Read My Slips: Speech Errors Show How Language Is Processed

Researchers are analyzing spoonerisms and other slips of the tongue to help understand how humans--and even apes--can comprehend and use language

Kanzi, a 27-year-old bonobo, knows the difference between a blackberry and a hot dog. But sometimes, when researchers asked him to touch the abstract visual symbol, called a lexigram, that means blackberry, he touched the lexigram for hot dog, blueberries, or cherries instead.

Kanzi's errors weren't random mistakes, nor an indication of apes' language limitations, says Heidi Lyn, a comparative cognitive scientist at the University of St. Andrews in Fife, U.K. Rather, they show the complex way in which his mind had organized the lexigrams. For example, if Kanzi made a mistake when asked for "blackberry," he was more likely than chance to choose a lexigram for another fruit, much as you or I might say "red" instead of "black," says Lyn, whose paper on Kanzi's mistakes was published online in Animal Cognition in April and will appear in print later this year or early next.

Analyzing errors for insight into the covert mental processes of animals is a new direction for a technique that language scientists have used for 40 years to study language processing in humans. For all its power, human language remains something of a scientific mystery. Researchers are still struggling to understand exactly how humans hear, comprehend, and produce words and sentences. Slips of the tongue, or linguistic mistakes made inadvertently by speakers who do know the correct form, offer potent clues about language processing in the brain. Speech error research is currently on the upswing with new methods and theories and increased attention to groups such as children and users of sign language--and, now, animals. "We have a long way to go before we understand how to put the multiple pieces of language systems together in the seamless way that we experience it," says psycholinguist Merrill Garrett of the University of Arizona, Tucson, who has studied slips of the tongue since the 1970s. "Error profiles that arise during spontaneous conversation are going to be an important part of the agenda."

Barn doors and darn bores

Early in the 20th century, collecting speech errors was chiefly a hobby, especially for people who found Freud's emotional explanations lacking. (Psychoanalysis had no way to account for the diverse, often mundane slips of the tongue that people make.) In the 1960s, Noam Chomsky sparked a wave of grammatical theorizing that transformed speech errors into theoretical gold. Linguist Victoria Fromkin, among others, argued in the late 1960s that speech errors showed that abstract mental units of sounds and words were also concrete symbols in speakers' minds.

Using speech errors as scientific data posed some problems: Waiting for speakers to make an error required an inordinate amount of time, and some questioned the reliability of what listeners heard. But the field got a boost in the 1970s when researchers created ways to elicit many (but not all) types of speech errors in the lab. One method involved giving people word pairs like "duck bill," "dart board," and "dust bin," then asking them to say "barn door." About 10% of the time, subjects said "darn bore." By eliciting speech errors, researchers can control for higher frequency sounds (in English, "s" is more frequent than "k") and words ("latrine" is more frequent than "tureen"). Words used more frequently are less likely to be involved in speech errors. For example, more errors occur with content words ("cat," "hat") than grammatical words ("the," "in"), because grammatical words are used more frequently. The effect of frequency also implies that what one usually talks about affects how one slips.

Lyn was the first to apply the study of errors to bonobos. Kanzi and a female bonobo, Panbanisha, who now live at the Great Ape Trust in Des Moines, Iowa, can comprehend instructions and descriptions in spoken English, and they can respond by using 384 lexigrams, which they touch on a keyboard. From 1990 to 2001, researchers tested the bonobos thousands of times, showing them a photo or lexigram or saying an English word. The bonobos then had to select the matching lexigram. The apes chose correctly 12,157 times and made 1497 incorrect choices, although no one thought to consider the errors as data until now.

Lyn found that Kanzi and Panbanisha have arranged hundreds of lexigrams in their minds in a complex, hierarchical manner based mainly on their meaning. She coded the relations between all 1497 sample-error pairs along seven dimensions, including whether the lexigrams looked alike, had English words that sounded alike, or referred to objects in the same category. She found that the errors were not random but patterned. If the lexigram stood for "blackberry," the error was more likely than chance to sound like blackberry, be edible, be a fruit, or be physically similar. Errors were also more likely to be associated with more than one category. For example, "cherries" are both edibles and fruits, and the word sounds like the correct one, "blackberries." All this indicated to Lyn that mental representations of the lexigrams must be stored not as simple one-to-one associations but in more complex arrangements. This suggests that, given the chance, bonobos and other apes can acquire systems of meaning that are closer than anyone has thought to what humans do, and that some aspects of language acquisition are not unique to humans. "We begin to see that the biological or species variable is far less important than we thought," says Susan Savage-Rumbaugh of the Great Ape Trust.

Out of the mouths of babes

Lyn's analysis is not the first to study errors in creatures that haven't mastered all the complexities of human speech: For about 20 years, researchers have also used speech errors to study language acquisition in children. Kids do say the darnedest things, but by definition, the true errors are the ones they make with linguistic levels and units they know, explains linguist Jeri Jaeger of the University at Buffalo in New York state, who in 2005 published a book that capped 20 years of collecting kids' slips, many of them from her three children. It was the first study of the same children's speech errors over a long period, allowing her to match their errors with their stages of language development. Jaeger's collection is "unique," says linguist Annette Hohenberger of Middle East Technical University in Ankara, Turkey, and shows how slips change over time.

Distinguishing true slips took a linguist's ear and a mother's patience. Jaeger's youngest daughter's exclamation that "She already showed me tomorrow!" wasn't a true slip, because she didn't yet know the meaning of "yesterday." On the other hand, at 16 months, her eldest daughter said "one two three, one two three, one tuwee"--a fusion of "two" and "three," which was a true slip because she knew the two words were distinct and had regularly pronounced them correctly. This anchors Jaeger's point that children only make slips with what they know.

Analysis of such speech errors can provide a novel perspective on how children acquire language. Linguists have debated, for instance, whether children need syntactic knowledge to speak in two-word clumps. Jaeger says no. Her data show that when children begin to combine words, at about age 2, they don't blend phrases or confuse intonations. Such slips require a mature knowledge of syntax. Not until children speak in sentences of three or more words do syntactic errors, such as "sit down this immediately!" (a blend of "sit down this minute" and "sit down immediately") appear.

It's long been known that children make more speech errors than adults, but it wasn't known how or if aging affected error rates. In 2006, Janet Vousden and Elizabeth Maylor at the University of Warwick in the U.K. published the first study tracking speech errors across the life span and reported no significant increase in total errors between young and older adults. However, compared to children, adults made proportionately more errors in which a sound segment was anticipated (frive frantic fat frogs) rather than perseverated (five frantic fat fogs).

That fits with a widely used model of speech errors developed in the 1980s by cognitive scientist Gary Dell of the University of Illinois, Urbana-Champaign. Most linguists think that words and sounds are stored in a kind of network in the brain, connected by variables such as how they sound, their parts of speech, and their meaning. Dell proposed that when sounds or words stored in such a network are selected, this also strengthens or "activates" neighboring words or sounds, which may be misread as the right ones. In his model, people forced to speak quickly make more errors not because they have more opportunities to do so but because the stimulation of neighboring units has less opportunity to fade. Dell also proposes that practice tends to activate present and future units more than past ones. As a result, the more practice a speaker has, the higher the proportion of anticipatory errors, although overall errors decrease. "Whatever makes you more error-prone makes your errors more perseveratory," explains Dell. Caroline Palmer, a psychologist at McGill University in Montreal, Canada, has found the same effect (among others) in piano performances.

Language need not be spoken, and linguists have long been interested in whether speech and sign are processed the same way. German linguists Hohenberger and Daniela Happ and Helen Leuninger at the University of Frankfurt used a newer method for eliciting slips from German speakers and signers of Deutsche Gebärdensprache (DGS, or German Sign Language), in the first slip study of signers in a language other than American Sign Language. In a series of papers, the most recent published in 2007, they asked speakers and signers to narrate a series of pictures under various stress conditions, such as putting pictures out of order.

They found that all types of slips found in spoken German are also present in DGS, although in different frequencies. The slips also occur with the same basic units. This indicates that signs and words are both stored in the brain as clusters of primary elements that can be flexibly recombined, and it underscores that humans possess a single language faculty regardless of how they deploy it, says Hohenberger.

But there are some differences. For instance, both signers and speakers catch and repair utterances that include mistakes. But signing is relatively slower, so signers catch more errors involving exchanges of individual signing elements, such as hand shapes or location of the sign.

Because of this, Hohenberger speculates that slips of the hand may next contribute to an emerging question in slip-of-the-tongue research. Based on ultrasound studies of speakers' tongues as they make sound exchanges (better known as spoonerisms, such as "jeef berky" instead of "beef jerky"), phonetician Marianne Pouplier of the University of Munich, Germany, has suggested in several recent papers that speakers don't substitute one whole sound segment for another as was previously thought. Rather, they attempt to pronounce the two sounds at the same time. This way of thinking about speech errors--as a collision of motor commands rather than as substitutions of mental symbols--might be more reliably investigated in slips of the hand, Hohenberger says, because researchers can capture the slower hand movements more clearly than tongue movements.

Although error studies offer intriguing data, their implications are not always clear. Take the bonobo findings. The apes confused fewer target-error pairs that were either both nouns or both verbs, implying that they don't take note of parts of speech. "This result argues against the claims made elsewhere that Kanzi has spontaneously developed an elementary grammar," says primatologist Robert Seyfarth of the University of Pennsylvania.

But Lyn says the error results don't directly address the question of grammar and don't contradict earlier findings in which bonobos appeared to prefer certain semantic sequences. Instead, she says, "the results support the idea that [apes'] representation of semantic information is much more complex than has been shown to date."

Still, the study of bonobo errors does rebut two frequent criticisms of ape language research: that the apes have simply been trained to respond, and that researchers may inadvertently shape the bonobos' responses. Errors can't be trained, nor can patterns of errors be deliberately produced. And if researchers were subtly guiding the apes by eye gaze or body posture, Kanzi and Panbanisha might have made far more errors based on simple proximity in the keyboard.

Lyn plans to continue analyzing the error data for other insights into the bonobos'conceptual world. "For me, the error analysis was not to just study one aspect of their symbolic representation," Lyn says, "but to get a glimpse of how it all hangs together." Such a big question hasn't been answered for human language, either, but speech errors will likely be central to the search. Says the University of Arizona's Garrett: "We have most certainly not reached the limits of that kind of research."

I have one word for you glib, fluent people, you who sound smooth, scripted, and rehearsed, who execute each sentence with crisp precision, because your success may be at stake:

“Um.”

Here and there you can catch a new attitude about this and other hesitations to the ideal, uninterrupted flow of speaking. Barack Obama’s main political consultant, David Axelrod, likes to record video of people on the street for political ads; they inevitably say “uh” or “um,” which he likes, he says, because it’s more authentic.

In the 2005 business language jeremiad, Why Business People Speak Like Idiots, the authors advocate that speakers be more relaxed. “It’s not a bad thing to open the kimono in celebration of those ‘ums’ and ‘ahs’ for all the world to see,” they write. (For those of you who are not idiots, “open the kimono” is corporate jargon for “letting down your guard” or “showing one’s hand.”)

When Condoleezza Rice first became Secretary of State, she uttered “uh” and “um” in public statements with surprising frequency, given her experience and position. At first glance (or listen) she was warming up to the new role, but the style has persisted. Whether intended or not, the effect seems to be that Rice isn’t reciting talking points, she’s opening the kimono of U.S. foreign policy.

Americans began to prefer um-less speech in the early 20th century, as an increasingly complex and urbanizing society boosted the value of planning and efficiency. Saying “um” was a sign that the bureaucracy in your head, of your self, was breaking down. Burgeoning technologies drew attention to verbal embarrassment, as the phonograph and the radio captured evidence of the fleeting inefficiencies of everyday speaking.

Teachers of public speaking contributed to this newly valued um-lessness: They boiled eloquence down to a set of traits—among them, not to say “uh” and “um.” By the 1920s, after decades of silence on the matter, books of etiquette urged young men and women to mind their ums, too. In 1928, author Helen Hathaway warned in Manners: American Etiquette that “puncturing our sentences with ah’s and er’s, mincing our words, [and] employing affected tones and gestures, are tricks annoying even to our friends and positively repellant to strangers.” Even psychotherapists frowned on so-called filled pauses. In 1959, psychiatrist Sandor Feldman wrote about patients whose “moaning-like ‘er…er…er…’ is annoying,” in Mannerisms of Speech and Gesture in Everyday Life. The only rule no one seems to have written was that you shouldn’t say “um” when talking to yourself.

Uttering “uh” and “um” is the sign of thought, a symptom of an additional mental load, rather than the absence of thinking. While the fashions of public speaking change, the psychological facts about language remain the same. For instance, we conventionally believe that people who say “um” a lot when they speak are unsure, over-careful, or nervous. However, in the 1950s George Mahl, a Yale psychologist, overturned that theory by measuring the number of times his subjects said “uh” or “um” against their levels of stomach acid—and found no correlation. What he did find was that for more a reliable marker of anxiety, we should notice how frequently a speaker restarts a sentence, repeats a word, or utters a fragment of a word.

One of these “speech disturbances” (as Mahl called them) occurs once in every 4.4 seconds of speaking, on average. About six percent of the words that normal speakers say aren’t pronounced without some sort of interruption, which has probably been true for the 100,000 years that humans have had language. Why? Because everyone has to plan what to say next. Some of us opt for “uh” and “um” at those junctures. It’s as if we can either think or speak—but not both at once. Nearly all of the psychological and cognitive research that has since been done in this area, which was inspired by George Mahl, says that uttering “uh” and “um” is the sign of thought, a symptom of an additional mental load, rather than the absence of thinking. But while it’s rare for someone to be able to have novel ideas and speak perfectly fluently, many of us continue to equate fluency of speech with fluency of thought.

So why the new lenience toward “uh” and “um?” Perhaps it represents a rhetorical resource that hasn’t been exhausted yet. Someone less cynical might suggest that American audiences simply want more variety in their broadcasted voices. It could also be related to the architectural impulse that exposes conduit and pipe in buildings or opens restaurant kitchens—a desire to see the backstages of life where the action really happens—like people on the street talking about Barack Obama. Or it could be an assertion of uniquely human qualities to counteract technological values that have pushed too far. A verbal dissent, so to speak.

But beware celebrating the open kimono. Not all people who say “um” are spontaneous and authentic, just as not all those who speak smoothly and fluently are more intelligent or competent. The phone message that starts, “Uh, yeah, hi, this is Bob,” may still be a telemarketer. “This business of keepin’ it real—it can be carried too far, and it can come across as arrogant,” writes Nancy Franklin in The New Yorker about the television version of This American Life: “Real is earned.”

So let me offer another way out of the style wars that will stand the test of time. A study by Nicholas Christenfeld, a psychologist at the University of California at San Diego, showed that about half the listeners to a speech naturally pay attention to the content. The other half listen to the style. But when the content gets boring, obvious, or offensive, the content listeners switch to listening for style. The solution for getting fewer people to notice your speaking style—and those pesky “ums”—is pretty clear:

Be more interesting.

The original layout from The Morning News can be seen here.

Jan Freeman, the regular language columnist for the Boston Globe, handed me her space when she went on vacation. This was my piece...

Wildness. We go outdoors, to the mountains or the ocean, to encounter the untamed and untameable. But this quality can be found closer to home, too -- our spoken sentences are full of wildness, right under the threshold of our attention.

I'm talking, of course, about verbal blunders.

By "verbal blunders," I mean one of two things. They can be slips of the tongue -- any moment where a speaker gets sounds out of order or selects the wrong word. The other day, my wife said "the fook I --" (stopping herself on the way) instead of "the food I cook." The same week a colleague pronounced her affairs to be in a "stad sate." Speech errors like these are unintentional accidents; linguists figure that a person makes about one or two of them every 1,000 words.

A verbal blunder can also be what's known as a "speech disfluency": "uh" and "um," repeated sounds and words, fragments of words, and sentence repairs, all of which occur when you're planning what to say next or realize you want to say something else. This isn't the same thing as stuttering, a disorder with neurological roots. I often speak disfluently, and so do we all: 5 to 8 percent of the words we say are somehow disfluent. So if American men and women say about 16,000 words a day (as a recent paper in Science calculated -- and no, women apparently don't talk more than men) that makes for some 800 to 1,280 disfluencies daily.

It's typical to think of verbal blunders as embarrassing slip-ups that we should avoid. But I've just written a whole book about verbal blunders, and I find them fascinating. Why? Because they're signs of the wild. Not in the sense of rough or savage, but because they're pure and untameable. They provide a window into what humans really are: biological organisms who live in complex groups and have really amazing brains. Blunders of the verbal sort may seem like violations of the order of language, but in fact they're spontaneous eruptions of the qualities that gave us this order in the first place.

Verbal blunders are uniquely human: If other species don't have language like humans do, then only we make verbal blunders. There's some evidence that the bonobos Kanzi and Panbanisha, who are part of a language program at Great Ape Trust in Des Moines, occasionally make errors in selecting the visual symbols with which they communicate, but these aren't really slips. And though some zebra finches sometimes stutter, these aren't disfluencies in the same sense as the ones humans produce, because the phrases of their songs aren't as long or varied as human speech.

Verbal blunders are also universal -- they occur in every language, both spoken and signed. Every human language provides its speakers with a way to signal some delay, either with a filler word, a pause, or a repeated word. Slips of the tongue in a particular language will follow the sound patterns and word structures of that language; a malapropism or a spoonerism in English will always sound like English, not Cherokee.

Though we might make 7 to 22 slips of the tongue a day, we typically only notice about one a week, and the majority of disfluencies don't even register to our ears. This suggests that we're evolved to perceive a message despite any "noise" in an utterance, and that we filter out most distractions automatically. So noticing someone's "uhs" and "ums" says as much about your filter as about the speaker's style.

Instead of sitting in judgment over blunders, linguists and psychologists use them as tools to pry language open. Slips of the tongue are useful in this way because they possess a surprising amount of order -- for instance, accidental word blends such as "behortment" (which mixes "behavior" and "deportment") will always have the same number of syllables and stress patterns as the original two words. Such patterns suggest, among other things, that words exist as skeletons into which sounds are slotted. Slips of the tongue were the original way linguists came to understand this, before experiments and instruments confirmed it.

Not all interpretations of verbal wildness hold up over time. Sigmund Freud listened to verbal blunders and heard people losing control of their unconscious desires, but contemporary psycholinguists now think that so-called Freudian slips -- despite their hold on the popular imagination -- were fashionable interpretations after the fact, not adequate explanations of why slips happened. And there's never been enough evidence to suggest that slips are rooted in something about the speaker rather than in the act of speaking itself.

Though slips of the tongue and speech disfluencies have different causes, I put them together because of their wildness -- both disrupt the way we want to present ourselves. That's why people either laugh at verbal blunders or try (in the case of disfluencies) to eradicate them. But I like what Elizabeth Zwicky, who grew up with linguists for parents, said about paying attention to verbal blunders. Not only are you always going to be amused wherever you go, but you also become less stressed out by them. It's like being a birdwatcher. "Birdwatchers," she said, "have a richer experience of birds than anybody else."

To see the original layout, go here.