IN THE DECADES SINCE SOCIOBIOLOGISTS first argued that human nature stands in the way of some human aspirations, their critics have repeatedly emphasized an important and obvious point: that we cannot really know the limits of human possibility without knowing the relevant response functions (or what biologists sometimes call “norms of reaction”) in their entirety—that is, without knowing how every possible environment will affect the expression of each trait of interest—and we do not know this now and never shall. But though the point is correct if taken literally, we can nonetheless have very good grounds for concluding that certain phenotypes really are unachievable for organisms with the genetic makeup of today’s humans, and that other phenotypes require environments that are difficult to create or maintain or that have unacceptable effects on other traits. To take a mundane example, we have reasonable grounds for believing that a world in which everyone lives to the age of five hundred is impossible to achieve without genetic modification or quite radical environmental manipulation, and that a world in which almost all people live to the age of one hundred might be achievable but would require an extraordinary investment of resources and involve obvious trade-offs, including some that would result in losses of freedom and happiness. We can make these claims with confidence because of what we know about the human response functions relevant to longevity and their relationships to some others. We know that maximum longevity is fairly consistent across a wide range of favorable environments; that it diminishes in environments with certain kinds of nutritional deficits or excesses, infectious agents, and environmental toxins; and that it increases in environments of certain kinds but only within quite predictable bounds. (For most of these patterns we know a good deal about the underlying mechanisms that make them so reliable.) We also know that there are trade-offs between longevity and certain other phenotypic traits and behaviors—freedom to engage in hazardous sports, to take the performance-enhancing steroids that many athletes use, or to consume alcohol or unhealthy foods, for example, would be curtailed in a society that assigned the highest priority to maximizing longevity. Such a society would thus have lower peaks than ours does on some measures of athletic performance and perhaps of creative output, and its parties would be duller. (New prospects for developing medical techniques for intervening directly in the mechanisms that control aging do open up other possibilities here. We’ll recur to this idea later on, but for now note that even these possibilities seem likely to involve considerable investment and significant trade-offs, should we attempt to realize them at a global scale.)
These assumptions about the human response functions relevant to longevity seem unexceptionable. Mainstream evolutionary psychologists think that they can make similarly well-supported assumptions about the human response functions relevant to sex-differentiated cognitive capacities and behaviors, and about many other human cognitive and behavioral features as well. Understanding the assumptions these thinkers make is thus crucial to assessing the claim that we are approaching at least some of the limits of reasonable human aspiration.
Internal and External Causes
To ask what is possible for humans, and at what cost, is in part to ask about the shape of human response functions: that is, about how internal (genetic) and external (environmental) causes would combine to produce the properties of the organism, including its behavior, across the range of possible environments in which it could develop. The response function for each genotype maps its possible environments onto the phenotypes that would result. Our suppositions about the shapes of human response functions (i.e., the range of human possibilities) and about the interrelations among the response functions of different human traits (i.e., the trade-offs among different desirable human possibilities) thus depend on how we think about genetic and environmental causes and how they combine.
The problem of how to think about internal and external factors in explaining how things change or remain stable is both ancient and general. The philosopher Peter Godfrey-Smith sketches three basic ways of explaining change or stability in an object’s properties:
Internalist explanations appeal to causes originating inside the object.
Externalist explanations appeal to causes originating outside the object.
Interactionist explanations appeal to a combination of internal and external causes.
What counts as internal (or as external) obviously depends on the object under consideration, so a theorist who takes an internalist view in one context may (indeed, must) take an externalist view in others. In evolutionary explanations, the idea that something internal to biological lineages fully determines their evolutionary path would be pure internalism. Lamarck’s pre-Darwinian theory that organisms have a sort of inherent drive toward greater perfection that expresses itself in evolutionary change would be close. The view that external forces such as natural selection do all the work, unimpeded by internal constraints, would be pure externalism. Most evolutionists today are closer to the externalist end of the spectrum, though internal factors have been getting more attention lately in the work of evolutionary developmental biologists and (especially) “process structuralists,” who argue that the structure of an organism itself creates the conditions for particular kinds of evolutionary change (Buss 1987; Goodwin and Saunders 1992; Webster and Goodwin 1996; A. Wagner 2014; G. P. Wagner 2014).
In explanations of human development and behavior, the situation seems less settled. Absolute genetic determinism would be a purely internalist position and absolute environmental determinism a purely externalist one. In fact, of course, everybody agrees that both genes and environment influence phenotypic and behavioral outcomes, but theorists nonetheless disagree substantially about the roles played by these components. If we are all interactionists, then interactionism seems to encompass a spectrum of positions ranging across a good deal of the broad span between the poles of pure internalism and pure externalism.
In fact, even the image of a spectrum here is a serious oversimplification. It would be very difficult to defend as uniquely correct any particular way of arranging the various possible combinations of genetic and environmental influence that lie between the poles in a simple one-dimensional order. It is possible, though, to discern some broad categories among them that should be considered more closely.
The simplest way for two factors to combine is additively, so that the same environmental difference has the same effect on different genotypes and in combination with various other environmental differences, and the same genetic difference plays out in the same way in different environments and in combination with various other genetic differences. The magnitude of each effect is therefore proportional to the magnitude of the cause.1 (For example, a large phenotypic difference between two organisms indicates a large genetic or environmental difference, or a combination of lesser differences in both.2) To take a very simple example, consider a plant species that has two genotypes, G1 and G2, and lives in two different environments, E1 and E2. The plants vary in height. For each genotype, plants will always grow taller in E1 than in E2. In each environment, a G1 plant will always grow taller than a G2 plant. A G1 plant in E1 will be the tallest of all, and a G2 plant in E2 the shortest.
Additive combinations of causal factors are easy to understand and model, but genetic and environmental factors do not in general combine additively. In our simple example, for instance, it is perfectly likely that the each genotype would thrive best in a different environment, so that G1 produced taller plants in E1, say, and G2 in E2. Addition does provide an important benchmark for understanding more complex combinations, yet it is difficult even to conceptualize the vast array of possible nonadditive combinations that result when we consider genotypes, environments, and phenotypes with far more than two possible states each. To understand these possibilities, we need some constraints to tame their multiplicity. Views about the interaction of biological causes are often expressed in terms of simple conceptual models or more concrete metaphors that focus attention on important subsets among the possible combinations. The well-known metaphors of “malleability” or the “blank slate” are used to represent a kind of interaction close to the externalist extreme in which the genotype offers little resistance to the influence of the environment (Pinker 2002). The metaphors of “programming” or “hardwiring” represent one close to the internalist extreme in which the genotype fixes an internal “nature” that is resistant to modification by outside forces.
Some thinkers have argued that we need to abandon all such metaphors in order to think clearly about development and evolution (West-Eberhard 2003). I am not sure that this is feasible, and it is clear in any case that these metaphors continue to shape the thinking of both evolutionary psychologists and their critics, affecting the kinds of empirical evidence they seek and how they interpret it. Indeed, a stronger claim might be made here. Cognitive psychologists have shown that our thinking about abstract relations among ideas, and about complex causal relations among objects, are often organized by simple metaphors based on our experience of living in the kind of bodies we have, in the kind of world we live in (Lakoff and Johnson, 1980). These metaphors and the larger conceptual “frames” that they bring along with them are essential but tricky cognitive tools. They help us to reason about abstract, complex, and unfamiliar things, but they also carry systematic biases. In particular, they can invoke emotional responses, importing a sense of value into our thinking in a way that may be very difficult to for us to override. Certainly, as the geneticist Richard Lewontin noted long ago, “the price of metaphor is eternal vigilance” (Lewontin 2002, 4).3 We need, at the least, to scrutinize these metaphors carefully, to consider their limitations and the unintended implications they may carry, to abandon any that are systematically misleading, and perhaps to find new ones that help reveal neglected features of the causal processes of behavior, development, and evolution.
FIGURE 3.1 Response functions or norms of reaction are given partial representation by simple diagrams that show for each genotype a curve corresponding to the phenotypes it produces across the range of its environments. These diagrams are very selective: they plot phenotypic outcomes for one or more genotypes on a single phenotypic measure (e.g., height, number of leaves, age at first mating, mathematical skill) against a single environmental variable (e.g., average temperature, nitrogen levels, mother’s educational level, prenatal testosterone exposure). They normally show only the actual range of environments, not the full range of possible-but-unrealized ones.
A. Genetic determinism or internal causation—Each genotype (G1 and G2) yields the same phenotypic outcome irrespective of the environment.
B. Environmental determinism or external causation—Phenotypes vary with environment, but in any given environment both genotypes produce identical phenotypic outcomes. (Note that the curve here could be any shape—what matters is that the G1 and G2 curves are the same. The sort of function shown—linear, with a nonzero slope—corresponds to the sort of relation between environmental variation and phenotype that people most often have in mind when they talk about environmental determinism, however.)
C. Addition—Phenotypes vary with both genotype and environment. Response functions for both genotypes are linear, and their slopes are the same: the same environmental change has the same effect on phenotype irrespective of the genotype or current environment.
Begin with the basic conceptual models associated with external and internal causes. Externalist explanations employ two basic models of change, one for change in individual objects and one for change in populations of objects. The first is direct action or direct impression. Here some element in the environment directly modifies some property of the object—as impact with another ball changes the velocity of a billiard ball or as pressure from a signet changes the shape of a wax seal. In biological development, the instances that best fit the metaphor of direct impression are cases like the shaping of a tree by constant wind pressure or deliberate training by a gardener. For humans, instances might include injury, body modifications such as piercing or foot binding, or “instructional learning” such as classical conditioning. Importantly, the metaphor of direct impression suggests that any modification is equally easy to achieve, just as the wax seal can take any impression equally readily. The second metaphor for external influence is that of sorting or sifting. Here some structure in the environment systematically removes (or preserves) members of a collection that have certain properties, thus changing the composition of the population. All selection processes fit this basic model, including evolutionary selection but also importantly including selective processes in individual development, such as neuronal selection in brain development, or trial-and-error learning. Here again the metaphor suggests that any selection is equally easy to achieve.
Internalist explanations employ a core conceptual model that takes a variety of metaphorical forms: that of the step-by-step unfolding of some preexisting structure or following-out of a preexisting plan, as in the literal unfolding of a newly emerged butterfly’s wing or the execution of a “program” or function for which the organism is “hardwired.” Here only one outcome is possible—the full expression of what is already there, laid down and ready to go—unless some outside influence diverts or obstructs the process of unfolding or execution. As the language of “diversion” and “obstruction” indicates, the unique preferred outcome can easily seem to take on a normative status as the correct fulfillment of the object’s potential.
These basic conceptual models are adjusted or modified to produce some of the most common ways of thinking about how internal and external factors interact. Wright’s claim that “human nature is malleable, but not infinitely malleable” suggests that the internal causes ensure that human development tends toward a particular outcome—an outcome that can be modified by external forces but only up to a point. Variants are possible, but not all are equally easy to achieve, and some are not achievable at all; human nature appears not as wax or clay but perhaps as a rather elastic putty with a resistant core. A related view sees development as internalist unfolding that is subject to significant constraints or interference from the environment. The butterfly’s unfolding wings may be bent or torn, the program may suffer from some minor glitches in execution, but the basic pattern remains. A version made popular by Richard Dawkins is the metaphor of the recipe: the genes provide the recipe, the environment provides ingredients and equipment; and the phenotype emerges from their interaction like a finished cake.
These various metaphors of interaction are importantly different from one another, but they share a strong normative flavor: the role of the external factors is either to enable the internal causes to unfold as they ought or to prevent them from doing so. A recipe carries information about how the cake is supposed to come out and about what the ingredients, pan, and oven should be like. If the cake falls, failing to come out as it should, the oven was the wrong temperature. The metaphor of genome-as-recipe implies that the genome similarly picks out a preferred phenotype and thus also picks out the preferred range of environmental specifications that will produce that phenotype. The conceptions of interaction expressed by these intermediate metaphors thus inherit from internalism a normative attitude to both phenotype and environment. From externalism they inherit a tacit assumption that the same external causes will have similar effects on outcomes even when the internal causes are different—too hot an oven will scorch any cake (this assumption is further encouraged by the internalist idea that the role of external causes is merely to enable or obstruct the action of internal ones). Internally caused differences will therefore tend to carry over from one environment to another—if one recipe produces a sweeter cake than another, that difference will tend to remain despite variations in cooking conditions and ingredient quality, for those variations affect both cakes more or less alike.
The common core of these widely used metaphors of interaction can be captured in a general description of what I will call conservative interactionism. Conservative interactionism explains an object’s features or behavior by referring to causes originating both inside and outside the object. It assumes that internal causes determine a “preferred” outcome and external causes modify it, but not additively: the internal causes tend to keep the outcome close to the preferred state, and large external forces are required to produce significant divergence from it. Extremely divergent outcomes cannot be reached no matter how powerful the external causes. The same external causes modify outcomes similarly for different internal causes.
Here the interaction behaves more like pure internal causation than like an additive combination—this is the sense in which the interaction is conservative. A picture like this is clearly consonant with the kind of thinking about the limits and costs of change that evolutionary psychologists have made popular, with their treatment of certain phenotypes as “natural” or “typical,” and of divergence from these as deviation requiring special explanation. But Lewontin (again, long ago) has pointed out that a different kind of interaction is possible and may be common (Lewontin 1974). I call the explanatory approach that he articulated radical interactionism. Radical interactionism explains an object’s features or behavior by referring to causes originating both inside and outside the object. It assumes that nonadditive interaction between internal and external causes may be both strong and multistranded so that external causes could have radically different effects (and effects of dramatically different magnitudes) depending on the internal causes with which they were interacting.
Such a view is not ruled out by anything we have seen so far. But we also have not seen any strong reason to adopt it, and the abstract description just offered is incomplete. Why should we expect organisms to work like this? What kind of patterns might we expect to find among such interactions?
Some important instances of radical interaction can be understood through a conceptual model that has been available to us all along in the form of another metaphor: response. Where the other metaphors cast organisms as passively impressionable lumps of wax or rigid mechanisms, the metaphor of response sees organisms as agents capable of sensing the state of particular external variables and actively modifying themselves in accordance with what they sense. The active self-shaping of a houseplant in a window “reaching” for the light is a simple example of this sort of interaction. Many kinds of learning seem to have this structure, in which the organism is especially sensitive to (and may actively seek) stimuli of a certain sort and uses them to fix certain aspects of its own behavior or cognitive state. Mammals easily learn to associate particular scents or sounds with feeding and to associate particular flavors with nausea and use these associations to guide their behavior. Young children are acutely sensitive to the vocalizations of the older humans around them and learn to imitate the words they hear. Where these sorts of interactions are in play, multiple outcomes—possibly quite different from each other—are possible; which one is realized may depend on subtle environmental cues. Plants in different circumstances grow differently to maximize their access to sunlight; children in different communities learn different languages. There seems little room for normativity here; it is difficult to see on what basis one outcome (and its triggering environment) could be picked out as the “right” one unless on the basis of a measure of fitness, and different outcomes may be equally fit. Diversity of outcomes thus appears as the result of diverse but (often) equally good responses to diverse environments.
FIGURE 3.2 Three pairs of response functions that fit the pattern of conservative interaction. The shaded areas indicate environmental states that are rarely or never observed. The phenotypes for both genotypes produce curves that are fairly flat through the range of common (“normal”) environments, approximating the internal causation of figure 3.1A.
A. Both genotypes give maximum phenotype values toward the middle of the common environmental range and fall off together in more extreme environments. Changing the environment cannot produce phenotypes that are higher or more equal than they are in currently common environments.
B. The genotypes give quite different phenotypic outcomes across the range of common environments but converge in more extreme environments. Phenotypes far above and below the “normal” range are possible, but only in environments that are not often encountered. Dramatic change to phenotypes and greater equality between the outcomes for the two genotypes can both be achieved, but only by imposing large environmental changes to extreme values.
C. Phenotypes for the two genotypes are close together through the range of common environments but diverge sharply in extreme environments, crossing at one end. Large changes in phenotype can be achieved with large environmental modification, but only for one genotype, so maximizing the peak outcome (or minimizing the lowest) means increasing inequality between the genotypes.
FIGURE 3.3 Three pairs of response functions that fit the pattern of radical interaction. The examples suggest the range of very different forms that radical interaction can take. What they have in common is that the effect of a given environmental change depends on both the genotype and the existing environment.
A. Two genotypes respond consistently, but reciprocally, to change in the same environmental variable.
Example: One ivy genotype grows larger leaves in shadier environments, the other in sunnier environments.
B. Two genotypes switch between two robust pathways in response to changes in the same variable but at different thresholds.
Example: Two alligator genotypes switch between male and female phenotypes at different temperatures.
C. Two genotypes have different nonlinear responses to the same environmental variable.
Example: One salmon genotype has a higher survival rate of fry at moderately high or low water temperatures; the other maximizes its fry survival rate at medium temperatures, but both have lower survival rates in extremely warm or cold water.
Evolutionary psychologists often use a different metaphor for talking about interactions of this sort, describing them as the playing out of programs that have a complex branching structure whose various conditional routines have environmental triggers. This way of thinking interprets response as a sort of complex unfolding; it lacks the imputation of agency that the metaphor of response carries but does greatly weaken the normative flavor of the metaphors of conservative interaction.
Some theorists, such as Susan Oyama and Mary Jane West-Eberhard, travel further down the path Lewontin indicated, going beyond radical interactionism to reject the basic metaphor of interaction entirely (Oyama 1985, 2000; Oyama, Griffiths and Gray 2001; West-Eberhard 2003). They argue that this metaphor mistakenly pictures the process of development as an ongoing confrontation between two interactants (the genotype and the environment) that retain their identities unchanged throughout. These theorists advocate, instead, a view of development as true ontogeny or coming into being: a sequence of responses, indeed, but one in which what responds at each stage—a functioning genome or a living organism with particular properties—is itself a product of the last stage’s response.
The question is not which of these conceptual models captures “the truth” about the roles that genes and environments play in development and behavior. If patterns of change that fit the metaphors of impression, sorting, unfolding, response, or ontogeny occur among the processes giving rise to full-grown organisms and fully realized behaviors—and I think they all do—they all work by means of simple physical causation at the micro level. Moreover, phenomena well described in terms of one of these metaphors may be the result of component processes well described in terms of another. Thus, response and ontogeny, where they occur, are often realized by means of lower-level impressions and sortings that themselves are ultimately realized by means of simple physical causes at the molecular level. (For example, the impressive capacity of a mammal’s immune system to mount an active response to a particular microbial invader works by means of a process of selection [sorting] acting on the cells of the immune system, but this in turn is ultimately achieved by means of simple impressions at the chemical level—the change in a receptor caused by its target molecule binding to it, and that is simply caused by the electromagnetic forces linking the relevant molecules.) What matters for our purposes is which—if any—of these metaphors is successful in capturing significant aspects of the high-level patterns found in organismal development and behavior, and especially in human cognitive development and social behavior.
If human development and behavior mainly follow the pattern of conservative interaction, it seems plausible to say that although reformers might strive for substantial social changes, these are likely to come only at an unacceptable cost. In a world dominated by conservative interaction, large-scale changes in human behavior would seem to demand large-scale environmental interventions that would be difficult to generate and could be expected to have extensive additional effects, “forcing” human developmental and behavioral patterns beyond their normal (or even their normative) bounds.
If the pattern of radical interaction is common, however, the prospects for social change look quite different and far more hopeful. The metaphor of response provides a conceptual starting point from which to begin examining some theoretical and empirical results that support and fill in this alternative picture.
