Darwin on the Mind
Evolutionary psychology is in fashion-but is any of it true?
Jeremy C. Ahouse and Robert C. Berwick
How the Mind Works
W. W. Norton, $29.95
"Of course the theory would be vacuous if it offered a glib explanation for every inexplicable act." --Steven Pinker
Steven Pinker picked an ambitious title, How the Mind Works (HTMW), and set off in the right direction, insisting that all human behavior should not be explained by a single wonder principle. Unfortunately, Pinker's opposition to oversimplification fades quickly, at two wonder principles: "the mind is a system of organs of computation" . . . "designed by natural selection." What is more, we arrive at the end of his long book without ever having heard how the mind works.
Pinker's approach to human behavior weds a particular account of cognition-a "modular computational theory"-to a particular account of how the mind emerged-"evolutionary psychology." According to the modular theory, the mind comprises a set of discrete faculties or "mental organs" that perform separate tasks and operate on distinct principles: bite-sized recipes for figuring out, say, whether that wispy patch lurking above is a cloud or weasel, that draw on distinctive kinds of information, distinctively processed. Cognition emerges as the joint product of these separate modules, each operating in its own domain, not as the reflex of Reason or general problem-solving intelligence. According to the evolutionary psychologist, the discrete problem-solving devices that constitute the human mind evolved because they solved problems faced by small nomadic bands of hunter-gatherers in the environment in which humans evolved-problems of mate selection and navigating one's way through the veldt. In short, your behavior reflects a mind that formed by adapting to the environment, 100,000 years ago on the African savanna.
As evolutionary past becomes present prologue, natural selection gives us a variety of genetically-based subroutines for solving specific problems-what to eat, whom to spend time with (and what to do with that time), and how to get around at night. We retain these evolutionary residues, sometimes in inappropriate modern contexts. When you find yourself reaching for the potato chips, it is because high-caloric fatty foods were once scarce but valuable in the Pleistocene, so our hominid ancestors who developed a genetically-grounded taste for fats did better, on average, at getting their potato-chip genes into future generations.
The unhappy results of these efforts to marry cognitive science to an adaptationist brand of evolutionary biology are a credulous conception about how the mind works (misrepresented as scientific consensus), an uncritical genetic determinism, and a borrowed evolutionary biology used not to generate hypotheses, but to rationalize Pinker's own opinions.
Pinker's analysis ranges far and wide, covering perception, reasoning, emotion, social relations, and the higher activities of art, religion, and philosophy that Hegel assigned to the realm of Absolute Spirit. Pinker makes the spirit flesh, or more precisely, gene. Our yen for green suburban lawns is an evolutionary memory of the green hills of Africa; we love abstract art (or do not) because shapes that please us are good art; mathematical thinking is all bootstrapped on primate visualization; parent/child rivalry and wicked stepparents result from promoting one's genes at the expense of others, be they siblings, parents, or step-children; older men prefer to marry younger women, and women prefer wealthier men because youth and wealth are visible proxies for fitness.
If you are thinking that you have read such near-tabloid headlines before, you are right. They descend directly from early 1970s "pop" sociobiology, with but a few differences. In the late 1970s, Richard Alexander (Darwinism and Human Affairs) and Donald Symons (Human Evolution and Sexuality), among others, turned sociobiology from focusing on the adaptive value of individual behaviors-such as "men marry younger women"-to a handful of modular mechanisms such as "mate selection" that result in those behaviors. For men, this means a module for identifying fertile and healthy females, and for women, a module to snag males who will be the most stable long-term providers of resources. This shift in the unit of selection from behavior to mechanism paralleled a sea-change in cognitive science; from a conception of the mind as a general-purpose computer to the Cartesian picture of the mind as a bundle of mental "organs," with separate units for language, theory-formation, visual perception, morality, and so forth. Pinker attempts to reanimate the pop sociobiology arguments by fusing Darwin with claims that the mind is modular.
In this project, he follows a new breed of sociobiologists, led by Leda Cosmides and John Tooby, who aim to explain human culture via evolutionary history, and undermine something they call the Standard Social Science Model, which (they say) ignores human evolutionary history and personal psychology, uses the category "culture" to explain sociological observations, and so on-eleven sins in all.1 In pursuit of this ambition, they argue that the mind is a set of problem-solving modules, each of which can be independently optimized; that there is less human variance and cultural diversity and less learning from our cultures than you might think; and, correspondingly, that more of human life reflects the intrinsic features of our mental modules than we might wish to suppose. Following Alexander and Symons, they focus on the adaptive value of the mechanisms that result in behaviors, rather than the behaviors themselves. Moreover, they hope to avoid political objections by concentrating on common traits, not individual, sexual, and racial differences.
All the evolutionary storytelling that has emerged from this line of work is interesting; reductive, counter-intuitive explanation always is. But is any of it true? Telling "just so" stories is too easy. We can always rationalize a particular behavior or trait by inventing a past that must have selected for it, and plausibility just does not narrow the field. Indeed, Philip Kitcher's criticism of the earlier generation of work applies with full force to the latest round:
"Again and again . . . assertions about human nature begin with unrigorous analyses of fitness, . . . deal loosely with data about human and animal behavior, . . . employ problematic concepts, . . . rely on dubious connections between optimality and selection, and . . . offer spurious arguments for the inflexibility of the phenotype."2
What is worse, "good old" sociobiology at least had the virtue of testability. Its claims were about optimal behaviors in the here-and-now. By recognizing that it takes time for optimal "solutions" to stabilize and that our present culture is in flux, the new sociobiology pushes the adjudicating evidence far back into the past, sometimes 1-2 million years to the Pleistocene and Homo erectus, sometimes 100,000 years to early Homo sapiens. It is not clear how we will ever test these claims of optimal behavior. It is hard enough to tell today whether one warbler's preference for nest building has a selective advantage over another's-so hard that the textbooks usually cite only a handful of cases, like Kettlewell's black moths that could hide better against British smog-darkened tree trunks (even this story has become more difficult to defend).3 As the evolutionary biologist Elizabeth Vrba reminds us, critical evidence about past environmental and selectional regimes simply do not leave strong traces. Worse, how does one track those selection pressures that modify behavior and leave nothing to fossilize behind?
Amidst all these unknowns, HTMW offers none of the cold skeptical eye that readers deserve. These still-contentious fields-and the still-more tendentious details sketched above-do not provide the confident prediction and understanding that we find in modern physics, chemistry, or even molecular biology. Indeed, HTMW goes the "just so" stories one better, transforming them into what might be dubbed "just say" stories. Pinker "just says" that suburban lawns serve as evocative evolutionary memories, like Proust's madeleines reminding us of our distant past. HTMW "just says" that 99 percent of our evolution took place on the savanna. It should be unnecessary to point out that savannas with five- to fifteen-foot-tall grasses and one-foot-short grasses punctuated by acacia trees do not quite look like suburban lawns. One might just as well say that our love for hot-tubs follows from what is really true, that over 99 percent of our evolutionary history was spent in (and most of our genes arose in) a warm, salty, sea.
And so it goes, breathless tale after tale. The "best theory" about sex, we are told, is that it arose "as a defense against parasites and pathogens." A recent report by Evelyn J. Lyons should temper such conviction.4 Scientists disagree about the origins of sex, and the parasite/pathogen theory is just one among several. In the green algae Chlamydomonas reinhardtii, "sex can serve both for mutation clearance and mutation assembly although no parasites are involved." We simply do not know enough to reach a solid conclusion and there is no strong consensus to report.
Pinker also tells us that "much of the variation in personality-about fifty percent-has genetic causes"; and "the biggest influence that parents have on their children is at the moment of conception." Really? Technical measures of the "heritability" of personality-the tendency for offspring to resemble their parents-show correlations, not "genetic causes."5 Contrast Marcus Feldman's entry in Keywords in Evolutionary Biology: "heritability is statistical in nature and does not involve a detailed specification of genetic or environmental transmission."6 Feldman goes on to note that statements in the popular press, such as Deborah Franklin's 1989 claim in the New York Times that "genes are 50 percent to 70 percent responsible for an individual's IQ," not only over-state the numbers, but also, just like HTMW, "incorrectly infer causation from correlation." 7 Our goal here is not to resolve the nature/nurture controversy. It has not been resolved, and it is bizarre for a scientist and irresponsible for a reporter to suggest otherwise. Pinker's assertion is simply the authority of modern science pressed into the service of speculative fictions-truly biology as ideology.
Once we begin to challenge Pinker's reporting, the natural question is whether any of the stories in HTMW have empirical support. We will not unpack each example in the book, but here is a sampling. Are wicked stepparents and parent-child conflict the natural outcome of genetic-payoff investment calculations? HTMW relies on readers' willingness to supply examples from their own lives and takes direction from evolutionary anthropologist Robert Trivers's calculus of "parental investment" and I'll-scratch-your-back-and-you-scratch-mine "reciprocal altruism." Performing these calculations is difficult. Even if you suspend biological intuition long enough to imagine that a single mutation in the DNA coding for a protein can turn selfish behavior altruistic, or that a population has simply segregating genetic materials that carry selfishness or altruism (a requirement for many of these models), the calculation of what is called "inclusive fitness" is difficult for organisms that are diploid-that is, whose chromosomes carry two alleles (copies) of every gene. The calculation is simplified by assuming that people are like ants or bees where sisters (queens and workers) are haploid, having only one allele or the other, while the brothers are diploid. This means that sisters share more genes with each other than they do with their brothers, and daughters (a sterile caste) would pass on more of their genes if their mother produced more sisters than if they themselves went on to mate. Pinker relies on Trivers's assumptions which do not carry over as easily for organisms in which both sexes are diploid-like people.8
The point is that the dirty details of population genetics matter-what geneticist Richard Lewontin dubs "the auto mechanics of evolution." Sadly, HTMW never even pops the hood. Again, we are not proposing to make a definitive statement on the dynamics of altruistic gene types in a population, but to insist that blissfully blinkered reporting on a non-existent consensus is irresponsible.
Even HTMW's centerpiece argument for the modularity of mind-an otherwise skillful retelling of David Marr's computational theory of vision, the "2 dimensional sketch"-falls prey to reporting finality. In Marr's approach, when we "see" a cup, the eyes and brain independently compute bits of information about the cup's orientation, surface texture, slant, depth, and tilt, using different algorithms-the very essence of modularity. The resulting sketch is 2-H-D because we interpret and apply more information than just the 2-D shapes that fall onto our retinas. HTMW trumpets Marr's theory as Gospel: "The 2-H-D sketch is the masterwork of the ingeniously designed, harmoniously running machinery of the visual system." There is a hitch: practically no one believes in the 2-H-D sketch anymore! The reason it has fallen from favor cuts to the very heart of Pinker's game plan. Nobody in the computational vision business figured out how to combine the various local bits of orientation, texture, slant, or depth information to pitch up a global answer to what has actually been seen-especially when the local pieces of information disagree. (This is a very different problem from the one Pinker goes on to worry about, the "reference frame" the observer uses.) Attempts to weight the various local pieces of evidence and assign relative costs to build a trade-off schedule have not succeeded. For nearly twenty years scientists at MIT have worked on this problem. The three most serious computational efforts at complete 2-H>-D sketch implementations-the MIT Vision Project, Ballard and Brown's at the University of Rochester, and Tennebaum and Barrow's at the Stanford Research Institute-have failed to produce "harmoniously running machinery." 9
You would never know any of this from reading HTMW. At every turn, those familiar with the topic would find alternatives, but Pinker supplies no corresponding doubts. Pinker takes refuge from this responsibility, suggesting he is participant, not reporter: "I have not given comprehensive literature reviews or an airing of all sides of every debate." But this defense misses an important point. Some of HTMW's speculations might be true, but we have no way of finding out. Part of the problem of marshaling the necessary evidence is that in adaptationist histories/fictions that Pinker so fancies there is no end to plausible story telling. In addition a commitment to a modular brain does not stop us from continuously inventing new modules. Pinker knows that we can always replace one adaptive story by another or add more modules as required, but offers us no way to reasonably refrain from doing so.
HTMW is also in the grip of genetic reductionism, this era's most relentlessly reductionist recipe. A behavior like "incest avoidance" gets boiled down to a simple heritable trait. From atomic trait we run a "reverse engineered" straight line to an equally atomic, modular recipe or "algorithm" that computes the trait. From this modular algorithm, we take a beeline to the fundamental heredity particles themselves-genes. At this level, evolution by natural selection boosts some gene frequencies and lowers others. Roughly speaking, a gene leading to a behavior that increases that gene's representation in succeeding generations will be "more fit" than alternatives that (in this example) permit incest, presumably because the incest-admitting gene-behavior leads to some deleterious side-effect that lowers the number of "incest-permitting genes." Over time, the "incest-avoidance gene" will tend to replace its alternative. For psychologists, the gene-to-behavior bait is especially tempting, because connect-the-dots atomicity and determinism makes cognitive science easy.
You need not believe a word of it. The long trek from patterns of external behavior to genes has many steps and at each one HTMW stumbles. If genes are to serve as accurate bookkeeping chits for maximizing fitness, all the way to love's blushes, then the dotted lines from genes to behaviors must run straight and true. Any deviation, any non-determinism or interaction between and among the stepping-stones, and our explanatory hold slips. Such slippage is found at every step. For every behavior there are many possible traits underlying it; for every trait, many possible "algorithms"; for every algorithm, many possible modules (or perhaps a generalized multipurpose module and not specific ones); for every module, many genes that interact, and so no simple way to "maximize fitness." In reality we do not have a story for any causal chain from genes to behavior-molecular biologists simply talk as if they do. Going the other direction, from traits to genes, is harder still. Even if there were a completely deterministic relationship between genes and phenotypes (behavior, morphology, etc.), we would still have no guarantee that similar phenotypes are produced by the same genes. To map from a behavior like "incest avoidance" to a "trait" that might have a coherent evolutionary history like eye color, one must assume that behaviors segregate just like the wrinkled or yellow peas in Mendel's garden-that is, independently-and that those gene frequencies reached equilibrium 100,000 years ago. Natural selection demands such independence so it can "see" features individually and pluck them out for attempted optimization. But as soon as one moves beyond the simplest case-one trait with two gene types like "wrinkled peas/non-wrinkled peas"-to three possible trait-determinants, then natural selection need not maximize fitness anymore-selection might home in on the most unfit combination of all.10 We are not claiming that unfit combinations will always win out, just that interesting complications can occur and that Pinker never alerts readers to this possibility.
Do "mate selection" or "aggression," for example, really possess the required Mendelian "atomistic&quo independence to be plucked out and optimized without reference to other features of an organism? Probably not. They are more properly an interconnected bundle of behaviors and strategies. For example, do we know that "mate selection" really depends on just two genetic variants? Mate selection may be the result of a large number of interconnected factors that makes knee-jerk fitness-optimization difficult. Pinker does not emphasize this possibility at all, assuming that the traits he would like to naturalize are independent and unconstrained, waiting only for a good story to be explained away.
How does Pinker's use of selectionism go so wrong? Selectionist explanations come in two modes. In the first, selection is invoked simply to rationalize a situation. Because anything that exists must in some sense have out-survived what has not, we have a survival of the fittest where the fittest are precisely those who survive. This kind of selectionist explanation is too powerful for its own good, as any status quo becomes the inevitable but trivial outcome of selection.
Selectionism as used by evolutionists demands a bit more. In "natural selection" there must be variance among the objects under consideration, heritability of some of the features and a winnowing of the objects across the generations so that some subset remain to represent the population and reproduce: random variation, heritability and selective retention. When you hear "natural selection designed this" or "created that," all of the credit seems to accrue to the selection part of the process. However, laying the credit at selection's door is possible only when the variation is not correlated with the needs of the population, when the variation is copious, when there is no bias in the process of heritability, and when the selection is not carried out with a plan in mind.
If variation is meager, then the short menu of options-the constraints-must get substantial credit for the results. Is anyone surprised when they leave McDonald's with a burger and fries? Was Bill Clinton elected president thanks solely to popular selection? No. Many factors were involved in providing a near binary choice at the polls and any responsible assessment of election results discusses those as well. If the process of inheritance was itself biased, then it would demand some of the credit for outcomes. If the selection was done with an agenda, as it is done in animal and plant breeding, then the selector should get much of the credit. Selection becomes "natural selection" when there is no breeder making decisions with a future goal in mind. A selectionist account must always implicitly include all of these elements: broad and undirected variance, heritability, and selection.
In evolutionary terms, every trait we examine is an admixture of physical constraints, natural selection and chance with history. These features are constitutive, not optional. Pinker presents a mutually exclusive conception pitting these factors against each other. This may result in part from his mixing up the two modes just described. There is little room for a more elaborate explanation once you are on the selection-as-rationalization track. For example, when explaining why rabbits have precocious babies whose eyes are set within faces that grow very little, Pinker offers only the story that they tend to be prey animals and implies that eye position is due solely to a reasonable need for an early adult vision system. The lagomorphs (rabbits, hares, and pikas) are one of the earliest branches in the mammalian radiation, possibly as early as the late Cretaceous (more than 65 million years ago).11 Surely any claims about the details of rabbit development need to make reference to this history. Evolutionists care deeply about which characters are ancestral and which are derived in a lineage. Front-facing eyes may be a feature that marked the lineage early, part of what it is to be a rabbit and not an adaptation toa current environmental pressure. All traits do not lend a current selective advantage, the origin of traits and their maintenance do not demand the same selectionist account. Pinker never embraces this distinction as a vivid part of his adaptationism.
Stumbles with rabbits may not seem like much, but this kind of mistake appears repeatedly in HTMW. If a trait is shared by all the individuals who share a common ancestor, we generally infer that the trait was present in the ancestor rather than having evolved independently in each separate lineage. (Of course, it did not have to go this way, but that is the starting hypothesis.) In contrast, when we have similar traits arising in lineages whose shared common ancestor did not have this trait-for example, flying structures in bats, birds, and pterosaurs-the result is often explained by an appeal to convergence. Independent lineages with similar solutions give comparative biologists a basis for reconstructing the pressure that were at work. Clearly, explaining the existence of wings in chickens, ducks, and eagles separately seems like a question at the wrong level of resolution. If the common ancestor lacked the trait and the populations under consideration acquired the same traits over time, this may reflect common response to the same selection pressure, or it may reflect shared constraints on possible variants imposed on the different lineages. We discussed the importance of menu size earlier. If we assume that selection was the sole explanation for parallel evolution we might miss some important parts of the overall dynamics of the changes in traits.
Consider an example: Pinker asks, "Why don't women give virgin birth?" Certainly the correct answer will not make particular reference to humans. Mammals, including humans, just do not have this as a developmental option. Put otherwise, it would take more than just a shift in selection regimes to make humans start reproducing asexually. For these reasons it is hollow bluster to talk about the selective advantage of sex in humans if the traits we are discussing evolved and became established long before the human lineage branched.
If we are interested in a trait unique to a lineage-e.g., language-it becomes very difficult indeed to make good inferences about selection. Pinker seems to be aware of the pitfalls of looking at a single lineage and making a fetish of a particular trait. He compares the search for extra-terrestrial intelligence with an analogous research program among elephants, a search for extra-terrestrials who have evolved unique and wonderful trait-the trunk. This striking and lucid example helps to lay bare our obsession with our own traits and the mistaken belief that they are inevitable. Like a light rain in the desert, it offers only fleeting relief, as we are then immediately subjected to hundreds of pages that ignore the warning. The difficulty of inferring selection pressures from lineages that have unique traits (technically called "autapomorphies") certainly does not stem our ability to tell stories, but it does create real problems in evaluating them.
Another part of challenge of evolutionary reconstruction is deciding what exactly the relevant traits are. How we pick out traits of an organism is critical. Central to the evolutionary psychology project is the possibility of inferring the function of biological structures and behaviors by "reverse engineering." This naming makes it sound as if there is a hard-nosed approach to deconstructing functions. Clever naming aside, this is not a simple task. What is the function of the eye; to see, to cry, to witness? Even when there is a known design agenda (for example, the BIOS or "basic input output operating system" that runs IBM PCs) engineers can find this job very difficult. The engineers at Phoenix Technologies who "cloned" (copied) the IBM BIOS had all the trump cards that evolutionary scientists do not: they already knew what the program code was for, and had IBM's published specifications for every subpart (i.e., they knew for every subroutine, what the input and outputs should look like). In addition they knew about previous designs, and were operating system designers themselves. Still, it took many people several years to do the job; and, most importantly for the evolutionary story, they came up with a very different "reverse-engineered" recipe or algorithm from IBM's. The vulgar adaptationism hawked in HTMW depends on all of the traits under consideration being independent enough of each other so that selection can pick them out and winnow without affecting others. It is not at all obvious that this is true for the kinds of mental abilities that HTMW discusses. Thus, as indicated earlier, we do not know whether vision, emotions, or language are modular in the way that Pinker "just says" they are.
For analytical convenience individuals can be seen as atomized parts. In a careful atomization of an organism into its traits we notice that the parts are constrained due to the many other connected systems. The constraints and opportunities to change are due to the details of development, chance associations with other features that may be strongly selected for or against, and finally selection on the trait itself. Darwin noted that such laws of "correlation and balance" would be of utmost importance to his theory-"cats with blue eyes are invariably deaf."12 Even when we operate with care, after the analysis we must remember that it is we who atomized and reified the parts of the organism. They are not actually separate modules of replaceable parts in an assembly line.
In HTMW Pinker avoids attending to contingent, intertwined, interdependent features of an organism. He simply asserts that "complex adaptations" are the result of selection. But this separation does not happen easily. Features mix; the contingent, the constrained, the selected are part of the story all the way down. If Pinker cared more about evolutionary biology, he would see that even the examples he favors have this property. Eyes-while superficially very different (compound vs. various kinds of single lens eyes) and which presumably evolved independently-share important developmental regulatory elements, so much so that a fly gene involved early in the cascade of eye development can be rescued by a similar protein (its homolog) from mice. If function can be conserved in proteins that have been on independent evolution trajectories for hundreds of millions of years, what does this tell us about the process of evolution? This is a deep and open question-one never raised in HTMW.
The problem of how we atomize an organism and its behaviors links back to Pinker's (ambivalent) embrace of mental modules. With all of his talk that the mind is computationally modular, you might think that Pinker is committed to a reflection of these brain modules in the structures of the brain, but he distances himself from this hypothesis. He commits to just the bare necessities required to think of behavior as broken down into mental modules. He does claim that "modules" are independent enough to allow natural selection to pluck them out and optimize them independently-and this is where talk of "gene(s) for" a particular behavior comes from. Genes
arranged as beads on a string (the chromosomes) sort out independently, and after a few generations of cell division and chromosomal crossing over, lose any correlation with their original neighbors.
Pinker proposes this same kind of independence for abstract mental modules. Notice that it would burden the storytelling to allow for strong correlations between modules. If the brain is too interdependent, then changes in one part demand changes in others, and the notion of unrestrained optimization of a single module (mate selection, cheater detectors, language, responses to landscapes, or other specialized psychological mechanisms) runs into trouble. HTMW does not suggest how to show that mental modules are independent, or that they segregate in the way that, say, Mendel showed how the green and wrinkled peas assorted independently. Moreover, this kind of evidence will be difficult to get. Since you can inherit all of the traits Pinker mentions from any human, it is difficult to say anything about their independence. Further, to get from segregating genes to evolutionary algorithms one must further assume that these traits were heritable in the postulated past environment. However we construe heritability, it is not clear how we could calculate it 100,000 years ago-where are we going to get the identical and fraternal twins?
Pinker's commitment to selectionist explanation at all costs comes to the fore when he discusses Frank Sulloway's dubious suggestion that birth order is the main factor that explains our willingness to accept new ideas. However you come down on Sulloway's thesis, "birth order" is a clear example of environment, not inheritance, as a prime mover (later-born children could not be expected on average to receive more gullibility genes from their parents). With the wind of adaptationism filling his sails, Pinker must claim that the whole behavioral repertoire response of first-borns or the inverse behavior of later-borns is all selected for. This path through the garden allows us to absorb any context into a trivial circular notion of genic contribution to behavior.
We do not mean to suggest that some behaviors are not adaptive. Surely the constraints of the physical world have found their way inside our heads-after all, there are distinct, measured, neural channels for computing what and where objects are, and we do make use of shading and texture separately, somehow, to distinguish shape. The problem is that we are at the beginning, not the end, of this story, and it is not yet clear how to make sense of all the evidence.
We want to answer explicitly a concern that always comes from adaptationists. There is an intuition that natural selection is the only imaginable "force" that can create exquisite and intricate adaptations (all those wonderful examples of teeming life that fill David Attenborough's TV specials). This deep intuition results in sincere puzzlement at criticisms of natural selection as sweeping explanation. After all, if natural selection is dismissed, that leaves only chance, magic, and the intentionality of the gods as counter explanations. Given these choices, natural selection surely must get the nod, and so it does. It should be clear however that an embrace of natural selection in principle does not commit us to an enthusiasm for a promiscuous rationalization procedure that naturalizes every feature of an organism as an optimal and inevitable creation of natural selection. Our complaint is with Pinker's use of natural selection as retrospective naturalization of the status quo and currently fashionable categories.
To give you a sense of the enormous challenge that we face in building a selectionist account that explains the current distribution and abundance of living forms imagine a parallel problem. Anyone who has experienced the end of snow-filled winter has noticed the gradual way that a few warmer spring days start to uncover the snow-covered ground. From this vantage you can try to imagine what it would take to have a theory of snow melt. We might include variables like the heat capacity of snow as a function of how densely it was packed, how much salt was dissolved in it, and so on. We would also need to know something about the local temperature variations across the covered area. From this we could in principle construct a detailed predictive account of which areas will remain covered with snow during a melt. Actually applying our theory would be a real challenge. Measuring some of our variables would require destroying the integrity of the snow pack itself and even then some of the measurements may be difficult. This is the magnitude of the challenge as many evolutionists see the problem of applying natural selection to account for distribution and abundance of living forms. Yet a biological account is even more challenging. There are no obvious external measurements that provide us with fitness values. So even as fitness is supposed to explain future abundance patterns we are constantly faced with measuring it in a way that is not just the trivial claim that future patterns are precisely the outcome of relative earlier fitness values.
The response to this problem in much evolutionary biology is to step back from a causal theory and simply define fitness in terms of survival and reproduction-which in our snow analogy would be to define snow left on the ground in terms of not melting. Biologists have long said a lot about evolution without any detailed knowledge of the steps from genotype to phenotype. All you need are statistical correlations to move from gene to phenotype without a full causal story. One could point out that selection has worked in plant and animal breeding. The parallel in our example would be that snow melts first over heating grates and we could arrange heat sources in the ground so that we could generate a particular pattern (by overwhelming other variables). While plant breeding is routine and general claims in evolutionary biology are possible this in itself does not provide what we need to explain the actual distribution and abundance of life at a given time and place, or in Pinker's case particular behavioral dispositions.
It is important to distinguish how we study a part of the world from how we believe it to be. There is a difference between choosing a particular point of view for exploring a problem (single-gene-based altruism, modular algorithmic brains, etc.) and insisting that this point of view is the only correct description. There may even be good methodological reasons to argue for caricature in a given instance. One of the things that frustrates non-scientists about the practitioners of science is their willing, but fickle embrace of caricature. This methodological stance-simplify until forced to do something else-is generally justified by pointing to the productivity of a given simplification: it generates testable hypotheses. So while gene-centric adaptationists would agree that theirs is a gross simplification, they also urge that adaptationism is a powerful explanatory scheme. The same might be said for a modular brain, with each "module" independently optimized. In either case, we can add nuance by including ever more constraints and connections between "genes" or allowing the modules to proliferate and interconnect, thus placing epicycles on epicycles in an attempt to recover actual organisms or brains. In this direction lies a top-heavy and unworkable theory, so we return to the simpleton version.
Our final and most serious criticism of Pinker's use of adaptationism is that the simple theory, in this case, is not even a hypothesis-generating machine: it is just a story-rationalizing machine. It is not powerful enough to show that a trait is due to optimization in the face of external pressure. It is especially weak when applied to a lineage that does not vary in the trait of interest. If all humans share the language instinct how are we to uncover the details that lead to this? On the other hand, when faced with abundant variation, the notion of a single optimal solution is in trouble. These are not just Pinker's problems: all evolutionists face this challenge. If we are going to insist on only using all-traits-are-optimal adaptationism the challenge then becomes how to use the simple model and avoid a self-enchanting reflection of our own prejudices.
Pinker may protest that we are demanding too much, that science without simplification stultifies. We might agree if HTMW did not present its story as the latest scientific truth. There are many stories we tell ourselves. HTMW presents one that allows Pinker to hammer current high school pedagogy, explain his affection for lawns over glaciers, and rationalize his particular view of relationships. Read as autobiography this may provide some insight, but as storytelling there are certainly more interesting organizing myths. Pinker's attenuated view of humanity is not one we need to adopt and certainly not because current science demands it.
What of the suggestion that science can not make headway without simplification? This is where we return to the interesting methodological question: is it sufficient to focus on an atomized organism (or brain) optimized one gene or module at a time? How important are chance and history in the story we tell? Is there anything to attend to in the emergent properties of development? In recent conversation, James Crow, our foremost population geneticist, has insisted to us that if there were not some trait independence, evolution would grind to a halt, because any change would change all the traits in an organism and so nothing of lasting substance could be built. There is surely something to this. Bird wings and legs do seem to have their own evolutionary trajectories. But then there is Darwin's notion of "correlation and balance" again. The contingent fact that we have five fingers and five toes may be better explained by an appeal to how toes and fingers develop than that five is optimal for their function.13 There really is a middle ground here, and most biologists strive for it.
To be sure, without our tools, metaphors, and simplifications, we are overrun. Without them we are left with awe, a canyon that invites us to ask only the grandest questions and offers only echoes in return. Our response is to embrace the power of tools to manage the unknown. We should be careful to acknowledge the constraints that arrive with each metaphor and model, and avoid the temptation of believing that our theories are somehow indicative of all that can be. The tools are not the world, though we use the tools to explore. Darwinian selection has been a marvelous way to organize and interrogate the complicated and interconnected "tangled bank" of nature. We can celebrate this achievement, while rejecting the inversion that places Darwinism at the center and builds from it a cartoon world of psychological motivation and limp moral theory.
HTMW's difficulties remind us of an old proverb: "button a shirt properly at the bottom, or it won't come out right at the top." Pinker misses too many of the lower buttons. This is exasperating in a book of this length. HTMW contains nothing-literally not one thing-resembling either evolutionary modeling, explicit fitness calculations, or the basics of population or behavioral genetics. It is a grab bag of assertions that could have been made without any appeal to neuroscience, computation, Darwinian psychology, or genetics. To paraphrase Freeland Judson, there is a precept here. More is not always more. Indeed it is sometimes disastrously less. Despite its 600 pages, HTMW's systematic omission of alternatives and detail creates a burden that readers should not have to shoulder.
1 The organizing document of evolutionary psychology is the 110-page introduction ("The Psychological Foundations of Culture") to a much longer collection of commissioned papers titled The Adapted Mind, edited by Jerome H. Barkow, Leda Cosmides, and John Tooby (Oxford: Oxford University Press, 1992). Pinker borrows heavily from this book, and in some sense, HTMW is a pop version of it.
2 Philip Kitcher,Vaulting Ambition:
Sociobiology and the Quest for Human
Nature (Cambridge, Mass.: MIT Press, 1985), p. 436.
3 The simple moth story has become more interesting over time. The traditional interpretation is that the light-colored moths were sheltered from predators by resting on lichen-encrusted trees. Peppered moths in Michigan went through similar changes (to a dark form and then back again) at the same time as British populations without tracking changes in lichen flora. Some of the pale forms returned long before the lichen reinvaded. There is a correlation with pollution but not with the background color that has been traditionally emphasized. It has also not been shown that moth predators are taking the light forms preferentially from soot covered trunks, the habitat is quite diverse and the moths spend much of their time in the crowns of trees. So it may be that saltational shift to darker forms is a response to pollution not the result of being more obvious to predators. B. S. Grant, D. F. Owen and C. A. Clark, "Parallel rise and fall of melanic peppered moths in America and Britain," Journal of Heredity 87 (1996): 351-357 .
4 Nature, 6 November 1997
5 For discussion, see Race, Genes, and IQ by Ned Block, Boston Review, December/January 1995-96, pp. 30-35.
6 Evelyn Fox Keller and Elizabeth A. Lloyd, eds., Keywords in Evolutionary Biology (Cambridge, Mass.: Harvard University Press, 1992), p. 156.
7 Heritability remains highly contentious. For HTMW to cheerfully report personality "heritability" as "50 percent determined" and settled is, well, unsettling. We assume that here Pinker here means "narrow-sense" heritability because that (partly) reflects the additive gene effects and so is the relevant number for evolutionary arguments (it's the number used in artificial selection, as in plant and animal breeding-if environmental conditions can be carefully controlled.) If this is so, then in the 31 July 1997 issue of Nature, Bruce Devlin, Michael Daniels, and Kathryn Roeder once again found a roughly one-third (34 percent) narrow-sense IQ heritability-with a strong effect due to "maternal environment" (the placental environment) of accounting for about 20 percent of the covariation among twins. Throughout HTMW Pinker evidently relies on Thomas J. Bouchard's Minnesota twin studies-on 25 monozygotic (identical twins. But as the Stanford geneticist Marcus W. Feldman noted to us, Bouchard still has not publicly supplied all the data for dizygotic (fraternal) twins raised together and apart so as to properly evaluate this evidence statistically. Pinker shows little appreciation of the hazards in this literature-e.g., none of the studies even begin to account for such factors as the differences between monozgotic twins sharing the same placenta or not; this cannot generally be determined after the fact in twin studies, so it is not clear how to deal with this factor, a matter that Devlin et al. consider at length.
8 Specifically, Robert Trivers's "The evolution of reciprocal altruism," Quarterly Review of Biology 46 (1971): 35-57, drew on W. D. Hamilton's classic population genetics analysis that an apparent reduction in personal fitness, an "altruistic behavioral cost" C, can be maintained under natural selection if that cost is less than the net gain in the expected number of offspring, a benefit B, produced by that same behavior-say, by aiding your sisters-discounted by a proxy for the fact that your sister won't share exactly all your genes, the "coefficient of relatedness", r. In short: C<Br, or B/C>1/r. However, Hamilton's original calculation was done only for haploid-diploid organisms where sisters have only half the usual number of chromosomes; this makes the bookkeeping easy, allowing us to conflate genes and genotypes. Suppose there are two alleles at a single chromosomal locus, a (100 percent dominant) allele, A, for "altruistic", and a (recessive) allele, a, for selfish. For most organisms-aside from social insects-chromosomes come in pairs, so the possible genotypes become AA (altruistic), Aa (also altruistic, assuming A is 100 percent dominant), and aa (selfish). Social insect sisters are haploid, with half the usual chromosomes, so either all the sisters have allele A, or they all have a. (In effect, the trait is "sex-linked".) Then we don't have to worry about genotypes like Aa at all. If, however, we tote up diploid gene frequencies directly, then not only can we dispense entirely with the problematic notion 'inclusive fitness'-but also, alas, the ratio B/C no longer tells us whether a particular "altruistic" investment strategy can be maintained under natural selection. The exact size of the benefit matters-in general, the benefit must be very small-a surprising result. And under such conditions natural selection doesn't maximize fitness at all. In this interpretation, the entire kin-selection-maximize fitness edifice collapses. See Luigi L. Cavalli-Sforza and Marcus W. Feldman, "Darwinian Selection and 'Altruism'," Theoretical Population Biology 14 (1978): 268-280. Hamilton's papers have recently been collected in a handy paperback collection, Narrow Roads of Gene Land (Nw York: W. H. Freeman, 1996).
9 Evidently, something else is going on beyond "local" modular 21/2-D computations and the frame-of-reference "global computation": for instance, as Marr's collaborator Tomaso Poggio and Poggio's student Pawan Sinha showed in last year's Nature, if one takes a newsphoto of President Clinton standing at a podium in front of Vice President Gore, and then grafts an exact copy of Clinton's head and shoulders, somehow one still perperceives the person standing behind Clinton as Gore, not Clinmton. The global pose and stance matter-an integration not considered in the 21/2-D sketch philosophy.
10 Philip Kitcher, in Vaulting Ambition, p.215, borrows this example from Alan Templeton. If you have three alleles, A, S, C in diploid organism (each individual has two alleles). Assume that the relative fitness of the allele combinations is as follows; AS is fitter than AA, SS is lethal, CC is fittest, CS is less fit than AA and C is recessive to A. The result of these assumptions for a large population, in which there is random mating and all else is equal, is that C is eliminated, so CC is driven out. This mathematical exercise captures something of what is thought to happen in the case of Bantu sickle-cell anemia population genetics. Pinker mentions the sickle-cell anemia example but does not seem to realize its implications for the simple notions of "fitness maximization."
11 A friend who spent eight years in 4-H reminds us that rabbit babies are not actually precocious: "They are little pinkies, just like rats, and don't even open their eyes for days after they've been born. They start to hop at two and a half weeks."
12 Charles Darwin, Origin of Species; originally from his notebook entry of 1839.
13 Five fingers and toes were not the original number of digits in tetrapods (see the discussion by M. I. Coates and J. A. Clark in Nature 347 (1990): 66-69), and amphibians probably never had more than four digits (and generally have three) on their front and back feet. There is a clever explanation from molecular developmental genetics that rationalizes why there are at most five different types of digits even if some are duplicated; see C. J. Tabin, "Why we have (only) five fingers per hand: Hox genes and the evolution of paired limbs," Development 116 (1992): 289-96. Still, this is an argument for constraints, not for five being an optimal solution.