The Language Instinct: How the Mind Creates Language

by Steven Pinker

  The second principle allows phrases to refer not just to single things or actions in the world but to sets of players that interact with each other in a particular way, each with a specific role. For example, the sentence Sergey gave the documents to the spy is not just about any old act of giving. It choreographs three entities: Sergey (the giver), documents (the gift), and a spy (the recipient). These role-players are usually called “arguments,” which has nothing to do with bickering; it’s the term used in logic and mathematics for a participant in a relationship. A noun phrase, too, can assign roles to one or more players, as in picture of John, governor of California, and sex with Dick Cavett, each defining one role. The head and its role-players—other than the subject role, which is special—are joined together in a subphrase, smaller than an NP or a VP, that has the kind of non-mnemonic label that has made generative linguistics so uninviting, “N-bar” and “V-bar,” named after the way they are written, and :

  The third ingredient of a phrase is one or more modifiers (usually called “adjuncts”). A modifier is different from a role-player. Take the phrase The man from Illinois. Being a man from Illinois is not like being a governor of California. To be a governor, you have to be a governor of something; the Californianess plays a role in what it means for someone to be governor of California. In contrast, from Illinois is just a bit of information that we add on to help identify which man we are talking about; being from one state or another is not an inherent part of what it means to be a man. This distinction in meaning between role-players and modifiers (“arguments” and “adjuncts,” in lingo) dictates the geometry of the phrase structure tree. The role-player stays next to the head noun inside the N-bar, but the modifier goes upstairs, though still inside the NP house:

  This restriction of the geometry of phrase structure trees is not just playing with notation; it is a hypothesis about how the rules of language are set up in our brains, governing the way we talk. It dictates that if a phrase contains both a role-player and a modifier, the role-player has to be closer to the head than the modifier is—there’s no way the modifier could get between the head noun and the role-player without crossing branches in the tree (that is, sticking extraneous words in among the bits of the N-bar), which is illegal. Consider Ronald Reagan. He used to be the governor of California, but he was born in Tampico, Illinois. When he was in office, he could have been referred to as the governor of California from Illinois (role-player, then modifier). It would have sounded odd to refer to him as the governor from Illinois of California (modifier, then role-player). More pointedly, in 1964 Robert F. Kennedy’s senatorial ambitions ran up against the inconvenient fact that both Massachusetts seats were already occupied (one by his younger brother Edward). So he simply took up residence in New York and ran for the U.S. Senate from there, soon becoming the senator from New York from Massachusetts. Not the senator from Massachusetts from New York—though that does come close to the joke that Bay Staters used to tell at the time, that they lived in the only state entitled to three senators.

  Interestingly, what is true of N-bars and noun phrases is true of V-bars and verb phrases. Say that Sergey gave those documents to the spy in a hotel. The phrase to the spy is one of the role-players of the verb give—there is no such thing as giving without a getter. Therefore to the spy lives with the head verb inside the V-bar. But in a hotel is a modifier, a comment, an afterthought, and is kept outside the V-bar, in the VP. Thus the phrases are inherently ordered: we can say gave the documents to the spy in a hotel, but not gave in a hotel the documents to the spy. When a head is accompanied by just one phrase, however, that phrase can be either a role-player (inside the V-bar) or a modifier (outside the V-bar but inside the VP), and the actual order of the words is the same. Consider the following newspaper report:

  One witness told the commissioners that she had seen sexual intercourse taking place between two parked cars in front of her house.

  The aggrieved woman had a modifier interpretation in mind for between two parked cars, but twisted readers give it a role-player interpretation.

  The fourth and final component of a phrase is a special position reserved for subjects (which linguists call “SPEC,” pronounced “speck,” short for “specifier”; don’t ask). The subject is a special role-player, usually the causal agent if there is one. For example, in the verb phrase the guitarists destroy the hotel room, the phrase the guitarists is the subject; it is the causal agent of the event consisting of the hotel room being destroyed. Actually, noun phrases can have subjects too, as in the parallel NP the guitarists’ destruction of the hotel room. Here, then, is the full anatomy of a VP and of an NP:

  Now the story begins to get interesting. You must have noticed that noun phrases and verb phrases have a lot in common: (1) a head, which gives the phrase its name and determines what it is about, (2) some role-players, which are grouped with the head inside a subphrase (the N-bar or V-bar), (3) modifiers, which appear outside the N- or V-bar, and (4) a subject. The orderings inside a noun phrase and inside a verb phrase are the same: the noun comes before its role-players (the destruction of the hotel room, not the of the hotel room destruction), and the verb comes before its role-players (to destroy the hotel room, not to the hotel room destroy). The modifiers go to the right in both cases, the subject to the left. It seems as if there is a standard design to the two phrases.

  In fact, the design pops up all over the place. Take, for example, the prepositional phrase (PP) in the hotel. It has a head, the preposition in, which means something like “interior region,” and then a role, the thing whose interior region is being picked out, in this case a hotel. And the same goes for the adjective phrase (AP): in afraid of the wolf, the head adjective, afraid, occurs before its role-player, the source of the fear.

  With this common design, there is no need to write out a long list of rules to capture what is inside a speaker’s head. There may be just one pair of super-rules for the entire language, where the distinctions among nouns, verbs, prepositions, and adjectives are collapsed and all four are specified with a variable like “X.” Since a phrase just inherits the properties of its head (a tall man is a kind of man), it’s redundant to call a phrase headed by a noun a “noun phrase”—we could just call it an “X phrase,” since the nounhood of the head noun, like the manhood of the head noun and all the other information in the head noun, percolates up to characterize the whole phrase. Here is what the super-rules look like (as before, focus on the summary of the rule, not the rule itself):

  XP (SPEC) YP*

  “A phrase consists of an optional subject, followed by an X-bar, followed by any number of modifiers.”

  X ZP*

  “An X-bar consists of a head word, followed by any number of role-players.”

  Just plug in noun, verb, adjective, or preposition for X, Y, and Z, and you have the actual phrase structure rules that spell the phrases. This streamlined version of phrase structure is called “the X-bar theory.”

  This general blueprint for phrases extends even farther, to other languages. In English, the head of a phrase comes before its role-players. In many languages, it is the other way around—but it is the other way around across the board, across all the kinds of phrases in the language. For example, in Japanese, the verb comes after its object, not before: they say Kenji sushi ate, not Kenji ate sushi. The preposition comes after its noun phrase: Kenji to, not to Kenji (so they are actually called “postpositions”). The adjective comes after its complement: Kenji than taller, not taller than Kenji. Even the words marking questions are flipped: they say, roughly, Kenji eat did?, not Did Kenji eat? Japanese and English are looking-glass versions of each other. And such consistency has been found in scores of languages: if a language has the verb before the object, as in English, it will also have prepositions; if it has the verb after the object, as in Japanese, it will have postpositions.

  This is a remarkable discovery. It means that the super-rules suffice not only for all phrases in Engli
sh but for all phrases in all languages, with one modification: removing the left-to-right order from each super-rule. The trees become mobiles. One of the rules would say:

  {ZP*, X}

  “An X-bar is composed of a head X and any number of role-players, in either order.”

  To get English, one appends a single bit of information saying that the order within an X-bar is “head-first.” To get Japanese, that bit of information would say that the order is “head-last.” Similarly, the other super-rule (the one for phrases) can be distilled so that left-to-right order boils away, and an ordered phrase in a particular language can be reconstituted by adding back either “X-bar-first” or “X-bar-last.” The piece of information that makes one language different from another is called a parameter.

  In fact, the super-rule is beginning to look less like an exact blueprint for a particular phrase and more like a general guideline or principle for what phrases must look like. The principle is usable only after you combine it with a language’s particular setting for the order parameter. This general conception of grammar, first proposed by Chomsky, is called the “principles and parameters” theory.

  Chomsky suggests that the unordered super-rules (principles) are universal and innate, and that when children learn a particular language, they do not have to learn a long list of rules, because they were born knowing the super-rules. All they have to learn is whether their particular language has the parameter value head-first, as in English, or head-last, as in Japanese. They can do that merely by noticing whether a verb comes before or after its object in any sentence in their parents’ speech. If the verb comes before the object, as in Eat your spinach!, the child concludes that the language is head-first; if it comes after, as in Your spinach eat!, the child concludes that the language is head-last. Huge chunks of grammar are then available to the child, all at once, as if the child were merely flipping a switch to one of two possible positions. If this theory of language learning is true, it would help solve the mystery of how children’s grammar explodes into adultlike complexity in so short a time. They are not acquiring dozens or hundreds of rules; they are just setting a few mental switches.

  The principles and parameters of phrase structure specify only what kinds of ingredients may go into a phrase in what order. They do not spell out any particular phrase. Left to themselves, they would run amok and produce all kinds of mischief. Take a look at the following sentences, which all conform to the principles or super-rules. The ones I have marked with an asterisk do not sound right.

  Melvin dined.

  * Melvin dined the pizza.

  Melvin devoured the pizza.

  *Melvin devoured.

  Melvin put the car in the garage.

  * Melvin put.

  * Melvin put the car.

  * Melvin put in the garage.

  Sheila alleged that Bill is a liar.

  * Sheila alleged the claim.

  * Sheila alleged.

  It must be the verb’s fault. Some verbs, like dine, refuse to appear in the company of a direct object noun phrase. Others, like devour, won’t appear without one. This is true even though dine and devour are very close in meaning, both being ways of eating. You may dimly recall from grammar lessons that verbs like dine are called “intransitive” and verbs like devour are called “transitive.” But verbs come in many flavors, not just these two. The verb put is not content unless it has both an object NP (the car) and a prepositional phrase (in the garage). The verb allege requires an embedded sentence (that Bill is a liar) and nothing else.

  Within a phrase, then, the verb is a little despot, dictating which of the slots made available by the super-rules are to be filled. These demands are stored in the verb’s entry in the mental dictionary, more or less as follows:

  dine:

  verb

  means “to eat a meal in a refined setting”

  eater = subject

  devour:

  verb

  means “to eat something ravenously”

  eater = subject

  thing eaten = object

  put:

  verb

  means “to cause something to go to some place”

  putter = subject

  thing put = object

  place = prepositional object

  allege:

  verb

  means “to declare without proof”

  declarer = subject

  declaration = complement sentence

  Each of these entries lists a definition (in mentalese) of some kind of event, followed by the players that have roles in the event. The entry indicates how each role-player may be plugged into the sentence—as a subject, an object, a prepositional object, an embedded sentence, and so on. For a sentence to feel grammatical, the verb’s demands must be satisfied. Melvin devoured is bad because devour’s desire for a “thing eaten” role is left unfulfilled. Melvin dined the pizza is bad because dine didn’t order pizza or any other object.

  Because verbs have the power to dictate how a sentence conveys who did what to whom, one cannot sort out the roles in a sentence without looking up the verb. That is why your grammar teacher got it wrong when she told you that the subject of the sentence is the “doer of the action.” The subject of the sentence is often the doer, but only when the verb says so; the verb can also assign it other roles:

  The big bad wolf frightened the three little pigs. [The subject is doing the frightening.]

  The three little pigs feared the big bad wolf. [The subject is being frightened.]

  My true love gave me a partridge in a pear tree. [The subject is doing the giving.]

  I received a partridge in a pear tree from my true love. [The subject is being given to.]

  Dr. Nussbaum performed plastic surgery. [The subject is operating on someone.]

  Cheryl underwent plastic surgery. [The subject is being operated on.]

  In fact, many verbs have two distinct entries, each casting a different set of roles. This can give rise to a common kind of ambiguity, as in the old joke: “Call me a taxi.” “OK, you’re a taxi.” In one of the Harlem Globetrotters’ routines, the referee tells Meadowlark Lemon to shoot the ball. Lemon points his finger at the ball and shouts, “Bang!” The comedian Dick Gregory tells of walking up to a lunch counter in Mississippi during the days of racial segregation. The waitress said to him, “We don’t serve colored people.” “That’s fine,” he replied, “I don’t eat colored people. I’d like a piece of chicken.”

  So how do we actually distinguish Man bites dog from Dog bites man? The dictionary entry for bite says “The biter is the subject; the bitten thing is the object.” But how do we find subjects and objects in the tree? Grammar puts little tags on the noun phrases that can be matched up with the roles laid out in a verb’s dictionary entry. These tags are called cases. In many languages, cases appear as prefixes or suffixes on the nouns. For example, in Latin, the nouns for man and dog, homo and canis, change their endings depending on who is biting whom:

  Canis hominem mordet. [not news]

  Homo canem mordet. [news]

  Julius Caesar knew who bit whom because the noun corresponding to the bitee appeared with -em at the end. Indeed, this allowed Caesar to find the biter and bitee even when the order of the two was flipped, which Latin allows: Hominem canis mordet means the same thing as Canis hominem mordet, and Canem homo mordet means the same thing as Homo canem mordet. Thanks to case markers, verbs’ dictionary entries can be relieved of the duty of keeping track of where their role-players actually appear in the sentence. A verb need only indicate that, say, the doer is a subject; whether the subject is in first or third or fourth position in the sentence is up to the rest of the grammar, and the interpretation is the same. Indeed, in what are called “scrambling” languages, case markers are exploited even further: the article, adjective, and noun inside a phrase are each tagged with a particular case marker, and the speaker can scramble the words of the phrase all over the sentence (say, put the a
djective at the end for emphasis), knowing that the listener can mentally join them back up. This process, called agreement or concord, is a second engineering solution (aside from phrase structure itself) to the problem of encoding a tangle of interconnected thoughts into strings of words that appear one after the other.

  Centuries ago, English, like Latin, had suffixes that marked case overtly. But the suffixes have all eroded, and overt case survives only in the personal pronouns—I, he, she, we, they are used for the subject role; my, his, her, our, their are used for the possessor role; me, him, her, us, them are used for all other roles. (The who/whom distinction could be added to this list, but it is on the way out; in the United States, whom is used consistently only by careful writers and pretentious speakers.) Interestingly, since we all know to say He saw us but never Him saw we, the syntax of case must still be alive and well in English. Though nouns appear physically unchanged no matter what role they play, they are tagged with silent cases. Alice realized this after spotting a mouse swimming nearby in her pool of tears: