A pair of talkative UCLA husband-wife linguistic researchers probe the mysteries of how and why we speak — and what we're really saying about ourselves when we open our mouths
The Science of Linguistics
Pot-ty. Par-ty . . . The woman speaks the words into a machine that could be out of the latest high-tech thriller. As she speaks, a custom-fitted "pseudopalate" rests snugly against the speaker's palate. Ninety-six gold contact electrodes line the instrument's surface, so that the portions of the roof of the speaker's mouth touched by her tongue can be displayed on a palate-shaped computer visual, at 10-millisecond intervals. Over and over again, she repeats the words.
This is no movie set — just one of the projects at the UCLA Phonetics Laboratory. "We're interested in how individual sounds are made, and how they change depending on what sounds they're next to," explains Patricia Keating, professor of linguistics and director of the lab. In this study, performed by one of Keating's students, the subject is the consonant "t." Previously, phoneticians had believed we made the "t" sound by briefly holding the tongue against the front of the palate. Using the visualization technique, UCLA researchers found that the formation of the quick "t" is dependent on what precedes it. When the "t" follows an "r" as in "party," for instance, many speakers curl their tongue back, starting in the middle of the palate, then brush forward to the front.
So what's it matter? Well, if you're a linguist interested in the cause of all things phonetic, a lot. "A typical linguist's attitude is that any language represents a gigantic puzzle," says Bruce Hayes, professor of linguistics at UCLA, who also happens to be Keating's husband. "We receive pleasure from putting the pieces together and finding new ones." But it's not all knowledge purely for the sake of knowledge. The work also has practical value, particularly with the development of speech synthesis and speech recognition. A major industrial effort is under way to get computers to speak, listen, understand and synthesize. "But," notes Hayes, "this can be done well only if you have a good handle on the structure of the language in question."
While researchers elsewhere in the department examine everything from syntax to semantics, Keating and Hayes represent a branch of UCLA's linguistics department that studies the sound of language. Phonetics has all the earmarks of a laboratory science — quantitative data, sophisticated equipment, even invasive procedures to get at how sound is formed and used in language. The two make a good team: Keating is a phonetician, Hayes a phonologist. While phoneticians view speech as a physical event, phonologists are grammarians of sound — they listen (generally, what the ear can detect suffices) and document how pitch and other aspects of sound are controlled by a system of linguistic rules, testing theories about the structures and functions of languages.
Both groups study languages ranging from the common to the obscure. Recently, Hayes and a Bengali-speaking colleague examined the intonation patterns of that little-studied language. They composed various syntactic structures, the Bengali speaker pronounced them, and Hayes charted the pitch contours. While some aspects of the analysis were easy to determine, the researchers' theoretical breakthrough came when they compared sentences in which the words were the same but varying the stresses changed the meaning. Finally, a coherent picture emerged, with the researchers concluding that the turning points — the pitch peaks and valleys — represent the structurally relevant sentence portions in Bengali.
"It's like any other experimental work," Hayes notes. "You take two sentences that differ only in the aspect you're trying to test, and then use the outcome in an effort to infer the general principles that apply."
Whether the language in question is one spoken in a large, industrialized nation or by a tribal group with a materially simple lifestyle, Hayes says that the intricacy of any grammatical and phonological system, if looked at with sufficient care, is overwhelming.
In that sense, he notes, linguistics research remains at a relatively early stage. While other sciences have advanced to the point where progress requires extremely technical work and a mix of skills seldom found in a single individual — hence, large-scale collaborations — linguistics researchers can still go it alone. "We've progressed far enough to show that we're asking legitimate questions," says Hayes. "But until we know more about the actual computations in the brain that manipulate language, it will remain difficult to prove many of our working hypotheses."
Countless linguists have studied English over the years, and yet, Hayes marvels, researchers continue to gain new knowledge. "Every English speaker has a tacit understanding of what sounds right and what doesn't, but in terms of ferreting out and understanding it explicitly, the job is far from complete," he says. "If you consider that an estimated 6,000 distinct languages are spoken today, the task of obtaining a complete description of the structure of every language begins to look impressive."
Meanwhile, Keating and her students have concentrated on documenting variations in the pronunciation of English words — work that is directly applicable to speech recognition, especially relevant in computer development. "If a computer is programmed to comprehend only one pronunciation of each word, you're going to have an ineffective system," she notes. So Keating and other linguists, working with databases containing phonetic renderings of words spoken by hundreds of English speakers, conduct computerized searches of the pronunciations and determine where groupings are possible. From a commercial standpoint, the question is how many programmed pronunciations for a given word will cover enough variations to make the system viable. For some words like "about," Keating has found as many as 15 pronunciations are needed to cover the bulk of what the computer will hear; for others like "farmers," only one is necessary. Keating has shared her findings with SRI, a company that manufactures speech-recognition products.
In addition to using pseudopalates to investigate how sounds change within a word, Keating has employed the instruments to observe how individual consonants are articulated differently, depending on their position within a sentence. Among other things, she discovered that English speakers mark the beginnings of phrases with a stronger articulation than if the same sound were produced at the end of the phrase.
"Most people believe that speech is all in your mouth, strictly a mechanical process of moving the articulators," Keating says. "Linguists have demonstrated that the knowledge of patterns in one's language is what's most important. The meaning of the sentence that we're trying to get across makes us articulate differently. This grouping into phrases is part of our higher-level linguistic knowledge."
Which gets to the core of where computer synthesizers could stand to improve. "A lot of them still sound mechanical," Keating says. "We're showing small effects that can produce a more natural sound." Typically, the synthesizers utter all sounds as if they're in the sentence's initial position — pronouncing them strongly. By weakening the articulation as the sentence progresses, Keating's findings suggest, a synthesizer could render a sound closer to natural speech. Closer, but so far, a long way from ideal — an ironic testament, Hayes suggests, to how much about everyday language remains a mystery.
In a nearby campus building, another husband-and-wife team, Elinor Ochs and Alessandro Duranti, are interested not just in a language's physical features, but in how it's used. Duranti and Ochs, the former a professor of anthropology, the latter a professor of applied linguistics, study communicative forms as they're used in real-life settings and as they relate to cultures and social systems. The questions they ask are contextual: How do same-language speakers communicate with each other when they come from vastly different cultures? How does one person shift linguistic gears when going from a business meeting to a dinner with friends, and then home to his or her family?
A language is much more than a system of grammatical rules," Duranti says. "It includes strategies for interacting with people in different situations. Language becomes a tool, a resource." Adds Ochs: "We're looking at language in context, and in that sense, this is part of a much larger intellectual movement to study how context affects behavior."
Examining language and communication in context is nothing new to many applied linguists, anthropologists, sociologists, educators and others across the UCLA campus, but Ochs and Duranti have spearheaded an effort to bring these researchers together for the first time. "We're all in different buildings and subscribing to different journals but, in fact, we have similar interests in this unnamed field," Duranti says. The field now has a home at UCLA: the Center for Language, Interaction and Culture (CLIC). Founded by Ochs, Duranti and a group of scholars, including sociology professors Emanuel Schegloff and John Heritage, the recently established center will document how languages are used in Los Angeles. "There's a big treasure out there," Duranti says. The center's founders also intend to develop an undergraduate course on language, interaction and culture in Los Angeles, taught across departments. Also on the agenda: an archive of languages and genres used in the city, including how the same language is used differently. And a project on miscommunication in Los Angeles.
The center's researchers focus not just on how people communicate, but on how they don't. One graduate student has conducted interviews and watched hours of videotape to examine interactions between Korean American retail merchants and their African American customers, focusing on socioculturally conflicting notions of respect. "It's about what's not said as well as what's said," Ochs explains. "An eye gaze, body posture or missing word can all have important connotations."
Duranti recalls feeling as if he had entered a Samoan village upon attending a church in Carson, CA, at the outset of a study focusing on Samoan children growing up in Los Angeles. "I thought, 'How is this possible? What does it mean?'" Duranti says.
One of Ochs' most interesting studies has described how American families communicate — for better or worse — over dinner. In the best cases, she found children learn about problem-solving best in the dinner setting, where family members recount the events of the day. "The person telling the story for the first time doesn't quite have it figured out, and it becomes a collective process," says Ochs, who was struck by the resemblance between these machinations and the scientific process. To test her new hypothesis about collective problem-solving, Ochs chose a quintessentially scientific field — physics — and has begun to look at the methods used by UCLA physicists to solve problems, and how their students become apprenticed through the language of the science.
Sometimes, as in Ochs' family-dinner study, the numbers are large — Ochs studied 20 families over four nights of dinners at which hundreds of stories were told — but often CLIC researchers opt instead for qualitative descriptions of smaller groups. In one case, Ochs and a student were able to make daily observations about a mother who had agoraphobia, a panic disorder typically reported from questionnaires and clinical interviews. "We were able to record this woman's words and to capture her interactions with family members in a way that gave us tremendous insight into anxiety," Ochs says. "Was she representative of all women with agoraphobia? I don't know. But we learned far more about this one person than we could have through a survey." Even in the larger family-dinner study, she adds, "I don't assume those 20 families are representative of all American families. I just think they're emblematic of something. They resonate with the experiences of many other families."
"We have a very strong empirical bias," says Duranti. "People talk about miscommunication, stereotypes and identity, but few actually try to see how that works. When a student comes in and says, ÔI'm interested in cross-cultural miscommunication,' we say, ÔOK, go and look, then come back and tell us what you saw.' You get into trouble when you rely on intuition rather than observation."
In that sense, advances in video and computer technology have added a dimension to this work that could not possibly be captured when researchers simply used audio tapes, which miss many of the facial and body nuances of communication. The CLIC researchers have a laboratory with sophisticated multimedia technology, which they use to analyze verbal and nonverbal behavior, isolating significant images and integrating them into their presentation or publication. "There's so much information in a single freeze-frame," says Duranti.
CLIC researchers can't study every interaction, he notes, but they attempt to examine the specific situations and communities that will help explain how different people make meaningful connections while maintaining their individuality, and why some are successful at it while others aren't.
We acquire a form of cultural knowledge along with the grammar of a language," says Ochs. "It's very deep and difficult to access. It's terribly important, and yet we don't have rules for it. Teaching an immigrant how to form a sentence is a lot easier than explaining the norms of social interaction." But for all our problems with miscommunication, says Ochs — whether between students and their teachers, communities and the police, or across generational lines in a single family — what's remarkable, considering the vastly different life experiences participants bring to an interaction, is that exchanges almost always work.
"We do pretty well 99.9 percent of the time," she says. "People are able to share things, even when there's not a perfect match."
— D.G.
Return to Front Page || RESEARCH@UCLA
Copyright 1996 Regents of the University of California - All Rights Reserved.