (Cooper & Zubek, 1958). When raised in an enriched setting, however, bright and dull rats perform equally well. Recent studies suggest possible physiologic bases for these results. Rampon and colleagues (2000) produced mice with the complete deletion of the NMDAR1 gene, which controls neuronal receptors in the CA1 region in the hippocampus ("CA1-KO mice"). Prior research shows that damage to this region of the brain is associated with deficits in declarative memory (concerning people, places, objects, and events) in humans and that N-methyl-D-aspartate receptors are associated with learning and memory in mice.

Initial findings confirmed the importance of this gene for cognitive function: When compared with control mice, CA1-KO mice showed profound deficits in novel object recognition, significantly lower preference for food smelled during a prior training session, and less contextual fear memory, which is a hippocampus-dependent trait. Further, no differences were observed between the two groups of mice in hippocampus-independent tasks. When both groups of mice were raised in enriched settings, however, these differences disappeared completely (in the case of food preference and contextual fear memory) or lessened significantly (in the case of object recognition). Further analyses suggested that dendritic spine density explained at least some of these differences. After enrichment training, controls and CA1-KO mice showed significant increases in the density of (nonperforated) synapses, and, indeed, the two groups did not differ significantly in this respect. That is, an enriched context compensated for a genetic predisposition toward poor hippocampus-dependent memory (see also Rampon & Tsien [2000] and Shimuzu et al. [2000]; for an example of how enrichment moderated the effects of strain on anxiety, see Chapillon et al. [1999]).

Thus, an intriguing line of animal research suggests that some genetic risks may express themselves in "normal" settings but not in enriched settings.

Social Context as Social Control
Social control is similar to compensation in that both interactions involve a genetic diathesis and a context that squelches or completely prevents genetic expression. They differ, however, in their substantive meaning: Social control refers to social norms and structural constraints that are placed on people to limit their behavior and their choices, whereas compensation refers to the avoidance of low levels of functioning through the absence of a stressor (weak variant) or the provision of enriched settings (strong variant).

Broadly conceived, social control refers to any social structure or process that maintains the social order. Through ties to significant others and social institutions, people are socialized to engage in behaviors that tend to promote the stability of social arrangements or socially desired outcomes at both the micro and the macro level (e.g., Sampson and Laub, 1990, 1993). Many biometric studies of heritability (i.e., h²) demonstrate a GE interaction that can be explained by a social control model. These studies typically show that in settings marked by high levels of social control, h² attenuates, whereas in contexts marked by low levels of social control, h² increases. In other words, in circumstances marked by high levels of social control, a large percentage of the sample—irrespective of their genetic diversity—exhibits the same phenotype; in settings marked by low social control, people's choices and behaviors are more apt to reflect their genotype.

One would also expect mean level differences in the phenotype such that highly controlled groups exhibit higher levels of socially desirable outcomes when compared with less controlled groups. That is, assuming no appreciable GE correlations, the interaction suggests: (a) a relatively small group of genetically similar people who will exhibit the phenotype in either setting, (b) a relatively small group of genetically similar people who will not exhibit the phenotype in either setting, and (c) a larger group of genetically similar people who will exhibit the phenotype only under circumstances of low social control.

For example, Dunne and colleagues (1997) report that birth cohort moderated the heritability of age at first intercourse among Australian youth. Specifically, h² accounted for 32% and 0% of the variance in age at first intercourse for women and men, respectively, born between 1922 and 1952 (earlier-born cohort) but 49% and 72% of the variance for women and men born between 1952 and 1965 (later-born cohort). Further, the groups differed substantially in their mean levels of age at first sexual intercourse: Among the later-born cohort, the mean age was 18.9 years (SD = 3.2), which was significantly earlier than the mean age of 21.1 (SD = 4.3) for the earlier-born cohort (t_4695.7 = 20.1, p =.0001). Assuming that the two groups have the same genetic diversity, these statistics imply greater concordance among genetically similar or identical people in the earlier-born cohort, such that among twins who are genetically inclined to earlier intercourse, both are likely to engage in earlier intercourse. (We return to problems of interpretation below but assume this interpretation to be reasonable in constructing our typology.)

The authors suggest that earlier-born cohorts were constrained by the higher levels of social control when compared with the content and force of social controls encountered by youth in later-born cohorts. Unfortunately, how social controls of sexual behavior changed between the earlier and later historical periods is unknown. It may be, for example, that when compared with the later period, the earlier period's proscriptions against precocious sexual behavior were more uniformly observed, that adult monitoring of youth was more thorough, and that engagement in nonfamilial adult socializing agents (school, extracurricular activities, voluntary associations) was more commonplace. In any event, the results suggest that changes in mechanisms of social control can lead some youth who are otherwise prone to engage in early sexual activity to delay intercourse.

Studies of the use of alcohol also illustrate social control mechanisms in GE interactions (e.g., Koopmans et al., 1999). Thus, for example, Higuchi, Matsushita, Imazeki, Kinoshita, Takagi, and Kono (1994) show that whereas the suppressive effect of the ALDH2*2 genotype (i.e., homozygous for the null allele) inhibits alcoholism among Japanese people, the suppressive effect of the heterozygous genotype (i.e., one null and one normal allele) has waned with successive cohorts. The authors speculate that social controls on drunkenness have loosened in Japanese society through the 20th century. The functional polymorphism ADH2*2 also protects against alcoholism, but its effects may be contingent on context (unlike ALDH2*2, which is, apparently, virtually completely protective; see Chen et al., 1999). The fact that Jews drink less than other Caucasians is thought to reflect the fact that ADH2*2 is more prevalent in the former group. Among Jews, however, the inhibitory effect of ADH2*2 may be contingent on environmental factors. Although drawing on a small sample, Hasin and colleagues (2002) report that the effect of ADH2*2 in suppressing alcohol consumption was lesser among Russian Jews—who had been exposed to an environment of heavy drinking prior to immigration—than among Israeli Ashkenazi and Sephardic Jews, who had not been exposed to such an environment.

Many studies of alcoholism, however, have studied social control mechanisms more directly than by cohort or ethnic comparisons. Drawing on a twin study of Finnish youth, Rose and colleagues (2001) observed that genetic factors were more prominent in urban than in rural settings from age 16 to 18.5 years, whereas common environmental factors accounted for more variation in phenotype in rural areas. Dick and colleagues (2001) examined whether this finding could be accounted for by specific dimensions of the urban–rural difference, including percentage migration into and out of municipalities, and per capita expenditure on alcohol in each region. The authors reasoned that as migration in an area increases, community monitoring and personal accountability decrease. The percentage of migration varied from 2% to 16% across the municipalities, and migration was strongly, positively correlated with urban areas. The results show that the average change in h² between each of the 15 levels of migration was.03, and the average change for the shared environment was –.03. For the area with the highest migration (defined by the 95% confidence interval), the additive genetic effect was.60, and the shared environmental effect was.01. In contrast, for the area with the lowest migration, the additive genetic effect was.16, and the shared environmental effect was.48. That is, as expected, in circumstances of high social control (i.e., low migration), the additive genetic effect is low and the shared environmental effect is high when compared with the areas marked by low social control (i.e., high migration).

Religion has often been conceptualized as an important source of social control (e.g., Udry, 1988): Through religious affiliation and observances, people internalize expectations and norms that encourage controlled behaviors. In fact, Koopmans and colleagues (1999) reported that genetic association with the initiation of alcohol use among Dutch youth was moderated by religious upbringing. Among females who were raised religiously, genetic influences accounted for 0% of the variance in whether they had or had not initiated alcohol use compared with 40% among females who were not raised religiously. The prevalence of alcohol use was slightly higher among nonreligious girls (64% vs 61%). This pattern of heritabilities was similar for boys, although not statistically different between the groups.

Disinhibition may explain some of these results. Boomsma and colleagues (1999) examined the links among religiousness and disinhibition, which assesses the "desire to find release through social disinhibition, drinking, going to parties, and having a variety of sexual partners" (p. 120). The results suggest a greater genetic association with disinhibition among youth who were not religiously raised. Among twins with a religious upbringing, the male monozygotic and dizygotic correlations are both.62, and the female monozygotic correlation (.61) is only somewhat greater than the dizygotic correlation (.50). Among twins without a religious upbringing, the male and female monozygotic correlations (.62 and.58, respectively) are greater than the dizygotic correlations (.35 for both males and females). Statistical comparisons revealed a significant GE interaction for males; for boys with a religious upbringing, genetic variation makes no contribution to individual differences in disinhibition.

Thus, social control mechanisms reflect norms and other social forces that "canalize" (i.e., restrict variability in the phenotype of) genetically diverse people. As these canalization forces increase (i.e., norms are more effective and choices are minimal), genetic differences are of diminishing consequence.

Social Context as Enhancement
Among individuals without a genetic diathesis, social context can also interact with genes to facilitate higher levels of developmental functioning. That is, given a genetic predisposition (which would not be considered a diathesis for a poor behavioral outcome), some contexts can lead to significantly higher levels of functioning. A GE enhancement interaction refers to the accentuation of "positive" genetic predispositions. Although triggering and enhancement interactions both refer to situations in which a genetic propensity is expressed, they are distinct in their substantive meanings. A trigger refers to the activation of a diathesis by stressors, whereas an enhancement refers to constructive sources of growth that canalize behavior to the positive end of the behavioral spectrum. Bronfenbrenner and Ceci's (1999) bioecological model suggests that proximal processes—which they define as enduring forms of social interactions characterized by progressive complexity—encourage the actualization of genetic potential: As proximal processes improve, the genetic potential for positive development is increasingly actualized. That is, heritability (as indicated by h²) will increase as proximal processes are enhanced (but see Rowe et al. [1999] for a critique). One should also observe mean level differences such that positive functioning is enhanced by better proximal settings. Extending this model suggests interactions involving social context. Bronfenbrenner and Ceci (1999) hypothesize that the efficacy of proximal processes is contingent on the broader context of those relationships. For example, in social settings marked by "advantage" and "organization," proximal processes will be more effective in realizing positive genetic potential than in settings marked by disadvantage and disorganization.

Several studies are consistent with the idea that heritability for positive development increases and predicted mean level differences are observed when more resources are provided to youth. Heath and colleagues (1985) examined the heritability of educational attainment in Norway before and after educational reforms that made it easier to continue one's education. They compared the h² values across groups defined by birth cohorts that correspond to the progressive opening up of the educational system to larger percentages of the population. That is, earlier-born cohorts have restricted choice with respect to education when compared with later cohorts. Heath and colleagues report an h² for educational attainment for Norwegians born between 1915 and 1939 of.41, between 1940 and 1949 of.74, and between 1950 and 1960 of.67.

The evidence therefore suggests that the heritability of educational attainment increases as more people face a realistic choice between continuing school or entering the labor market. As the authors conclude, "increased educational opportunity has led to an increased dependence of educational attainment on innate ability" (Heath et al., 1985, p. 736). Assuming that proximal processes were not significantly different between the age cohorts, the results suggest that a broader context of enriched resources is more likely to actualize genetic potential for positive development (see also, e.g., Scarr-Salapatek, 1971).

Focusing on a research topic directly linked to proximal process, Rowe and colleagues (1999) focused on how the family of origin influences vocabulary intelligence. In a study of twins and nontwin siblings in the Add Health data set, they report that the heritability (assessed as h²) of verbal intelligence is significantly greater among high-education households (i.e., the average educational level attained by both residential parents exceeded high school) than among low-education households (.74 and.26, respectively). Heritability increases from about 25% in offspring whose parents have less than a high school degree to about 74% in offspring whose parents have more than a high school degree. These results suggest that the genetic potential for verbal intelligence is more fully realized in homes of better-educated parents, who are assumed to provide better proximal processes (see also, e.g., Guo & Stearns, 2002).

Thus, although no research has examined how within-family proximal processes moderate gene expression, research does suggest that proximal family processes realize genetic potential for intellectual development. Enhancement is also likely to explain secular trends in height, weight, and age of the pubertal transition. In all of these instances, recent historical changes are thought to reflect improvements in nutrition and other health-related factors, not changes in the gene pool. That is, for example, the same genetic pool is associated with increasing height because of changes in context that may be construed as enhancers.

	CRITIQUE OF STUDIES OF GENE–CONTEXT INTERACTIONS

TOP Abstract A Typology of Genotype-Social... Critique of Studies of... Implications for Future Research References

 
This typology suggests four basic mechanisms by which social context moderates the influence of a genotype on a phenotype: Contexts can (1) trigger or (2) compensate for a genetic diathesis, contexts can (3) control phenotypes despite genetic propensities to the contrary, and contexts can (4) help actualize genetic potential. When viewed with these mechanisms in mind, however, prior empirical studies are open to several criticisms, which have implications for how research problems are conceptualized, how social context is measured, and how data are analyzed.

Life Course Trajectories of Social Experiences
Whereas behavioral geneticists appreciate that genes may express themselves at specific times in development (e.g., Lu et al., 2004; McClearn et al., 2001) and that phenotypes have dynamic qualities (e.g., see Harris, 2003; McGue & Christensen, 2002; Wadsworth, Corley, Hewitt, & DeFries, 2001), the dynamic nature of context has been overlooked. This neglect is hard to reconcile with a large and growing body of empirical research demonstrating that the dynamic properties of context often determine the meaning of social experience (e.g., Mortimer & Shanahan, 2003). That is, the conceptualization and measurement of context as a cross-sectional phenomenon—characteristic of behavioral genetic studies—typically fail to capture the necessary level of detail for determining the meaning of social experiences.

The case of GE triggering interactions involving life events illustrates this point. The magnitude of the relationship between life events and indicators of distress like depression or depressive symptoms is often modest: In reviewing the evidence, Turner, Wheaton, and Lloyd (1995) report that life events typically account for <10% of the variance in indices of mental health and well-being. The cumulative evidence shows that these modest associations are observed because, in part, the invidious nature of stressors is contingent on the prior, contemporaneous, and subsequent experiences of the person (Elder, George, & Shanahan, 1996). In effect, life events—like all potential sources of stress, compensation, enhancement, and control—can be understood only as parts of the life course trajectories that embed them.

Indeed, whereas the vast majority of studies examine the negative implications of life events for well-being or the conditions in which such negative effects are attenuated, life events can actually have positive effects on well-being, depending on prior circumstances. For example, Wheaton (1990) shows that severe life events have a positive effect on psychological well-being if they resolve an antecedent source of chronic distress. Thus, among adults who have lost a spouse, persons with high levels of prior marital problems report significantly less distress than persons with low levels of prior marital problems. Similar patterns are observed for earlier divorce, premarital break-up, and child moving out, with qualified evidence for recent divorce, job loss, retirement, and getting married. Wheaton concludes that role histories often determine the meaning of a life event (for other examples of the positive effects of life events, see also Amato, Loomis, & Booth [1995] and Sweeney & Horwitz [2001]).

Furthermore, the experiences of early traumas (including life events) on later well-being are likely to be contingent on complex patterns of cumulative stressors. Turner and Lloyd (1995) report that number of cumulative lifetime traumas significantly predicts onset of disorder (major depression or substance abuse) but not relapses of disorders. The number of traumas experienced since the first onset, however, significantly predicts relapses. Controlling recent life events, the authors find that the number of post-onset traumas and chronic stressors increases major depression and substance use, whereas the number of pre-onset traumas decreases the risk of major depression. The authors conclude that the effects of life events on distress will be significantly underestimated if lifetime patterns of both traumas and episodes of disorder are not taken into account (for the independent effects of earlier and later stressors, see also, e.g., Ensel & Lin [2000], Hayward & Gorman [2004], and Poulton et al. [2002]).

With respect to experiences subsequent to the life event, research shows that the effects of life events are often contingent on their implications for later life course patterns. For example, in their overview of research on childhood adversity and its effects on adult adjustment, McLeod and Almazan (2003) note that much of the effect of parental loss (other than by parental separation) is mediated by subsequent experiences. Studies suggest that the provision of good child care, integration into and achievements at school, good peer relations, and supportive, intimate relationships can all act to break the link between parental loss in childhood and poor psychosocial outcomes in adulthood (e.g., Quinton & Rutter, 1988; Rutter, 1989). In adulthood, the effects of life events are often contingent on how the events are resolved.

Although our review of the life-events literature is not meant to be comprehensive, it does underscore that the cross-sectional measurement of exposure to life events represents a crude proxy for increases in stress load that could, in turn, actually trigger a genetic diathesis for psychosocial distress. Likewise, variables that represent other forms of stressors, as well as potential sources of social compensation, control, and enhancement, are likely to acquire their meaning and impact only when viewed as part of a life course trajectory. In an effort to enhance the accuracy and validity of their models, behavioral geneticists are beginning to assess behavior in developmental terms; a similarly dynamic orientation is necessary, however, to capture the full significance of social context.

Studying Correlated Systems
All of the empirical studies that demonstrate GE interactions involve the interplay between an indicator of genetic risk and one dimension of social experience. A focus on one dimension of context, however, is likely to underestimate the effect of contextual factors, which often will operate as sets of "correlated constraints," or groups of variables that co-occur and work interactively. The possibility that correlated constraints are more realistic views of context than single indicators is particularly great given the complex, multifaceted nature of the trigger, compensation, social control, and enhancement mechanisms. That is, highly stressful circumstances (capable of producing GE triggering interactions), constraining circumstances (capable of producing GE social control interactions), and enhanced circumstances (capable of producing GE enhancement or compensation interactions) are likely to reflect manifold aspects of context that exert their influence on the person as a set of variables, not individually.

For example, Rutter (1990) suggests that the presence of three or more risk factors predicts maladjustment in an interactive fashion. Likewise, the developmental challenges posed by neighborhoods marked by concentrated disadvantage are numerous, intercorrelated, and likely to exert their negative influences in nonadditive ways. Indeed, in their overview of research on childhood adversities and their implications for adulthood, McLeod and Almazan (2003) observed that

"attempts to disaggregate the effects of clustered adversities may offer relatively little insight into processes of risk and resilience. The different clusters of events that children experience have different meanings that are lost when those events are studied in isolation." (p. 401)

While behavioral geneticists have understandably been interested in GG (gene–gene) and GE (gene–environment) interactions, environmental factors may create EE interactions, whereby groups of contextual factors have nonadditive effects on behavior. Particularly if only extreme settings will moderate genetic expression (Scarr, 1992), then such environments are likely to involve EE interactions. Regardless of the presence of EE interactions, the high association observed between contextual variables warrants caution when interpreting bivariate studies that interrelate one genotype and one contextual factor (e.g., Caspi et al., 2002, 2003).

The case of virginity pledges among adolescents illustrates the highly interactive nature of a form of social control in delaying age at first sexual intercourse. Drawing on data from Add Health, Bearman and Bruckner (2001) observe that the risk of sexual initiation is 34% lower among youth who took a virginity pledge than among nonpledgers. The effect of pledging is contingent, however, on other contextual features, which, taken as a whole, create controlling circumstances. First, the pledging effect is stronger in early and middle adolescence, but it does not prevent sexual intercourse until marriage. Second, the effect of pledging depends on the type of school that the student attends and the percentage of pledgers within the school. In "socially open schools," where many of the students report friendships and romantic relationships with students from other schools, pledging has no effect if no other pledgers are present. In such schools, for every 1% additional same-sex pledgers, the rate of the transition to first intercourse is delayed by 2%. In socially closed schools, where friendships and romantic relationships are contained within the school, the opposite is observed: With no other pledgers present, pledgers are much less likely to experience their sexual debut. When other pledgers are present, the pledgers' transition rate is higher than that of pledgers in schools with few pledgers. For adolescents in schools with >30% pledgers, a threshold is reached whereby pledging has no effect. That is, by itself, pledging status tells little of the story of how sexual behavior is controlled. When this form of social control is viewed as a constellation of variables that create EE interactions, however—encompassing status, type of school, and percentage of pledgers in the school—the controlling nature of the context can be more fully appreciated.

The role of context in genetic expression may be further complicated if the population is heterogeneous in the type and (possibly complexity) of GE interactions. That is, some interactions may apply only to subgroups of the population, or the specific function of the interaction may change across subgroups. These subgroups could be defined, for example, by other genes or by distinctive social experiences. Such a "population mixture" alerts the analyst to the possibility of subtypes of GE interactions and encourages a "person-centered" analytic strategy.

Joining Distal Levels of Analysis
Many scholars advocate the adoption of a developmental systems perspective in the study of genetic expression (e.g., Johnston & Edwards, 2002; McClearn et al., 2001; Wahlsten, 1999). According to this view, genes appear as one among many contributors to complex networks of interactions involving molecular, cellular, physiologic, behavioral, and environmental components (Gottlieb, 1991). Each level of analysis is governed by unique properties and mechanisms, and yet behavior is fully understood only by taking into account exchanges among the levels. Johnston and Edwards (2002) provide a useful map of the dynamic, nonrecursive pathways among these levels of analysis. Accordingly, contextually based experiences (what they call "sensory stimulation," which is unfortunately the model's sole reference to contextual effects) have dynamically reciprocal relationships with patterned neural and individual nerve cell activity, which are, in turn, reciprocally interrelated with intracellular processes by way of the cell membrane. The implication of such a model is that the link between social context and the genotype will often involve multifaceted mediating processes marked by contingency.

While further elaborations of how the multiple levels of a developmental system are interrelated are exciting, they also suggest a caveat: Context is not likely to co-act in a simple, direct way with genes. Rather, social context is part of a "cascade of associations" (Johnston & Edwards, 2002) or mediating mechanisms that makes certain behaviors more likely than others. As Rutter and Pickles (1991) observe, a GE interaction is not an explanation, but rather something to be explained. For example, Link and Phelan (1995) suggest that socioeconomic status is a "fundamental cause" of well-being, meaning that, as a rule, high socioeconomic status is associated with good health. Nevertheless, the mechanisms by which high socioeconomic status has these salutary effects vary considerably depending on the time and place. Socioeconomic status may promote manifold dimensions of well-being and health through preventative behaviors, monitoring and treatment, the amelioration of stressors, and/or the provision of stimulating, healthy environments, for example.

The study of mediating processes is complicated by two problems. First, in addition to genetic information, conceptual models and empirical research must encompass both social processes and their biological interface. What are the immediate points of connection between biological functioning and contextual sources of triggering, compensation, enhancement, and social control? Johnston and Edwards (2002) suggest that neural processes represent the major link between social experience and genetic expression. Accordingly, the experience–neural link involves immediate-early genes: "[immediate-early gene] induction is involved in all responses by the nervous system to sensory stimulation, including instances of behavioral development that have been shown to depend on experiences"(p. 29). At the same time, the immune and endocrine systems may also serve as important mediational pathways (e.g., Kiecolt-Glaser, McGuire, Robles, & Glaser, 2002).

Second, the principles of equifinality and multifinality—which state that the end phenotype may be the product of multiple, distinct causal pathways and that single causes may lead to multiple, distinct phenotypic expressions—may apply to links between social context and gene expression. For example, Kendler, Gardner, and Prescott (2002) examined the interrelationships among 18 risk factors over the life course (some of which were retrospectively recalled) and depressive episodes. The best-fitting, most parsimonious model included 64 paths among risk factors and the occurrence of major depression, a simple but dramatic example of how multifaceted mediational pathways can be.

Unfortunately, the possibility of separate causal pathways applies to both social and biological processes. At the social level, for example, many specific forms of social control that inhibit antisocial behavior have been identified, including warm and nurturant parenting, positive connections with schools and nonrelated adults, intimate interpersonal relations, marriage, parenthood, and meaningful ties to the labor market (e.g., Sampson and Laub, 1993). Similarly, distress is known to reflect many different types of stressors, which can, in turn, take on many different forms. At the biological level, for example, distinct pathways to colorectal and pharyngeal cancers have been identified (Brennan, 2002). These considerations suggest causal pathways involving social context, and biological substrates may involve complex combinations of different factors that lead to the same outcome.

Population Models and Individual Development
Virtually all studies of GE interactions to date have relied on estimates of heritability, which refers to the proportion of phenotypic variance (i.e., individual differences) that can be accounted for by genetic variability in a population (Falconer & Mackay, 1996). These estimates are based on between-person differences and derive mainly from cross-sectional studies. Yet Molenaar, Huizenga, and Nesselroade (2003) provide evidence that such estimates are unlikely to describe developmental processes accurately because the structure of the within-individual variation is often not the same as the structure of the between-individual variation. According to the ergodicity hypothesis, the parameters describing multiple observations in one person through time are equal to the parameters derived from the comparison of many individuals, each at a different phase of their own development (but observed at a single time). Thus, if ergodicity holds true, a cross-sectional interindividual mean accurately describes a longitudinal intraindividual pattern of observations.

Whereas ergodicity often works well when describing physical systems that change toward an equilibrium (e.g., gases, liquids, crystals, Mendelian populations) and when the observations span the initial condition of the system up to the point of equilibrium, Molenaar and colleagues (2003) argue, on both conceptual and empirical grounds, that it is unlikely to apply to developmental systems. Conceptually, developmental systems are characterized by dynamic patterns of behavior that may be nonequilibrating to a significant degree and that are likely to have high levels of time-varying complexity. For example, Molenaar (1999) conducted a simulation study showing that a standard longitudinal factor model fits data well even though those data reflect a group of people, each of whom has a different dynamic factor model. Indeed, despite the satisfactory model fit, the correspondence between the standard and dynamic factor scores was very low and, in some cases, actually negative. Similar discrepancies are observed with simulated data modeled with the standard Martin–Eaves model of heritability (Molenaar, 1999).

Ergodicity seems especially implausible in the presence of interactions, including GE interactions, because they create heterogeneity in the population with respect to developmental trajectories. The problem is likewise compounded by equifinality (i.e., diverse pathways may lead to the same phenotype), which is likely in developmental systems and also creates distinct subgroups that, when viewed with sample-wide, population-level statistics, obscure within-person processes. The essential problem is that genotypes vary across people and not within people, and the range of potential environments occurring within individuals is but a small sample of environments that vary between individuals. Thus, whereas GE interactions refer to processes occurring in individual development, studies of heritability allow for inferences only at the level of the population.

One implication of the foregoing is that while population-level trends suggesting an association between a context and phenotype (or gene and phenotype) are worthy of further study, population statistics suggesting no such associations may be misleading because they are "averaging" across estimates that describe distinct subgroups in the population. The magnitude of this problem is likely to depend on the phenotype, but a simple example is suggested by studies on polychlorinated biphenyls (PCBs) and breast cancer. Laden and associates (2001) pooled data from five studies of PCB exposure and breast cancer with a combined sample size of 1,400 case patients and 1,642 control subjects. Results revealed that women in the fifth and first quintiles of lipid-adjusted values for PCBs did not differ in their odds of breast cancer. The authors concluded that exposure to PCBs is not likely to explain rates of breast cancer. Drawing on one of the five samples, however, the same team of researchers examined whether the exon 7 polymorphism of CYP1A1—a gene associated with a biotransformation enzyme that metabolizes carcinogenic, exogenous substances—moderated the association between PCB exposure and breast cancer (Laden et al., 2002). In fact, whereas neither exposure to environmental PCBs nor the CYP1A1 exon 7 polymorphism alone is associated with an increased risk of breast cancer, women exposed to high levels of PCBs and having this polymorphism run an elevated risk of breast cancer (Laden et al., 2002). The population-wide, additive estimates were misleading.

	IMPLICATIONS FOR FUTURE RESEARCH

TOP Abstract A Typology of Genotype-Social... Critique of Studies of... Implications for Future Research References

 
Our review of studies that report gene–context interactions suggests four ways by which social context moderates genetic expression: the triggering of a genetic diathesis, compensation for a genetic diathesis, social control of a genetic propensity, and the maximal realization of genetic potential. Furthermore, our review of these mechanisms identifies several limitations that characterize previous empirical studies of GE interactions, including a failure to consider social context from a dynamic perspective, inattention to the multifaceted nature of social context, failure to specify mediating mechanisms that link context with the immediate biological substrate, and overreliance on population-based statistical models that require untenable assumptions to make statements about individual development.

These critiques may contribute to explaining why, despite expectations to the contrary, few GE interactions have been observed. All of these limitations are probably serious shortcomings of prior research because the four GE mechanisms are likely to reflect complex combinations of variables that work interactively through the life course. That is, a life course view of context and a developmental/aging view of person are critical conditions for any valid test of GE interactions. We can identify three conceptual and two empirical implications of this view for the study of phenotypes.

First, behavior reflects a genetic predisposition coupled with social experiences that amount to triggering agents, a lack of social controls, a lack of compensatory experiences, and an environment lacking enrichment agents. That is, the mechanisms are not mutually exclusive, and synergies among them could lead to the same observed phenotype. That GE interactions reflect these multiple processes is suggested, for example, by a study by Heath, Jardiner, and Martin (1989). They observe that, among young twins, genetic differences between individuals account for 31% of the variance in the alcohol consumption of married respondents but for 60% of the variance in unmarried respondents. Research on marital relations suggests that spouses likely serve as sources of both social control and compensation. Thus, conceptual models should consider how all four mechanisms jointly operate to make a specific phenotype more or less likely.

Second, people may exhibit the same phenotype in the presence of different variations of the four processes. For example, numerous types of stressors have been identified and studied—including daily hassles, chronic stressors, and life events in all domains of life—and different combinations of them could produce the same stress load. The point is implicit in life-events research, which typically sums the number of events experience based on the premise that any combination of the same number of life events is functionally equivalent. But combinations of other forms of stressors may also produce the same stress load. Similarly, a myriad of specific social control, enhancement, and compensation mechanisms are recognized, and different combinations of these could produce the same effective level of control, enhancement, and compensation, respectively.

Third, social experiences acquire their meaning as triggers, controls, compensators, and enrichers only when viewed as part of a life course trajectory. Thus, it is unlikely, for example, that a person's stress load can be accurately assessed with a cross-sectional assessment of exposure to life events given that life events take on their meaning only as part of a trajectory of experience. Life events represent but one form of stressors. The stress–distress process refers to mechanisms that operate at the level of the person and can result from any number of causal pathways with the potential for distressful effects of stressors to be neutralized via compensators and enrichers.

With respect to empirical implications: First, new analytic strategies are needed, given the highly interactive, contingent nature of context and behavioral development. Given an interactive, equifinal or multifinal system, population-level statistics are likely to be misleading because they fail to capture heterogeneity within the population. A promising approach to understanding complex pathways of multiple influences on common outcomes is that of the configural or person-centered strategy, which refers to a wide range of statistical models that identify a set of discrete configurations of variables based on their distributions (Cairns, Bergman, & Kagan, 1998). These strategies are particularly useful for studying processes marked by equifinality and high levels of nonlinear interaction.

In the social sciences, Ragin (1987) initially proposed that the highly contingent processes be studied with Boolean algorithms that identify sets of variables joined by "and" and "or" statements that lead to a behavior. The method was subsequently extended to include "nor" statements and then applied to life-history data (Singer & Ryff, 1999, 2001; Singer, Ryff, Carr, & Magee, 1998) to reveal different combinations of social experiences across the life course that can lead to depression among elderly women. Although there are limitations to many person-centered approaches—often including their exploratory nature, awkwardness with longitudinal data, and the loss of information through categorization of continuously measured variables—interest in configural approaches to genetic analysis is quickly increasing and is likely to represent a promising avenue for the study of GE interactions (e.g., Bushel et al., 2002; Nguyen and Rocke, 2002; Ooi & Tan, 2003; Zhang, Yu, Singer, & Ziong, 2001).

Second, data collection efforts must encompass the sampling and measurement of social context as a trajectory of experience as well as the genotype. To date, behavioral genetic studies have focused on the measurement of the person over time, in marked distinction to a life course perspective that emphasizes the formative nature of context over time. Comparatively holistic data collection efforts in turn require more comprehensive, inclusive theorizing that spans what is known and conjectured about social context and biology.

In the final analysis, empirical research based on unidimensional, cross-sectional measures of context and comparisons among groups defined by degrees of relatedness reveal little about the lived experiences of people and how these experiences interact with their biological make-up. What are needed are new conceptual models and empirical strategies that focus on the complexity of social experiences and how they interact with the individual's genetic make-up to make behaviors more or less likely.

	Acknowledgments

 
Earlier versions of this article were presented at the National Institute on Aging Environmental Workshop for Genetically Informative Studies of Aging, February 26–27, 2003, Bethesda, MD, and the Annual Meeting of the American Sociological Association, August 14–17, San Francisco, CA. The authors thank Michael Rutter, Harrison Cleveland, and anonymous reviewers for insightful comments; Jason Langberg and S. T. A. Wilkens for help with bibliographic sources; and Thomas Johnson for his valuable editorial comments.

	References

TOP Abstract A Typology of Genotype-Social... Critique of Studies of... Implications for Future Research References

 

• Amato, P. R., Loomis, S. L., Booth, A. (1995). Parental divorce, marital conflict, and off-spring well-being during early adulthood. Social Forces, 73,895-915.
• Bearman, P. S., Bruckner, H. (2001). Promising the future: Virginity pledges and first intercourse. American Journal of Sociology, 106,859-912.
• Boomsma, D. I., de Geus, E., van Baal, G. C. M., Koopmans, J. R. (1999). A religious upbringing reduces the influence of genetic factors on disinhibition: Evidence for interaction between genotype and environment on personality. Twin Research, 2,115-125.
• Brennan, P. (2002). Gene–environment interaction and aetiology of cancer: What does it mean and how can we measure it? Carcinogenesis, 23,381-387.[Abstract/Free Full Text]
• Bronfenbrenner, U., Ceci, S. J. (1994). Nature–nurture reconceptualized in developmental perspective: A bioecological model. Psychological Review, 101,568-586.[Medline]
• Bushel, P. R., Hamadeh, H. K., Bennett, L., Green, J., Ableson, A., Misener, S., et al. (2002). Computational selection of distinct class- and subclass-specific gene expression signatures. Journal of Biomedical Informatics, 35,160-170.[Medline]
• Cadoret, R. J., Yates, W. R., Troughton, E., Woodworth, G., Stewart, M. A. (1995). Genetic–environmental interactions in the genesis of aggressivity and conduct disorders. Archives of General Psychiatry, 52,916-924.[Abstract/Free Full Text]
• Cairns, R. B., Bergman, L. R., Kagan, J. (1998). Methods and models for studying the individual: Essays in honor of Marian-Radke Yarrow. Thousand Oaks, CA: Sage.
• Caspi, A., McClay, J., Moffitt, T. E., Mill, J., Martin, J., Craig, I. W., et al. (2002). Role of genotype in the cycle of violence in maltreated children. Science, 297,851-854.[Abstract/Free Full Text]
• Caspi, A, Sugden, K., Moffitt, T. E., Taylor, A., Craig, I. W., Harrington, H., et al. (2003). Influence of life stress on depression: Moderation in the 5-HTT gene. Science, 301,386-389.[Abstract/Free Full Text]
• Chapillon, P., Manneche, C., Belzung, C., Caston, J. (1999). Rearing environmental enrichment in two inbred strains of mice: 1. Effects on emotional reactivity. Behavior Genetics, 29,41-46.[Medline]
• Chen, Y. C., Lu, R. B., Peng, G. S., Wang, M. F., Wang, H. K., Ko, H. C., et al. (1999). Alcohol metabolism and cardiovascular response in an alcoholic patient homozygous for the ALDH2*2 variant gene allele. Alcoholism: Clinical Experimental Research, 23,1853-1860.[Medline]
• Collins, R. (1981). On the microfoundations of macrosociology. American Journal of Sociology, 86,984-1014.
• Cooper, R. M., Zubek, J. P. (1958). Effects of enriched and restricted early environments on the learning ability of bright and dull rats. Canadian Journal of Psychology, 12,159-164.[Medline]
• Crabbe, J. C. (2002). Genetic contributions to addiction. Annual Review of Psychology, 53,435-462.[Medline]
• Cutrona, C. E., Cadoret, R. J., Suhr, J. A., Richards, C. C., Troughton, E., Schutte, K., et al. (1994). Interpersonal variables in the prediction of alcoholism among adoptees: Evidence for gene–environment interactions. Comprehensive Psychiatry, 35,171-179.[Medline]
• Dick, D. M., Rose, R. J., Viken, R. J., Kaprio, J., Koshenvuo, M. (2001). Exploring gene–environment interactions: Socioregional moderation of alcohol use. Journal of Abnormal Psychology, 110,625-632.[Medline]
• Dunne, M. P., Martin, N. G., Statham, D. J., Slutske, W. S., Dinwiddie, S. H., et al. (1997). Genetic and environmental contributions to variance in age at first sexual intercourse. Psychological Science, 8,211-216.
• Elder, G. H., Jr., George, L., Shanahan, M. J. (1996). Psychosocial stress over the life course. In H. B. Kaplan (Ed.), Psychosocial stress: Perspectives on structure, theory, life-course, and methods (pp. 247–292). San Diego: Academic.
• Elder, G. H., Jr., Shanahan, M. J., in press The life course and child development. In R. M. Lerner (Ed.), Handbook of child psychology, volume 1: Theory. New York: Wiley.
• Ensel, M., Lin, N. (2000). Age, the stress process, and physical distress: The role of distal stressors. Journal of Aging and Health, 12,139-168.[Abstract/Free Full Text]
• Evans, W. E., McLeod, H. L. (2003). Pharmacogenomics—Drug disposition, drug targets, and side effects. New England Journal of Medicine, 348,538-549.[Free Full Text]
• Falconer, D. S., Mackay, T. F. C. (1996). Introduction to quantitative genetics (4^th ed.). Essex, UK: Longman.
• Festing, M. F. W. (1979). Inbred strains in biomedical research. London: Macmillan.
• Giddens, A. (1984). The constitution of society: An outline of the theory of structuration. Los Angeles: University of California Press.
• Gottlieb, G. (1991). Experiential canalization of behavioral development: Theory. Developmental Psychology, 27,4-13.
• Grigorenko, E. L. (2005). The inherent complexities of gene–environment interactions. Journals of Gerontology Series B (Special Issue I), 60,53-64.
• Guo, G., Stearns, E. (2002). The social influences on the realization of genetic potential for intellectual development. Social Forces, 80,881-910.
• Hamer, D. (2002). Rethinking behavior genetics. Science, 298,71-72.[Free Full Text]
• Harris, J. R. (2003). Introduction to the special issue on aging. Behavioral Genetics, 33,79-82.
• Hasin, D., Aharonovich, E., Liu, X., Mamman, Z., Matseoane, K., Carr, L., et al. (2002). Alcohol and ADH2 in Israel: Ashkenazis, Sephardic, and recent Russian immigrants. American Journal of Psychiatry, 159,1432-1434.[Abstract/Free Full Text]
• Hayward, M. D., Gorman, B. K. (2004). The long arm of childhood: The influence of early-life social condition on men's mortality. Demography, 41,87-107.[Medline]
• Heath, A. C., Berg, K., Eaves, L. J., Salaas, M. H., Corey, L. A., Sundet, J., et al. (1985). Education policy and the heritability of educational attainment. Nature, 314,734-736.[Medline]
• Heath, A. C., Jardiner, R., Martin, N. G. (1989). Interactive effects of genotype and social-environment on alcohol consumption in female twins. Journal of Studies on Alcohol, 50,38-48.[Medline]
• Heath, A. C., Todorov, A. A., Nelson, E. C., Madden, P. A. F., Bucholz, K. K., Martin, N. G. (2002). Gene–environment interaction effects on behavioral variation and risk of complex disorders: The example of alcoholism and other psychiatric disorders. Twin Research, 5,30-37.
• Higuchi, S., Matsushita, S., Imazeki, H., Kinoshita, T., Takagi, S., Kono, H. (1994). Aldehyde dehydrogenase genotypes in Japanese alcoholics. Lancet, 343,741-742.
• Hofer, S. M., Sliwinski, M. J. (2002). Understanding ageing: An evaluation of reseach designs for assessing the interdependence of ageing-related changes. Gerontology, 47,341-352.
• Johnston, T. D., Edwards, L. (2002). Genes, interactions, and the development of behavior. Psychological Review, 109,26-34.[Medline]
• Kendler, K. S., Eaves, L. (1986). Models for the joint effect of genotype and environment on liability to psychiatric illness. American Journal of Psychiatry, 143,279-289.[Abstract/Free Full Text]
• Kendler, K. S., Gardner, C. O., Prescott, C. A. (2002). Toward a comprehensive developmental model for major depression in women. American Journal of Psychiatry, 159,1133-1145.[Abstract/Free Full Text]
• Kendler, K. S., Kessler, R. C., Walters, E. E., Maclean, C., Neale, N. C., Heath, A. C., Eaves, L. J. (1995). Stressful life events, genetic liability, and onset of an episode of major depression in women. American Journal of Psychiatry, 152,833-842.[Abstract/Free Full Text]
• Kiecolt-Glaser, J. K., McGuire, L., Robles, T. F., Glaser, R. (2002). Psychoneuroimmunology: Psychological influences on immune function and health. Journal of Consulting and Clinical Psychology, 70,537-547.[Medline]
• Koopmans, J. R., Slutske, W. S., van Baal, G. C. M., Boomsma, D. I. (1999). The influence of religion on alcohol use initiation: Evidence for genotype x environment interaction. Behavior Genetics, 29,445-453.[Medline]
• Laden, F., Collman, G., Iwamoto, K., Alberg, A. J., Berkowitz, G. S., Freundenheim, J. L., et al. (2001). 1,1-Dichloro-2,2-bis(p-chloro-phenyl)ethylene and polychlorinated biphenyls and breast cancer: Combined analysis of five studies. Journal of the National Cancer Institute, 93,768-775.[Abstract/Free Full Text]
• Laden, F., Ishibe, N., Hankinson, S. E., Wolff, M. S., Gortig, D. M., Hunter, D. J., et al. (2002). Polychlorinated biphenyls, cytochrome p450 1A1, and breast cancer risk in the Nurses' Health Study. Cancer Epidemiology, Biomarkers and Prevention, 11,1560-1565.[Abstract/Free Full Text]
• Levinson, D. F., Holmans, P. A., Laurent, C., Riley, B., Pulver, A. E., Gejman, P. V., et al. (2002). No major schizophrenia locus detected on chromosome 1q in a large multicenter sample. Science, 296,739-741.[Abstract/Free Full Text]
• Li, S.-C. (2003). Biocultural orchestration of developmental plasticity across levels: The interplay of biology and culture in shaping the mind and behavior across the life-span. Psychological Bulletin, 129,171-194.[Medline]
• Link, B. G., Phelan, J. (1995). Social conditions as fundamental causes of disease. Journal of Health and Social Behavior, SI,80-94.
• Lu, T., Pan, Y., Kao, S.-Y., Li, C., Kohane, I., Chan, J., et al. (2004). Gene regulation and DNA damage in the ageing human brain. Nature., 429,883-891.[Medline]
• Mackay, T. C. (2001). Quantitative trait loci in Drosophila. Nature Review Genetics, 2,11-20.[Medline]
• McCall, R. B. (1991). So many interactions, so little evidence. Why? In T. E. Wachs and R. Plomin (Eds.), Conceptualization and measurement of organism–environment interaction (pp. 142–161). Washington, DC: American Psychological Association.
• McClearn, G. E. (2004). Nature and nuture: Interaction and coaction. American Journal of Medical Genetics Part B (Neuropsychiatric Genetics), 124B,124-130.[Medline]
• McClearn, G. E., Vogler, G. M., Hofer, S. M. (2001). Gene–gene and gene–environment interactions. In Handbook of the biology of aging (pp. 423–444). San Diego: Academic Press.
• McGue, M. (1999). The behavioral genetics of alcoholism. American Psychological Society, 8,109-115.
• McGue, M., Bouchard, T. J. (1998). Genetic and environmental influences on human behavioral differences. Annual Review of Neuroscience, 21,1-24.[Medline]
• McGue, M., Christensen, K. (2002). The heritability of level and rate-of-change in cognitive functioning in Danish twins aged 70 years and older. Experimental Aging Research, 28,435-451.[Medline]
• McLeod, J. D., Almazan, E. P. (2003). Connections between childhood and adulthood. In J. T. Mortimer & M. J. Shanahan (Eds.), Handbook of the life course (pp. 391–411). New York: Kluwer.
• Mednick, S., Gabrielli, W., Hutchings, B. (1984). Genetic influences in criminal convictions. Science, 224,891-893.[Abstract/Free Full Text]
• Molenaar, P. C. M. (1997). Time series analysis and its relationship with longitudinal analysis. International Journal of Sports Medicine, 18,232-237.
• Molenaar, P. C. M. (1999). Longitudinal analysis. In H. J. Ader & G. J. Mellenbergh (Eds.), Research methodology in the social, behavioral, and life sciences (pp. 143–167). London: Sage.
• Molenaar, P. C. M., Huizenga, H. M., Nesselroade, J. R. (2003). The relationship between the structure of intra-individual and inter-individual variability: A theoretical and empirical vindication of developmental systems theory. In U. M. Staudinger & U. Lindenberger (Eds.), Understanding human development (pp. 339–360). Boston: Kluwer.
• Mortimer, J. T., Shanahan, M. J., (Eds.). (2003). Handbook of the life course. New York: Kluwer.
• Murphy, G. M., Kremer, C., Rodrigues, H. E., Schatzberg, A. F. (2003a). Pharmacogenetics of antidepressant medication intolerance. American Journal of Psychiatry, 160,1830-1835.[Abstract/Free Full Text]
• Murphy, G. M., Kremer, C., Rodrigues, H. E., Schatzberg, A. F. (2003b). The apolipoprotein E 4 allele and antidepressant efficacy in cognitively intact elderly depressed patients. Biological Psychiatry, 54,665-673.[Medline]
• Nguyen, D. V., Rocke, D. M. (2002). Multi-class cancer classification via partial least squares with gene expression profiles. Bioinformatics, 18,1216-1226.[Abstract/Free Full Text]
• Ooi, C. H., Tan, P. (2003). Genetic algorithms applied to multi-class prediction for the analysis of gene expression data. Bioinformatics, 19,37-44.[Abstract/Free Full Text]
• Plomin, R., DeFries, J. C., McClearn, G. E., McGuffin, P. (2001). Behavioral genetics (4^th ed.). New York: Worth.
• Plomin, R., McGuffin, P. (2003). Psychopathology in the postgenomic era. Annual Review of Psychology, 54,205-228.[Medline]
• Poulton, R., Caspi, A., Milne, B. J., Thomson, W. M., Taylor, A., Sears, M. R., et al. (2002). Association between children's experience of socioeconomic disadvantage and adult health: A life course study. Lancet, 360,1640-1645.[Medline]
• Quinton, D., Rutter, M. (1988). Parental breakdown: The making and breaking of intergenerational links. Adlershot: Gower.
• Ragin, C. (1987). The qualitative-comparative method: Moving beyond qualitative and quantitative strategies. Berkeley: University of California Press.
• Rampon, C., Tang, Y.-P., Goodhouse, J., Shimizu, M., Tsien, J. Z. (2000). Enrichment induces structural changes and recovery from nonspatial memory deficits in CA1 NMDAR1-knockout mice. Nature Neuroscience, 3,238-244.[Medline]
• Rampon, C., Tsien, J. Z. (2000). Genetic analysis of learning behavior-induced structural plasticity. Hippocampus, 10,605-609.[Medline]
• Rende, R., Plomin, R. (1992). Diathesis–stress models of psychopathology: A quantitative genetic perspective. Applied and Preventative Psychology, 1,177-182.
• Rose, R. J. (1995). Genes and human behavior. Annual Review of Psychology, 46,625-654.[Medline]
• Rose, R. J., Dick, D. M., Viken, R. J., Kaprio, J. (2001). Gene–environment interaction in patterns of adolescent drinking: Regional residency moderates longitudinal influences on alcohol use. Alcoholism: Clinical and Experimental Research, 25,637-643.[Medline]
• Rose, R. J., Kaprio, J., Winter, T., Koskenvuo, M., Viken, R. J. (1999). Familial and socioregional environmental effects on abstinence from alcohol at age sixteen. Journal of Studies on Alcohol, 13,63-74.
• Rowe, D. C. (2001). Assessing genotype–environment interactions and correlations in the postgenomic era. In R. Plomin, J. C. DeFries, G. E. McClearn, & P. McGuffin (Eds.), Behavioral genetics (4^th ed., pp. 71–86). New York: Worth.
• Rowe, D. C., Jacobson, K. C., van den Oord, E. J. C. G. (1999). Genetic and environmental influences on vocabulary IQ: Parental education level as moderator. Child Development, 70,1151-1162.[Medline]
• Rutter, M. (1989). Pathways from childhood to adult life. Journal of Child Psychology and Psychiatry, 30,23-51.[Medline]
• Rutter, M. (1990). Psychosocial resilience and protective mechanisms. In J. E. Rolf, A. S. Masten, D. Cicchetti, K. H. Neuchterlein, & S. Weintraub (Eds.), Risk and protective factors in the development of psychopathology (pp. 181–214). New York: Cambridge.
• Rutter, M., Pickles, A. (1991). Person–environment interactions: Concepts, mechanisms, and implications for data analysis. In T. E. Wachs and R. Plomin (Eds.), Conceptualization and measurement of organism–environment interaction (pp. 105–141). Washington, DC: American Psychological Association.
• Rutter, M., Silberg, J. (2002). Gene–environment interplay in relation to emotional and behavioral disturbance. Annual Review of Psychology, 53,463-490.[Medline]
• Sampson, R. J., Laub, J. H. (1990). Crime and deviance over the life course: The salience of adult social bonds. American Sociological Review, 55,609-627.
• Sampson, R. J., Laub, J. H. (1993). Crime in the making: Pathways and turning points in life. Cambridge, MA: Harvard University Press.
• Sampson, R. J., Morenoff, J. D., Gannon-Rowley, T. (2002). Assessing neighborhood effects: Social processes and new directions. Annual Review of Sociology, 28,443-478.
• Sawa, A., Snyder, S. H. (2002). Schizophrenia: Diverse approaches to a complex disease. Science, 296,692-695.[Abstract/Free Full Text]
• Scarr, S. (1992). Developmental theories for the 1990s: Development and individual differences. Child Development, 63,1-19.[Medline]
• Scarr-Salapatek, S. (1971). Race, social class, and IQ. Science, 174,1285-1295.[Free Full Text]
• Sewell, W. H. (1992). A theory of structure: Duality, agency and transformation. American Journal of Sociology, 98,1-29.
• Shanahan, M. J., Sulloway, F. J., Hofer, S. M. (2000). Conceptual models of context in developmental studies. International Journal of Behavioral Development, 24,421-427.[Abstract/Free Full Text]
• Shimizu, E., Tang, Y.-P., Rampon, C., Tsien, J. Z. (2000). NMDA receptor-dependent synaptic reinforcement as a crucial process for memory consolidation. Science, 290,1170-1174.[Abstract/Free Full Text]
• Sigvardsson, S., Bohman, M., Cloninger, C. R. (1996). Replication of the Stockholm Adoption Study of Alcoholism. Archives of General Psychiatry, 53,681-687.[Abstract/Free Full Text]
• Silberg, J., Rutter, M., Neale, M., Eaves, L. (2001). Genetic moderation of environmental risk for depression and anxiety in adolescent girls. British Journal of Psychiatry, 179,116-121.[Abstract/Free Full Text]
• Singer, B., Ryff, C. D. (1999). Hierarchies of life histories and associated health risks. Annals of the New York Academy of Sciences, 896,96-115.[Medline]
• Singer, B., Ryff, C. D. (2001). Person-centered methods for understanding aging. In R. H. Binstock and L. K. George (Eds.), Handbook of aging and the social sciences (pp. 44–65). San Diego: Academic Press.
• Singer, B., Ryff, C. D., Carr, D., Magee, W. (1998). Linking life histories and mental health: A person-centered strategy. In Sociological methodology (pp. 1–58). Washington, DC: American Sociological Association.
• 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, 125,392-407.[Medline]
• Suomi, S. J. (2004). How gene–environment interactions shape biobehavioral development: Lessons from the studies of Rhesus monkeys. Research in Human Development, 1,205-222.
• Sweeney, M. M., Horwitz, A. V. (2001). Infidelity, initiation, and the emotional climate of divorce: Are there implications for mental health? Journal of Health and Social Behavior, 42,295-309.[Medline]
• Tienari, P., Wynne, L. C., Moring, J., Lahti, I., Naarala, M., Sorri, A., et al. (1994). The Finnish adoptive family study of schizophrenia: Implications for family research. British Journal of Psychiatry, 164,20-26.
• Turner, R. J., Lloyd, D. A. (1995). Lifetime traumas and mental health: The significance of cumulative adversity. Journal of Health and Social Behavior, 36,360-376.[Medline]
• Turner, R. J., Wheaton, B., Lloyd, D. A. (1995). The epidemiology of social stress. American Sociological Review, 60,104-125.
• Udry, J. R. (1988). Biological predispositions and social control in adolescent sexual behavior. American Sociological Review, 53,709-722.
• van Ommen, B., Stierum, R. (2002). Nutrigenomics: Exploiting systems biology in the nutrition and health arena. Current Opinion in Biotechnology, 13,517-521.[Medline]
• van Os, J., Marcelis, M. (1998). The ecogenetics of schizophrenia: A review. Schizophrenia Research, 32,127-135.[Medline]
• Vieira, C., Pasyukona, E. G., Zeng, Z. B., Hackett, J. B., Lyman, R. F., MacKay, T. F. C. (2000). Genotype–environment interaction for quantitative trait loci affecting life span in Drosophila melanogaster. Genetics, 154,213-227.[Abstract/Free Full Text]
• Wadsworth, S. J., Corley, R. P., Hewitt, J. K., DeFries, J. C. (2001). Stability of genetic and environmental influences on reading performance at 7, 12, and 16 years of age in the Colorado Adoption Project. Behavior Genetics, 31,353-359.[Medline]
• Wahlsten, D. (1999). Single-gene influences on brain and behavior. Annual Review of Psychology, 50,599-624.[Medline]
• Wheaton, B. (1990). Life transitions, role histories, and mental health. American Sociological Review, 55,209-223.
• Wohlwill, J. B. (1973). The study of behavioral development. New York: Academic Press.
• Zhang, H., Yu, C.-Y., Singer, B., Xiong, M. (2001). Recursive partitioning for tumor classification with gene expression microarray data. Proceedings of the National Academy of Sciences U S A, 98,6730-6735.[Abstract/Free Full Text]

This article has been cited by other articles:

				M. Jokela, L. Keltikangas-Jarvinen, M. Kivimaki, S. Puttonen, M. Elovainio, R. Rontu, and T. Lehtimaki Serotonin Receptor 2A Gene and the Influence of Childhood Maternal Nurturance on Adulthood Depressive Symptoms Arch Gen Psychiatry, March 1, 2007; 64(3): 356 - 360. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) Services Download to citation manager Citing Articles Citing Articles via HighWire PubMed PubMed Citation