The Language Instinct: How the Mind Creates Language (38 page)

The Watergate tapes are the most famous and extensive transcripts of real-life speech ever published. When they were released, Americans were shocked, though not all for the same reason. Some people—a very small number—were surprised that Nixon had taken part in a conspiracy to obstruct justice. A few were surprised that the leader of the free world cussed like a stevedore. But one thing that surprised everyone was what ordinary conversation looks like when it is written down verbatim. Conversation out of context is virtually opaque.

Part of the problem comes from the circumstances of transcription: the intonation and timing that delineate phrases is lost, and a transcription from anything but the highest-fidelity tape is unreliable. Indeed, in the White House’s independent transcription of this low-quality recording, many puzzling passages are rendered more sensibly. For example,
I want the, uh, uh, to go
is transcribed as
I want them, uh, uh, to go
.

But even when transcribed perfectly, conversation is hard to interpret. People often speak in fragments, interrupting themselves in mid-sentence to reformulate the thought or change the subject. It’s often unclear who or what is being talked about, because conversers use pronouns (
him, them, this, that, we, they, it, one
), generic words (
do, happen, the thing, the situation, that score, these people, whatever
), and ellipses (
The U.S. Attorney’s Office will
and
That’s why
). Intentions are expressed indirectly. In this episode, whether a man would end the year as president of the United States or as a convicted criminal literally hinged on the meaning of
get it
and on whether
What is it that you need?
was meant as a request for information or as an implicit offer to provide something.

Not everyone was shocked by the unintelligibility of transcribed speech. Journalists know all about it, and it is a routine practice to edit quotations and interviews heavily before they are published. For many years the temperamental Boston Red Sox pitcher Roger Clemens complained bitterly that the press misquoted him. The
Boston Herald
, in what they must have known was a cruel trick, responded by running a daily feature in which his post-game comments were reproduced word for word.

Journalists’ editing of conversations became a legal issue in 1983, when the writer Janet Malcolm published an unflattering
New Yorker
series about the psychoanalyst Jeffrey Masson. Masson had written a book accusing Freud of dishonesty and cowardice in retracting his observation that neurosis is caused by sexual abuse in childhood, and was fired as the curator of the Freud archives in London. According to Malcolm, Masson described himself in her interviews as “an intellectual gigolo” and “after Freud, the greatest analyst who’s ever lived,” and as planning to turn Anna Freud’s house after her death into “a place of sex, women, and fun.” Masson sued Malcolm and the
New Yorker
for ten million dollars, claiming that he had never said these things and that other quotations had been altered to make him look ridiculous. Though Malcolm could not document the quotations from her tapes and handwritten notes, she denied having manufactured them, and her lawyers argued that even if she had, they were a “rational interpretation” of what Masson had said. Doctored quotes, they argued, are standard journalistic practice and are not examples of printing something with knowledge that it is false or with reckless disregard for whether it is false, part of the definition of libel.

Several courts threw out the case on First Amendment grounds, but in June 1991 the Supreme Court unanimously reinstated it. In a closely watched opinion, the majority defined a middle ground for journalists’ treatment of quotations. (Requiring them to publish quotes verbatim was not even considered.) Justice Kennedy, writing for the majority, said that the “deliberate alteration of the words uttered by a plaintiff does not equate with knowledge of falsity,” and that “If an author alters a speaker’s words, but effects no material change in meaning, the speaker suffers no injury to reputation. We reject any special test of falsity for quotations, including one which would draw the line at correction of grammar or syntax.” If the Supreme Court had asked me, I would have sided with Justices White and Scalia in calling for some such line to be drawn. Like many linguists, I doubt that it is possible to alter a speaker’s words—including most grammar and syntax—without materially changing the meaning.

These incidents show that real speech is very far from
The dog likes ice cream
and that there is much more to understanding a sentence than parsing it. Comprehension uses the semantic information recovered from a tree as just one premise in a complex chain of inference to the speaker’s intentions. Why is this so? Why is it that even honest speakers rarely articulate the truth, the whole truth, and nothing but the truth?

The first reason is air time. Conversation would bog down if one had to refer to the United States Senate Select Committee on the Watergate Break-In and Related Sabotage Efforts by uttering that full description every time. Once alluded to,
the Ervin thing
, or just
it
, will suffice. For the same reason it is wasteful to spell out the following chain of logic:

It is more efficient to say, “For your immediate thing you’ve got no choice with Hunt but the hundred and twenty or whatever it is.”

The efficiency, though, depends on the participants’ sharing a lot of background knowledge about the events and about the psychology of human behavior. They must use this knowledge to cross-reference the names, pronouns, and descriptions with a single cast of characters, and to fill in the logical steps that connect each sentence with the next. If background assumptions are not shared—for example, if one’s conversational partner is from a very different culture, or is schizophrenic, or is a machine—then the best parsing in the world will fail to deliver the full meaning of a sentence. Some computer scientists have tried to equip programs with little “scripts” of stereotyped settings like restaurants and birthday parties to help their programs fill in the missing parts of texts while understanding them. Another team is trying to teach a computer the basics of human common sense, which they estimate to comprise about ten million facts. To see how formidable the task is, consider how much knowledge about human behavior must be interpolated to understand what
he
means in a simple dialogue like this:

Understanding, then, requires integrating the fragments gleaned from a sentence into a vast mental database. For that to work, speakers cannot just toss one fact after another into a listener’s head. Knowledge is not like a list of facts in a trivia column but is organized into a complex network. When a series of facts comes in succession, as in a dialogue or text, the language must be structured so that the listener can place each fact into an existing framework. Thus information about the old, the given, the understood, the topic, should go early in the sentence, usually as the subject, and information about the new, the focus, the comment, should go at the end. Putting the topic early in the sentence is another function of the maligned passive construction. In his book on style, Williams notes that the usual advice “Avoid passives” should be flouted when the topic being discussed has the role connected with the deep-structure object of the verb. For example, read the following two-sentence discussion:

The second sentence feels like a non sequitur. It is much better to put it in the passive voice:

The second sentence now fits in smoothly, because its subject,
a black hole
, is the topic, and its predicate adds new information to that topic. In an extended conversation or essay, a good writer or speaker will make the focus of one sentence the topic of the next one, linking propositions into an orderly train.

The study of how sentences are woven into a discourse and interpreted in context (sometimes called “pragmatics”) has made an interesting discovery, first pointed out by the philosopher Paul Grice and recently refined by the anthropologist Dan Sperber and the linguist Deirdre Wilson. The act of communicating relies on a mutual expectation of cooperation between speaker and listener. The speaker, having made a claim on the precious ear of the listener, implicitly guarantees that the information to be conveyed is relevant: that it is not already known, and that it is sufficiently connected to what the listener is thinking that he or she can make inferences to new conclusions with little extra mental effort. Thus listeners tacitly expect speakers to be informative, truthful, relevant, clear, unambiguous, brief, and orderly. These expectations help to winnow out the inappropriate readings of an ambiguous sentence, to piece together fractured utterances, to excuse slips of the tongue, to guess the referents of pronouns and descriptions, and to fill in the missing steps of an argument. (When a receiver of a message is not cooperative but adversarial, all of this missing information must be stated explicitly, which is why we have the tortuous language of legal contracts with their “party of the first part” and “all rights under said copyright and all renewals thereof subject to the terms of this Agreement.”)

The interesting discovery is that the maxims of relevant conversation are often observed in the breach. Speakers deliberately flout them in the literal content of their speech so that listeners can interpolate assumptions that would restore the conversation to relevance. Those assumptions then serve as the real message. A familiar example is the following kind of letter of recommendation:

Though the letter contains nothing but positive, factual statements, it guarantees that Mr. Smith will not get the position he is seeking. The letter contains no information relevant to the reader’s needs, and thereby violates the maxim that speakers be informative. The reader works on the tacit assumption that the communicative act as a whole
is
relevant, even if the content of the letter itself is not, so he infers a premise that together with the letter makes the act relevant: that the writer has no relevant positive information to convey. Why does the writer demand this minuet, rather than just saying “Stay away from Smith; he’s dumb as a tree”? It is because of another premise that the reader can interpolate: the writer is the kind of person who does not casually injure those who put their trust in him.

It is natural that people exploit the expectations necessary for successful conversation as a way of slipping their real intentions into covert layers of meaning. Human communication is not just a transfer of information like two fax machines connected with a wire; it is a series of alternating displays of behavior by sensitive, scheming, second-guessing, social animals. When we put words into people’s ears we are impinging on them and revealing our own intentions, honorable or not, just as surely as if we were touching them. Nowhere is this more apparent than in the convoluted departures from plain speaking found in every society that are called politeness. Taken literally, the statement “I was wondering if you would be able to drive me to the airport” is a prolix string of incongruities. Why notify me of the contents of your ruminations? Why are you pondering my competence to drive you to the airport, and under which hypothetical circumstances? Of course the real intent—“Drive me to the airport”—is easily inferred, but because it was never stated, I have an out. Neither of us has to live with the face-threatening consequences of your issuing a command that presupposes you could coerce my compliance. Intentional violations of the unstated norms of conversation are also the trigger for many of the less pedestrian forms of nonliteral language, such as irony, humor, metaphor, sarcasm, putdowns, ripostes, rhetoric, persuasion, and poetry.