“My uncle had his throat cut,” the patient said. “And my uncle…ah, he wasn’t killed, but they tortured him. They took him and they slit his throat right…the skin…just cut the skin around like this.”
“Who did that?” the therapist asked.
“Mmm…hulligan…holigan…huligans, whatever you call them.”
“The who?”
“The…ah…you know the…ah…sss…he was in…the…he…was dri…traveling from one town to the oth…next and—”
“Mhm,” the therapist said.
“He was stopped by a couple…ah…if you know the situation in the south…”
“Were…were they…were they Russians or were they Bolsheviks?”
“Yes. Yeah…mujik…mujiks…peasants.”
Mahl related this scene in “Everyday Disturbances in Spontaneous Speech,” an essay that explored how flustered speech (not only slips, as Freud would have it, but other hesitations and self-interruptions, too) might be therapeutically useful. A therapist measuring disturbances at the start of therapy and again at the end might have an objective measure of the therapy’s success. More immediately, a therapist could note that when a patient spoke in a flustered way, it meant he or she had ventured too near a repressed emotion. That would be a clue to the therapist that a mother lode of therapeutically useful emotion lay nearby.
Once, Mahl asked a patient where he was going for vacation. “Oh, to B,” the patient replied. “I have a friend…. well, I planned first to go to A,” then rambled on about vacation choices. Then Mahl commented, “You said, ‘I have a friend.’”
“Well,” the patient said, “I was referring to L., a woman at Z. We bedded together and became close friends.” By noting the sentence repair, Mahl had opened the patient to talking more about the woman, who turned out to be ten years older than the patient, and whom he called “prudish, stern, and remote.” Mahl reported that this conversation led to a breakthrough when the young man realized that his attachment to his mother had shaped his attractions to other women. “Left unremarked,” Mahl wrote, “the sentence completion would have enabled the patient to avoid this oedipal material.”
Though it wasn’t an original way of using verbal blundering, Mahl thought if he could collect enough of these flustered moments he could correlate them with other instrumental measures. Yet “flustered speech” was too crude a category to be useful for such a grand project. What’s the difference between the nephew of the tortured uncle who said “uh…uh…uh” and the oedipal son who changed his mind about vacation spots? Were they anxious in different ways? Were they anxious at all?
Mahl realized that the patient’s speech swarmed with interesting bits and pieces: the repetitions of words, the attempts to pronounce a word correctly, the pause fillers, the sentences that start in one place but head in another direction. So he developed a more detailed taxonomy of speech disturbances, which is still used today and is his most enduring contribution to the study of disfluencies. His list included the “ah” (for consistency I’ll relabel these “uh” and “um,” except in direct quotes from Mahl), the sentence change, the word repetition, the stutter (which was to be a repetition of a single sound or syllable), the omission of a word or part of a word, the incomplete sentence, the slip of the tongue, and the “incoherent intruding sound” (one such sound made by the patient with the uncle was “sss”).
With financial support from the U.S. Public Health Service, Mahl recorded interactions between psychiatrists and patients on twelve-inch tape reels and set up a room where trained secretaries turned the recorded speech into text. Later Mahl coded the transcripts, counting all the disfluencies and marking them up with a red pencil. His plan: correlate speech disturbances with stomach acid levels due to anxiety, then develop a diagnostic tool for dealing with stress.
Long before he concluded his research, however, something interesting happened. The patients, the interviewers, the research assistants, the secretaries, and even Mahl himself were shocked that perfect prose didn’t come out of their mouths when they spoke. Some became angry or ashamed. Others disavowed their own speaking or questioned the secretaries’ accuracy. It was as if Mahl had revealed them to be dirty or ill—indeed, much of speech education in the early twentieth century had taught the need for “verbal hygiene.” People not only hadn’t been aware of how they spoke; they believed that they spoke in prose, all of their sentences starting with a capital letter, finishing with a punctuation mark, and flowing unceasingly between letter and punctuation like a river to the sea. That, Mahl had to tell them, was a fantasy, a dream.
Mahl became so absorbed by the realities of speech as people interacted with one another that, in 1952, he began recording a radio talk show hosted by Tex McCrary and his wife, a former bathing beauty, Jinx Falkenburg. Tex and Jinx would interview celebrities live from the Waldorf-Astoria, which gave Mahl access to the most interesting blunderers of all: the celebrity blunderers.
One time Jinx asked Edward R. Murrow, the radio broadcaster, what six people he would invite to dinner.
“One…one would certainly be Sir Winston Churchill. A second would be H. G. Wells. A third one would be Heinrich Heine. The fourth one, Marcus Aurelius. The fifth, Mr. Lincoln. And the sixth, now it gets really difficult, Jinx…ah…the sixth, I think, would be Beethoven,” Murrow said.
“Murrow had been talking along very fluently,” Mahl remembered, “and then Jinx said, ‘They’re all men,’ and his speech fell apart.”
“You know something? You selected all men,” Jinx told Murrow.
“I wonder why that was,” Murrow said. “I haven’t a clue. I don’t know why. I—”
“No woman in history, or currently, would interest you?”
“No, no. Th…this…ah…ah…you trapped me somehow. I don’t know how you did it, but…ah…ah…this—”
Murrow worried about what “some psychiatrist” would make of his confusion. Mahl believes that Murrow must have been afraid that he’d tipped people off to his male chauvinism or even some latent homosexuality. “I’m embarrassed,” Murrow admitted. Then he regained what Mahl called his “characteristic fluent speech.”
With his recordings and a room full of transcribing secretaries, Mahl was the first to discover that people say “uh” more frequently when they’re talking to an interviewer on a telephone than when they’re face-to-face. It’s not because of any inherent property of the technology. Rather, the speaker’s disfluency is an ingredient of the face-to-face interactions.
Mahl also recognized very early in his research that disfluent individuals fall into one of two groups: they either change their sentences, or they say “uh” and “um” and repeat words. Intuitively, you can see what personality types correspond with these two patterns. Sentence changers are the people who burst ahead, confident, speaking in a fluid stream of words. If they don’t like the way the sentence is going, no problem—they rethink it and start over as fluidly. Even if there’s nothing wrong, they often rush to double back and dive into rephrasing as easily as they first phrased it.
Then there are people who say “uh” and “um.” These are the planners, the crafters, the forestallers. Mahl called them “ah-ers.” He told me that he had once studied his own speech patterns, thinking himself a frequent uh-er and repeater. To his surprise, he found that he’s a sentence changer. (After listening to the tape of our conversation, I concur—Mahl, mostly umless, spoke in rolling, ongoing sentences that he would restart now and then but invariably managed to keep comprehensible.)
Because he had also asked his patients to fill out standard personality questionnaires, Mahl could sort the traits of the uh-ers and sentence changers. Uh-ers reported being “unusually self-conscious” and “worry quite a bit over possible misfortunes.” They rated themselves as distractable and careful, and reported that they’d been weaned early. They tended to come from strict families, with a preponderance of strict fathers. Meanwhile, sentence-changers considered themselves shy public spe
akers and easily embarrassed. Individuals from this group also agreed with the statement that “It is not always a good thing to be frank,” and thought that it was easy for others to see what was going on with their emotions.
Subsequent to Mahl, researchers have confirmed in more sophisticated ways that the verbal blundering clusters in several types, almost as if there are species of talkers. When Liz Shriberg (the speech researcher who gave the robot Flakey his “uh”) was working on her dissertation at Berkeley in the 1990s, she plotted how many of each type of disfluency each of her subjects made. At first she couldn’t make sense of the resulting graph, whose random-seeming data points only seemed to confirm that individuals varied widely in the ways they showed their delays. Both groups were equally disfluent overall and spoke sentences of the same length.
But when she inspected her data, two distinct groups emerged. Those in one group tended to say “uh” and “um” and also repeat words but didn’t restart their sentences or accidentally choose the wrong word. She called them “repeaters.” Members of the second group were exactly the opposite: they tended to substitute the wrong words and restart their sentences, but they didn’t fill pauses or repeat words quite as frequently. Shriberg called them “deleters.” Deleters differed from repeaters in one other regard: they spoke more words per second and, as a result, were more disfluent per unit of time.
This dichotomy is useful, especially if one were, like Shriberg, designing automatic speech recognition software. It’s not enough to lump all speakers together, averaging their disfluencies. We might want the systems to “profile” a speaker as one or the other. Beyond this immediate application, the dichotomy might point to cognitive qualities we might someday be able to make sense of and measure. When Shriberg noticed that the deleters repaired and restarted their sentences more often, she mused about their cognitive profiles—how the spikes in certain blunders and the absence of others might disclose something about deleters’ cognitive traits and how they went about solving problems. To justify an approach in which blundering style may indicate cognitive traits, she noted that timing, planning, and executing occur in many other activities. If you were to ask someone to draw a picture, for example, some will stop to scratch their heads and gaze out the window. Others will dive in, begin drawing, and quickly exhaust the erasers on their pencils.
This difference may reflect variations of cognitive style—in how a person solves problems, how she accepts new information, and how she monitors her own behavior. If you speak quickly and repair more often, there’s something about you that causes you to monitor less while you’re speaking. This allows you to sound stylistically as if you’re moving ahead even though you’re actually moving backward. If you’re speaking more slowly, you could be monitoring yourself more. Does that reflect some underlying personality trait, say, of being a more responsible person?
Such inferences are intriguing, but beyond a plethora of anecdotes, psychologists so far have little solid evidence that such variation in blundering points to personal traits. In the only comprehensive study to date, college students were tape-recorded at random intervals for three days with computer-controlled microcassette recorders they wore on their belts. The researchers, at the University of Texas at Austin, James Pennebaker and Matthias Mehl, counted students’ word choices, mostly nouns, adjectives, negative words, personal pronouns, and other content items. They also transcribed “uhs” and “ums,” swear words, and filler words (such as “like”). Before the taping, the students took a personality test that measures what psychologists call the “big five” personality traits: extraversion (how much stimulus from people and situations a person can handle), agreeableness (to what degree a person defers to others), conscientiousness (how organized and careful a person is), openness (how open a person is to new experiences and ideas), and emotional stability (how a person responds to stress). They also took tests to measure their levels of depression and self-esteem.
Of the disfluencies, Pennebaker and Mehl counted only “uhs” and “ums.” Hence, the only correlations they calculated were between personality variables and those pause fillers, the most common disfluencies. Contrary to the stereotypical expectations, the students who rated themselves depressed or introverted didn’t say “uh” or “um” more frequently. And when snippets of their recordings were played to independent judges, the speakers who said “uh” and “um” more often weren’t rated more depressed, either. Otherwise, there were no meaningful correlations between the frequency of “uh” and “um” and personality traits. Whether they were extroverts, introverts, neurotics, nice kids, open kids, or responsible kids, their personality traits didn’t correlate with more (or less) use of “uh” and “um” at all. Only in one dimension did anything significant turn up: people who said “uh” and “um” were rated as more conscientious by the judges, as were those who used fillers like “well” and “so.”
Pennebaker and Mehl did, however, find evidence that supports an observation that George Mahl also made: that each person consistently blunders in a way that’s unique to him or her. Over the course of the three days, the recordings showed that each student tended to say the same number of “uhs” and “ums.” If their pause fillers were counted at Time 1 and then again four weeks later, the students would be blundering at roughly the same rate. There were only two other features of their speech that were more stable than this: their swearing and their use of filler words (“well,” “like,” “so”). In casual listening we can often hear speakers’ distinctive speech patterns when their disfluencies become excessive—the novelist who repeats every word at the beginning of each phrase (“the the the the book has been received well and and and and it generated a lot of discussion…”) or the computer scientist who begins each sentence with a long string of “uhs” (“uh uh uh uh uh I’m always entertained by the dogs in the park”).
The scientific work that would link disfluency patterns with personality traits hasn’t yet been done—it would require experiments that can measure reaction times in milliseconds, not a crude instrument like a personality questionnaire. Until then, the question of some link remains wide open.
Perhaps George Mahl’s most interesting finding in those early days of disfluency research was that anxious people don’t say “uh” or “um” more. That is, a person’s use of pause fillers didn’t correlate with his or her level of anxiety. At first he was disappointed—he had hoped to measure anxiety with pause fillers. Then he made another finding: if he counted all eight of his speech disturbance types, subtracted the “uhs” and “ums” from the total, then calculated this number relative to the total words a person had spoken, the ratio was highly correlated with anxious moments. In other words, “uh” and “um” didn’t correlate with anxiety, but all the other disfluencies did.
This has been Mahl’s most enduring finding. It isn’t widely known, though, and flies against the conventional wisdom. If you ask a typical person why speakers say “uh,” he or she will probably reply that it’s because they’re anxious. Not so, said Mahl. To measure anxiety accurately, listen to all of their other verbal blunders.
Heather Bortfeld, a psychologist at Texas A&M University, and her colleagues have confirmed this finding. In a study published in 2001, they analyzed the conversations between pairs of people and drew a stunning picture of all the factors that influence a person’s disfluency. Each pair was assigned the task of arranging a set of picture cards, with one person directing the other person to do the actual arranging. The pairs were mixed by gender, age, and familiarity—that is, there were males partnered with males, females with females, and females with males; some of the pairs were strangers, while others were married, and some were age peers while others were young-old pairs.
Bortfeld found that the oldest people in the group, the seventy-year-olds, spoke more disfluently than the people in their twenties. This wasn’t surprising—it’s long been known that older adults are more disfluent than younger adults. Bortfeld also
found that men said “uh” and “um” much more than women did, making them more disfluent than women overall. Men also restarted more sentences and repeated more words. Surprisingly, the married couples blundered as much as the strangers. If you’ve cherished the notion that you blunder more with people you don’t know, you’ll have to give that up.
So what determines how much someone blunders? In the study Bortfeld found that the more active member of the pair, the person directing the other person, made the most blunders. Presumably this was because they were trying to manage thinking and speaking simultaneously. Both men and women in the director’s role were more disfluent than their partners. She found, too, that their disfluencies (the pause fillers, silent pauses, repetitions, and repairs) were more frequent at the beginning of the sentences. This is where the brain is between planning and executing. Thus, people are more disfluent when there’s an additional cognitive load, such as a distraction, an additional task, or an emotional response. The anxious state of an uh-er may be the local, most visible reason for his hesitation. The deeper reason is a cognitive one.
In Bortfeld’s study, this heavier cognitive load came not only from the social roles but from some aspects of the tasks themselves. Each pair of speakers had two sets of cards to arrange, one showing abstract shapes and one with children’s faces. Each director spoke more disfluently when arranging cards with abstract shapes than the ones with faces. Bortfeld concluded that the more serious problems with talking and acting at the same time are indicated by word repeats and sentence restarts. As Mahl had also found, “uh” and “um” aren’t reliable symptoms of a brain under stress.
Bortfeld says she first noticed disfluencies as a junior in college when she was living in Madrid, where people said “eh,” not “uh.” She realized she was going to have to learn how to fill her pauses like the natives did in order to be treated as a native speaker. (In fact, children must learn how to use pause fillers as adults do—it doesn’t come naturally.) There is a right way and a wrong way to be disfluent, and the right way seemed to be a matter of linguistic ease as well as convention. “Uh” is one of the easiest sounds to make in English, so it makes sense that the Spanish equivalent, “eh,” would be the sound of the neutral vowel in that language as well, the one that can be pronounced with a relaxed tongue and jaw. The pause fillers of many languages sound so similar because the relaxed tongues and jaws that make them are also similar. Yet each language trains its speakers in a slightly different set of vowels, making for slight phonetic differences among its pause fillers, too.
Um-- Slips, Stumbles, and Verbal Blunders, and What They Mean Page 8