Steven Pinker [5.7.96]

George C. Williams:I'm very favorably impressed with Steven Pinker. He's going to be a superstar well into the twenty-first century. What's particularly notable is his work on the evolution of our language capability, and being able to talk about this in specific terms. There are features there that have been evolving, and that we can interpret with respect to why they evolved. I remember speculating in my 1966 book about what it is that makes the human species special. There have been all sorts of suggestions: bipedalism, tool use, that sort of thing, but it struck me at the time that the one defining capability is language.


STEVEN PINKER is an experimental psychologist; professor in the Department of Brain and Cognitive Sciences at MIT; director of the McDonnell-Pew Center for Cognitive Neuroscience at MIT; author of Language Learnability and Language Development (1984), Learnability and Cognition (1989), The Language Instinct (1994), and How the Mind Works, forthcoming,1997.

Steven Pinker's Edge Bio Page

[Steven Pinker:] I call language an "instinct," an admittedly quaint term for what other cognitive scientists have called a mental organ, a faculty, or a module. Language is a complex, specialized skill, which develops in the child spontaneously without conscious effort or formal instruction, is deployed without awareness of its underlying logic, is qualitatively the same in every individual, and is distinct from more general abilities to process information or behave intelligently. (One corollary is that most of the complexity in language comes from the mind of a child, not from the schools or from grammar books.) All this suggests that language is caused by dedicated circuitry that has evolved in the human brain. It then raises the question of what other aspects of the human intellect are instincts coming from specialized neural circuitry.

I'm interested in all aspects of human language. I'm an experimental psychologist who studies language for a living: how children learn language, how people put sentences together in their minds and understand sentences in conversation, where language is situated in the brain, and how it changes over history.

My work concentrates on what science has discovered about language since 1950. In answering those questions, other questions repeatedly come up. Why is the hockey team in Toronto called the Maple Leafs instead of the Maple Leaves? Why do we say, "He flied out to center field" in baseball — why has no mere mortal ever "flown out" to center field? Why do immigrants labor with lessons and tapes and homework and English classes, while their four-year-old kids learn the language so quickly that they can make fun of their parents' grammatical errors? What language would a child speak if he was raised by wolves? I also look at what we know about how language works, how children acquire it, how people use it, and how it breaks down after injury or disease of the brain.

I unify this knowledge with three key ideas. One responds to the fact that what people do know about language is often wrong. The view of language that suffuses public discourse — that people assume both in the sciences and in the humanities — is that language is a cultural artifact that was invented at a certain point in history and that gets transmitted to children by the example of role models or by explicit instruction in schools. The corollary is that now that the schools are going to pot and people get their language from rock stars and athletes, language will steadily deteriorate, and if current trends continue we're all going to be grunting like Tarzan. I argue instead that language is a human instinct.

The second idea comes from the following: If language is a mental organ, where did it come from? I believe it came from the same source as physical organs. It's an adaptation, a product of natural selection in the evolution of the human species. Depending on how you look at it, this is either an incredibly boring conclusion or a wildly controversial conclusion. On the one hand, most people, after hearing evidence that language is an innate faculty of humans, would not be surprised to learn that it comes from the same source that every other complex innate aspect of the human brain and body comes from — namely, natural selection. But two very prominent people deny this conclusion, and they aren't just any old prominent people, but Stephen Jay Gould, probably the most famous person who has written on evolution, and Noam Chomsky, the most famous person who has written on language. They've suggested that language appeared as a by-product of the laws of growth and form of the human brain, or perhaps as an accidental by-product of selection for something else, and they deny that language is an adaptation. I disagree with both of them.

The third idea comes from the question, "Why should we be so interested in the details of language in the first place?" Language is interesting because, of course, it's distinctly human, and because we all depend on it. For centuries, language has been the centerpiece of discussions of the human mind and human nature, because it's considered the most accessible part of the human mind. The reason people are likely to get exercised by technical disagreements over the proper syntax of relative clauses in Choctaw, say, is that everyone has an opinion on human nature, and lurking beneath such discussions of language is the belief that language is the aspect of science where human nature is going to be understood first.

If language is an instinct, what does it say about the rest of the mind? I think the rest of the mind is a set of instincts as well. There's no such thing as intelligence, a capacity for learning, or a general ability to imitate role models. The mind is more like a Swiss Army knife: a large set of gadgets, language being one of them, shaped by natural selection to accomplish the kinds of tasks that our ancestors faced in the Pleistocene.

Why do I call language an instinct? Why not a manifestation of an ability to acquire culture, or to use symbols? There are four kinds of evidence that have been gathered over the last century.

One of them is universality. Universality, by itself, doesn't indicate that the ability in question is innate. For all I know, VCRs and fax machines are now close to universal across human societies. But universality is a first step to establishing innateness, and it was a remarkable and unexpected discovery — early in the century, when anthropologists first started exploring societies in far-flung parts of the globe — that without exception, every human society has complex grammar.

There's no such thing as a Stone Age language. Often you'll find that the most materially primitive culture has a fantastically sophisticated, complex language. Likewise, within a society, complex grammar is universal. To appreciate this, you first have to put aside "prescriptive grammar" — the grammar of schoolmarms and copyeditors (don't split infinitives, watch how you use "hopefully," don't let your participles dangle, don't say "them books"). That has nothing to do with what I'm talking about; it's in large part conformity to a set of conventions for a standard written dialect — something that all literate people have to master, but separate from ordinary conversation. The grammar of the vernacular, in the sense of the unconscious rules that string the words together into phrases and sentences when we converse, is far more sophisticated. If you simply try to determine what kind of mental software it would take to generate the speech of a typical person in the street, or a typical four- year-old, you'll find that it's always extremely complex and has the same overall design within a society and across societies. All languages use things like nouns and verbs, subjects and objects, cases and agreement and auxiliaries, and a vocabulary in the thousands or tens of thousands.

Those are the first two bits of evidence, the universality of language and the universality of the design of language — that is, the kinds of mental algorithms that underlie people's ability to talk. The third bit of evidence is from my own professional specialty, language development in children. We see language development proceed the same way in all the world's cultures. It's remarkably rapid, as any parent can attest. Children begin to babble in their first year of life. First words appear at about one year of age. First word combinations, things like "more milk" and "all gone doggy," happen at about eighteen months. Then around the age of two, there's a burst of about six months — even less for some children — in which one sees a flowering of virtually the entire grammar of English: relative clauses, passives, questions with "WH" words, and constructions so complex that the researchers in artificial intelligence haven't been able to duplicate them in computer systems that would allow us to converse with a computer in English. Nonetheless, children have mastered these constructions before the age of three, and you have the impression at a certain point that you're having conversations with your child, whereas the child a short time before could produce no more than one or two words of baby talk.

And what the child has done is solve a remarkably difficult computational problem. The problem can be stated as an engineering task: design an algorithm that will take a sample of sentences and their contexts from any of the five thousand languages on the planet, and after crunching through a number of these sentences — say, a couple of hundred thousand — come out with a grammar for the language, regardless of what the language is. That is, Japanese sentences in, Japanese grammar out; Swahili sentences in, Swahili grammar out. This problem is way beyond the capability of any current artificial-intelligence system. Current natural-language processing systems can't even use a single language, let alone learn to use any language. Nonetheless that's what the child does in those six months, despite the lack of grammar lessons or even feedback from parents. Moreover, if you crank up the microscope on baby talk, you often find that it conforms to universal constraints that characterize language across the planet. In the kind of experiments I do in my day-to- day work, in which you get a child in a situation where he has to use some construction he hasn't been challenged with before, the child often gets it perfect on the first shot, as if he had all the pieces and just had to let them fall together.

Children also have a remarkable ability to avoid errors. Our ears do perk up when we hear things like "breaked" and "comed" and "goed." But if we were to look at the much larger set of errors that a computer would make, because the errors would be natural conclusions to jump to about the logic of the language, in most cases it never occurs to children to make that error, even though it's the first thing a logician or a cryptographer or a computer program would guess.

Language development isn't driven by general communicative utility. The child doesn't talk better and better just to get more cookies, or to get more TV, or to be allowed to play outside more often. A lot of the changes you see in children's development simply make their speech conform better to the grammar of the language they're acquiring. Here's an example. Take a verb like "to cut," "to hit," or "to put." Children go through a stage in which they make errors like "cutted," "hitted," and "putted." A child at that stage is simply making distinctions that we adults don't. If I say "On Wednesday I cut the grass," it could mean that I cut the grass every Wednesday or that I cut the grass last Wednesday, because in English the past tense and present tense of "cut" are identical. A child who says "cutted" can distinguish the two, even though in some sense he is making a grammatical error. Children outgrow that "error," and in doing so they make their language worse in terms of the ability to communicate thoughts. What's going on in the mind of the child isn't like a hill-climbing procedure, where the better you're communicating the more you stick with what you have, but an unconscious program that synchronizes the child's language with the language of the community.

There are exotic circumstances where one can show that children are injecting complexity into the language. They're not simply repeating or reproducing imperfectly what they hear, but making the language more complex. These situations are referred to as creolization. They were first documented in cases where children in plantation or slave colonies were exposed to a mishmash or "pidgin" of choppy, ungrammatical strings of words that served as a lingua franca among the adults, who had come from different language communities. The first generation of children who were exposed to a pidgin did not reproduce that pidgin but converted it into a language with a systematic grammar called a creole. There are several cases where creolization can be seen happening today. These are cases in which deaf children are either exposed to a defective version of sign language, because their parents didn't learn it properly, or, in the case of Nicaraguan sign language, because no sign language exists and the children were recently put together in schools for the first time and are inventing, in front of our eyes, a language with a systematic grammar.

The final bit of evidence is that language seems to have neurological and perhaps even genetic specificity. That is, the brain is not a meatloaf, such that the less brain you have the worse you talk and the stupider you are, but seems to be organized into subsystems. Using brain damage and genetic deficits as tools, we can see how the brain fractionates into subcomponents.

First, there are cases in which language is impaired but intelligence is intact. For example, there are forms of aphasia, caused by strokes, in which people lose the ability to speak or understand but retain the rest of their intelligence. A slightly less extreme condition is called "specific language impairment," or SLI, in which children don't develop language on schedule or in a normal way: the language appears late and the children have to struggle with it. Pronunciation improves in adulthood, with the help of lots of therapy and practice, but the victims speak slowly, hesitantly, and with many grammatical errors. They have trouble doing certain language tasks that any five year-old can do. For example, a tester shows a picture of a man doing something for which there doesn't exist a word, like swinging a rope over his head, and says, "Here's a man who likes to `wug.' He did the same thing yesterday. Yesterday he..." A five-year-old will say "wugged," even though he's never heard "wugged" before. Presumably he creates it by applying the mental equivalent of the rule of grammar: "Add `ed' to form the past tense." If you give this task to a language-impaired victim, very often he'll say, "Well, how should I know? I've never heard the word before." Or he'll sit and think, and reason it out as if you'd given him a calculus problem to solve; the answer doesn't come naturally.

This is despite the fact that victims of SLI are, by diagnostic definition, normal in intelligence — that is, if they weren't normal in intelligence they wouldn't have been classified as "specifically language-impaired." They aren't deaf, and they aren't autistic or socially abnormal. Often, in fact, they can be superior in intelligence. There are some children with SLI who are excellent in math but who find speaking a pain. Specific language impairment seems to run in families — something that language therapists have known for years, because they'll treat Johnny and then a few years later they'll treat Johnny's sister and Johnny's cousin. In the last few years, large-scale familial and twin studies have shown that SLI is highly heritable. The crucial study — identical twins reared apart — has not been done, because only about seventy of these pairs in the whole world have been studied, and none of them happens to have SLI.

In cases where you find a bad gene or an injured brain, and language suffers but the rest of the brain is all right, there's always the objection that perhaps language is the most mentally demanding thing we do. If there is any compromise in processing power, language will suffer the most, but that doesn't indicate that language is somehow separate from the rest of cognition; it may be just quantitatively different. The clincher is what people in my field call a double dissociation, where one sees the opposite kind of impairment; these are syndromes in which language is intact but the rest of intelligence suffers — a linguistic idiot savant, who can speak, and speak well, but is retarded. There are a number of syndromes in which that can happen, including spina bifida and Williams syndrome. In those cases, you have what therapists call chatterboxes or blatherers; a child goes on and on in beautifully formed sentences that often have no connection to reality. This can happen in children with an IQ of 50, who cannot tie their shoes or handle money. That's evidence for the claim that language is a separate mental system, an instinct.

Why do I call language an adaptation? What is the alternative? Gould and Chomsky suggest that language is a by- product. Perhaps, as we developed a big brain in our evolutionary history, language came automatically, the same way that when we adopted upright posture our backs took on an S-shaped curve. Perhaps we have language for the same reason we have white bones. No one would look for an adaptive explanation for why bones are white as opposed to green. They're white as a side consequence of the fact that bones were selected for rigidity; calcium is one way to make bones rigid, and calcium is white. The whiteness is simply an epiphenomenon, an accident.

The argument from Chomsky and Gould is that maybe language was an unavoidable physical consequence of selection for something else, perhaps analytical processing, hemispheric specialization, or an enlarged brain. No one who was around when language evolved is here to tell us about it, and words don't fossilize, so the arguments have to be indirect. However, there's a standard set of criteria in biology for when to attribute something to natural selection — that is, when it may be called an adaptation — and when to look at it as a by-product, or what Gould and Lewontin call a "spandrel." Ironically, what Gould and Chomsky have not done is apply these standard criteria to the case of language. They've noted the logical possibility that language doesn't have to be an adaptation, but they haven't said, "Let us now pull out the test kit, apply it to language the way we apply it to any other biological system, and see what the answer is."

The test is articulated very well by George Williams and Richard Dawkins, and that test is complex adaptive design. The fundamental problem in biology is to explain biological organization: why animals are complex arrangements of matter that do unlikely but interesting things. Dawkins and Williams noted that before Darwin, complex design was recognized as the fundamental puzzle of life, even by theologians. In fact, for them, it was an argument for the existence of God. The Reverend William Paley put it best: Imagine that you're walking across a field and you come across a rock, and you ask someone, "How did the rock get there?" and they say, "Well, the rock's always been there." You'd probably accept that as about as good an explanation as you had any right to expect. But now let's say you're walking across a field and you come across a watch, and you ask, "How did the watch get there?" and someone says, "Well, it's always been there." You wouldn't accept that explanation, because a watch is an inherently improbable arrangement of matter. You can rule out the possibility that some pattern of wind and earthquakes just happened to throw together a bunch of matter that fell into the exact configuration of springs and gears and hands and dials that you find in a watch. The watch shows uncanny signs of having been designed for the purpose of telling time, which implies some intelligent creator.

Paley's argument in the nineteenth century was that any biological organ, like the eye, is much more complex than a watch. The eye has a retina, and a lens, and muscles that move it in precise convergence, an iris that closes in response to light, and many other delicate parts. Just as a watch implies a watchmaker by virtue of its complex design, an eye implies an eyemaker — namely, God. What Darwin did was not to deny that complex design was a serious problem that needed a solution but to change the solution. The brilliance of Darwin's idea, natural selection, is that it's the only physical process ever proposed that can explain the emergence of complex design. The reason you have eyes that are uncannily designed for vision is that they're at the end of a long series of replicators, such that the better the eyes worked, the more likely the design would have made it into the next generation.

One can distinguish between the eye, which all biologists agree is the product of natural selection, and features like the whiteness of bones or the S-shape of our spine, which aren't complex gadgets or seemingly engineered systems or low- probability arrangements of matter. We don't have to invent some scenario in which animals were selected by the whiteness of their bones. There, a by product explanation rather than adaptation is perfectly plausible.

That's the test. Apply it now to language. What we've discovered in recent studies of language is that it, too, is an improbably complex biological system. It's improbable in the sense that it's found only in one species, and improbable also in the sense that most of the things you do to a brain will disrupt the ability to use language. Moreover, like a watch or an eye, it has many finely meshing parts. There is the mental dictionary, which in a typical high-school graduate contains about sixty thousand words. There are the unconscious rules of syntax, which allow us to put words together into sentences. There are the rules of morphology, which allow us to combine bits of words, like prefixes and suffixes and stems, into words. There are the rules and processes of phonology, which massage sequences of words into a pronounceable sound pattern — what we informally call an accent. There are the mechanisms of speech production, including the shape and placement of the tongue and the larynx, which seem to have been built for speech production at the expense of another biological function, like being able to breathe while you're swallowing — which other mammals can do. There's speech perception, in which the ear can decode speech at the rate of between 15 and 45 sound units per second, faster than it can decode any other kind of signal. This is almost a miracle, because at a frequency of about 20 units per second sound merges into a low pitched buzz, so the mouth and the ear are doing a kind of multiplexing, or information compressing and unpacking. And there is the ability of a child to learn all this in a very short period of time.

These facts suggest that the anatomy of language is complex, like the anatomy of the eye. Moreover, language is quite clearly adaptive, in the sense of inherently serving the goals of reproduction. All societies use language for patently useful things like sharing technology and inventions. Language is a major means by which people share what they have learned about the local environment. Also, social relations in the human species are largely mediated by language. We rise to power, manipulate people, find mates, keep mates, win friends and influence people by language. Moreover we, and every human society, value people who are articulate and persuasive, which certainly sets up pressures for better language.

Those two lines of evidence suggest that language meets the criteria for an adaptation and a product of natural selection. We can also test the alternative — that there's some way in which language could have arisen through another route, just as whiteness comes from making bones out of calcium. Chomsky, and many anthropologists, have speculated that a big brain was sufficient to give us language. We can test that idea, because there are people with small brains. There are dwarfs, and there is normal variation within the human species, and it's certainly not the case that people with smaller brains have more trouble with language. There are some syndromes of dwarfism where the brain is not much bigger than that of a chimpanzee. Those people are retarded, but nonetheless they have language.

Brain shape is another possibility that we can rule out as the ultimate source of language. Could it be that a generally spherical brain with a certain kind of neuron packing, through complex laws of physics we don't understand, somehow gives rise to language? Again, over the range of normal variation and of pathology, there are reports of grotesquely distorted brains, usually from hydrocephalus, sometimes cases in which the brain lines the inside of the skull like the flesh of a coconut. It's possible for a person to have that condition and nonetheless develop language on schedule. One reported case was an undergraduate student at Oxford.

If we applied those criteria to any organ we weren't as fond of — and hence likely to have strong preconceptions about — as language, we'd come to the same conclusion that we do for the eye: namely, that it's a product of natural selection.

What about the rest of the mind? In this century, starting in the 1920s, there has been a pervasive, enormous intellectual movement that treated the human mind as a general-purpose learning device and attributed its complexity to the surrounding culture. There's an obvious political motivation for this idea, in that it was a reaction to some of the racist doctrines of the nineteenth century; it seems consonant with ideals of human equality and perfectibility. One can take any infant and make him or her into anything, given the right society. People who take issue with this view have often been tarred with the epithet "biological determinist" — someone who, according to the stereotype, believes that women are biologically designed for child rearing, say, or that the poor are biologically inferior. This is a specter that hovers in the background of these discussions; both in the academy and in polite intellectual discourse, the politically correct position is that the mind is a lump of wax or a blank slate.

Carl Degler, in his book on the history of Darwinism in the social sciences, traces this credo back to two sources in the academy. One is anthropology, which contributed the idea that human cultures can vary freely and without limit and that one can therefore say nothing definitive about the human species, because somewhere there will be a tribe that demonstrates the opposite. The other is psychology, which contributed the idea of the general all-purpose learning mechanism. But both ideas have now been discredited.

The impression from anthropology that humanity is a carnival where anything is possible came in part from a tourist mentality: when you come back from a trip, you remember what was different about where you went, otherwise you might as well have stayed at home. That is, many anthropologists exaggerated the degree to which the tribes they studied were exotic and strange, both to justify their profession and to raise people's consciousness about human potential. But many of their claims have turned out either to be canards, like Margaret Mead's claims about Samoa, or to miss the forest for the trees: the anthropologists spent so much time looking for differences that they didn't notice basic categories of human experience that are found in every culture, like humor, love, jealousy, and a sense of responsibility. Language is simply the most famous example of a human universal. Donald Brown, an anthropologist at UC Santa Barbara, wrote a book called Human Universals, in which he scoured the archives of ethnography for well substantiated human universals. He came up with a list of about a hundred and fifty, covering every sphere of human experience. That's my interpretation of the main lessons of anthropology. The interesting discoveries aren't about this kinship system or that form of shamanry. Underneath it all — just as, in the case of language, there's a universal design Chomsky called universal grammar — there is in the rest of culture what Donald Brown calls the universal people. He characterized the human species much the way a biologist would characterize any other species.

There has also been disillusionment with the idea that came from psychology and the study of learning — including the attempt to engineer artificial intelligence — that there's a magical learning mechanism that can acquire anything. It's an idea that sounds plausible, until you start to build one.

The main discovery of cognitive science and artificial intelligence is that ordinary people are apt to be blasé about abilities that are, upon closer examination, remarkable engineering feats, like seeing in color, picking up a pencil, walking, talking, recognizing a face, and reasoning in ordinary conversation. These are fantastically complex tasks that require their own special kinds of software. When one builds a learning system, one doesn't build a system that can learn anything; one has to build a system that can learn something very special, like a system that learns large territories, a system that learns grammar, a system that learns plant and animal species, or a system that learns particular kinds of social interactions. The only way a brain could possibly work is to have this large set of learning mechanisms, tailored to specific aspects of knowledge and experience. A general-purpose learning device is like a general-purpose tool: rather than a box full of hammers, screw drivers, and saws, one would have a single tool that does everything. That possibility is inconceivable in hardware engineering and equally inconceivable in the mental software engineering we call psychology.

If language is innate, then how much else is? Is carburetor repair innate? Is innateness a slippery slope? Of course not! The idea of a general-purpose learning device in an otherwise blank mind is so deeply entrenched that for many people it is inconceivable that there could be anything other than the two extremes: at one end, nothing is innate; on the other end, even the ability to repair carburetors is innate.

But research in psychology, linguistics, and AI have shown that there can be an interesting intermediate position. All the wonderful complex things that people do — repairing carburetors, following soap-opera plots, finding cures for diseases — might come out of the interactions among a smaller number of basic modules. The mind might have, among other things, the following: a system for intuitive mechanics — that is, our understanding of how physical objects behave, how things fall, and so forth; an intuitive biology — that is, expectations about how plants and animals work; a sense of number, the basis of mathematics and arithmetic; mental maps, the knowledge of large territories; a habitat-selection module, recognizing the kinds of environments we feel comfortable in; a sense of danger, including the emotion of fear and a set of phobias all humans have, like fear of heights and of venomous and predatory animals; intuitions about food, about contamination, about disease and spoilage and what is icky and disgusting. Monitoring of current well being: is my life going right? Is it all O.K., or should I change something? An intuitive psychology — that is, an ability to predict people's behavior from knowledge about their beliefs and desires (which, incidentally, seems to be the module that is defective in autism). A mental Rolodex, in which we store knowledge of other people and their talents and abilities. The self concept: our knowledge of ourselves and how to package our identity for others. A sense of justice, rights, obligations. A sense of kinship, including the tendency towards nepotism. A system concerned with mating, including sexual attraction, love, and feelings of fidelity and desertion.

So, with regard to the question "Why should we care so much about language?," one answer might be that language is a human intellectual instinct, and there might be many more.


Back to Contents

Excerpted from The Third Culture: Beyond the Scientific Revolution by John Brockman (Simon & Schuster, 1995) . Copyright © 1995 by John Brockman. All rights reserved.