A review of The Stuff of Thought: Language as a Window into Human Nature, by Steven Pinker
One of my favorite science books… no, wait… one of my favorite books altogether, is a shortish volume by Steven Pinker entitled Words and Rules. (I cannot remember how many copies of that book I have bought for various friends over the years, 
but I can pretty safely say that Pinker owes me a drink or two from his royalties.) I admired Pinker before I had read this book because I had already admired other books he had written. The first of these was the first book Pinker wrote for a wide audience: The Language Instinct. I read this book while I was still a very serious young student of analytical philosophy of language and mind in a Ph.D. program at Columbia University. Some of my philosophy professors didn’t like the book, but I did. Here’s why: Pinker knew a lot about the philosophical issues we were worrying about in our seminars, and he had empirically verifiable things to say about them. In fact, he had identified important and deep linguistic issues which had testable implications. And he always backed up what he said with a lot of footnotes (meaning he always cited studies to back up whatever it was he was asserting). This was very exciting and pleasing to my sciency heart. (My undergraduate degree is in Electrical Engineering and Computer Science.) What he was saying in The Language Instinct actually made predictions and retrodictions (explaining what we already know to be true from past observation is just as important in science as any soothsaying of the future) about very concrete patterns in how language is actually acquired by children, and used by adults.
In any case, the reason Words and Rules is such a favorite of mine is that in it, Pinker manages to squeeze a shocking amount of intellectual juice out of something seemingly quite dry: the nature of regular and irregular verbs (walk–walked/go–went) and regular and irregular noun plurals (kid–kids/child–children). It is truly a tour de force: one of those rare small books (like Language, Truth, and Logic by A.J. Ayer, Fact, Fiction, and Forecast by Nelson Goodman, or The Idea of a Critical Theory by Raymond Geuss) that changes how we think about something very important. But I really don’t have space here to tell you why that book is so wonderful. On the other hand, before we get to The Stuff of Thought, we can and should try to answer this: why is language and how we actually use it so important? It’s because of nothing less than this: we want to know what the meaning of life is.
I’m going to make this story very simple: In 1879 a man in Germany named Gottlob Frege wrote a paper entitled “Über Sinn und Bedeutung.” (That means “On Sense and Meaning.”) For more than two thousand years before Frege, the Western world had been worrying about all kinds of philosophical questions: What is the nature of justice? What is the nature of beauty? What is the nature of truth? And, of course: What is the meaning of life? After Frege, we (at least Anglo-American analytical philosophy) have spent the last century-and-a-quarter mostly wondering whether it makes sense to even ask such questions, and to answer that, focusing on language itself. From Bertrand Russell’s attempts to model natural languages with formal ones such as the predicate calculus, to Wittgenstein’s language games, to the verificationism of logical positivism and the Vienna Circle, to Rudolf Carnap’s confirmation theory, to Gilbert Ryle and J. L. Austin, to W.V.O. Quine, to, in more recent times, Hilary Putnam, Donald Davidson, and my own Ph.D. adviser (and Davidson’s student) Akeel Bilgrami, the struggle to elucidate the workings of language, and therefore the meaning of meaning, has been the primary focus of philosophers, as well, of course, as of linguists. Suppose for a second that we had been struggling with the question “What is the color of love?” for all that time. Wouldn’t that have been silly? Is it not obvious that to ask, “What is the color of love?” is a category mistake? Purple, after all, is not a predicate that applies to the category “love,” just as “brittle” is not a predicate that applies to something like the number 17, say. Noam Chomsky famously coined the grammatically perfect but nevertheless meaningless sentence “Colorless green ideas sleep furiously” as an illustration (partly) of this point. (And this is also the basis of Douglas Adams’ joke that the meaning of life is 42.) What if the basic questions we have been grappling with for millennia are so intractable precisely because they are nonsensical? (I say all this by way of motivating the minute attention to details of language that is soon to absorb us.)
Things become especially interesting when we come to the predicate “true.” What does that apply to? Clearly not to words, as it seems obviously stupid to ask if “cat” is true or not. Clearly it also does not apply to very long collections of words, as it seems equally nonsensical to ask whether “Hamlet” is true. So, what does “true” apply to, properly? Basically: propositions, or more loosely, sentences. Something like “Snow is white” can actually be true or false. It happens, in this case, to be true. And it is truth which connects philosophy through language to science, because science is concerned with representations of the world which are true. Not beautiful, not good, true. So a map can be “true” to the degree that it correctly represents a given terrain. Similarly, “Snow is white” is a bona fide scientific statement. It is a representation in language of a state of affairs in the world. But we represent reality in our minds in other ways besides language and those representations are not all available to our conscious selves by simple introspection. What Steven Pinker is out to do in The Stuff of Thought is to tease out what our patterns of language use can tell us about how we think and the very nature of our minds. This linguistic approach to cognitive science turns out to be very fertile indeed, and combines and connects the subjects of Pinker’s previous books on language, which I have already mentioned, with some of the ideas expressed in his How the Mind Works. In fact, if it weren’t so unwieldy, the title What the Language Instinct and Words and Rules Tell Us About How the Mind Works could have been pressed into service.
Rather than make a futile attempt at summarizing 439 idea-crammed pages, what I’d like to try to do here is give you a flavor of the kinds of things the book is about by briefly explaining one of the many fascinating stories that Pinker tells about language and what it entails for “conceptual semantics”–the concepts and schemes that we use to think–indeed, the language of thought itself. Let’s jump right in: we begin by considering what one of Pinker’s colleagues once jokingly referred to as one of Pinker’s “little friends”: the verb “to load”. Take a sentence like Hal loaded hay into the wagon. [All linguistic examples used in this review are Pinker’s own.] This is what linguists call a content-locative construction because it is the contents being moved that are the object of the sentence. Notice that this sentence is indistinguishable in meaning from Hal loaded the wagon with hay. This latter sentence is known as a container-locative construction, since it is the container which is the object here. One can do also perform this operation (call it the locative rule) with other transitive verbs:
Jared sprayed water on the roses.
Jared sprayed the roses with water.Betsy splashed paint on the wall.
Betsy splashed the wall with paint.Jeremy rubbed oil into the wood.
Jeremy rubbed the wood with oil.
The mind of a child might absorb such a pattern (linguists call it an alternation) as a generalization. So now, if you heard someone say brush paint onto the fence you might guess that brush the fence with paint is also fine. So far so good. But now consider a different sentence: Hal poured water into the glass. It cannot be transformed in a similar manner: Hal poured the glass with water sounds immediately wrong to a normal speaker of English. Similarly, problems arise in the other direction with other verbs like fill: while the container-locative construction Bobby filled the glass with water is fine, the content-locative Bobby filled water into the glass is not grammatical English. Why? As Pinker puts it, “How do children succeed in acquiring an infinite language when the rules they are tempted to postulate just get them into trouble by generating constructions that other speakers choke on? How do they figure out that certain verbs can’t appear in perfectly good constructions?” (p. 37)
Pinker now considers and rejects three possibilities: First, maybe we have over-generalized the rule. Maybe verbs have some trait that children can sense that indicates that they resist this alternation. But if such a trait exists, it is not very obvious what it could be since load, pour, and fill are all ways of moving something to another place, but pour only allows the content-locative (pour water), fill only allows the container locative (fill the glass), and load allows both (load the hay, load the wagon).
Second, it might be that children simply memorize which constructions are allowed for which verbs, one at a time. This is unlikely because children have to master an infinite language and only have a very limited set of samples to learn from. Also consider that when new words (or new senses of words) enter the language, such as burning songs onto a CD, no one has trouble generalizing to the container-locative burning a CD with songs. Indeed children do generalize to the container locative even when they could not have heard the usage from their parents. Many examples can be found in children’s speech which has been recorded by psychologists, such as “I hitted this into my neck.”
The third possibility is that children do make generalizations, but are corrected by their parents (or others) when the generalization leads to a construction which, like “I hitted this into my neck”, is not allowed. Well, even attempts to show that parents react differently to their children’s deviant sentences, much less correct them, have not come up with anything. And there is a bigger problem: Even if parents were trying their best to always correct their children, this would not be enough to explain the strong intuitions people have about what verbs can and can’t do: “People sense that they would never say They festooned ribbons onto the stage or She siphoned the bottle with gasoline, yet word-frequency counts show that these verbs are literally one in a million. It is unlikely that every English speaker uttered each of the obdurate verbs in each of the offending constructions at some point in childhood (or, for that matter, adulthood), was corrected, and now finds the usage strange on account of that episode.” (p. 40)
So where does that leave us? Pinker lists four apparent facts that can’t be all true at the same time:
- people generalize
- they avoid some exceptions
- the exceptions are unpredictable
- children don’t get corrected for every mistake
One of these, at least, must be false, and indeed when we examine them carefully, the one that seems weakest is that the exceptions are not predictable. What if they are somehow predictable? “Often a linguistic pattern that seems haphazard turns out to have a stipulation that divides the sheep from the goats. For example, the mystery of why you can’t apply —er and —est to certain adjectives, as in specialer and beautifullest, was solved when someone noticed that the suffixes apply only to words that are monosyllabic (redder, nicer, older) or have at most an insubstantial second syllable (prettier, simpler, narrower). Perhaps there is also a subtle criterion that distinguishes the verbs enlisted into the locative construction from the draft dodgers.” (p. 42)
The breakthrough came in a paper by Malka Rappaport Hovav and Beth Levin who realized that it is not just a Chomskian matter of cutting and pasting phrases, such as moving a prepositional phrase leftward into the position of a direct object (in the case of changing a content-locative into a container-locative construction) or moving the direct object rightward into a prepositional phrase (in changing from container-locative to content-locative construction), with the meanings left indistinguishable. It is something more abstract: the rule actually transforms the mental framing of events that goes into a construction. Pinker explains:
Imagine that the meaning of the content-locative construction is “A causes B to go to C,” but the meaning of the container-locative construction is “A causes C to change state (by means of causing B to go to C).” In other words, loading hay onto the wagon is something you do to hay (namely, cause it to go to the wagon), whereas loading the wagon with hay is something you do to the wagon (namely, cause it to become loaded with hay). These are two different construals of the same event, a bit like the gestalt shift in the classic face-vase illusion in which the figure and ground switch places in one’s consciousness.
In the sentences with the hay and the wagon, the flip between figure and ground is not in the mind’s eye but in the mind itself–the interpretation of what the event is really about….
When conceived as a conceptual gestalt shift, the locative rule is no longer a matter of cutting and pasting phrases in complicated ways for no particular reason. It can now be factored into two very general and useful rules:
- A rule of semantic reconstrual (the gestalt shift): If a verb means “A causes B to move to C,” it can also mean “A causes C to change state by moving B to it.”
- A rule for linking meaning to form: Express the affected entity as the direct object. (p. 44)
The really interesting bit is that this gestalt-shift theory implies that the two constructions might not be completely synonymous (they are two different construals of an event, after all), and when we think about it carefully, that is indeed the case:
When one loads hay onto a wagon, it can be any amount, even a couple of pitchforkfuls. But when one loads the wagon with hay, the implication is that the wagon is full. This subtle difference, which linguists call the holism effect, can be seen with the other locative verbs: to spray the roses with water implies that they all got sprayed (as opposed to merely spraying water onto the roses), and to stuff the turkey with breadcrumbs implies that it is completely stuffed.
The holism effect is not an arbitrary stipulation tacked onto the rule, like a pork-barrel amendment on a spending bill. It falls out of the nature of what the rule does, namely, construe the container as the thing that is affected. And that, in turn, reveals an interesting feature of the way the mind conceives what things are and how they change. The holism effect turns out not to be restricted to the locative construction; it applies to direct objects in general. For instance, the sentence Moondog drank from the glass of beer (where the glass is an oblique object of from) is consistent with his taking a few sips. But the sentence Moondog drank the glass of beer (where the glass is a direct object) implies that he chugged down the whole thing.
But the holism effect has even wider applicability. It is really not even a property of the direct object, but of the affected entity which normally happens to get expressed as a direct object. So in constructions where the entity affected is the subject, you have still constructions displaying the holism effect, such as:
Bees are swarming in the garden.
The garden is swarming with bees.
So then why is the content interpreted as a whole in these container-locative constructions? I’ll let Pinker explain again:
The reason is that English treats a changing entity (a loaded wagon, sprayed roses, a painted door) in the same way as it treats a moving entity (pitched hay, sprayed water, slopped paint). A state is conceived as a location in a space of possible states, and change is equated with moving from one location to another in that state space… (p. 47)
And also:
When the mind conceptualizes an entity in a location or in motion, it tends to ignore the internal geometry of the object and treat it as a dimensionless point or a featureless blob…. So, the figure being positioned and the place where it is said to be located are treated differently in language: the first is reduced to a dimensionless speck, whose internal geometry is ignored; the second is diagrammed, at least schematically. Take the English phrases on your hand, under your hand, and in your hand. Each picks out an aspect of the geometry of the hand, namely its top, its bottom, and a cavity it can form…. This leads us to a deeper explanation of the holism effect. In the locative alternation, when the container (such as the wagon in load hay into the wagon) gets promoted to direct object, it is also conceptually reanalyzed as something that has been moved in state-space (from the “empty” slot to the “full” slot). And in this reconstrual, it gets compacted into a single point, its internal geometry obliterated. Wagons become loaded, flowerbeds sprayed, turkeys stuffed, not as arrangements of matter in space with niches and hidey-holes that may separately accommodate bits of matter, but as entities that are, taken as a whole, now ready for carting, blooming, or cooking…. But if an object can be thought of as changing state even when it has stuff in just one part, then the container locative may be used there, too. Thus we can say that a graffiti artist has sprayed a statue with paint even if he has colored just one part of it, because a single splotch is enough for people to consider it defaced. (p. 49)
We have been discussing the holism effect to show that what we have come to realize is that the way the gestalt-shift theory of the locative explains why some verbs allow the shift while others don’t is that it establishes a relationship between the meaning of the construction and the meaning of the verb. As Pinker points out, one can throw a cat into the room, but one cannot throw the room with a cat because throwing a cat into a room cannot be construed as a way of significantly changing the state of the room. And this same kind of reasoning applies to all the other cases we have discussed. As a last example, let us return to why one can’t pour a glass with water:
Verbs that differ in their syntactic fussiness, like pour, fill, and load, all pertain to moving something somewhere, giving us the casual impression that they are birds of a feather. But on closer examination each of these verbs turns out to have a distinct kind of semantic fussiness–they differ in which aspect of the motion they care about.
Take the verb pour, and think about when you can use it. To pour means, more or less, to allow a liquid to move downward in a continuous stream. It specifies a causal relation of “letting” rather than “forcing,” and it specifies a manner of motion; these are the bits of meaning that differentiate it from other ways in which liquid moves, such as spray, splash, and spew. Since pour says something about the motion, it can be used in the construction that is about motion; hence we can say pour water into the glass. But pour doesn’t care about how or where the liquid ends up. You can pour water into a glass, all over the floor, or out the window of an airplane, dispersing it into a mist. Nothing predictable happens to the destination of a poured liquid, and so the verb is inconsistent with a construction that specifies how the state of a container has changed. And thus we can’t say she poured the glass with water. (p. 50)
Other verbs which, like pour, do not allow the locative alternation (you can’t dump a truck with iron) are: dribble, drip, drop, dump, funnel, ladle, shake, siphon, slop, slosh, spill, and spoon. On the other hand, here are some seemingly similar verbs that do allow the alternation (you can smear grease on the axle, or you can smear the axle with grease): brush, dab, daub, plaster, rub, slather, smear, smudge, spread, streak, and swab. To see why they are different, we can once again look at the physics underlying their meanings: in the first set of pour-like verbs, we let gravity do the work, while in the second set, the agent applies force to the substance and pushes it actively onto it. And the mind makes these fine distinctions when deciding whether the alternation should apply or not.
Pinker gives many more examples, and cites many experiments to confirm the theory that I do not have the space here to convey. As it is, I have distilled this brief exposition from over twenty pages of flavorful prose, peppered with interesting facts such as the one I mentioned above about why specialer and beautifullest are not proper words, and full of Pinker’s delightfully wry sense of humor which made my wife wonder why I kept laughing as I read a serious book on language and mind. I can’t resist just one of many examples:
Even the most palpable cognitive distinction–who did something, and who had something done to him–can be mentally flip-flopped, as when a hockey player shouts, “Kiss my elbow!” or when Woody Allen in Play it Again, Sam gets roughed up by some bikers and tells his friends, “I snapped my chin down on some guy’s fist and hit another on the knee with my nose.”
I hope I have managed to give some sense of the content and tone of the book. The rest of it is just as jam-packed with facts and ideas about how and what the structure of language can tell us about how our minds work, as the small part I have presented. Pinker also discusses ideas that he thinks are wrong. One of my favorites was his destruction of Jerry Fodor’s Mentalese, but he is equally effective in dismissing other interesting but ultimately fruitless ideas. It seems conventional when reviewing a book favorably to trot out a few petty criticisms to give the appearance of objectivity and balance. I shall commit no such crime and recommend the book as highly as I can recommend any book, without reservation. It ships on September 11th, but you can order it now. Buy it. And read it. You’ll find yourself educated and entertained at the same time.
Full disclosure: when the publisher sent me a review copy of the book, I was pleased to find my own name cozily nestled in the list of those thanked in the acknowledgments section, to which after reading the book I can only say: no, Steve, thank you.
All my previous Monday Musings can be seen here.
Have a good week!