Um-- Slips, Stumbles, and Verbal Blunders, and What They Mean

by Erard, Michael.

  Intriguingly, people who are lucky enough to live one hundred years are no more disfluent than they are at seventy. Jeffrey Searl, a speech pathologist, listened to oral history recordings of seven centenarians who had no speech disorders and hadn’t suffered strokes. As the centenarians answered questions about their lives, Searl counted their disfluencies and found them well within the normal range of 6.2 per hundred words. They spoke more slowly than younger adults but weren’t remarkably more disfluent.

  We may grudgingly admit that, like it or not, verbal blunders have the inevitability of gravity. The next question is why? It’s because speaking is one of the most complicated human activities that we do, at any age. The average adult English speaker has a vocabulary of around thirty thousand words and speaks ten to twelve sounds per second. Most of us in modern America, apart from the very solitary and the very garrulous, speak anywhere from 7,500 to 22,500 words a day.*20 Grabbing these words, one every four hundred milliseconds on average, and arranging them in sequences that are edited and reviewed for grammar and appropriateness before they’re spoken requires a symphony of neurons working quickly and precisely. Pronouncing (or signing) words in any language requires that your brain coordinate with your body in order to turn the electricity of nerve impulses into waves of sound (or, if you sign, of gesture and motion). So far, scientists have been able to draw only simple models of how the control of language toggles back and forth between the brain and the body.

  Given the speeds involved, why aren’t we better equipped to put units of language in the right order? The problem is that sounds, words, and grammatical items aren’t arranged in our brains as though on a library’s shelves, with all the items ordered and catalogued by topics and authors. Rather, they’re associated with one another in a matrix or a web. Some of these connections relate to the meanings of words; others relate to the sounds of those words. The Web offers multiple ways to arrange these elements and many ways to arrange and use them in a linear fashion. They have to be arranged linearly because we have a mouth through which only one word at a time can come, in the same way that you have a lot of friends, all of whom know one another to varying degrees, but who can come into your house only one at a time through your front door.

  In the 1950s, the psychologist Karl Lashley called the challenge of turning a disarrayed jumble of mental elements into a straight sequence the “problem of serial order in behavior.” This was a challenge for many aspects of human behavior. He mentioned slips of the tongue as a good example of how getting things in order in time had to involve more than a linear set of responses to a sequence of stimuli. This was good news. Slips meant we weren’t rats in mazes. In speaking, the serial ordering can fall apart across stretches of time. The other day I was listening to the radio and heard the expert being interviewed say, “You’re not hearing over the phoine everybody’s true voice.” Instead of saying “phone” (which he clearly planned to do), he accidentally anticipated the vowel sound of “voice” and placed it in “phone.” The problem of serial order suggests that many behaviors, including talking, are composed of continual efforts to plan and execute our actions. A full picture will be more complicated, but the simplest explanation for all verbal blunders, both the slips and the disfluencies, is that they occur as the brain shifts from planning to executing or back again.

  Sentences also hide their internal structures, in the same way that skyscrapers hide their girders, their air-conditioning systems, their concrete pillars. We tend to think of the utterance that comes out of a person’s mouth as a unitary thing, in the same way that a skyscraper appears monolithic. Yet in both cases what we perceive are the external characteristics. The building has thirty floors and it’s made of glass and has odd balconies. The sentence has fifteen words of English, two of which are borrowed from Latin, and describes something that happened in the past.

  If you ask an engineer what’s going on, she’ll tell you that the floor slabs inside the skyscraper are made of concrete, and that those slabs rest on steel beams that are connected to the central structural spine of the building, and that the slabs and beams and spine help stiffen the building and keep it from swaying during strong winds. Our brains work like those engineers. They make sense of a sentence’s slabs and steel beams and the pressures on them. They know that sentences are composed of chunks of words that appear, on the surface, to be connected to each other in sequence, one after the other. They are actually arranged so that some chunks determine what other chunks mean. For instance, the sentence “I am not going to exploit for political purposes my opponent’s youth and inexperience” is more than just a string of words. Though the phrases “for political purposes” and “my opponent’s youth and experience” can be reversed in the sentence, neither can be moved to the front of the sentence—the resulting sentence sounds too odd. Thus, there’s something about the verb “exploit” that has power over those phrases that requires them to follow, not precede.

  Such chunks are the real building blocks of language; when you get enough chunks of the right kinds in the right relationship with one another, you’ve made a sentence. A speaker plans a sentence in these chunks, not in sentences or in single words. On the surface the sentence may look like a single unit. Yet most talking amounts to the accumulation of chunks. These natural chunks of language are called “constituents.” In a sentence like “We think that a sentence exists as a unitary thing,” the chunk “We think” seems to belong together more than the words “that a” do. So “We think” makes up a constituent.

  Because we know something about this chunk-by-chunk accumulation, we know something about slips and disfluencies. For instance, slips are remarkably local in nature because we don’t plan what to say too far in advance. Remember that spoonerisms don’t happen over more distance than a phrase. It’s also been noticed that disfluencies tend to occur more frequently at the edges of such chunks, not in the middle. And they occur before content words more often than before grammatical words (thus “the uh boat,” not “uh the boat”). Why? Because at such moments a speaker is transitioning between planning and executing, which is where a pause is more necessary.

  Conversely, because we know something about slips and disfluencies, we know something about the accumulating chunks. For instance, when you accidentally swap one word for another, you always choose a word of the same part of speech, which is why I heard someone say, “That’s the cake on the icing” (where two nouns swap) and not, “That’s the on icing the cake” (where a noun swaps with a preposition). Linguists use this regularity as evidence that before a person utters something, the brain is constructing a sort of frame with empty slots for words, and words can fit only into the slots designed for them. If they don’t fit, they’re kicked out, or the speaker notices the impending error and stops the sentence to revise it.

  Slips never pop out willy-nilly; they follow the rules of English, or French, or Chinese, or whatever language the person is speaking. They are, in the term of the actuarial tables, “normal accidents.” This has an intuitive appeal. If someone were trying to say “think quickly” and confused the two words, she would surely say “thick quinkly” and not “fxhamdx.” Another example of a constraint on slips is that when a speaker swaps sounds between words, the sounds tend to come from the same place in the two words. This is why he’d say “chlodium soride,” not “sochlo rideium.”

  Despite how widespread verbal blunders are, we recognize that each person blunders in a different way. We also know that while we can predict what blunders will sound like, we can’t predict precisely when they’ll occur. So it’s tempting—and easy—to read all sorts of things into the bumps and hitches of a person’s speaking. Thanks to Freud and Morelli, we take it as a given that a person’s self has ways of getting noticed that he or she can’t consciously control. We interpret all sorts of verbal blunders as if we were intellectual descendants of Freud. But we also owe a secret debt to Morelli, who also helped us extract information from noise, interru
ption, error, and silence. How someone talks provides precious information about their intentions and abilities in the future. Potential mates, along with voters, potential business partners, and new employers face the same dilemma. Before we commit our confidence, we want to know: will the client who pauses a bit too long in her speaking be late with her payments? Can you trust the boyfriend who can’t decisively start a sentence to pack the heirloom china without tragedy? If you—if you—if you repeat words, would you be able to lead the men over the hill and take out the enemy pillbox? What do the actuarial tables say about the guy who forgets the punch line to the joke?

  Informally, people try to use verbal blunders as information about traits—personality traits, intelligence, emotional tendencies, thinking style, and behavioral habits. We make these judgments all the time and probably don’t realize we’re doing it. When these judgments get turned into generalizations, they reinforce stereotypes. Take, for instance, the categories of “powerful” and “powerless” speaking styles.*21 As one might expect, the features of “powerless” speaking are “uh” and “um,” as well as tag questions (“I think so, don’t you?”), frequent use of intensifying adjectives (“so,” “very”), hedges (“kinda,” “I guess”), and politenesses. But certain styles of language are not intrinsically powerless or powerful, and such generalizations can be easily contradicted. For instance, Danielle Duez once showed that François Mitterand spoke more slowly as president than he did as a candidate—that is, the powerful man spoke “powerlessly.” She deduced that challengers to power speak quickly in order to fit more ideas and critiques into a limited time span. The powerful, on the other hand, speak slowly, keeping the conversational ball and running down the proverbial clock.

  Though verbal blunders may be unreliable measures of traits, they are fairly reliable gauges of states—states of mind, emotional states, and physical states. For instance, people who are forced to devote mental resources to other activities quickly jettison fluent speaking. To see how acting under time pressure changed a person’s speech, a group of German scientists gave people a video game to play in which they had to navigate through an airport while asking questions of a robot helper device at the same time. As expected, whenever the players had to navigate through particularly tricky or distracting areas, they repaired sentences more frequently and used more word fragments. Because navigating increased the demands on their attention, it was more difficult to think and speak, and disfluency was a cognitive response, not an emotional one.

  In law enforcement, verbal blunders are often used informally to assess states of mind. One snowy winter morning, I met Jerry Giorgio, a retired New York Police department homicide detective, to learn what the law enforcement descendants of Sherlock Holmes owe to Freud and Morelli.*22 Giorgio’s a large man, barely contained behind the desk at the assistant district attorney’s office where he now works, his tie neatly knotted, his irrepressible hands perpetually involved in telling his stories about bad guys and investigations on the streets of New York City. As a young man, before he became a cop in 1959, he pursued a film career. He still has a handsome ruggedness that would have suited him well depicting a small-town sheriff or a sea captain with a heart of gold. He became a detective in 1966 and worked on many homicide cases, becoming renowned for extracting confessions—without force but subtly manipulating the suspects nonetheless.

  Giorgio didn’t sit someone down unless he was certain about what the facts were, what jibed, what didn’t fit—in other words, that the person he was talking to was the guy. But there was invariably still one crucial connection or piece of evidence missing, which was the most important one: the suspect with a weapon in hand at the scene of the crime. Giorgio would have already built a mosaic from other facts; he needed the suspect to confirm that this mosaic was, indeed, true. So Giorgio laid out the facts, piece by piece, then asked the person to reconcile the anomalies. Surprisingly, the suspect often obliged. Giorgio did it once, then again. Can I go? The guy would ask. “I’d say, just let me clear this one thing up, it won’t take very long.” The suspects didn’t know it, but Giorgio was pushing them, slowly, to the point where two mosaics, two versions of the story, would emerge, two versions that were irreconcilable. He’d push the discrepancies in their face. And when he’d pushed far enough, they’d break. They’d cry. They’d confess.

  At that moment of confession, a moment that always seemed to be approaching ineluctably, Giorgio says he listened to the suspects’ voices most closely. Up to that point, many of them weren’t under arrest, so technically they were free and could have walked out of the police station at any time. Giorgio wasn’t listening to where a suspect stood in relation to the truth; he was trying to judge how close they were to spilling everything—or to walking out the door. He watched their faces and their movements. Is he agitated? Is he starting to move around in his chair? Does he uncross and cross his legs? Is he fidgeting with his hands in a way he wasn’t doing before?

  “Or he gets angry. ‘You callin’ me a fuckin’ liar? You sayin’ I’m lyin’ to you?’ I say, ‘Ah, no, this is what I have here,’” and Giorgio gestured with his hands toward an imaginary pile of facts on his left, “‘and this is what we have here,’” and he gestured to a file on his right, “‘and I’m trying to make the two fit. And they don’t fit.’”

  Jerry Giorgio has a knack, an interpersonal gift. But police departments have begun to expect their staffs to have some acquaintance with the techniques that Giorgio uses, even if skills like his can’t be entirely reproduced. Some experts have attempted to codify this know-how and train others in it. A number of training firms have developed specialized services: some focus on law enforcement; others specialize in shoplifting or retail store theft. One such firm is Wicklander-Zulawski, based near Chicago, which offers seminars on interviewing and interrogation.

  Observing verbal blunders plays an important role in the company’s approach. Liars are most under stress on their first attempts to lie—David Zulawski says it’s like the opening day of a play. The words aren’t scripted; the props are misplaced; no one knows where to stand. So liars, he says, will blunder at the beginning of their story. The further they get into the tale, the more confident they get, and the fewer blunders they make. If they have a chance to tell the lie several times, the telling gets smoother. That may be why interrogators isolate suspects and don’t talk to them too much: the less practice suspects have in telling lies, the fresher their attempts will be.

  Zulawski trains interrogators to listen for rising volume and increased pitch, which matches rising stress levels. In his model, a change in word choice is also important. For instance, “a gun” versus “the gun” indicates a varying level of familiarity with the gun. Repetitions of words and silences can indicate mental struggle. When a person diverges from his or her own speech style, then you know something’s up. As Zulawski acknowledges, the interrogator still doesn’t know what such interruptions mean. “All I know is that it was an oddity,” Zulawski says, “and I want to go back to that point and see if it happens again. Is there a consistent response? A fluke? It might be. It could be a deception. It could mean a lot of little things.”

  But, he says, verbal blunders in and of themselves aren’t significant, and the interrogator must interpret them. For instance, he has to consider the context of the conversation. “If I asked you right now, do you cheat on your taxes, and you say, ‘Uh, uh,’ I wouldn’t think that was a cover for a lie,” Zulawski told me. “I’d think it was a cover for confusion.” On the other hand, “If we’d been talking about taxes and ethical dilemmas and I asked you the same question, then your ‘uh’ would more likely be a cover for deception.”

  A number of experiments have tried to map the typical behavior of liars, who toss their hands, twirl their pens, scratch their heads, talk more quickly, and keep eye contact a little longer. Why not use verbal blunders, as well as these other behaviors, as markers of deception? Because other subjects in the same experime
nts who weren’t lying but whose emotions had been aroused when they had been blasted by noise also tossed their hands, twirled their hair, scratched their heads, and spoke in a different pitch. They also made verbal blunders. In other words, they sounded and acted like the liars did. So what the interrogator may be seeing and hearing is merely emotional arousal, not a speaker’s self-indictment. If you rely only on Morelliesque tics and details, you’ll uncover the inexperienced, nervous, and leaky liars—not the skilled ones. You might also pull in some innocent people who’ve never been in the hotseat before.

  Despite the conflicting evidence, some researchers still maintain that stutters, pauses, “uh” and “um,” and restarted sentences might be indicators of lying. In 1997, the Italian psychologists Luigi Anolli and Rita Ciceri wrote that blunders were vocal signs that a person was lying, because they “reveal [the subject’s] lack of preparation and his/her weakness in coping with his/her own emotional reactions.” In their view, a person fundamentally doesn’t like to lie, and the blunders are signs of this emotional conflict. This approach to “vocal leakage” may assume too much. For one thing, do people actually prefer to tell the truth? One could argue convincingly that people are as likely to prefer social cohesion and tranquility and will tolerate telling and hearing omissions, white lies, and everyday spin to maintain that calm. Anolli and Ciceri’s claim also assumes that liars are anxious because lying breaches a social contract, which causes shame, hence anxiety. But this explanation can’t possibly be a universal across cultures, because such everyday social contracts are not universal.