Steven Pinker
Steven Pinker; drawing by David Levine

The Stuff of Thought is Steven Pinker’s fifth popular book in thirteen years, and by now we know what to expect. It is long, packed with information, clear, witty, attractively written, and generally persuasive. The topic, as earlier, is language and the mind—specifically, how language reflects human psychological nature. What can we learn about the mind by examining, with the help of linguistics and experimental psychology, the language we use to express ourselves?

Pinker ranges widely, from the verb system of English, to the idea of an innate language of thought, to metaphor, to naming, obscenity, and politeness. He is unfailingly engaging to read, with his aptly chosen cartoons, his amusing examples, and his bracing theoretical rigor. Yet there are signs of fatigue, not so much in the energy and enthusiasm he has put into the book as in the sometimes less than satisfying quality of the underlying ideas. I don’t blame the author for this: it is very hard to write anything deep, surprising, and true in psychology—especially when it comes to the most interesting aspects of our nature (such as our use of metaphor). A popular book on biology or physics will reliably deli-ver well-grounded information about things you don’t already know; in psychology the risk of banality dressed up as science is far greater. Sometimes in Pinker’s book the ratio of solid ideas to sparkling formulations is uncomfortably low (I found this particularly in the lively and amusing chapter on obscenity). He has decided to be ambitious, and there is no doubt of his ability to keep the show on the road, but it is possible to finish a long chapter of The Stuff of Thought and wonder what you have really learned—enjoyable as the experience of reading it may have been.

To my mind, by far the most interesting chapter of the book is the lengthy discussion of verbs—which may well appear the driest to some readers. Verbs are the linguistic keyhole to the mind’s secrets, it turns out. When children learn verbs they are confronted with a problem of induction: Can the syntactic rules that govern one verb be projected to another verb that has a similar meaning? Suppose you have already learned how to use the verb “load” in various syntactic combinations; you know that you can say both Hal loaded the wagon with hay and Hal loaded hay into the wagon. Linguists call the first kind of sentence a “container locative” and the second a “content locative,” because of the way they focus attention on certain aspects of the event reported—the wagon (container) or the hay (content), respectively (the word “locative” referring here to the way words express location). The two sentences seem very close in meaning, and the verb load slots naturally into the sentence frame surrounding it. So, can other verbs like fill and pour enter into the same combinations? The child learning English verbs might well suppose that they can, thus instantiating a rule of grammar that licenses certain syntactic transformations—to the effect that you can always rewrite a content locative as a container locative and vice versa. But if we look at how pour and fill actually work we quickly see that they violate any such rule. You can say John poured water into the glass (content locative) but you can’t say John poured the glass with water (container locative); whereas you can say John filled the glass with water (container locative) but you can’t say John filled water into the glass (content locative).

Somehow a child has to learn these syntactic facts about the verbs load, pour, and fill—and the rules governing them are very different. Why does one verb figure in one kind of construction but not in another? They all look like verbs that specify the movement of a type of stuff into a type of container, and yet they behave differently with respect to the syntactic structures in question. It’s puzzling.

The answer Pinker favors to this and similar puzzles is that the different verbs subtly vary in the way they construe the event they report: pour focuses on the type of movement that is involved in the transfer of the stuff, while neglecting the end result; fill by contrast specifies the final state and omits to say how that state precisely came about (and it might not have been by pouring). But load tells you both things: the type of movement and what it led to. Hence the verbs combine differently with constructions that focus on the state of the container and constructions that focus on the manner by which the container was affected.

The syntactic rules that control the verbs are thus sensitive to the precise meaning of the specific verb and how it depicts a certain event. And this means that someone who understands these verbs must tacitly grasp how this meaning plays out in the construction of sentences; thus the child has to pick up on just such subtle differences of meaning if she is to infer the right syntactic rule for the verb in question. Not consciously, of course; her brain must perform this work below the level of conscious awareness. She must implicitly analyze the verb—exposing its deep semantic structure. Moreover, these verbs form natural families, united by the way they conceive of actions—whether by their manner or by their end result. In the same class as pour, for example, we have dribble, drip, funnel, slosh, spill, and spoon.


This kind of example—and there is a considerable range of them—leads Pinker to a general hypothesis about the verb system of English (as well as other languages): the speaker must possess a language of thought that represents the world according to basic abstract categories like space, time, substance, and motion, and these categories constitute the meaning of the verb. When we use a particular verb in a sentence, we bring to bear this abstract system to “frame” reality in certain ways, thus imposing an optional grid on the flux of experience. We observe some liquid moving into a container and we describe it either as an act of pouring or as the state of being filled: a single event is construed in different ways, each reflecting the aspect we choose to focus on. None of this is conscious or explicit; indeed, it took linguists a long time to figure out why some verbs work one way and some another (Pinker credits the MIT linguists Malka Rappaport Hovav and Beth Levin). We are born with an implicit set of innate categories that organize events according to a kind of primitive physics, dealing with substance, motion, causality, and purpose, and we combine these to generate a meaning for a particular verb that we understand. The grammar of our language reflects this innate system of concepts.

As Pinker is aware, this is a very Kantian picture of human cognition. Kant regarded the mind as innately stocked with the basic concepts that make up Newtonian mechanics—though he didn’t reach that conclusion from a consideration of the syntax of verbs. And the view is not in itself terribly surprising: many philosophers have observed that the human conceptual scheme is essentially a matter of substances in space and time, causally interacting, moving and changing, obeying laws and subject to forces—with some of those substances being agents—i.e., conscious, acting human beings—with intentions and desires. What else might compose it? Here is a case where the conclusion reached by the dedicated psycholinguist is perhaps less revolutionary than he would like to think. The chief interest of Pinker’s discussion is the kind of evidence he adduces to justify such a hypothesis, rather than the hypothesis itself—evidence leading from syntax to cosmology, we might say. Of course the mind must stock basic concepts for the general structure of the universe if it is to grasp the nature of particular things within it; but it is still striking to learn that this intuitive physics shapes the very syntax of our language.

Not that everyone will agree with the general hypothesis itself—and Pinker has a whole chapter on innateness and the language of thought. Here he steers deftly between the extreme nativism of Jerry Fodor, according to which virtually every concept is innate, including trombone and opera (despite the fact that the concepts must therefore have preceded the invention of what they denote, being merely triggered into consciousness by experience of trombones and operas), and the kind of pragmatism that refuses to assign a fixed meaning to any word. Pinker sees that something conceptual has to be innate if language learning is to be possible at all, but he doesn’t believe it can be anything parochial and specific; so he concludes that only the most general categories of the world are present in the genes—the categories that any human being (or animal) needs to use if he or she is to survive at all. Among such categories, for example, are: event, thing, path, place, manner, acting, going, having, animate, rigid, flexible, past, present and future, causality, enabling and preventing, means and ends.

The picture then is that these innate abstract concepts mesh with the individual’s experience to yield the specific conceptual scheme that eventually flowers in the mind. The innate concepts pre-date language acquisition and make it possible; they are not the products of language. Thus Pinker rejects the doctrine of “linguistic determinism,” which holds that thought is nothing other than the result of the language we happen to speak—as in the infamous hypothesis of the linguists Benjamin Whorf and Harold Sapir that our thoughts are puppets of our words (as with the Eskimos who use many different words for snow). The point Pinker makes here—and it is a good one—is that we mustn’t mistake correlation for causation, assuming that because concepts and words go together the latter are the causes of the former. Indeed, it is far more plausible to suppose that our language is caused by our thoughts—that we can only introduce words for which we already have concepts. Words express concepts; they don’t create them.


Let’s suppose, then, that Pinker and others are right to credit the mind with an original system of basic physical concepts, supplemented with some concepts for number, agency, logic, and the like. We innately conceive of the world as containing what he calls “force dynamics”—substances moving through space, under forces, and impinging on other objects, changing their state. How do we get from this to the full panoply of human thought? How do we get to science, art, politics, economics, ethics, and so on? His answer is that we do it by judicious use of metaphor and the combinatorial power of language, as when words combine to produce the unlimited expressions of a human language. Language has infinite potential, because of its ability to combine words and phrases into sentences without limit: this is by now a well-worn point.

More controversial is the suggestion that metaphor is the way we transcend the merely mechanical—the bridge by which physics leads us to more abstract domains. Pinker notes, as many have before, that we routinely use spatial expressions to describe time (“he moved the meeting to Tuesday,” “don’t look backward“), as well as employ words like rise, fall, went, and send to capture events that are not literally spatial (prices rising, messages sent, and so on). Science itself is often powered by analogies, as when heat was conceived as a fluid and its laws derived accordingly. Our language is transparently shot through with meta-phors of one kind or another. But it is far from clear that everything we do with concepts and language can be accounted for in this way; consider how we think and talk about consciousness and the mind, or our moral thinking. The concept of pain, say, is not explicable as a metaphorical variation on some sort of physical concept.

It just doesn’t seem true that everything nonphysical that we think about is metaphorical; for example, our legal concepts such as “rights” are surely not all mere metaphors, introduced on the shoulders of the concepts of intuitive physics. So there is a question how Pinker’s alleged language of thought, restricted as it is, can suffice to generate our total conceptual scheme; in which case we will need to count more concepts as innate (what about contract or punishment?)—or else rethink the whole innateness question. Not that I have any good suggestions about how human concepts come to be; my point is just that Pinker’s set of basic Kantian concepts seems too exiguous to do the job.

If the Kantian categories are supposed to make thought and language possible, then they also, for Pinker, impose limits on our mental functioning. This is a second main theme of his book: the human mind, for all its rich innate endowment, is fallible, prone to confusion, easily foiled. The very concepts that enable us to think coherently about the world can lead us astray when we try to extend them beyond their natural domain. Pinker discusses the concepts of space and time, exposing the paradoxes that result from asking whether these are finite or infinite; either way, human thought reels. As he says, we can’t think without these concepts, but we can’t make sense of them—not when we start to think hard about what they involve. For example, if space is bounded, what lies on the other side of the boundary? But if it’s not bounded, we seem saddled with an infinite amount of matter—which implies multiple identical universes.

The concept of free will poses similar paradoxes: either human choices are caused or they are not, but either way we can’t seem to make sense of free will. A lot of philosophy is like that; a familiar concept we use all the time turns puzzling and paradoxical once we try to make systematic sense of it. Pinker has fun detailing the natural errors to which the human mind is prone when trying to reason statistically or economically; human specimens are notoriously poor at reasoning in these matters. Even more mortifying, our prized intuitive physics, foundation of all our thought, is pretty bad as physics: projectiles don’t need impetus to keep them in steady motion, no matter what Aristotle and common sense may say. As Newton taught us, motion, once it begins, is preserved without the pressure of a continuously applied force—as when a meteor keeps moving in a straight line, though no force maintains this motion. And relativity and quantum theory violate commonsense physics at every turn.

Our natural concepts are as much a hindrance to thought as they are a springboard for it. When we try to turn our minds away from their primitive biological tasks toward modern science and industrial-electronic society we struggle and fall into fallacies; it’s an uphill battle to keep our concepts on track. Our innate “common sense” is riddled with error and confusion—not all of it harmless (as with the economically naive ideas about what constitutes a “fair price”).

Pinker also has three bulky chap-ters on the social aspects of language, dealing with naming and linguistic innovation in general, with obscenity and taboo words, and with politeness and authority relations in speech. The chapter on naming achieves something I thought was impossible: it gives an accurate exposition of the philosopher Saul Kripke’s classic discussion of proper names by a nonphilosopher—the gist of which is that the reference of a name is fixed not by the descriptive information in the mind of the speaker but by a chain of uses stretching back to an initial identification. For example, I refer to a certain Greek philosopher with the name “Plato” in virtue of the chain of uses that link my present use with that of ancient Greeks who knew him, not in virtue of having in my mind some description that picks him out uniquely from every other Greek philosopher.

Apart from this, Pinker worries at the question of fashions in names and how they change. He refutes such popular theories as that names are taken from public figures or celebrities; usually, the trend is already in place—and anyway the name “Humphrey” never took off, despite the star of Casablanca. It is fascinating to read that in the early part of the twentieth century the following names were reserved primarily for men: Beverly, Dana, Evelyn, Gail, Leslie, Meredith, Robin, and Shirley. But not much emerges about why names change as they do, besides some platitudes about the need for elites to stand out by adopting fashions different from the common herd.

I very much enjoyed the chapter on obscenity, which asks the difficult question of how words deemed taboo differ from their inoffensive syn-onyms (e.g., shit and feces). It can’t obviously be the referent of the term, since that is the same, and it isn’t merely that the taboo words are more accurately descriptive (excre-ment is equally accurate, but it isn’t taboo). Pinker reports, no doubt correctly, that swearing forces the hearer to entertain thoughts he’d rather not, but that too fails to distinguish taboo words from their nontaboo synonyms. The phenomenon is especially puzzling when we note that words can vary over time in their taboo value: damn used to be unutterable in polite society, while cunt was once quite inoffensive (Pinker reports a fifteenth-century medical textbook that reads “in women the neck of the bladder is short, and is made fast to the cunt”).

Of particular interest to the grammarian is the fact that in English all the impolite words for the sexual act are transitive verbs, while all the polite forms involve intransitive verbs: fuck, screw, hump, shag, bang versus have sex, make love, sleep together, go to bed, copulate. As Pinker astutely observes, the transitive sexual verbs, like other verbs in English, bluntly connote the nature of the motion involved in the reported action with an agent and a receiver of that motion, whereas the intransitive forms are discreetly silent about exactly how the engaged objects move in space. The physical forcefulness of the act is thus underlined in the transitive forms but not in the intransitive ones. None of this explains why some verbs for intercourse are offensive while others are not, but it’s surely significant that different physical images are conjured up by the different sexual locutions—with fuck semantically and syntactically like stab and have sex like have lunch.

Pinker’s discussion of politeness verges closest to platitude—noting, for example, that bribes cannot usually afford to be overt and that authority relations are sometimes encoded in speech acts, as with tu and vous in French. Here he relies heavily on lively examples and pop culture references, but the ideas at play are thin and rather forced. But, as I say, he has a tough assignment here—trying to extract theoretical substance from something both familiar and unsystematic. Laying out a game theory matrix, with its rows and columns of payoffs, for a potential bribe to a traffic cop adds little to the obvious description of such a situation.

The book returns to its core themes in the final chapter, “Escaping the Cave.” Pinker sums up:

Human characterizations of reality are built out of a recognizable inventory of thoughts. The inventory begins with some basic units, like events, states, things, substances, places, and goals. It specifies the basic ways in which these units can do things: acting, going, changing, being, having. One event may be seen as impinging on another, by causing or enabling or preventing it. An action can be initiated with a goal in mind, in particular, the destination of a motion (as in loading hay) or the state resulting from a change (as in loading a wagon). Objects are differentiated by whether they are human or nonhuman, animate or inanimate, solid or aggregate, and how they are laid out along the three dimensions of space. Events are conceived as taking up stretches of time and as being ordered with respect to one another.

If that strikes you as a bit platitudinous, then such is the lot of much psychology—usually the good sort. What is interesting is the kind of evidence that can be given for these claims and the way they play out in language and behavior—not the content of the claims themselves.

But Pinker is also anxious to reiterate his thesis that our conceptual scheme is like Plato’s cave, in giving us only a partial and distorted vision of reality. We need to escape our natural way of seeing things, as well as appreciate its (limited) scope. Plato himself regarded a philosophical education as the only way to escape the illusions and errors of common sense—the cave in which we naturally dwell. Pinker too believes that education is necessary in order to correct and transcend our innate cognitive slant on the world. This means, unavoidably, using a part of our mind to get beyond the rest of our mind, so that there must be a part that is capable of distancing itself from the rest. He says little about how this might be possible—how that liberating part might operate—beyond what he has said about metaphors and the infinity of language. And the question is indeed difficult: How could the mind ever have the ability to step outside of itself? Aren’t we always trapped inside our given conceptual scheme? How do we bootstrap ourselves to real wisdom from the morass of innate confusion?

One reason it is hard to answer this question is that it is obscure what a concept is to start with. And here there is a real lacuna in Pinker’s book: no account is given of the nature of the basic concepts that are held to constitute the mind’s powers. He tells us at one point that the theory of conceptual semantics “proposes that word senses are mentally represented as expressions in a richer and more abstract language of thought,” as if concepts could literally be symbols in the language of thought. The idea then is that when we understand a verb like pour we translate it into a complex of symbols in the brain’s innate code (rather like the code used by a computer), mental counterparts of public words like move, cause, change. But that leaves wide open the question of how those inner words have meaning; they can’t just be bits of code, devoid of semantic content. We need to credit people with full-blown concepts at the foundation of their conceptual scheme—not just words for concepts.

Pinker has listed the types of concepts that may be supposed to lie at the foundation, but he hasn’t told us what those concepts consist in—what they are. So we don’t yet know what the stuff of thought is—only that it must have a certain form and content. Nowhere in the course of a long book on concepts does Pinker ever confront the really hard question of what a concept might be. Some theorists have supposed concepts to be mental images, others that they are capacities to discriminate objects, others dispositions to use words, others that they are mythical entities.

The problem is not just that this is a question Pinker fails to answer or even acknowledge; it is that without an answer it is difficult to see how we can make headway with questions about what our concepts do and do not permit. Is it our concepts themselves that shackle us in the cave or is it rather our interpretations of them, or maybe our associated theories of what they denote? Where exactly might a concept end and its interpretation begin? Is our concept of something identical to our conception of it—the things we believe about it? Do our concepts intrinsically blind us or is it just what we do with them in thought and speech that causes us to fail to grasp them? Concepts are the material that constitutes thought and makes language meaningful, but we are very far from understanding what kind of thing they are—and Pinker’s otherwise admirable book takes us no further with this fundamental question.

This Issue

September 27, 2007