Optimal versus adaptive development

Much theory, research, and writing about development and psychopathology today is based on what has been described recently as the “mental health model” (Belsky, Reference Belsky2008) or “developmental psychopathology model” (Ellis et al., Reference Ellis, Del Giudice, Dishion, Figueredo, Gray and Griskevicius2012; Frankenhuis & Del Giudice, Reference Frankenhuis and Del Giudice2012). Central to this theoretical and empirical framework is the implicit if not explicit assumption that there is something ubiquitously referred to in the developmental literature as “optimal” development. The notion of optimal development more or less implies that it is the natural condition of humans from youth to later in life to become secure, autonomous, self-controlling, prosocial, achievement striving, intimate in the context of pair bonds, hardworking, sensitively responsive in parenting, and generative in the next generation, as well as physically healthy, happy, and long lived. Development goes awry (i.e., becomes dysregulated) when forces direct it from its natural or otherwise anticipated course. Thus, being insecure, dependent, impulsive, antisocial, a risk taker, an insensitive parent, and/or depressed are implicitly if not explicitly regarded as developmental outcomes that nature did not intend.

However, an evolutionary perspective challenges this prevailing view of so-called dysfunctional, dysregulated, or maladaptive development, especially when, consistent with diathesis–stress thinking, it arises within settings of adversity (Belsky, Reference Belsky2008; Daly & Wilson, Reference Daly and Wilson2005; Ellis, Reference Ellis2004; Ellis, Boyce, et al., Reference Ellis and Boyce2011; Ellis et al., Reference Ellis, Del Giudice, Dishion, Figueredo, Gray and Griskevicius2012; Hinde & Stevenson-Hinde, Reference Hinde and Stevenson-Hinde1990). Because stressful and supportive environments have been part of human experience throughout our evolutionary history, evolutionists conceptualize developmental systems as shaped by natural selection to respond adaptively to both kinds of contexts. From this perspective, the experience of encountering stressful environments does not so much disturb, or dysregulate, development as direct, or regulate, it toward strategies that are or at least once were biologically adaptive under conditions of adversity. Moreover, by adaptive we mean functional and perhaps even optimal, given the circumstances in which the individual finds him/herself, in that they directly or indirectly promoted the individual's reproductive fitness. (See below for further discussion of this fundamental goal of all living things.)

Some of the most compelling evidence highlighting the validity of the adaptational perspective just outlined can be found in the much-heralded rodent work showing that what is regarded by many as “low-quality” maternal care in the rat, as reflected in limited licking and grooming of the newborn pup, modifies stress physiology and brain morphology, but in functional and strategic rather than maladaptive ways. Even though evidence of higher corticosterone levels, shorter dendritic branch lengths, and lower spine density in hippocampal neurons would appear to provide indisputable evidence of negative effects of early experience, such presumptively nonoptimal developments appear strikingly functional, given the conditions that gave rise to them. This is because the developmental experiences in question enhance learning and memory processes under stressful conditions (Champagne et al., Reference Champagne, Bagot, van Hasselt, Ramakers, Meaney and de Kloet2008). Moreover, such putatively disturbed or dysregulated physiological and morphological changes promote central features of defensive and reproductive strategies (behavior under threat, open-field exploration, pubertal development, sexual behavior, and parenting; Cameron et al., Reference Cameron, Champagne, Parent, Fish, Ozaki-Kuroda and Meaney2005) in ways consistent with evolutionary models of adaptive reproductive strategies (Belsky et al., Reference Belsky, Steinberg and Draper1991; Chisholm, Reference Chisholm1993), a subject we consider in greater detail below.

Ultimately, developmental adaptations to high-stress environments enable individuals to make the best of a bad situation (i.e., mitigate inevitable fitness costs), even though “the best” may still constitute a high-risk strategy that jeopardizes the person's health and survival (e.g., Mulvihill, Reference Mulvihill2005; Shonkoff, Boyce, & McEwen, Reference Shonkoff, Boyce and McEwen2009). It is certainly possible to view many of the developmental sequelae of child maltreatment in this manner. After all, it is not difficult to appreciate the developmental “wisdom” in being especially vigilant to threat, engaging in biased attributions or behaving aggressively if the lessons of a painful life have taught one to “act first and ask questions later” and “presume the worst rather than the best of people.” By the same token, the indiscriminant friendliness of institutionally reared children (Bakermans-Kranenburg et al., Reference Bakermans-Kranenburg, Steele, Zeanah, Muhamedrahimov, Vorria and Dobrova-Krol2011) could be regarded as a highly functional strategy for seeking help and support in a world that has offered little, even if it looks disturbed or disordered from the perspective of family-reared children.

Reflection on these observations should clarify why natural selection may have crafted human development to respond to certain conditions by making individuals behave in ways we now regard as disturbed, dysfunctional, dysregulated, psychopathological, and nonoptimal. As such, this analysis perhaps strengthens the theoretical foundations of diathesis–stress thinking that we questioned earlier. That is, one can now understand why nature may have crafted at least some individuals to respond to adversity by developing in putatively disturbed and dysregulated ways. Nevertheless, it remains our view that the potential costs of not responding to adversity in a putatively dysfunctional, dysregulated, and disordered manner merits more consideration by students of development and psychopathology than it has been given. Thus, in addition to considering the possible benefits of behavior regarded by many as dysregulated and problematic, students of development and psychopathology would be well advised to consider what the costs may be of behaving, given the circumstances some live in, in those ways that mainstream thinking labels as optimal.

The argument advanced here ultimately calls into question the use of the term optimal when discussing development and suggests that there could be substantial merit in thinking more in terms of why it may actually make good sense for children, adolescents, and adults to behave in the dysregulated and troubled ways in which they do. However, such a perspective does not mean abandoning efforts to reduce psychological pain and suffering. What it suggests is that to do so most effectively may require consideration of the payoffs and trade-offs that may have resulted in natural selection crafting an organism to develop and behave in the way we observe them to, especially under conditions of adversity. We suspect that intervention efficacy could be enhanced if interventionists seriously considered “biological gravity” (natural selection has crafted phenotypes that serve to enhance fitness) when designing programs rather than think only in traditional mental health terms about what behavior is optimal and what is not. Much like the efforts of would-be aviators who predated the modern ones who eventually came to grips with physical gravity, thereby recognizing the need for lift and propulsion before they could fly, some intervention efforts reflect little more than attaching wings to the arms and jumping off cliffs. Without understanding the nature, origin, and function of disordered and dysregulated behavior, including its evolutionary basis (i.e., biological gravity), the likelihood of preventing or remediating it may be seriously compromised. Ellis and associates (Reference Ellis and Bjorklund2012) develop this point further in the case of adolescent risk taking, highlighting existing interventions that appear consistent or inconsistent with evolutionary thinking as well as offering novel, evolutionary-inspired ones.

Probabilism versus determinism

It is widely appreciated by developmentalists today that factors that shape human development typically do so in a probabilistic rather than deterministic fashion. That is, forces that we expect to enhance or undermine human functioning, in the traditional mental health sense, will not always do so, even if they sometimes, or even often, do so. This probabilistic nature of development also needs to be appreciated in its evolutionary basis, something that seems forgotten when it is argued that disturbances in development could not reflect evolutionary adaptations because they so often lead to failure, including perhaps failure to disburse genes in future generations in the modern western world. Two points merit consideration in response to this seemingly reasonable observation.

The first is that natural selection is backward not forward looking, selecting phenotypes that have fostered fitness in the past. This means that what was once evolutionarily adaptive may no longer be so today. However, even if that is true, which should not necessarily be the default assumption, it would not “biologically” follow that the neurobiological and psychological processes that gave rise to them in the first place would have been short circuited in order to no longer engender the behavioral functioning that is typically regarded as problematic today. If that were the case, it would be easy to override the human proclivity to consume sugary foods. After all, this penchant for sweets evolved at a time when such substances were rarely or intermittently available, yet in the modern world today they are continuously available and abundant. Thus, even if a developmental process no longer achieves the goal it was selected to realize but instead engenders developmental problems in the modern world (e.g., obesity), the process itself likely remains operative because of the time it takes for evolution to engineer genetic change.

The second point regarding probabilism and nonoptimal development is that, for selection to take place, a phenotype does not have to always or often succeed in enhancing fitness; it only needs to have done so more often than alternative phenotypes once did (Frankenhuis & Del Giudice, Reference Frankenhuis and Del Giudice2012). Moreover, the costs of failure need to be weighed against the benefits of success (Ellis et al., Reference Ellis, Del Giudice, Dishion, Figueredo, Gray and Griskevicius2012; Figueredo & Jacobs, Reference Figueredo, Jacobs, Frias-Armenta and Corral-Verdugo2010; Frankenhuis & Del Giudice, Reference Frankenhuis and Del Giudice2012). Thus, even if antisocial behavior or depression or teenage pregnancy seem counterproductive from either a mental health or fitness standpoint and could have proven so in the past (failing more often than succeeding), such putative disturbances in development could still have been selected, especially when a low-frequency success engendered a substantial payoff. What this implies, of course, is that some putatively dysregulated behavior will be of the high-stakes’ variety: unlikely to succeed yet succeeding hugely when it does.

What is development trying to achieve?

The concept of optimal development is most certainly intended to capture phenomena like health, happiness, security, longevity, and other indisputably highly valued life conditions. On the basis of much developmental writing, it would not seem misguided to infer that nature has crafted human development to achieve such outcomes. From an evolutionary perspective, however, this is certainly not (exclusively) the case. Although security, good health, happiness, and a long life may be the means by which some or many individuals successfully pass genes on to future generations and promote their more general inclusive fitness, especially when growing up and living under benign or benevolent conditions, there is no reason to presume that these phenotypes achieved or can achieve these same ends under all conditions, especially contrasting conditions. This observation was central to Belsky's (in press; Belsky et al., Reference Belsky, Steinberg and Draper1991) effort to recast socialization theory in evolutionary perspective two decades ago, which was built on Draper and Harpending's (Reference Draper and Harpending1982) evolutionary insight about why father absence may be related to “promiscuous” sexual behavior on the part of females.

Rather than presuming that development was crafted by natural selection to yield phenotypes reflective of optimal development, Belsky et al.'s (Reference Belsky, Steinberg and Draper1991) psychosocial acceleration theory and its derivatives (Chisholm, Reference Chisholm1993; Del Giudice, Reference Del Giudice2009; Del Giudice, Angeleri, & Manera, Reference Del Giudice, Angeleri and Manera2009; Ellis, Reference Ellis2004; Ellis & Garber, Reference Ellis and Garber2000; Frankenhuis & Panchanathan, Reference Frankenhuis and Panchanathan2011; James, Ellis, Schlomer, & Garber, Reference James, Ellis, Schlomer and Garber2012) are founded on the premise that the ultimate goal of all living things, including humans, is to disperse their genes to future generations and that the means of achieving this fitness end is contextually contingent. Thus, what works under certain developmental conditions does not necessarily work under others. It is ultimately for this reason that the notion of optimal development is fundamentally problematic: what is optimal in one (modern and well resourced?) context may not be so in another.

We submit that students of development and psychopathology would do well to think in fitness terms when trying to understand phenotypes of concern. Upon doing so, what is regarded as dysregulated, dysfunctional, disturbed, or disordered may come to be seen in a new and informative light, a point recently made by Ellis and associates (Reference Ellis, Del Giudice, Dishion, Figueredo, Gray and Griskevicius2012) with respect to adolescent risk taking. Once again, however, an evolutionary perspective with its emphasis on selection of phenotypes that foster the dispersion of genes in future generations does not mean that nothing should be done. Such would reflect succumbing to the naturalistic fallacy, inferring what “ought to be” from what “is,” a long-derided perspective. However, we do suggest that seeing development the way natural selection views it may facilitate efforts to address contemporary concerns.

Fast versus slow development

Because of its grounding in evolutionary life-history theory (Kaplan & Gangestad, Reference Kaplan, Gangestad and Buss2005; Stearns, Reference Stearns1992), psychosocial acceleration theory appreciated that growing up fast and behaving in putatively nonoptimal ways was sometimes advantageous when it came to dispersing genes in future generations. Although maturing early is known to carry numerous risks for females, including early sexual debut, early first birth, sexually transmitted disease, and breast cancer (Ellis, Reference Ellis2004), such “ontogenic risk taking” was theorized to make biological sense under conditions of adversity. It was for this reason that Belsky et al. (Reference Belsky, Steinberg and Draper1991) proposed that rearing conditions that induced in the individual the sense that the future was risky and that supportive relationships were precarious would promote early maturation, early sexual debut, promiscuous sexual behavior, unstable pair bonds, limited parental investment, and the bearing of more rather than fewer offspring. After all, from the standpoint of reproductive fitness, it is better to “live fast and die young,” having offspring along the way, than to die (or become disabled) before getting the chance to reproduce (Chisholm, Reference Chisholm1993; Nettle, Reference Nettle2010). Thus, adolescents who, for example, respond to dangerous environments by developing insecure attachments; adopting opportunistic, advantage-taking interpersonal orientations; engaging in externalizing behavior; discounting the future; and experiencing early sexual debut are no less functional or even less regulated than are those responding to a well-resourced and supportive social environment by developing the opposing characteristics and orientations (Belsky et al., Reference Belsky, Steinberg and Draper1991; Ellis, Boyce, et al., Reference Ellis, Boyce, Belsky, Bakermans-Kranenburg and van IJzendoorn2011).

What this analysis highlights is that development involves trade-offs, a concept central to an evolutionary, life-history perspective. As just noted, a primary risk associated with waiting to mature, find a mate, and breed is that one may not survive long enough to achieve any or all of these ends. Additional risks include limits to ones’ capabilities and competencies, as well as poor health, thereby making it difficult to attract mates and/or protect and facilitate the development of the offspring one eventually produces. Ellis and associates (in press) recently pointed out in this regard that the concept of allostatic load, which informs ever more developmental inquiry (e.g., Evans & Kim, Reference Evans and Kim2012; Fuller-Rowell, Evans, & Ong, Reference Fuller-Rowell, Evans and Ong2012), may need rethinking. Instead of conceptualizing this construct as part of a disease process, it could be better to think of it in terms of evolutionary and life-history trade-offs. Although there are indisputable costs associated with developing in certain ways under conditions of ecological stress, costs that seem to undermine child well-being, long-term health and even longevity, these costs may be compensated for by the increased likelihood that the individual incurring these costs would make it to maturity, find a mate, and produce offspring.

It is intriguing that the aforementioned rodent work once again proves consistent with the view just advanced: life conditions can regulate the rate of development in the service of fitness goals. Moreover, this is because it is not true that stressful rearing conditions that take the form of limited maternal licking of the rat pup increase fearful and defensive behaviors via the epigenetic regulation of the stress-response system. All too unappreciated by many developmentalists who (appropriately) herald this work for its documentation of such effects is that such putatively limited maternal care also accelerates sexual maturation, increases sexual behavior, and reduces parental investment. It would appear then that parental “neglect” in the form of limited licking and grooming serves as a means of regulating (not dysregulating) development in a biologically strategic manner. This is consistent with Belsky et al.'s (Reference Belsky, Steinberg and Draper1991) psychosocial acceleration theory, as noted by Cameron and associates (Reference Cameron, Champagne, Parent, Fish, Ozaki-Kuroda and Meaney2005). In this rodent context, neglect can be regarded as a mechanism through which rat mothers facilitate their offspring's development by accelerating it toward survival and reproductive strategies that are optimal under conditions of adversity. Therefore, it would seem mistaken to view diminished licking and grooming as “poor maternal care” or the development induced by such care as disturbed, dysregulated, or nonoptimal. From an evolutionary perspective, the care provided by the putatively neglectful parents may serve as appropriate preparation of their offspring for the ecological conditions into which they are likely to mature.

Costs and limits of plasticity

Like many theories and perspectives on human development, psychosocial acceleration theory presumes that development is contextually conditional and developmentally plastic. Developmental experiences and environmental exposures regulate growth and development such that how an individual turns out is at least partly based on environmental influences, shared or unshared, to use the terminology of behavior genetics (Plomin & Daniels, Reference Plomin and Daniels1987). Evolutionists use the terminology adaptive developmental plasticity to characterize this process (Schlichting & Pigliucci, Reference Schlichting and Pigliucci1998; West-Eberhard, Reference West-Eberhard2003), as do evolutionary-minded developmentalists (Belsky et al., Reference Belsky, Steinberg and Draper1991; Chisholm, Reference Chisholm1993; Ellis, Reference Ellis2004; Kuzawa, Reference Kuzawa2005; Nettle, Reference Nettle2011; Quinlan, Reference Quinlan2007).

However, what many developmentalists have failed to appreciate (and we include ourselves among this group, at least until recently) is that plasticity carries costs (DeWitt, Sih, & Wilson, Reference DeWitt, Sih and Wilson1998; Frankenhuis & Del Giudice, Reference Frankenhuis and Del Giudice2012; Schlichting & Pigliucci, Reference Schlichting and Pigliucci1998; Sih, Reference Sih2011). There is the cost associated with the increased complexity that may be required of organisms in order to be capable of developing in diverse ways depending on contextual conditions. The costs derive from the idea that, as complexity increases, so does the risk of something going wrong, in the same way that a machine with more moving parts has more ways to break down than one with fewer parts. Consider the following and related costs of plasticity enumerated by DeWitt and associates (Reference DeWitt, Sih and Wilson1998): energetic costs of the sensory and regulatory mechanisms of plasticity; the production cost of environmentally inducible structures and processes; the cost of acquiring information about the environment, including the energy expended and potential risks incurred; and the imprecision in “manufacturing” a phenotype based on environmental experience. Collectively, these would seem to imply that some individuals may manifest limited malleability, with their development highly canalized, trading off the benefits of being able to thrive in more than a single ecological niche against costs of a breakdown in the developmental machinery and processes needed to do so.

It is also important to appreciate with respect to the costs of plasticity that modifying development early in life in response to developmental experiences and environmental exposures may result in a “mismatch” between the developed organism and the environment in which it finds itself later in life (Nederhof & Schmidt, Reference Nederhof and Schmidt2012; Sih, Reference Sih2011). In this way, seemingly adaptive developmental plasticity, selected to allow the organism to be “programmed” on the basis of early experience so as to “fit” the later environment, can undermine developmental opportunity and eventual fitness rather than enhance it. Thus, here is another reason why there seem to be grounds for expecting individual differences in developmental plasticity. After all, those individuals whose development is more canalized and less programmable by environmental exposures would be less likely to experience a mismatch resulting from developmental plasticity. Of course, such highly canalized specialists with more or less fixed characteristics could end up mismatched to their environment, but for different reasons from that of the plastic, programmable generalist, who may develop in a manner guided by experiences earlier in development.

What these observations imply is that the being developmentally malleable should not be regarded, as it typically is, as an unmitigated good for the developing organism and, thus, as something that should be pronounced in all individuals. In some cases it would appear wiser not to be plastic or to be less malleable, manifesting a more canalized pattern of development, than others and thus less susceptible to environmental conditions. What this suggests, as noted repeatedly, is that we should expect individual differences in developmental plasticity. In other words, developmental plasticity should be conceptualized as a phenotype in its own right, something that will vary across individuals for a variety of reasons. For the most part, this possibility has not been appreciated by those studying development and psychopathology, at least not until recently. Where it has, questions can be raised about how those studying development and psychopathology have conceptualized the issue.

Resilience or reduced plasticity?

Implicit in the diathesis–stress model is the notion of resilience. After all, the resilient individual develops in a manner opposite that of the vulnerable one according to this theoretical perspective, because he or she is, by definition, “protected” from the anticipated negative effects of adversity (Cicchetti & Rogosch, Reference Cicchetti and Rogosch2012; Garmezy, Masten, & Tellegen, Reference Garmezy, Masten and Tellegen1984; Masten, Best, & Garmezy, Reference Masten, Best and Garmezy1990; Rutter, Reference Rutter1987, Reference Rutter2012; Werner & Smith, Reference Werner and Smith1982). Because diathesis–stress thinking focuses upon response to adversity, it makes no claims about differences in how resilient and vulnerable individuals develop under benign or supportive conditions. In this regard it differs substantially from differential-susceptibility thinking, because differential susceptibility presumes that the putatively vulnerable are generally plastic in a “for-better-and for-worse manner” (Belsky et al., Reference Belsky, Bakermans-Kranenburg and van IJzendoorn2007); whereas the putatively resilient will be relatively unresponsive to supportive conditions, just like they are to unsupportive ones, with their developmental pathways highly canalized. This view leads us to take issue with the use of the terms buffer and protection when discussing how more developmentally plastic individuals respond to positive circumstances; such individuals are not so much protected (against what?) when they experience good conditions, as some contend (e.g., Zimmermann et al., Reference Zimmermann, Brueckl, Nocon, Pfister, Binder and Uhr2011), but they experience enhanced functioning, something not expected of those resilient to adversity (for a review, see Pluess & Belsky, Reference Pluess and Belsky2013).

To the extent that this is the case, it would suggest that resilience, or at least that which derives from endogenous characteristics of the individual (e.g., temperament, genotype), is something of a misnomer, reflecting a highly canalized developmental process with limited developmental plasticity more than anything else (Pluess & Belsky, Reference Pluess and Belsky2013). Unfortunately, too many investigations of risk and protection informed by diathesis–stress thinking have not only failed to entertain this possibility for theoretical reasons but have also not been well positioned to evaluate it for reasons previously outlined: a supportive environment in much resilience research often reflects little more than the absence of adversity (e.g., not impoverished, not maltreated), and positive functioning often reflects nothing more than the absence of dysfunction (e.g., not depressed, not aggressive). In order to be able to determine whether those resilient to adversity are also relatively unresponsive to environmental support and enrichment, as the differential-susceptibility framework presupposes, more will be required than merely expanding the range of measurements of environments and development.

Vantage sensitivity: Response to positive environmental influences

According to differential susceptibility reasoning, diathesis–stress thinking only addresses the “dark side” of the Person × Environment equation, ignoring the “bright side.” Recently, we adopted and promoted the term vantage sensitivity, originally introduced by Manuck and associates (Manuck, Reference Manuck2011; Sweitzer et al., Reference Sweitzer, Halder, Flory, Craig, Gianaros and Ferrell2013), to convey the notion that some individuals are more sensitive and positively responsive to the environmental advantages to which they are exposed and vantage resistance to describe the absence of such responsivity in other individuals (Pluess & Belsky, Reference Pluess and Belsky2013). In that effort, we also reviewed emerging empirical evidence for vantage sensitivity as a function of behavioral, physiological, and genetic factors previously associated with differential susceptibility and diathesis–stress. Although the same factors seem to moderate effects of environmental influences, irrespective of whether they are exclusively negative (i.e., diathesis–stress), exclusively positive (i.e., vantage sensitivity), or range from the negative to the positive (i.e., differential susceptibility), it is important to caution against inferring that every risk factor will also, by default, function as a vantage-sensitivity and thus differential-susceptibility factor. Ultimately, this is an empirical question, as there may very well be specific factors that play a predominant role in diathesis–stress but not in vantage sensitivity and vice versa. Future research should address whether there are psychological, behavioral, or neurobiological factors and mechanisms that are specific to diathesis–stress or vantage sensitivity.

Competitive evaluation of models of Person × Environment interaction

Much of the evidence cited in our earlier reviews of differential-susceptibility related evidence (Belsky et al., Reference Belsky, Jonassaint, Pluess, Stanton, Brummett and Williams2009; Belsky & Pluess, Reference Belsky and Pluess2009a), as well as in subsections to follow, is based on “eye-ball” tests of graphed interactions between an environmental predictor and developmental outcome moderated by a plasticity factor. The limits of this approach to “testing” the differential-susceptibility hypothesis led Belsky et al. (Reference Belsky, Bakermans-Kranenburg and van IJzendoorn2007) to propose explicit empirical criteria. Subsequently, Kochanska, Kim, Barry and Philibert (Reference Kochanska, Kim, Barry and Philibert2011) extended this effort by using the regions of significance approach (Preacher, Curran, & Bauer, Reference Preacher, Curran and Bauer2006; Takane & Cramer, Reference Takane and Cramer1975) to separately evaluate the for-better and for-worse sides of the differential-susceptibility hypothesis upon documenting the required cross-over interaction.

Roisman and associates (Reference Roisman, Newman, Fraley, Haltigan, Groh and Haydon2012) recently offered an even more demanding approach for evaluating differential susceptibility by providing two new quantitative metrics in combination with the regions of significance analysis: the proportion of interaction index, which represents the proportion of the total area of an interaction plot that is uniquely attributable to differential susceptibility, and the proportion affected index which quantifies the proportion of people in the sample who fall above the cross-over point (i.e., proportion of sample in the for-better condition). In addition, Roisman et al. (Reference Roisman, Newman, Fraley, Haltigan, Groh and Haydon2012) recommend testing for nonlinear relationships between predictor and outcome and correction for multiple comparisons.

All of the approaches offered to date to evaluate differential susceptibility have involved conducting essentially exploratory tests of Organism × Environment interactions as a first step before further interrogating the data at hand. Widaman and associates (Belsky, Pluess, & Widaman, Reference Belsky, Pluess and Widaman2013; Widaman et al., Reference Widaman, Helm, Castro-Schilo, Pluess, Stallings and Belsky2012) recently advanced a strategy that sidesteps such approaches, moving directly to test competing theoretical models. In a first step, the cross-over point between high- and low-susceptible individuals is estimated with the help of a reparameterized regression equation (Widaman et al., in press). If the estimated cross-over point lies within the observed range of the predictor, four additional reparameterized regression models are applied: two constrained to reflect a weak and a strong version of diathesis–stress and two to reflect a weak and a strong version of differential susceptibility. Fit indices of the four models are then compared to evaluate whether the data fits better with a diathesis–stress or differential-susceptibility model (Belsky et al., Reference Belsky, Pluess and Widaman2013). The strong version of differential susceptibility presupposes that there are those who are susceptible to positive and negative contextual conditions and others who are not; the weak version presupposes that, although many may be susceptible to environmental influences to some extent, some are more affected than others.

The time has thus come to move beyond exploratory tests of Organism × Environment interaction in regression and analysis of variance designs and to more formally evaluate in a competitive manner the a priori hypotheses central to diathesis–stress and differential-susceptibility models of environmental action.

Temperament/negative emotionality as plasticity factor

The first evidence Belsky (Reference Belsky1997, Reference Belsky, Ellis and Bjorklund2005) could point to consistent with his evolutionary-inspired, differential-susceptibility hypothesis surprisingly indicated that a difficult temperament or high levels of negative emotionality (e.g., fear, inhibition) early in life not only function as a risk factor that increases the likelihood of problematic functioning when coupled with contextual risks (e.g., poverty, maternal depression), as long presumed, but also seemed to predispose individuals to benefit from benign or supportive experiences. Since our 2009 review, even more evidence to this effect has emerged in the case of children, as revealed in research linking maternal empathy (Pitzer, Jennen-Steinmetz, Esser, Schmidt, & Laucht, Reference Pitzer, Jennen-Steinmetz, Esser, Schmidt and Laucht2011) and anger (Poehlmann et al., Reference Poehlmann, Hane, Burnson, Maleck, Hamburger and Shah2012) with externalizing problems; mutual responsiveness observed in the mother–child dyad with effortful control (Kim & Kochanska, Reference Kim and Kochanska2012); intrusive maternal behavior (Conway & Stifter, Reference Conway and Stifter2012) and poverty (Raver, Blair, & Willoughby, Reference Raver, Blair and Willoughby2012) with executive functioning; and sensitive parenting with social, emotional, and cognitive–academic development (Roisman et al., Reference Roisman, Newman, Fraley, Haltigan, Groh and Haydon2012).

However, it is not just responsiveness to parenting that negativity seems to moderate in a for better and for worse manner, as documented in recent work examining the influence of teacher–child conflict on change in symptomology during the primary-school years (Essex, Armstrong, Burk, Goldsmith, & Boyce, Reference Essex, Armstrong, Burk, Goldsmith and Boyce2011) and of marital quality on sensitive mothering (Jessee et al., Reference Jessee, Mangelsdorf, Brown, Schoppe-Sullivan, Shigeto and Wong2010). The latter work also makes it clear that it is not just children who prove differentially susceptible to environmental effects. We would be remiss, however, if we failed to call attention to studies showing it was less rather than more negatively emotional or reactive children who proved highly responsive to environmental experiences in a manner consistent with differential susceptibility (Essex et al., Reference Essex, Armstrong, Burk, Goldsmith and Boyce2011, see their figure 2, bottom; Du Rocher Schudlich, White, Fleischhauer, & Fitzgerald, Reference Du Rocher Schudlich, White, Fleischhauer and Fitzgerald2011) and to other evidence showing that sometimes difficult temperament operates only in a manner consistent with the diathesis–stress rather than differential-susceptibility framework (Kochanska & Kim, Reference Kochanska and Kim2013; Roisman et al., Reference Roisman, Newman, Fraley, Haltigan, Groh and Haydon2012; Yaman, Mesman, van IJzendoorn, & Bakermans-Kranenburg, Reference Yaman, Mesman, van IJzendoorn and Bakermans-Kranenburg2010).

The cited research reveals that diverse indices of negative emotionality function as plasticity factors (e.g., fear, inhibition, distress, difficult temperament, negative affect), which raises several issues that future research should address. For example, is there a particular component or feature of negativity that is principally responsible for the moderational effects detected (Buss, Reference Buss2011) or could more progress be made with regard to this issue by embracing the integrated, evolutionary, hawk–dove model of temperament (Korte, Koolhaas, Wingfield, & McEwen, Reference Korte, Koolhaas, Wingfield and McEwen2005) that Sturge-Apple, Davies, Martin, Cicchetti and Hentges (Reference Sturge-Apple, Davies, Martin, Cicchetti and Hentges2012) recently found to be empirically useful? Whereas hawklike strategies are characterized by approach, dominant–negative affect (e.g., anger), and activity, dovelike ones involving avoidance, inhibition, and vulnerable affect (e.g., fear) seem more consistent with heightened plasticity according to existing research.

We also wonder about the extent to which the moderational effect of negative emotionality depends on the origins of this plasticity factor, given evidence that negativity is heritable (e.g., Buss & Plomin, Reference Buss and Plomin1984; Rhee et al., Reference Rhee, Corley, Friedman, Hewitt, Hink and Johnson2012) but also shaped by prenatal and postnatal experience (Belsky & Pluess, Reference Belsky and Pluess2009b; Davis, Glynn, Waffarn, & Sandman, Reference Davis, Glynn, Waffarn and Sandman2011; Glynn & Sandman, Reference Glynn and Sandman2011; for a review, see Pluess & Belsky, Reference Pluess and Belsky2011), although sometimes more so in some genotypes than others (Ivorra et al., Reference Ivorra, Sanjuan, Jover, Carot, Frutos and Molto2010; Pluess et al., Reference Pluess, Velders, Belsky, van IJzendoorn, Bakermans-Kranenburg and Jaddoe2011). Thus, what may be particularly important is distinguishing negative emotionality that reflects an experience-induced failure to self-regulate and that which reflects a general hypersensitivity to the environment.

Physiological reactivity as a plasticity factor

Belsky's (Reference Belsky1997, Reference Belsky, Ellis and Bjorklund2005) differential-susceptibility hypothesis was based entirely on evolutionary reasoning, but Boyce and Ellis’ (Reference Boyce and Ellis2005) evolutionary-oriented biological sensitivity to context (BSC) proposal emerged in response to an unanticipated finding that Boyce and associates (Reference Boyce, Chesney, Alkon, Tschann, Adams and Chesterman1995) generated a decade earlier. Central to the BSC framework is the claim that physiological reactivity is a plasticity factor regulated by environmental experience. The recently promulgated adaptive calibration model represents the further development of these ideas (Del Giudice, Ellis, & Shirtcliff, Reference Del Giudice, Ellis and Shirtcliff2011).

The evidence indicating that physiological reactivity operates as a plasticity factor is diverse, with findings showing that more physiologically reactive children are more susceptible, in a for better and for worse fashion, to the effects of actual marital conflict (Obradovic, Bush, & Boyce, Reference Obradovic, Bush and Boyce2011), as well as simulated interparental aggression (Davies, Sturge-Apple, & Cicchetti, Reference Davies, Sturge-Apple and Cicchetti2011), on externalizing problems; of family adversity on school achievement (Obradovic, Bush, Stamperdahl, Adler, & Boyce, Reference Obradovic, Bush, Stamperdahl, Adler and Boyce2010); and of teacher–child conflict on change in symptom severity (Essex et al., Reference Essex, Armstrong, Burk, Goldsmith and Boyce2011). Work showing similar moderational effects of physiological reactivity with respect to family influences on female pubertal development by Ellis and associates (Reference Ellis, Boyce, Belsky, Bakermans-Kranenburg and van IJzendoorn2011) is especially noteworthy, given the roots of this work in psychosocial acceleration theory (Belsky et al., Reference Belsky, Steinberg and Draper1991) and the way in which its findings proved consistent with the differential-susceptibility related revision of the theory (Belsky, Reference Belsky, Rodgers, Rowe and Miller2000). After all, Belsky (Reference Belsky, Rodgers, Rowe and Miller2000) postulated that some females would be “alternative strategists” whose pubertal development proved unrelated to their rearing experiences, just as Ellis and associates (Reference Ellis, Shirtcliff, Boyce, Deardorff and Essex2011) found for girls low in physiological reactivity, whereas others would be “conditional strategists” susceptible to the accelerating and delaying effects of unsupportive and supportive rearing, respectively, just as Ellis et al. (Reference Ellis, Boyce, Belsky, Bakermans-Kranenburg and van IJzendoorn2011) observed in the case of highly reactive girls. Once again, though, we would be remiss if we did not highlight some evidence inconsistent with the BSC thinking, such as that from the aforementioned Obradovic et al. (Reference Obradovic, Bush and Boyce2011) investigation showing that it was young children with low rather than high respiratory sinus arrhythmia reactivity who seemed, in some analyses, to benefit from growing up under supportive family conditions characterized by little marital strife.

One of the key questions that still remains to be answered regarding BSC concerns its life span implications. More specifically, do those individuals who are induced to be highly physiologically reactive early in life, and thus developmentally plastic, remain that way over time? It seems hard to imagine that a child induced by supportive rearing to be highly physiologically reactive, in accordance with the theory (Boyce & Ellis, Reference Boyce and Ellis2005), would remain highly reactive as the years go by. Of course, this is an empirical question, one that has implications for our understanding of the role physiological reactivity does or does not play as a plasticity factor as the individual develops.

Genetic polymorphisms as plasticity factors

Perhaps nowhere has the diathesis–stress framework informed Person × Environment interaction research in recent years more than in the study of G × E interaction, launched by Caspi and associates (Reference Caspi, McClay, Moffitt, Mill, Martin and Craig2002, Reference Caspi, Sugden, Moffitt, Taylor, Craig and Harrington2003) a decade ago. In the time since our 2009 reviews of differential-susceptibility related evidence (Belsky et al., Reference Belsky, Jonassaint, Pluess, Stanton, Brummett and Williams2009; Belsky & Pluess, Reference Belsky and Pluess2009a), ever more G × E findings have appeared consistent with the notion that there are individual differences in developmental plasticity. Much of the evidence to this effect comes from studies involving the serotonin transporter linked polymorphic region (5-HTTLPR) and the dopamine receptor D4 gene (DRD4).

Consider in this regard evidence showing for better and for worse results in the case of those carrying one or more 5-HTTLPR short alleles when the rearing predictor and child outcome were maternal responsiveness and moral internalization (Kochanska et al., Reference Kochanska, Kim, Barry and Philibert2011), child maltreatment and antisocial behavior (Cicchetti, Rogosch, & Thibodeau, Reference Cicchetti, Rogosch and Thibodeau2012), and supportive parenting and positive affect (Hankin et al., Reference Hankin, Nederhof, Oppenheimer, Jenness, Young and Abela2011), respectively. Differential-susceptibility related findings also emerged (among male, African American adolescents) when perceived racial discrimination was used to predict conduct problems (Brody et al., Reference Brody, Beach, Chen, Obasi, Philibert, Kogan and Simons2011), when life events were used to predict neuroticism (Pluess, Belsky, Way, & Taylor, Reference Pluess, Belsky, Way and Taylor2010) and life satisfaction of young adults (Kuepper et al., Reference Kuepper, Wielpuetz, Alexander, Mueller, Grant and Hennig2012), and when retrospectively reported childhood adversity was used to explain aspects of impulsivity among college students (e.g., pervasive influence of feelings, feelings trigger action; Carver, Johnson, Joormann, Kim, & Nam, Reference Carver, Johnson, Joormann, Kim and Nam2011).

To clarify, other 5-HTTLPR related findings prove more consistent with the diathesis–stress than the differential-susceptibility framework (e.g., Bakermans-Kranenburg, Dobrova-Krol, & van IJzendoorn, in press; Brody et al., Reference Brody, Yu, Chen, Kogan, Evans and Windle2012), sometimes even in G × E studies that also chronicle differential susceptibility (e.g., Kochanska et al., Reference Kochanska, Kim, Barry and Philibert2011). Moreover, in apparently rare instances it is those with long rather than short alleles who prove most susceptible to environmental influences, even when differential-susceptibility related findings emerge (e.g., Cicchetti, Rogosch, & Oshri, Reference Cicchetti, Rogosch and Oshri2011; Sulik et al., Reference Sulik, Eisenberg, Lemery-Chalfant, Spinrad, Silva and Eggum2012). Therefore, the results of a recent meta-analysis of G × E findings pertaining to children under 18 years of age are especially noteworthy; they show that short allele carriers are more susceptible to the effects of positive and negative developmental experiences and environmental exposures, at least in Caucasians (van IJzendoorn, Belsky, & Bakermans-Kranenburg, Reference van IJzendoorn, Belsky and Bakermans-Kranenburg2012).

Turning to DRD4, it is worth highlighting that recent differential-susceptibility related evidence showing heightened or exclusive susceptibility of individuals carrying the seven-repeat (7R) allele when the environmental predictor and developmental outcome were maternal positivity and prosocial behavior (Knafo, Israel, & Ebstein, Reference Knafo, Israel and Ebstein2011), early nonfamilial childcare and social competence (Belsky & Pluess, Reference Belsky and Pluess2013), contextual stress and support and adolescent negative arousal (Beach et al., Reference Beach, Lei, Brody, Simons, Cutrona and Philibert2012), childhood adversity and young-adult persistent alcohol dependence (Park, Sher, Todorov, & Heath, Reference Park, Sher, Todorov and Heath2011), and newborn risk status (i.e., gestational age, birth weight for gestational age, length of stay in neonatal intensive care unit) and observed maternal sensitivity (Fortuna et al., Reference Fortuna, van IJzendoorn, Mankuta, Kaitz, Avinun and Ebstein2011), respectively. However, note that at least one investigatory team finds that it is those without the 7R allele who prove most responsive to childhood adversities in a diathesis–stress manner (Das, Cherbuin, Tan, Anstey, & Easteal, Reference Das, Cherbuin, Tan, Anstey and Easteal2011). Nevertheless, a meta-analysis of G × E research involving dopamine-related genes revealed that children 8 years and younger respond to positive and negative developmental experiences and environmental exposures in a manner consistent with differential susceptibility (Bakermans-Kranenburg & van IJzendoorn, Reference Bakermans-Kranenburg and van IJzendoorn2011).

Emerging evidence has also begun to suggest that the brain-derived neurotrophic factor gene (BDNF) may be a plasticity gene (e.g., Chen, Li, & McGue, Reference Chen, Li and McGue2012; Gunnar et al., Reference Gunnar, Wenner, Thomas, Glatt, McKenna and Clark2012; Juhasz et al., Reference Juhasz, Dunham, McKie, Thomas, Downey and Chase2011; Mata, Thompson, & Gotlib, Reference Mata, Thompson and Gotlib2010; Suzuki et al., Reference Suzuki, Matsumoto, Shibuya, Sadahiro, Kamata and Goto2011), as is the oxytocin receptor gene (OXTR; Johansson et al., Reference Johansson, Bergman, Corander, Waldman, Karrani and Salo2012; Poulin, Holman, & Buffone, Reference Poulin, Holman and Buffone2012; Sturge-Apple, Cicchetti, Davies, & Suor, Reference Sturge-Apple, Cicchetti, Davies and Suor2012) and the FK506 binding protein 5 gene (Bevilacqua et al., Reference Bevilacqua, Carli, Sarchiapone, George, Goldman, Roy and Enoch2012; Xie et al., Reference Xie, Kranzler, Poling, Stein, Anton and Farrer2010; Zimmermann et al., Reference Zimmermann, Brueckl, Nocon, Pfister, Binder and Uhr2011). There are perhaps less frequent indications of this in the catechol-O-methyltransferase gene (Laucht et al., Reference Laucht, Blomeyer, Buchmann, Treutlein, Schmidt and Esser2012; Nijmeijer et al., Reference Nijmeijer, Hartman, Rommelse, Altink, Buschgens and Fliers2010) and, in the previously reviewed literature, the monoamine oxidase A gene (MAOA; Enoch, Steer, Newman, Gibson, & Goldman, Reference Enoch, Steer, Newman, Gibson and Goldman2010; Wakschlag et al., Reference Wakschlag, Kistner, Pine, Biesecker, Pickett and Skol2010). It will be worth monitoring these and other polymorphisms as more work emerges to see whether additional evidence proves consistent with differential susceptibility.

Given that virtually all of the polymorphisms just considered have been studied in G × E perspective as a result of the original interests of psychiatric geneticists in simple genotype–phenotype linkages, we cannot but wonder whether an entirely different set of polymorphisms reflecting sensitivity to the environment would have emerged had the phenotype of interest been not some psychiatric disorder but developmental plasticity itself. It therefore behooves investigators to expand the list of genetic “suspects” beyond those thought to be related to disturbances in functioning by thinking biologically about genes that could be related to physiological processes instantiating plasticity. An excellent recent example of this approach is found in Grazioplene, DeYoung, Rogosch, and Cicchetti's (Reference Grazioplene, DeYoung, Rogosch and Cicchetti2013) work showing that the neuronal acetylcholine receptor subunit α-4 genotype interacts with childhood maltreatment to predict features of personality in a differential-susceptibility related manner. This polymorphism was selected for investigation because it is involved in the production of the neurotransmitter acetylcholine, a component of the cholinergic system that is strongly involved in neural plasticity and learning.

Polygenetic plasticity

Most G × E research has focused on one or another polymorphism. In recent years, however, work that we believe should be encouraged has emerged focusing on multiple polymorphisms, thus reflecting the operation of epistatic (i.e., G × G) interactions (e.g., Beaver, Sak, Vaske, & Nilsson, Reference Beaver, Sak, Vaske and Nilsson2010; Conner, Hellemann, Ritchie, & Noble, Reference Conner, Hellemann, Ritchie and Noble2010) as well as G × G × E polymorphisms. One can distinguish polygenetic G × E research in terms of the basis used for creating multigene composites. One strategy that seems fundamentally problematic to us involves identifying genes that show main effects and then compositing only these to then test an interaction with some environmental parameter (e.g., Docherty, Kovas, & Plomin, Reference Docherty, Kovas and Plomin2011). What leads us to question the wisdom of this approach is that genes or any other factors that do not exert main effects could still be involved in interactions with the environment, just as some recent G × E research reveals (e.g., Cicchetti & Rogosch, Reference Cicchetti and Rogosch2012; Cicchetti, Rogosch, & Oshri, Reference Cicchetti, Rogosch and Oshri2011; Pluess et al., Reference Pluess, Velders, Belsky, van IJzendoorn, Bakermans-Kranenburg and Jaddoe2011).

Another approach is to composite genes for a secondary, follow-up analysis that has been found in a first round of inquiry to generate significant G × E interactions (e.g., Brody, Chen, & Beach, Reference Brody, Chen and Beach2013; Sonuga-Barke et al., Reference Sonuga-Barke, Oades, Psychogiou, Chen, Franke and Buitelaar2009). When Cicchetti and Rogosch (Reference Cicchetti and Rogosch2012, see their figure 2) applied this approach using four different polymorphisms, they found that as the number of sensitivity to the environment alleles increased (i.e., short–short allele of 5-HTTLPR, zero copies of the corticotropin releasing hormone receptor 1 gene TAT haplotype, the TT genotype of DRD4-521C/T, and A carrier of OXTR), so did the degree to which maltreated and nonmaltreated low-income children differed on a composite measure of resilient functioning in a for better and for worse manner.

A third approach that has now been used successfully a number of times to chronicle differential susceptibility involves compositing a set of genes selected on an a priori basis before evaluating G × E (e.g., Brody et al., Reference Brody, Chen and Beach2013). Consider in this regard evidence indicating that two-gene composites moderate links between sexual abuse and adolescent depression/anxiety and somatic symptoms (Cicchetti, Rogosch, & Sturge-Apple, Reference Cicchetti, Rogosch and Sturge-Apple2007), between perceived racial discrimination and risk-related cognitions reflecting a fast versus slow life-history strategy (Gibbons et al., Reference Gibbons, Roberts, Gerrard, Li, Beach and Simons2012, see their figure 2), between contextual stress/support and aggression in young adulthood (Simons et al., Reference Simons, Lei, Beach, Brody, Philibert and Gibbons2011), and between social class and postpartum depression (Mitchell et al., Reference Mitchell, Notterman, Brooks-Gunn, Hobcraft, Garfinkel and Jaeger2011); that a three-gene composite moderates the relation between a hostile–demoralizing community and family environment and aggression in early adulthood (Simons et al., Reference Simons, Lei, Beach, Brody, Philibert and Gibbons2011); and that a five-gene composite moderates the relation between parenting and adolescent self-control (Belsky & Beaver, Reference Belsky and Beaver2011).

No matter how informative the approach to compositing polymorphisms on an a priori basis has proven to be when detecting differential susceptibility, we suspect that the wave of the future will be the “system-level genetic approach.” This involves combining genes considered to collectively influence a particular biological process or pathway, such as the dopaminergic or serotonergic system (e.g., Brody et al., Reference Brody, Chen and Beach2013; Conner et al., Reference Conner, Hellemann, Ritchie and Noble2010; Nikolova, Ferrell, Manuck, & Hariri, Reference Nikolova, Ferrell, Manuck and Hariri2011), or neurological endophenotypes for increased susceptibility, such as amygdala volume (Yap et al., Reference Yap, Whittle, Yucel, Sheeber, Pantelis and Simmons2008), hippocampus volume (Whittle et al., Reference Whittle, Yap, Sheeber, Dudgeon, Yuecel and Pantelis2011), and ventral striatum reactivity (Nikolova, Bogdan, Brigidi, & Hariri, Reference Nikolova, Bogdan, Brigidi and Hariri2012), to list just a few possibilities. Further insight could be gained, conceivably, by even more refined analysis of such subsystems, including those reflecting synthesis, degradation/transport, receptor, and modulation (Chen et al., Reference Chen, Moyzis, Stern, He, Li and Li2011). To date, these latter approaches have not been used in studies of G × E to our knowledge, whether evaluating or not evaluating the differential-susceptibility hypothesis. Of course, they could be.

G × E mechanisms

There are likely multiple explanations as to why the mechanisms in the G × E findings prove consistent with differential susceptibility and thus our claim that developmental plasticity should be regarded as a phenotype in its own right. When considering mechanisms, however, we must remember that this term means quite different things to different scholars. For example, mechanism may refer to brain structure or function to a brain scientist; for an endocrinologically oriented investigator it may refer to a hormonal phenomenon, for a geneticist to an epigenetic cascade, and for a cognitive neuroscientist an attentional process. Thus, what may qualify for a cognitive scientist as a “proximate mechanism” (e.g., attentional bias) may well be regarded by a geneticist focused on the epigenome as a “distal predictor.” Developmental and behavioral scientists should thus not be arrogant when stipulating what does and does not qualify as a mechanism. It ultimately depends on the level or levels of analysis at which one is working.

Based on recent empirical evidence, in the following we call attention to some possible differential-susceptibility related mechanisms at different levels of analysis. Instead of emphasizing the findings directly linking particular genes to particular psychological or biological processes as most genotype–phenotype and even G × E reports do, we highlight research showing that such processes are responsive to environmental exposures in a differential-susceptibility related manner.

In regard to cognitive processes, the following have been found to be moderated by 5-HTTLPR and an environmental factor in a for better and for worse manner: positive and negative attentional biases (Fox, Zougkou, Ridgewell, & Garner, Reference Fox, Zougkou, Ridgewell and Garner2011; see also Pergamin-Hight, Bakermans-Kranenburg, van IJzendoorn, & Bar-Haim, Reference Pergamin-Hight, Bakermans-Kranenburg, van IJzendoorn and Bar-Haim2012), reward sensitivity (Roiser, Rogers, Cook, & Sahakian, Reference Roiser, Rogers, Cook and Sahakian2006), and accurate processing of emotional faces (Jacobs et al., Reference Jacobs, Pine, Schoeny, Henry, Gollan and Moy2011). The same is true of inattention in DRD4 (Berry, Deater-Deckard, McCartney, Wang, & Petrill, Reference Berry, Deater-Deckard, McCartney, Wang and Petrill2013), working-memory accuracy and reaction time in BDNF (Gatt et al., Reference Gatt, Nemeroff, Dobson-Stone, Paul, Bryant and Schofield2009), and rumination in a two-gene composite (5-HTTLPR, BDNF; Clasen, Wells, Knopik, McGeary, & Beevers, Reference Clasen, Wells, Knopik, McGeary and Beevers2011) and “street code,” reflecting high value placed on being tough and aggressive, standing up for one's rights, and being ready to fight in the case of a three-gene composite (5-HTTLPR, DRD4, MAOA; Simons et al., Reference Simons, Lei, Stewart, Beach, Brody and Philibert2012).

In regard to physiological processes, 5-HTTLPR and an environmental factor have been found to interact in a differential-susceptibility related manner in predicting cortisol reactivity (Way & Taylor, Reference Way and Taylor2010), with the same being true of BDNF when the dependent variable is heart rate (Gatt et al., Reference Gatt, Nemeroff, Dobson-Stone, Paul, Bryant and Schofield2009) or evening salivary cortisol (Vinberg et al., Reference Vinberg, Trajkovska, Bennike, Knorr, Knudsen and Kessing2009). For brain measurements, Alexander and colleagues (Reference Alexander, Klucken, Koppe, Osinsky, Walter and Vaitl2012) present evidence that 5-HTTLPR interacts with stressful life events in a for better and for worse manner to predict left amygdala reactivity to fearful faces, whereas Gatt et al. (Reference Gatt, Nemeroff, Dobson-Stone, Paul, Bryant and Schofield2009) document much the same when using BDNF to forecast grey matter volume in the hippocampus and the amygdala. In addition, Klucken and associates (Reference Klucken, Wehrum, Schweckendiek, Merz, Hennig and Vaitl2013) report that BDNF interacts with positive and negative feedback in a for better and for worse manner during appetitive conditioning when predicting hemodynamic response in the amygdala. Evidence also suggests that individuals homozygous for the isoallele of a polymorphism in the mineralocorticoid receptor gene (rs5522) manifest the most and least right amygdala reactivity relative to valine carriers when they (retrospectively) reported experiencing high and low levels, respectively, of emotional neglect in childhood (Bogdan, Williamson, & Hariri, Reference Bogdan, Williamson and Hariri2012).

However, it may not be only adversity to which the amygdala of certain individuals proves especially sensitive. After all, an extensive meta-analysis showed that the amygdala responds to negative stimuli and even more strongly to positive stimuli (Sergerie, Chochol, & Armony, Reference Sergerie, Chochol and Armony2008). This suggests to us, as does the work cited in the preceding paragraph, that amygdala reactivity could be one of several central nervous mechanism by which differential susceptibility operates. This would certainly be consistent with the hypothesis proposed by the differential-susceptibility framework (Belsky & Pluess, Reference Belsky and Pluess2009a) and the concept of sensory-processing sensitivity (Aron & Aron, Reference Aron and Aron1997; Aron, Aron, & Jagiellowicz, Reference Aron, Aron and Jagiellowicz2012) that heightened susceptibility is the function of a more sensitive central nervous system on which experiences register more easily and deeply. According to this neurosensitivity hypothesis, specific gene variants (e.g., 5-HTTLPR short allele, DRD4 7R) contribute to the increased sensitivity and responsivity of specific brain regions that then manifests itself in increased negative emotionality and physiological reactivity (Pluess, Stevens, & Belsky, Reference Pluess, Stevens, Belsky, Legerstee, Haley and Bornstein2013), in part because highly sensitive individuals are more easily aroused.

What this brief summary of diverse findings related to putative plasticity genes should make clear is that a variety of plausible mechanisms at multiple levels of analysis may help to account for the G × E findings noted or referred to earlier. More such research is required, and the big challenge going forward will be to “put Humpty Dumpty together” (even if not “again”) so that a multilevel system of causation can be illuminated (Hyde, Bogdan, & Hariri, Reference Hyde, Bogdan and Hariri2011). This represents a huge agenda, perhaps the holy grail of inquiry.

Repeated measurements

All of the differential-susceptibility related work cited herein, as well as in our earlier reviews, essentially presumes that children and adults who share the same characteristics, be they temperamental, physiological or genetic plasticity factors, would function in a manner opposite to what was observed were they also observed under contrasting conditions. For example, we have presumed that children carrying DRD4 7R and showing high levels of externalizing problems in the face of insensitive parenting would score low on such problems if they were raised by skilled caregivers; however, there is almost no research validating this assumption, in part because it is unethical to provide a child growing up under for-better, supportive conditions with for-worse, adverse ones just to test the differential-susceptibility hypothesis. One of the few studies that systematically addressed the repeated-measurement issue under consideration, even if not in such differential-susceptibility related terms or with children, examined the probability of taking a risk in a gambling-decision-making game when the chances of winning were high (i.e., for-better condition) and when they were low (i.e., for-worse condition; Roiser et al., Reference Roiser, Rogers, Cook and Sahakian2006). The results revealed that the more short 5-HTTLPR alleles an individual carried, the more likely the person was to make a bet (i.e., take a risk) when the chance of winning was high, and the less likely the individual was to make such a bet when the probability of winning was low.

In addition to constructing repeated-measurement experiments, future researchers would be well advised to take advantage of experiments in nature that could serve this purpose. Consider in this regard the imaginative work of Verschoor and Markus (Reference Verschoor and Markus2011), who took advantage of days in which undergraduates did and did not have examinations in order to see how those students homozygous for long and short 5-HTTLPR alleles functioned. They found that the latter experienced the most tension, negative mood, and perceived stress on exam days, yet the least on nonexam days. Consider Schoebi, Way, Karney, and Bradbury's (Reference Schoebi, Way, Karney and Bradbury2012) marital research, which indicated that spouses homozygous for short alleles proved more sensitive to their partner's positive and negative emotion during marital interactions.

Experimental evaluation of plasticity

All of the research cited through this point has been observational, correlational, and often longitudinal in character. The common strategy employed in such work for discounting the possibility that organism–environment correlation is masquerading as Organism × Environment interaction is to insure that the temperamental, physiological, or genetic moderator under study is not correlated with the contextual predictor under consideration. However laudable such efforts are, they are not without limits. For example, just because the moderating polymorphism (e.g., DRD4) under investigation proves unrelated to the contextual predictor of interest (e.g., sensitive parenting) does not mean that this is true of all other unmeasured or unexamined organismic factors (e.g., BDNF). Ultimately, that environments have not been systematically manipulated in most plasticity-related research means that one cannot be certain, even in the case of the putatively malleable, that truly causal environmental effects are being detected (van IJzendoorn & Bakermans-Kranenburg, Reference van IJzendoorn and Bakermans-Kranenburg2012).

As pioneered by Dutch investigators in genetic plasticity factors (Bakermans-Kranenburg, van IJzendoorn, Mesman, Alink, & Juffer, Reference Bakermans-Kranenburg, van IJzendoorn, Mesman, Alink and Juffer2008; Bakermans-Kranenburg, van IJzendoorn, Pijlman, Mesman, & Juffer, Reference Bakermans-Kranenburg, van IJzendoorn, Pijlman, Mesman and Juffer2008), one way around this problem involves conducting intervention experiments in which individuals are randomly assigned to a treatment condition. Because it is unethical to assign individuals, especially children, to conditions of adversity, this approach has only been used to evaluate differential response, as a function of some hypothesized plasticity factor, to an experience presumed to promote positive functioning. Such a design is limited, because it can only evaluate the for-better side of developmental plasticity. Nevertheless, it represents an ideal way to determine whether those considered to be resilient to adversity on the basis of observational research also prove relatively unresponsive to supportive conditions and whether those considered vulnerable to adversity disproportionately benefit from environmental enrichment; there is ever-growing experimental evidence that this is the case.

With regard to the plasticity factor of negative emotionality, consider findings showing that it is infants who score relatively low on irritability as newborns who fail to benefit from an otherwise security-promoting intervention (Cassidy, Woodhouse, Sherman, Stupica, & Lejuez, Reference Cassidy, Woodhouse, Sherman, Stupica and Lejuez2011) and infants who show few, if any, mild perinatal adversities (known to be related to limited negative emotionality) who fail to benefit from computer-based instruction otherwise found to promote preschoolers’ phonemic awareness and early literacy (van der Kooy-Hofland, van der Kooy, Bus, van IJzendoorn, & Bonsel, Reference van der Kooy-Hofland, van der Kooy, Bus, van IJzendoorn and Bonsel2012). In other words, only the putatively “vulnerable.” those manifesting or likely to manifest high levels of negativity, experienced developmental enhancement as a function of the interventions cited. Similar results emerge among older children, as Scott and O'Connor's (Reference Scott and O'Connor2012) parenting intervention resulted in the most positive change in conduct among emotionally dysregulated children (i.e., loses temper, angry, touchy). Van de Wiel, van Goozen, Matthys, Snoek, and van Engeland (Reference van de Wiel, van Goozen, Matthys, Snoek and van Engeland2004) presented related results when the plasticity factor was physiological reactivity, finding that an intervention for children with disruptive behavior disorder proved effective but only for those displaying high cortisol related, stress reactivity.

Recent genetically informed intervention evaluations also indicate that alleles presumed to place individuals at risk in the face of adversity (5-HTTLPR short, DRD4 7R) or to promote resilience (i.e., not short/not 7R) are associated with them being respectively susceptible or not susceptible to the benefits of intervention (but, for counterevidence, see Cicchetti, Rogosch, & Toth, Reference Cicchetti, Rogosch and Toth2011). Consider in this regard Drury and associates’ (Reference Drury, Gleason, Theall, Smyke, Nelson and Fox2012) data showing that it was only children growing up in Romanian orphanages who carried 5-HTTLPR short alleles who benefited, in terms of reductions in the display of indiscriminant friendliness, from being randomly assigned to high-quality foster care. Eley and associates (Reference Eley, Hudson, Creswell, Tropeano, Lester and Cooper2012) also documented intervention benefits restricted to short allele carriers, but their design included only treated children (i.e., not a randomized clinical trial [RCT]). Turning to DRD4, Kegel, Bus, and van IJzendoorn (Reference Kegel, Bus and van IJzendoorn2011) tested and found support for the hypothesis that it would be DRD4 7R carriers who would benefit from specially designed computer games promoting phonemic awareness and, thereby, early literacy in their RCT. Other RCT results point in the same direction with regard to DRD4 7R, including research on African American teens in which substance use was the examined outcome (Beach, Brody, Lei, & Philibert, Reference Beach, Brody, Lei and Philibert2010; Brody et al., Reference Brody, Chen and Beach2013, in press).

Last to be considered is research examining genetically moderated intervention effects by considering multigene composites rather than singular candidate genes. The Drury et al. (Reference Drury, Gleason, Theall, Smyke, Nelson and Fox2012) findings show that even though BDNF did not operate as a plasticity factor when it came to distinguishing those who did and did not benefit from the foster-care intervention, the moderating effect of 5-HTTLPR was amplified if a child carried methionine alleles and short 5-HTTLPR alleles. In other words, the more plasticity alleles children carried, the more their indiscriminant friendliness declined over time when assigned to foster care and the more it increased if they remained institutionalized. Consider next Brody et al.'s (Reference Brody, Chen and Beach2013) confirmed prediction that the more GABAergic and dopaminergic genes African American teens carried, the more protected they were from increasing their alcohol use over time when enrolled in a whole-family prevention program. Such results once again call attention to the benefits of moving beyond single polymorphisms when operationalizing the plasticity phenotype. They also indicate that even if a single gene may not moderate an intervention (or other environmental) effect, in could still play a role in determining the degree to which an individual benefits. These are insights future investigators and interventionists should keep in mind when seeking to find “what works for whom?”

Domain specific versus domain general

Conceptualizing plasticity as a phenotype would seem to imply, particularly as we have discussed it, that some individual differences will be more and some less developmentally malleable. However, should this be broadly interpreted to mean that some individuals will be more and some less malleable to any and all environmental inputs and with respect to any and all developmental sequelae? Is it more likely that a narrow conceptualization holds, such that plasticity will prove domain specific rather than domain general? Thus, instead of some being more plastic in general, individuals will vary in terms of what they are susceptible to (e.g., parents, not peers; peers, not parents) and with regard to particular developmental outcomes but not others (e.g., social but not academic competence; Belsky et al., Reference Belsky, Bakermans-Kranenburg and van IJzendoorn2007; Belsky & Pluess, Reference Belsky and Pluess2009a, Reference Belsky and Pluess2009b)?

However more likely the latter prospect would seem, two new sets of evidence underscore the domain generality of plasticity. The first involves the nature and effects of two dramatically different Dutch interventions that used different indirect and direct approaches to influence children and that focused on different developmental outcomes yet revealed that children sharing the same plasticity factor were the primary beneficiaries of each. More specifically, even though one intervention promoted sensitive parenting via video feedback in order to reduce toddler's externalizing behavior (Bakermans-Kranenburg, van IJzendoorn, Pijlman, et al., Reference Bakermans-Kranenburg, van IJzendoorn, Pijlman, Mesman and Juffer2008) and cortisol-related stress reactivity (Bakermans-Kranenburg, van IJzendoorn, Mesman, et al., Reference Bakermans-Kranenburg, van IJzendoorn, Mesman, Alink and Juffer2008) and the other relied on a computerized instructional program to promote preschooler's phonemic awareness and thereby early literacy (Kegel et al., Reference Kegel, Bus and van IJzendoorn2011), it was children carrying the DRD4 7R allele who disproportionately or exclusively benefited in both cases. There is also the longitudinal data evaluated by Roisman and associates (Reference Roisman, Newman, Fraley, Haltigan, Groh and Haydon2012) showing that infants with difficult temperaments measured in the first 6 months of life proved more susceptible to the respective positive and negative effects of sensitive and insensitive mothering experienced during their first 3 years of life when it came to predicting teacher-reported social competence, behavior problems, and academic skills across the primary school years; this was so even with different teachers rating the children each and every year. Moreover, even though a diathesis–stress rather than differential-susceptibility model fit the data best when it came to predicting performance on objective tests of academic performance through 15 years of age, it once again proved to be children who as infants had difficult temperaments who were disproportionately susceptible to the shorter and longer term effects of early mothering. Of interest may be recent work showing that even though DRD4 and negative emotionality early in life are related (Holmboe, Nemoda, Fearon, Sasvari-Szekely, & Johnson, Reference Holmboe, Nemoda, Fearon, Sasvari-Szekely and Johnson2010; Ivorra et al., Reference Ivorra, Sanjuan, Jover, Carot, Frutos and Molto2010), the moderating effect of DRD4 does not account for a parallel moderating effect of negative emotionality, at least in a study of effects of child care on children's social functioning (Belsky & Pluess, Reference Pluess, Stevens, Belsky, Legerstee, Haley and Bornstein2013).

Despite such evidence, it still seems premature to embrace the broad domain-general interpretation of developmental plasticity. Nevertheless, the results just summarized caution against its premature dismissal. Ultimately, further work is called for before any firm conclusions can be drawn. Our suspicion is that there will be variation across individuals in that some will prove highly susceptible to many developmental experiences and environmental exposures (i.e., extremely plastic) and that some will prove susceptible to virtually none (i.e., extremely fixed) but that most will prove to be intermediate between these extremes. One could even envision a bell curve with regard to the degree of plasticity that individuals manifest.

Environmental cue reliability

Whether children develop in a generally supportive or unsupportive environment, it is likely that the reliability of environmental cues indicative of the valence of the rearing milieu will vary. One can imagine in this regard two different hostile environments: one in which conflict is constant and thus highly predictable and another in which conflict is episodic and unpredictable. Frankenhuis and Panchanathan (Reference Frankenhuis and Panchanathan2011) have insightfully observed that highly plastic individuals growing up in these two environments may differ in terms of when they “commit” to a developmental strategy, with the former “deciding” to specialize early, having high confidence about the state of the environment, but the latter deferring commitment and choosing to keep sampling the environment until a better estimate of the environment can be established. Whether this is actually the case is an issue that should be pursued empirically in the future. Frankenhuis and Panchanathan's (Reference Frankenhuis and Panchanathan2011) intriguing ideas lead us to wonder whether it is when cue reliability is low that individuals “call upon” the epigenome to infer from markings passed down across multiple generations what past environments were like and thus what is the best developmental pathway to pursue. Such a process would be consistent with Kuzawa's (Reference Kuzawa2005) notion of “intergenerational phenotypic inertia,” which highlights that it is the experiences of not only the immediately preceding generation to which developing organisms are sensitive but also generations preceding that.

Parent–child conflict of interest

Like many developmental scientists lacking foundational knowledge about evolutionary theory, virtually all of our thinking and writing about differential susceptibility to date has presumed that malleable children will have their development regulated by their rearing experiences. However, as Trivers (Reference Trivers1974) pointed out more than four decades ago, that parents and children share an average of 50% of their genes means that their biological interests are not the same and thus that they often experience a conflict of interest (see also Schlomer, Del Giudice, & Ellis, Reference Schlomer, Del Giudice and Ellis2011). In discussing the influence of prenatal stress on development, Del Giudice (Reference Del Giudice2012) recently pointed out that it is because of this that children, even highly malleable ones, may not simply take instructions from their parents. What are the implications of this observation for understanding developmental plasticity?

Family dynamics

However intrigued we have become with theory and evidence pertaining to differential susceptibility and the notion that individual differences exist in developmental plasticity, one issue has always caused us concern. We have never heard a parent characterize one of their children as more susceptible to their influence than another, which we regard as distinct from saying that their children are temperamentally different, with perhaps one being more “sensitive” to challenging or stressful situations than others or even being more “difficult” to care for. Why is that? Evolutionary reasoning would seem to imply that parents should detect for whom their socialization efforts, implicit or explicit, pay off and for whom they do not and so adjust those efforts accordingly. However, we are aware of no evidence suggesting or indicating this to be the case. Should not more malleable siblings seek to alert their parents that they are more responsive to rearing inputs in order to obtain more attention and other resources? However, could it be that such children do engage in such efforts but that they do not prove effective because less malleable children employ tactics to camouflage this characteristic of theirs in some way? Perhaps it is such children who prove most likely to claim “it's not fair” within families when siblings are perceived as being treated better as a way of “throwing their parents off track.” At the very least, it would be interesting to study parental awareness or perception of variation in susceptibility across siblings within a family.

Timing of susceptibility

It is widely assumed by many developmentalists that it is during the early years of life that human development is most susceptible to environmental influences, for better and for worse (e.g., Blair & Raver, Reference Blair and Raver2012; Ganzel & Morris, Reference Ganzel and Morris2011), as plasticity is presumed to be greatest when biological systems are being laid down. Even if this is true in general, does it mean that it is true in the case of each and every individual? It certainly seems conceivable that, at times in the ancestral past, individuals who were especially susceptible to environmental influence at some point in their life span could have experienced an adaptive advantage, leading perhaps to the selection of genes for later rather than earlier plasticity. If this is so, it implies that variation should exist in terms of when children and even adults are especially developmentally plastic. If this is the case, does it not imply that those who appear not to be especially malleable may simply not be so at the developmental point in time when developmental experiences and environmental exposures are measured? Could it be that some are especially susceptible to contextual regulation early, others later, others at all times, and still others more or less never?