Words and Rules

The New copy of this book will include any supplemental materials advertised. Please check the title of the book to determine if it should include any access cards, study guides, lab manuals, CDs, etc.

The Used, Rental and eBook copies of this book are not guaranteed to include any supplemental materials. Typically, only the book itself is included. This is true even if the title states it includes any access cards, study guides, lab manuals, CDs, etc.

Chapter One

THE INFINITE LIBRARY

Language comes so naturally to us that it is easy to forget what a strange and miraculous gift it is. All over the world members of our species fashion their breath into hisses and hums and squeaks and pops and listen to others do the same. We do this, of course, not only because we like the sounds but because details of the sounds contain information about the intentions of the person making them. We humans are fitted with a means of sharing our ideas, in all their unfathomable vastness. When we listen to speech, we can be led to think thoughts that have never been thought before and that never would have occurred to us on our own. Behold, the bush burned with fire, and the bush was not consumed. Man is born free, and everywhere he is in chains. Emma Woodhouse, handsome, clever, and rich, with a comfortable home and happy disposition, seemed to unite some of the best blessings of existence. Energy equals mass times the speed of light squared. I have found it impossible to carry the heavy burden of responsibility and to discharge my duties as King without the help and support of the woman I love.

    Language has fascinated people for thousands of years, and linguists have studied every detail, from the number of languages spoken in New Guinea to why we say razzle-dazzle instead of dazzle-razzle . Yet to me the first and deepest challenge in understanding language is accounting for its boundless expressive power. What is the trick behind our ability to fill one another's heads with so many different ideas?

    The premise of this book is that there are two tricks, words and rules. They work by different principles, are learned and used in different ways, and may even reside in different parts of the brain. Their border disputes shape and reshape languages over centuries, and make language not only a tool for communication but also a medium for wordplay and poetry and an heirloom of endless fascination.

* * *

The first trick, the word, is based on a memorized arbitrary pairing between a sound and a meaning. "What's in a name?" asks Juliet. "That which we call a rose by any other name would smell as sweet." What's in a name is that everyone in a language community tacitly agrees to use a particular sound to convey a particular idea. Although the word rose does not smell sweet or have thorns, we can use it to convey the idea of a rose because all of us have learned, at our mother's knee or in the playground, the same link between a noise and a thought. Now any of us can convey the thought by making the noise.

    The theory that words work by a conventional pairing of sound and meaning is not banal or uncontroversial. In the earliest surviving debate on linguistics, Plato has Hermogenes say, "Nothing has its name by nature, but only by usage and custom." Cratylus disagrees: "There is a correctness of name existing by nature for everything: a name is not simply that which a number of people jointly agree to call a thing." Cratylus is a creationist, and suggests that "a power greater than man assigned the first names to things." Today, those who see a correctness of names might attribute it instead to onomatopoeia (words such as crash and oink that sound like what they mean) or to sound symbolism (words such as sneer, cantankerous , and mellifluous that naturally call to mind the things they mean).

    Today this debate has been resolved in favor of Hermogenes' conventional pairing. Early in this century Ferdinand de Saussure, a founder of modern linguistics, called such pairing the arbitrary sign and made it a cornerstone of the study of language. Onomatopoeia and sound symbolism certainly exist, but they are asterisks to the far more important principle of the arbitrary sign--or else we would understand the words in every foreign language instinctively, and never need a dictionary for our own! Even the most obviously onomatopoeic words--those for animal sounds--are notoriously unpredictable, with pigs oinking boo-boo in Japan and dogs barking gong-gong in Indonesia. Sound symbolism, for its part, was no friend of the American woman in the throes of labor who overheard what struck her as the most beautiful word in the English language and named her newborn daughter Meconium , the medical term for fetal excrement.

    Though simple, the principle of the arbitrary sign is a powerful tool for getting thoughts from head to head. Children begin to learn words before their first birthday, and by their second they hoover them up at a rate of one every two hours. By the time they enter school children command 13,000 words, and then the pace picks up, because new words rain down on them from both speech and print. A typical high-school graduate knows about 60,000 words; a literate adult, perhaps twice that number. People recognize words swiftly. The meaning of a spoken word is accessed by a listener's brain in about a fifth of a second, before the speaker has finished pronouncing it. The meaning of a printed word is registered even more quickly, in about an eighth of a second. People produce words almost as rapidly: It takes the brain about a quarter of a second to find a word to name an object, and about another quarter of a second to program the mouth and tongue to pronounce it.

    The arbitrary sign works because a speaker and a listener can call on identical entries in their mental dictionaries. The speaker has a thought, makes a sound, and counts on the listener to hear the sound and recover that thought. To depict an entry in the mental dictionary we need a way of showing the entry itself, as well as its sound and meaning. The entry for a word is simply its address in one's memory, like the location of the boldfaced entry for a word in a real dictionary. It's convenient to use an English letter sequence such as r-o-s-e to stand for the entry, as long as we remember this is just a mnemonic tag that allows us to remember which word the entry corresponds to; any symbol, such as 42759, would do just as well. To depict the word's sound, we can use a phonetic notation, such as [roz]. The meaning of a word is a link to an entry in the person's mental encyclopedia , which captures the person's concept of a rose. For convenience we can symbolize it with a picture, such as [flower]. So a mental dictionary entry looks something like this:

    rose

sound: roz

meaning: [flower]

    A final component is the word's part of speech, or grammatical category, which for rose is noun (N):

    rose

sound: roz

meaning: [flower]

part of speech: N

And that brings us to the second trick behind the vast expressive power of language.

* * *

People do not just blurt out isolated words but rather combine them into phrases and sentences, in which the meaning of the combination can be inferred from the meanings of the words and the way they are arranged. We talk not merely of roses, but of the red rose, proud rose, sad rose of all my days. We can express our feelings about bread and roses, guns and roses, the War of the Roses, or days of wine and roses. We can say that lovely is the rose, roses are red, or a rose is a rose is a rose. When we combine words, their arrangement is crucial: Violets are red, roses are blue , though containing all the ingredients of the familiar verse, means something very different. We all know the difference between young women looking for husbands and husbands looking for young women , and that looking women husbands young for doesn't mean anything at all.

    Inside everyone's head there must be a code or protocol or set of rules that specifies how words may be arranged into meaningful combinations. Modern linguists call it a grammar , sometimes a generative grammar to distinguish it from the grammars used to teach foreign languages or to teach the dos and don'ts of formal prose.

    A grammar assembles words into phrases according to the words' part-of-speech categories, such as noun and verb. To highlight a word's category and reduce visual clutter often it is convenient to omit the sound and meaning and put the category label on top:

Similarly, the word a , an article or determiner , would look like this:

They can then be joined into the phrase a rose by a rule that joins a determiner to a noun to yield a noun phrase (NP). The rule can be shown as a set of connected branches; this one says "a noun phrase may be composed of a determiner followed by a noun":

The symbols at the bottom of the branches are like slots into which words may be plugged, as long as the words have the same labels growing out of their tops. Here is the result, the phrase a rose :

With just two more rules we can build a complete toy grammar. One rule defines a predicate or verb phrase (VP); the rule says that a verb phrase may consist of a verb followed by its direct object, a noun phrase:

The other rule defines the sentence itself (S). This rule says that a sentence may be composed from a noun phrase (the subject) followed by a verb phrase (the predicate):

When words are plugged into phrases according to these rules, and the phrases are plugged into bigger phrases, we get a complete sentence, such as A rose is a rose :

Other parts of the rules, not shown here, specify the meaning of the new combination. For example, the complete NP rule says that the meaning of the yellow rose of Texas is based on the meaning of rose , which is called the head of the phrase, and that the other words modify the head in various ways: yellow specifies a distinctive trait, Texas its location.

    These rules, though crude, illustrate the fantastic expressive power made available by grammar. First, the rules are productive . By specifying a string of kinds of words rather than a string of actual words, the rules allow us to assemble new sentences on the fly and not regurgitate preassembled clichés--and that allows us to convey unprecedented combinations of ideas. Though we often speak of roses being red, we could talk about violets being red if the desire came over us (perhaps to announce a new hybrid), because the rule allows us to insert violets into the N slot just as easily as roses .

    Second, the symbols contained by the rules are symbolic and hence abstract . The rule doesn't say, "A sentence may begin with a bunch of words referring to a kind of flower"; rather, it says, "A sentence may begin with an NP," where NP is a symbol or variable that can be replaced by any noun, just as x or y in a mathematical formula can be replaced by any number. We can use the rules to talk about flowers and their colors and smells, but we can just as easily use them to talk about karma or quarks or floob-boober-bab-boober-bubs (who, according to Dr. Seuss, bounce in the water like blubbery tubs).

    Third, the rules are combinatorial . They don't just have a single slot, like a fill-in-the-blank exam question; every position in the sentence offers a choice of words from a lengthy menu. Say everyday English has four determiners ( a, any, one, and the ) and ten thousand nouns. Then the rule for a noun phrase allows four choices for the determiner, followed by ten thousand choices for the head noun, yielding 4 x 10,000 = 40,000 ways to utter a noun phrase. The rule for a sentence allows these forty thousand subjects to be followed by any of four thousand verbs, providing 40,000 x 4,000 = 160,000,000 ways to utter the first three words of a sentence. Then there are four choices for the determiner of the object (640 million four-word beginnings) followed by ten thousand choices for the head noun of the object, or 640,000,000 x 10,000 = 6,400,000,000,000 (6.4 trillion) five-word sentences. Suppose it takes five seconds to produce one of these sentences. To crank them all out, from The abandonment abased the abbey and The abandonment abased the abbot , through The abandonment abased the zoologist , all the way to The zoologist zoned the zoo , would take a million years.

    Many such combinations are ungrammatical of course, owing to various complications I haven't mentioned--for example, you can't say The Aaron, a abandonment , or The abbot abase the abbey . And most of the combinations are nonsensical: Abandonments can't abbreviate, and abbeys can't abet. Yet even with these restrictions the expressive range of a grammar is astonishing. The psychologist George Miller once conservatively estimated that if speakers keep a sentence perfectly grammatical and sensible as they choose their words, their menu at each point offers an average of about ten choices (at some points there are many more than ten choices; at others, only one or two). That works out to one hundred thousand five-word sentences, one million six-word sentences, ten million seven-word sentences, and so on. A sentence of twenty words is not at all uncommon (the preceding sentence has twenty words before and so on ), and there are about one hundred million trillion of them in English. For comparison, that is about a hundred times the number of seconds since the birth of the universe.

    Grammar is an example of a combinatorial system, in which a small inventory of elements can be assembled by rules into an immense set of distinct objects. Combinatorial systems obey what Miller calls the Exponential Principle: The number of possible combinations grows exponentially (geometrically) with the size of the combination. Combinatorial systems can generate inconceivably vast numbers of products. Every kind of molecule in the universe is assembled from a hundred-odd chemical elements; every protein building block and catalyst in the living world is assembled from just twenty amino acids. Even when the number of products is smaller, a combinatorial system can capture them all and provide enormous savings in storage space. Eight bits define [2.sup.8] = 256 distinct bytes, which is more than enough for all the numerals, punctuation marks, and upper- and lowercase letters in our writing system. This allows computers to be built out of identical specks of silicon that can be in just two states, instead of the dozens of pieces of type that once filled typesetters' cases. Billions of years ago life on Earth settled on a code in which a string of three bases in a DNA molecule became the instruction for selecting one amino acid when assembling a protein. There are four kinds of bases, so a three-base string allows for 4 x 4 x 4 = 64 possibilities. That is enough to give each of the twenty amino acids its own string, with plenty left over for the start and stop instructions that begin and end the protein. Two bases would have been too few (4 x 4 = 16), four more than needed (4 x 4 x 4 x 4 = 256).

    Perhaps the most vivid description of the staggering power of a combinatorial system is in Jorge Luis Borges's story "The Library of Babel." The library is a vast network of galleries with books composed of all the combinations of twenty-two letters, the comma, the period, and the space. Somewhere in the library is a book that contains the true history of the future (including the story of your death), a book of prophecy that vindicates the acts of every man in the universe, and a book containing the clarification of the mysteries of humanity. People roamed the galleries in a futile search for those texts from among the untold number of books with false versions of each revelation, the millions of facsimiles of a given book differing by a character, and, of course, the miles and miles of gibberish. The narrator notes that even when the human species goes extinct, the library, that space of combinatorial possibilities, will endure: "illuminated, solitary, infinite, perfectly motionless, equipped with precious volumes, useless, incorruptible, secret."

    Technically, Borges needn't have described the library as "infinite." At eighty characters a line, forty lines a page, and 410 pages a book, the number of books is around [10.sup.1,800,000], or 1 followed by 1.8 million zeroes. That is, to be sure, a very large number--there are only [10.sup.70] particles in the visible universe--but it is a finite number.

    It is easy to make a toy grammar that is even more powerful than the scheme that generates The Library of Babel. Suppose our rule for the verb phrase is enriched to allow a sentence (S) to appear inside it, as in Mary he was a fool , in which he was a fool comes after the object NP I told Mary :

    Now our grammar is recursive : The rules create an entity that can contain an example of itself. In this case, a Sentence contains a Verb Phrase which in turn can contain a Sentence. An entity that contains an example of itself can just as easily contain an example of itself that contains an example of itself that contains an example of itself, and so on:

In this case a sentence can contain a verb phrase, which can contain a sentence, which can contain a verb phrase, which can contain a sentence, ad infinitum. For example, I think I'll tell you that I just read a news story that recounts that Stephen Brill reported that the press uncritically believed Kenneth Starr's announcement that Linda Tripp testified to him that Monica Lewinsky told Tripp that Bill Clinton told Vernon Jordan to advise Lewinsky not to testify to Starr that she had had a sexual relationship with Clinton. That statement is a Russian doll with thirteen sentences inside sentences inside sentences. A recursive grammar can generate sentences of any length, and thus can generate an infinite number of sentences. So a human being possessing a recursive grammar can express or understand an infinite number of distinct thoughts, limited in practice only by stamina and mortality.

* * *

The idea that the creativity inherent in language can be explained by a grammar of combinatorial rules is usually associated with the linguist Noam Chomsky. Chomsky traced the idea to Wilhelm von Humboldt, a nineteenth-century pioneer of linguistics, who explained language as "the infinite use of finite media." According to Chomsky, the idea is even older than that; Humboldt was the last in a tradition of "Cartesian linguists" dating back to the Enlightenment.

    Enlightenment philosophers were captivated by the dizzying range of thoughts made expressible by a combinatorial grammar. In his book The Search for the Perfect Language the semiotician Umberto Eco recounts the many Promethean schemes these philosophers came up with to perfect and harness their power. Descartes noticed that the decimal system allows a person to learn in a day the names of all the quantities to infinity, and he suggested that a universal artificial language built on similar principles could organize all human thoughts. Leibniz, too, dreamed of a universal logical grammar that would generate only valid sequences of ideas, banishing irrationality and error forever.

    Three hundred years later we still are fallible, and still take years to learn a Babel of local languages with their tens of thousands of arbitrary signs. Why has no modern language used the horsepower of combinatorial grammar to the fullest and abandoned the unprincipled, parochial, onerous-to-memorize laundry list called vocabulary? The answer becomes clear when we look at the most famous of the combinatorial schemes of the Enlightenment, the philosophical language of Bishop John Wilkins. The arbitrary name was an affront to Wilkins's sense of good design, and he strove for a way to eliminate it. He wrote, "We should, by learning ... the Names of things, be instructed likewise in their Natures ."

    Wilkins's system, laid out in a lengthy 1668 opus, offered the user a non-arbitrary name for every thing by dividing the universe into categories and subcategories and sub-subcategories, and assigning a vowel or consonant to every branch in the tree. The first syllable identified one of the forty categories into which Wilkins had sorted all thinkable thoughts. For example, Z stood for "sensitive species" (animals) and could be followed by i for "beasts" (quadrupeds). The next consonant picked out a subdivision; t , for example, stood for rapacious terrestrial European canines. A final vowel pinpointed the species, yielding Zita as the name for dogs. By similar computations one could deduce another two thousand names for things. Zana is a scaly river fish with reddish flesh, in other words, salmon. Siba is a type of public military relation, namely, defense. Deba is a portion of the first of the terrestrial elements (fire), to wit, flame. Coba is a consanguinous economic relation of direct ascendant, a.k.a, father.

    Wilkins's philosophical language has been analyzed insightfully by Borges and Eco, and we can see why no one today speaks Wilkish. For one thing, it forces users to perform a chain of computations in their heads every time they want to refer to a dog. Every vowel and consonant is laden with meaning and acts as a premise in a lengthening deduction. Speakers of the language would have to play a game of Twenty Questions, inferring an entity from a description, for every word in a sentence. They could of course simply memorize the answers, such as that a portion of the first of the terrestrial elements is a flame, but that is not much easier than memorizing that the word for flame is flame .

    A second problem is that there are more things in heaven and earth than were dreamt of in Wilkins's philosophy, which identified only two thousand concepts. Wilkins understood the exponential principle and tried to cope with the problem by lengthening the words. He provided suffixes and connectors that allowed calf , for example, to be expressed as cow + young , and astronomer to be expressed as artist + star . But eventually he gave up and resorted to using synonyms for concepts his language could not generate, such as box for coffin . Wilkins's dilemma was that he could either expand his system to embrace all concepts, which would require even longer and more unwieldy strings, or he could force his users to remember the nearest synonym, reintroducing the despised memorization process.

     A third problem is that in a logical language words are assembled purely on information-theoretic principles, with no regard to the problems that incarnate creatures might have in pronouncing and understanding the strings. A perfect combinatorial language is always in danger of generating mouthfuls like mxyzptlk or bftsplk , so Wilkins and other language-designers of the Enlightenment all had to make concessions to pronounceability and euphony. Sometimes they defiled their systems with irregularities, for example, reversing a vowel and consonant to make a word more pronounceable. At other times they hobbled the system with restrictions, such as that consonants and vowels must alternate. Every even-numbered position in a word had to be filled by one of the nine vowels of English, and that restricted many categories, such as species in a genus, to nine apiece, regardless of how many species exist in the world.

    Another problem is that Wilkins's words are packed tight with information and lack the safety factor provided by redundancy. The slightest slip of the tongue or pen guarantees misunderstanding. Eco catches Wilkins himself misusing Gade (barley) for Gape (tulip).

    Finally, all that power is not being put to any sensible use. The beauty of a combinatorial system is that it generates combinations that have never before been considered but that one might want to talk about some day. For example, the combinatorial system known as the periodic table of the elements inspired chemists to look for hitherto unknown chemical elements that should have occupied the empty slots in the table. Combinatorial grammar allows us to talk about a combinatorial world, a world in which violets could be red or a man could bite a dog. Yet familiar objects and actions around us often form a non combinatorial list of distinctive kinds. When we merely have to single out one of them, a combinatorial system is overkill. We never will have to refer to fish with an enmity to sheep or to military actions with scales and reddish flesh, and that's what a combinatorial system for words like Wilkins's allows us to do. To refer to everyday things it's easier to say dog or fish than to work through a complicated taxonomy that is just a fancy way of singling out dogs or fish anyway.

* * *

The languages of Wilkins and other Enlightenment thinkers show that combinatorial grammar has disadvantages as well as advantages, and that illuminates our understanding of the design of human language. No language works like Wilkins's contraption, with every word compiled out of meaningful vowels and consonants according to a master formula. All languages force their speakers to memorize thousands of arbitrary words, and now we can see why. Many bodily organ systems are made from several kinds of tissue optimized for jobs with contradictory specifications. Our eyes have rods for night vision and cones for day vision; our muscles have slow-twitch fibers for sustained action and fast-twitch fibers for bursts of speed. The human language system also appears to be built out of two kinds of mental tissue. It has a lexicon of words, which refer to common things such as people, places, objects, and actions, and which are handled by a mechanism for storing and retrieving items in memory. And it has a grammar of rules, which refer to novel relationships among things, and which is handled by a mechanism for combining and analyzing sequences of symbols.

    To a parsimonious scientific mind, however, two mental mechanisms can be one too many. The poet William Empson wrote of the Latin philosopher,

Lucretius could not credit centaurs;

Such bicycle he deemed asynchronous.

Today's skeptics also might wonder about a two-part design for language. Perhaps words and rules are two modes of operation of a single faculty. Simple, familiar thoughts need short noises, which we call words, and complicated, unfamiliar thoughts need long noises, which we call phrases and sentences. A single machine might make either short or long noises, depending on the kinds of thoughts it is asked to express. Or perhaps there is a gradual continuum between memory and combination rather than two distinct mechanisms, with words at the memory end of the continuum and sentences at the combination end.

    To show that words and rules are handled by different machines we need to hold the input and output of the putative machines constant. We need side-by-side specimens in which the same kind of thought is packed into the same kind of verbiage, but one specimen shows the handiwork of a word regurgitator and the other shows the handiwork of a rule amalgamator. I believe that languages do provide us with such specimens. They are called regular and irregular words.

    English verbs come in two flavors. Regular verbs have past tense forms that look like the verb with - ed on the end: Today I jog , yesterday I jogged . They are monotonously predictable: jog-jogged, walk-walked, play-played, kiss-kissed , and so on. (Regular nouns, whose plurals end in - s , such as cats and dogs , are similar.) The list of regular verbs is also open-ended. There are thousands, perhaps tens of thousands, of regular verbs in English (depending on how big a dictionary you consult), and new ones are being added to the language all the time. When fax came into common parlance a decade or so ago, no one had to inquire about its past-tense form; everyone knew it was faxed . Similarly, when other words enter the language such as spam (flood with E-mail), snarf (download a file), mung (damage something), mosh (dance in roughhouse fashion), and Bork (challenge a political nominee for partisan reasons), the past-tense forms do not need separate introductions: We all deduce that they are spammed, snarfed, munged, moshed , and Borked .

    Even young children do it. In 1958 the psychologist Jean Berko Gleason tested four- to seven-year-old children with the following procedure, now known as the wug -test:

[diagram omitted]

The children could have refused to answer on the grounds that they had never heard of a wug and had never been told how to talk about more than one of them. Instead, Berko Gleason wrote, "Answers were willingly, and often insistently, given." Three-quarters of the preschoolers and 99 percent of the first-graders filled in the blank with wugs . Similarly, when shown a picture of a man who knows how to rick or bing or gling and did the same thing yesterday, most children said that he ricked or binged or glinged .

    The children could not have heard their parents say wugs or binged before entering the lab, because these words had been coined especially for the experiment. Children therefore are not parrots who just play back what they hear. And the children could not have been previously rewarded by parents for uttering those forms, because the children did not know the words before entering the lab. Children therefore are not like pigeons in a Skinner box, who increase or decrease the frequency of responses in reaction to the contingencies of reinforcement. Noam Chomsky and Eric Lenneberg, pioneers of the modern study of language and contemporaries of Berko Gleason in the Harvard-MIT community, pointed to children's ability to generalize constructions such as the regular past tense in support of their theory that language is actively acquired by a special rule-forming mechanism in the mind of the child.

    As it happens, all children are subjects in a version of Berko Gleason's experiment. Children often make up words or mangle them and are happy to put their new verbs in the past tense. Here are some examples:

spidered

lightninged

smunched

poonked

speeched

broomed

byed (went by)

eat lunched

cut-upped egg

All children also make creative errors in their speech like these:

I buyed a fire dog for a grillion dollars. Hey, Horton heared a Who. My teacher holded the baby rabbits and we patted them. Daddy, I stealed some of the people out of the boat. Once upon a time a alligator was eating a dinosaur and the dinosaur was eating the alligator and the dinosaur was eaten by the alligator and the alligator goed kerplunk.

    Such errors bring us to the second flavor of a verb in English: irregular. The past-tense form of an irregular verb is not simply the verb decorated with an - ed ending. For example, the past tense of buy is not buyed , but bought . Similarly, the past tense of hear, hold, steal , and go are heard, held, stole , and went .

(Continues...)

Copyright © 1999 Steven Pinker. All rights reserved.