Stephen Pinker’s How the Mind Works is an ambitious attempt to bring recent developments in cognitive science to a non-specialist audience. Philosophers’ quibbles be damned, Pinker reaches right for the brass ring: his title refers to the mind, not just to gray matters like the brain. Pinker means to do for mentality what Stephen Jay Gould does for life or Carl Sagan did for the universe.
He’s got a lot of company. There’s been a stampede lately of would-be “re-discoverers” and “rethinkers” and “explainers” of the mind and consciousness, including John Searle, the Churchlands, John Eccles, Alwyn Scott, David Chalmers, Daniel Dennett, et al. Indeed, this field is becoming so crowded it may well take Pinkeresque cheek to, as the advertisers dream, “cut through the clutter.”
How the Mind Works reads like a more broadly-focused sequel to Pinker’s fast-selling The Language Instinct (1994). That book attempted a synthesis of Chomskian generative linguistics and Darwinian natural selection-a shotgun marriage if there ever was one, since Noam Chomsky is renowned for his evolutionary agnosticism. The new book seems to have been written in a spirit of “mopping up” remaining pockets of resistance to Pinker’s view of the mind as a bunch of specialized processing “organs,” like Chomsky’s language module (add a vision module, a physics module, a sex-getting module etc.) Two other leading proponents of this model, evolutionary psychologists Leda Cosmides and John Tooby, have famously advanced the simile of the mind being like a swiss Army knife: an all-in-one collection of purpose-built, content-rich devices, albeit in the mind’s case designed not by the Swiss, but by natural selection. Writes Pinker:
The mind is a system of organs of computation, designed by natural selection to solve the kinds of problems our ancestors faced in their foraging way of life, in particular, understanding and outmaneuvering objects, animals, plants, and other people…The mind is organized into modules or mental organs, each with a specialized design that makes it an expert in one arena of interaction with the world. The modules’ basic logic is specified by our genetic program.
He develops these ideas into a smooth but selective confection of experimental results, reasonable-sounding argument, and trenchant criticism, leavened frequently with humor and counter-intuitive but demonstrable observations of how the mind “really” works:
When Hamlet says, “What a piece of work is man! how noble in reason! how infinite in faculty! in form and moving how express and admirable!” we should direct our awe not at Shakespeare or Mozart or Einstein or Kareem Abdul-Jabbar but a four-year old carrying out a request to put a toy on a shelf…I want to convince you that our minds are not animated by some godly vapor or single wonder principle. The mind, like the Apollo spacecraft, is designed to solve many engineering problems, and thus is packed with high-tech systems each contrived to overcome its own obstacles.
The dish goes down surprisingly easily-Pinker makes hundreds of pages of technospeak go by about as quickly as Tom Clancy. Perhaps this is because, like any airport thriller, Pinker’s book has clear villains. These are perpetrators of what Cosmides and Tooby call the “Standard Social Science Model”: those folks (once based in philosophical behaviorism and psychology, with certain elements inherited by current cultural anthropology and literary studies) who insist on believing that the mind is primarily a social construction, entering the world as a blank slate that gets written upon by the environment and by culture.
As Pinker everywhere argues, “learning” as commonly conceived is simply too underpowered a process to explain the complex abilities the mind acquires and performs, all largely beneath our conscious awareness. Hence those “high tech” features our minds possess as “standard equipment;” hence the special scorn Pinker pours on those who continue to press the “folklore” that language, perception, etc. are the fruits of general learning mechanisms. “…the contents of the world are not just there for the knowing,” he asserts, “but have to be grasped with suitable mental machinery.”
High-tech and mechanistic imagery aside, a clear intellectual pedigree can be traced from Pinker to Chomsky to Descartes and the decidedly unmechanistic Plato. Chomsky himself has been more forthcoming in his debt to these earlier thinkers, on occasion allowing himself to be called “neo-Cartesian.”
Indeed, so enamored is Chomsky of his “law and order” view of mental life, he has denied the legitimacy of studying real-life utterances in a “properly” scientific linguistics. Instead, the Chomskian view of language study verges on medieval scholasticism, with colleges of closeted linguists hunched over their manuscripts, musing over rules like how many prepositions can dance on the head of a noun phrase. Actual language, meanwhile, rages on unstudied outside the monastery walls.
Pinker can never quite bring himself to go this far. Sometimes, he comes close:
Systems of [mental] rules are idealizations that abstract away from complicating aspects of reality. They are never visible in pure form, but are no less real for all that…[the idealizations] are masked by the complexity and finiteness of the world and by many layers of noise…Just as friction does not refute Newton, exotic disruptions of the idealized alignment of genetics, physiology, and law do not make “mother” any fuzzier within each of these systems.
That is, in Pinker’s mind the rules are just as real as the reality, though they are abstractions. Of course, what counts as a natural “law” and what counts as distracting “noise” is never so easily resolvable: while it is true that friction doesn’t refute Newtonian notions of gravitation, the “laws” of friction are also handy for keeping airplanes in the air and braking your car. What counts as law and what counts as “noise” therefore depends on context. Unfortunately, in the Chomskian case- Pinker included- these are suspiciously often matters of authority and/or selective attention.
The Dawn of the Chuck
As in The Language Instinct, Pinker has an unfortunate habit of making issues seem resolved that aren’t. As Cosmides and Tooby themselves note, there’s a problem with using “learning” as an explanation:
Advocates of the Standard Social Science Model have believed for nearly a century that they have a solid explanation for how the social world inserts organization into the psychology of the developing individual. They maintain that structure enters from the social (and physical) world by a process of “learning”- individuals “learn” their language, they “learn” their culture, they “learn” to walk, and so on…Of course, as most cognitive scientists know (and all should), “learning”…is not an explanation for anything, but is rather a phenomenon that itself requires explanation.
Cosmides and Tooby use their critique of learning to support their innativist views. However, it also suggests that it is not learning itself that is lacking, but our conception of learning. More to the point, there’s some evidence that Skinnerian stimulus-response, rats-pressing-levers-and-running-mazes-type learning is not the only kind of learning there is, especially in infants and children. This has led some fans of general intelligence to tell the evolutionary psychologists to put away their Swiss Army knives.
From outside academe, the differences between camps of cognitive scientists look positively trifling. Most within the field agree that the mind/brain does not come “out of the box”totally unstructured, a blank slate mostly “filled” by culture. Most agree that this innate structure is mediated to some degree by natural selection. There’s even some broad agreement that cultural factors (language, for instance), if they operate long enough and consistently enough (a few thousand generations, give or take), can also act as selective pressures, helping to reshape both the mind and body of what Jared Diamond has called “the third chimpanzee.”
From within the field, the remaining arguments look bitter. People like Pinker, Cosmides and Tooby, Dan Sperber, Nicholas Humphrey, and Elizabeth Spelke see modules, modules everywhere, each as innate and superbly adapted for their functions as the pancreas or an elephant’s trunk. Meanwhile, “domain-generalists” like Jeffrey Elman, Elizabeth Bates and Anna Karmiloff-Smith turn this reasoning on its head . That is, they don’t deny modularity per se (modules are, after all, a good way to package complex neural structures in the limited volume inside the skull) but maintain that our specialized abilities emerge from our predisposition to attend to certain regularities in the world. What’s innate is not the knowledge, but the capacity to observe the regularities and learn them quickly.
For example, Pinker invites us to marvel at the “software driver” that controls the human hand:
A still more remarkable feat is controlling the hand. . . It is a single tool that manipulates objects of an astonishing range of sizes, shapes, and weights, from a log to a millet seed. . . “A common man marvels at uncommon things; a wise man marvels at the commonplace . ” Keeping Confucius’ dictum in mind, let’s continue to look at commonplace human acts with the fresh eye of a robot designer seeking to duplicate them . . .
Typically, Pinker notes a complex adult capacity and wonders how to “reverse engineer” what natural selection hath wrought. He hardly considers the alternative, however: that the mind/brain is disposed to learn quickly and efficiently how to operate whatever appendage it happens to find at the end of its arm, whether it be a hand, a paw, or a flipper.
Neuroscientists have long known that when neurons fire, they not only can make muscles move and glands secrete, they also reinforce their own tendency to fire the same way in the future. (The principle is called Hebbian learning; its dictum is “Fire together, wire together.”) Conceivably, the act of using the hand reinforces the pattern of synaptic connections that control the thumb, the fingers, etc. The end result in the adult looks so well-designed and appropriate it might seem like an innate “program” for moving the hand was genetically “wired in”-but it wasn’t. We began only with a proper neural connection between hand and motor cortex, and a need to manipulate.
The point is not whether a Hebbian model fully explains motor learning. The point is that “gee whiz” explanations of complex adult abilities don’t necessarily prove full-blown innateness. (Most obviously because nobody starts off as an adult!) Furthermore, in evolutionary terms, the weaker model of predisposition, the “disposed to learn” model, is a more parsimonious explanation than inborn mental modules. This is because, in the case of the hand, it doesn’t require genes to somehow code the “hook grip,” “the five-jaw chuck”, “the two-jaw pad-to-side chuck,” “the scissors grip” etc. etc.. It only obliges our genes to motivate us to learn.
Indeed, exactly how genes might code things like language or “social intelligence” or “natural history intelligence” has never been too clear. DNA, after all, actually regulates nothing more than protein production and other DNA. In this sense, Cosmides and Tooby’s critique of learning might also be leveled at the catch-all notion of innateness. Like learning, innateness “…is not an explanation for anything, but is rather a phenomenon that itself requires explanation.”
Perhaps the most fascinating aspect of How the Mind Works is watching Pinker wrestle with the problem of connectionism. On the one hand, the fact that experimenters have succeeded in teaching artificial neural nets to do some pretty human-like things, like recognize written letters and put English verbs into the past tense, is a vindication of one of the pillars of his model: the computational theory of mind. On the other hand, the uncanny way neural nets have of learning the regularities of input data without set rules being programmed in is a challenge to Chomsky’s “poverty of the stimulus” arguments for innate knowledge. Some connectionist nets have shown modularity of function, and even human-like cognitive deficits when experimenters simulate “injuries” by removing parts of the system. Importantly, these mind-ish qualities have emerged with learning, and were not introduced pre-formed.
Pinker navigates this quandary by pure elan. After devoting 14 pages to the nature and advantages of “toy neurons” for understanding the mind, and thereby establishing his connectionist credentials, he suddenly takes to calling nets “connectoplasm” (a term clearly meant to evoke that other discredited substance, protoplasm), and asserts “neural networks alone cannot do the job” of accounting for human intelligence:
I do think that connectionism is oversold. Because networks are advertised as soft, parallel, analogical, biological and continuous, they have acquired a cuddly connotation and a diverse fan club. But neural networks don’t perform miracles, only some logical and statistical operations.
Of course, nobody thinks networks or neurons “perform miracles.” As a supporter of the computational theory of mind, Pinker must also believe that the mind/brain itself, at a certain basic level, performs “only some logical and statistical operations.” So what is he talking about?
The real sin of the “strong” version of connectionism- the argument that language, creativity, consciousness itself are all ultimately explicable along connectionist lines- is that it resurrects the associative model of learning. Connectionist nets, after all, learn by doing, and by crudely “associating” certain patterns of inputs and outputs. Following Chomsky, Pinker prefers to imagine a structure of cognitive rules and regulations is doing the real work of mindedness. These rules are not epiphenomenal artifacts of the learning process, or post hoc abstractions from regularities of behavior. Rather, they are ontologically “real.”
To debunk associationism, Pinker makes a list of human-like things that nets can’t do (yet). At least one of these is downright silly: Pinker claims that nets can’t distinguish individual examples of a class of things from each other. Within the connectionist paradigm, “there is no longer a way to tell apart individuals with identical properties. They are represented in one and the same way, and the system is blind to the fact that they are not the same hunk of matter.” People make such distinctions all the time; for instance, identical twins are different people, regardless of how much they look and seem alike: “The spouse of one identical twin feels no romantic attraction toward the other twin. Love locks our feelings in to another person as that person, not as a kind of person, no matter how narrow the kind.”
The fallacy here resides in the assumption that any two examples of any real-world class are really identical with each other. Distinguishing individuals has to do with noticing subtler and subtler kinds of variation. It often takes some time, for instance, for field ethologists to begin to see their subject animals as individuals- at first, they all look the same. To take a more commonplace example, my wife and I own two Himalayan cats that happen to be siblings. Despite the fact that the female is a tortoise-shell point, has a smaller head, and a completely different carriage and personality, houseguests invariably can’t distinguish her from her blue-point, big-headed, lay-about brother. My wife and I have had a longer amount of time to make the appropriate, fine associations.
Despite their genetic identity, not even monozygotic twins are phenotypically or behaviorally identical. Indeed, the one place where such identities do exist is the abstract mathematical world that inspired Chomskian linguistics- it’s a rule, for instance, that a line segment of length X is identical to any other of length X. This is an area where Pinker’s intellectual roots are exposed, and they mislead him. (In the non-mathematical world, incidentally, people do have an uncanny knack for associating certain romantic feelings with the same “types”- the same hair, same build, same foibles. It’s no secret. Maybe Pinker just doesn’t get out much.)
Some of his other objections are more persuasive. It is indeed hard to visualize how nets can handle complex combinatorics, or alter the quantification of elements in a problem when they’re the same-but-different, or process recursively unless specifically constructed to do so. (In such cases the connection weights have a tendency to interfere with each other.) On the other hand, all of these problems have the definite air of claims once made by reputable Victorian physicists who asserted the physical impossibility of heavier than-air flight. Above all, we should know by now that it’s not too smart to bet the farm on something(s) being technically impossible.
Pinker’s treatment of the other key concept in the book-evolution- is equally provocative. He’s clearly very much aware of the principles and objections to the reigning synthesis of Darwinian natural selection and Mendelian genetics. Steering clear of the pan-adaptationism decried by Gould and Richard Lewontin, he rightly observes that not everything about an organism is necessarily adaptive: “A sane person can believe that a complex organ is an adaptation, that is, a product of natural selection, while also believing that features of an organism that are not complex organs are a product of drift or a by-product of some other adaptation”
Gould and Lewontin once wrote of the “Panglossian paradigm”: the tendency among some evolutionary scientists to mistake how things actually happened for the optimal way things could have happened. Though Pinker disavows it, his work fits the paradigm anyway. For instance, in a discussion of whether the development of intelligent life is inevitable on any life-supporting planet, he compiles a list of the unlikely factors that “made it especially easy and worth their while [for organisms] to evolve better powers of causal reasoning.” First on the list is the primates’ fortunate dependence on the visual sense. Why? Because “Depth perception defines a three-dimensional space filled with movable solid objects…Our capacity for abstract thought has coopted the coordinate system and inventory of objects made available by a well-developed visual system.”
Compare this to other mammals, such a dogs, who rely more on olfactory information:
Rather than living in a three-dimensional coordinate space hung with movable objects, standard mammals [sic] live in a two-dimensional flatland [the ground] which they explore through a zero-dimensional peephole [the nose]…If most mammals think in a cognitive flatland, they would lack the mental models of movable solid objects in 3-D spatial and mechanical relationships that became so essential to our mental life.
Anyone who has seen an earthworm bury itself or a dog sniff his way up the trunk of a tree knows olfactory dependence is not synonymous with living in a “two-dimensional flatland.” Nor does Pinker take note of other 3-D modalities, such as echolocation in bats and cetaceans, which likewise represent a world of “solid objects in 3-D spatial and mechanical relationships.” Faced with such a poverty of imagination with respect to terrestrial creatures, it’s hard to take Pinker’s musings over the unlikelihood of extraterrestrial intelligence very seriously. What about an alien creature living in liquid methane that uses short-wave radar? Or one that lives underground and finds petrocarbon “food” by using seismic “thumps”? What’s wrong with 2-D intelligence anyway? Would a creature able to reason in 4-, 8-, or 1,000- dimensions be justified in denying the significance of our 3-D intelligence? Perhaps we shouldn’t give up on SETI just yet.
Pinker’s discussion falls prey to the Panglossian paradigm because he thinks a sufficient condition for human intelligence is a necessary one for all intelligence-the particular way we evolved, in other words, is established as “optimal” for invading the cognitive niche. He plays Pangloss again elsewhere, in his criticism of idea of “meme evolution.” This is the notion, notably suggested by Richard Dawkins, that ideas, like organisms, might reproduce and evolve in the “habitat” of human brains. Sensing an opening for the cultural constructivists, Pinker tries to slam the door by asserting “When ideas are passed around, they aren’t merely copied with occasional typographical errors; they are evaluated, discussed, improved on, or rejected. Indeed, a mind that passively accepted ambient memes would be a sitting duck for exploitation by others and would have quickly been selected against.”
Try telling that to a Scientologist. Unlike in Pinker’s cognitive symposium, real people are actually very good at “passively accepting ambient memes.” It might even be adaptive to do so: Pinker himself suggests the survival benefit of not standing out, of hanging with the herd. In fact, Pinker is telling a variation on a “just so” story here, using an argument for adaptation to justify a point he asserts to be true. This is precisely what Gould and Lewontin warned against when they observed how, wrongly applied, tales of adaptation could be concocted to justify virtually any position. They note “…Since the range of adaptive stories is as wide as our minds are fertile, new stories can always be postulated.” Though Pinker professes an understanding of non-adaptationist factors in evolution, his work clearly falls into that category where, as Gould and Lewontin lament, “Constraints upon the pervasive power of natural selection are recognized…But…are usually dismissed as unimportant or else, more frustratingly, simply acknowledged and then not taken to heart and invoked.”
All of these problems might be traced to the consequences of Pinker’s primary methodology. This is the idea that we can figure out the mind/brain by “reverse engineering” it:
…psychology is engineering in reverse. Reverse engineering is what the boffins at Sony do when a new product is announced by Panasonic, or vice versa. They buy one, bring it back to the lab, take a screwdriver to it, and try to figure out what all the parts are for and how they combine to make the device work.
Up to a point, this seems like a reasonable analogy. Bodies and brains are, after all, kinds of organo-chemical mechanisms, and as Dawkins has notably observed, natural selection is “the blind watchmaker.” Why not pry the back off the timepiece of the mind and take a look?
Trouble is, human engineers and natural selection work in quite different ways. Following C.G. Langton, Daniel Dennett explains in Consciousness Explained:
…human engineers, being farsighted but blinkered, tend to find their designs thwarted by unforeseen side effects and interactions, so they try to guard against them by giving each element in the system a single function, and insulating it from all the other elements. In contrast, Mother Nature…is famously myopic and lacking in goals. Since she doesn’t foresee at all, she has no way of worrying about unforeseen side effects. Not “trying” to avoid them, she tries out designs in which many side effects occur…[and] every now and then there is a serendipitous side effect: two or more unrelated functional systems interact to produce a bonus: multiple functions for single elements.
The difference in how human engineers and the natural one build mechanisms entails more than the obvious fact that organisms self-organize (they grow) and machines get built. It affects every stage of the “design” process. When some capacity evolves in nature (say, flight), Darwinian selection doesn’t start out with a dream and a blank piece of paper- it starts out with an existing, functional organism. If the Wright Brothers had worked this way, they wouldn’t have designed a new machine from scratch. Instead, they would have gradually “retrofitted” some existing vehicle, like a horseless carriage. The resulting “flying flivver” might have taken much longer to realize than a purpose-built flyer; it might have suffered many more failed test flights until it achieved a sustained glide, then powered flight; it might have taken longer to get the heavy weight of the car down and the wingspan just right. In any case, aeronautical history would have been quite different.
All of which goes to show the problem with “reverse engineering” natural mechanisms: you can never be sure a widget was designed for some function, only that it presently serves that function. In the case of the “flying flivver,” it would be useless to wonder how the fenders and the bumper help the car fly better. Those features have to do with the history of structure, not its present function.
Of course, Pinker and every informed adaptationist knows all this. Furthermore, they would argue that certain essential features (like the wings) are so directly necessary in the evolved function that we must invoke adaptation. All true enough. But this is not the same as saying the human mind is “like the Apollo spacecraft…packed with high tech systems, each contrived to overcome its own obstacles.” As Langton argues, each system may well overcome several obstacles, and it pays not to be too categorical in assigning roles to each widget. If I were asked whether the brain is more like the Apollo spacecraft or a more like a petunia, I’d have to confess I’m not sure.
The Nature of Nature
Pinker is a master rhetorician. When he is on firm ground, he’s a superbly articulate popularizer. When he isn’t, he spins beautifully, exploits what he can, and knows when to beat a tactical retreat. His wit can disarm criticism.
All of which makes it surprising when his sense of humor deserts him and he reverts to dull partisanship. The ceaseless drumbeat of distortion and belittlement of social scientists is one such puzzling element of his book. These people, we learn, are too dense to understand the problem with Lamarckianism; they’re wrong, wrong, wrong about associationism; they insist on believing in “folklore” about the mind because they’re either bent on “feel good” politics or distracted by moral straw-men like genetic determinism.
If cultural anthropologists agree on any human universal, it is the tendency of all cultures to justify their own cultural constructions by “naturalizing” them. As Cosmides and Tooby argue and Pinker agrees, this has led many anthropologists either to deny any “human nature” exists, or to declare the search for universals as unavoidably an exercise in Western ethnocentrism.
Yet human beings did have an origin, and do have some sort of nature. Dread or misunderstanding of these facts have too often resulted in an incurious particularism that prefers to celebrate, not to explain, difference. If anthropology is traditionally a boat powered by two oars-the study of difference and the study of commonality amongst peoples-then the modern discipline has an empty oarlock and is rowing in circles.
But none of this is to say that “naturalization” doesn’t happen, especially among thinkers who profess totalizing theories. When Pinker is spinning his synthesis with respect to stereoscopic vision and incest avoidance, he talks a good game. But when we are expected to believe that, for instance, most peoples’ taste in landscapes is a feature of Cosmides and Tooby’s Swiss Army Knife, he strays into the full-blown ridiculous. He argues, for instance, that we exhibit a “default habitat preference” for savannas-according to certain cross-cultural surveys, everybody likes “semi-open space…even ground cover, views to the horizon, large trees, changes in elevation, and multiple paths leading out…” Though the very idea that we evolved in savannas is fiercely debated, Pinker conclusively declares “No one likes the deserts and the rainforests.” (Color me weird, then.) Nor does Pinker shy from drawing the logical aesthetic conclusions from this bit of human standard equipment-“…we are designed to be dissatisfied by bleak, featureless scenes and attracted to colorful, patterned ones.” There, I knew there was a reason I prefer Henri Rousseau to Georgia O’Keefe.
This is naturalizing. Based on such arguments, and observations of the range of human variation, anthropologists et al. may still have quite defensible reservations about importing whole disciplinary paradigms like that of cognitive science into anthropology, history, linguistics, etc. As Pinker himself suggests, it is quite reasonable for people- and that does include social scientists- not to “passively accept ambient memes.”
How the Mind Works, Stephen Pinker, W.W. Norton, 565 pages
This review by Nick Nicastro