Editor’s note: This is the first in a series of articles linking the work of MIT’s emeritus faculty members with the current state of research in their given fields.
In 1978, a PhD candidate in linguistics named Donca Steriade arrived at MIT not long after leaving communist Romania. Steriade recalls that, excited about her studies and fearful of failure, she thought, “If I don’t know what they expect me to know, they’re going to send me back to Romania.”
Steriade took a class with Morris Halle and received a jolt. Halle, now 86 and an Institute Professor Emeritus at MIT, is one of the 20th century’s most influential academic linguists. He helped create modern phonology, the study of the production of sound in language.
“Morris was very stern with all of us,” says Steriade. “His first reaction, when he was not completely satisfied with our work, was to ask us if we wanted to stay in linguistics. He would say: ‘Do you really want to be in this field?’ My first year was rather traumatic, because Morris asked me that multiple times.” When Steriade handed in her first paper to Halle, it came back with a blunt comment: “A garden of horrors.”
Steriade had come to MIT to tackle a major problem in phonology: Determining the extent to which the broad laws that apply to sound patterns in all languages are due to the fact that all people share the same kind of vocal apparatus. Halle, for one, suspected that other factors played a role; Steriade wanted to see how much speech perception — our hearing — influenced the rules of spoken language.
In time, Steriade not only survived her first year of graduate school, but became an accomplished professor of linguistics at MIT. And now she even defends Halle’s prickly approach. “I think it was right to ask if we wanted this profession,” Steriade says. “It was difficult to find jobs, and you really had to be dedicated to the subject to seek this kind of life.” Moreover, students who did want to be in linguistics found themselves in a vibrant intellectual community.
“It is impossible not to get excited in Morris’ classes,” the linguists Stephen Anderson (now of Yale) and Paul Kiparsky (now of Stanford) once wrote in a published essay about Halle, noting, “what Morris says in class … is often outrageous, but it sets off something productive in his listeners.” Halle, they added, was “largely responsible for the special quality of life that has characterized linguistics at MIT.”
He is also something of an Institute grandfather. Two of Steriade’s own PhD students (whom she taught at UCLA), phonologists Adam Albright and Edward Flemming, are now MIT professors and have offices just down the hall in the Stata Center.
‘Noam, where’s your other office?’
Halle’s route to American academia was circuitous. He was born into a Latvian Jewish family in 1923. After Germany invaded Poland, Halle’s father, a businessman, moved the family to the United States in 1940. (Not all of Halle’s relatives left, however, and the Nazis soon decimated Latvia’s Jewish population.) Halle was drafted into the U.S. Army, served in France during World War II — “I didn’t have a glorious record; I just did what they told me” — then completed his graduate work at Harvard with the eminent linguist Roman Jakobson.
By 1951, Halle had also been hired by MIT. “I got a job at MIT mainly because I could teach languages,” he says. Linguistics at the Institute then consisted of language courses, not research, and Halle is fluent in English, German, French, Russian, Latvian and Hebrew.
MIT became the world center of linguistics research after Halle suggested it could use a promising young research fellow at Harvard: Noam Chomsky. “In the summer of 1955, Chomsky, with whom I was friends, needed a job,” says Halle. “So I went to the department head and I said, ‘Why don’t we hire Chomsky?’ ” MIT inaugurated its PhD program in linguistics in 1960.
In good MIT fashion, Halle and Chomsky had unglamorous offices next to each other. “When I came in 1978,” recalls Steriade, “the department was spread throughout Building 20, and Noam and Morris had offices that were the two most miserable holes in the whole place.”
As Halle now jokes: “I would say to Noam, ‘Where’s your other office?’”
In the 1950s, Chomsky upended linguistics by developing his theory of Universal Grammar, which holds that language is not simply acquired through social learning. Instead we have an innate faculty for language; we have to learn words and rules in any language, but all languages have underlying organizational commonalities. And while he focused on syntax — the principles governing the structure of language — Chomsky collaborated with Halle extensively to extend these ideas to phonology, culminating in their seminal 1968 book, The Sound Pattern of English.
In the work, Halle and Chomsky laid out for the first time the elaborate series of rules in English that turn written words into vocal utterances. These laws are linked to syntax. Our voices fall when we say “blackboard,” but rise when we say “black board.” The first is a noun, the second an adjective and noun; this difference in syntax underlies the difference in sound.
From Romania to Cambridge
Meanwhile, Steriade was clearing her own unusual path to MIT. Her father, Mircea Steriade, was a renowned neuroscientist who left Romania during the Ceausescu regime. Steriade studied classics in Bucharest, then left, too. “I wanted to pursue the project that Morris and Noam had defined,” Steriade says, referring to the MIT-based effort to find the universal elements of language. Her doctoral thesis examined the shared properties of syllables across languages, finding that, despite appearances, even ancient Greek syllables have important commonalities with those of modern languages.
Since then, Steriade has focused on modern Indo-European languages. Among other things, she used insights in speech perception to make new observations about a phenomenon called “neutralization,” in which distinct sounds in words become smoothed over to help speech flow. (One example in English is the way “writer” and “rider” sound alike.) Steriade has found that across languages, neutralization occurs where contrasts are hardest for listeners to detect. Thus the way we hear influences the evolution of the rules of speech. This is the kind of discovery Steriade hoped to make when she arrived at MIT in 1978: Spoken language is shaped by more than just our physical ability to speak.
Steriade has also become a valued mentor herself. “She has a great ability to make phonological theory exciting,” says Flemming. It seems Steriade has never called a student’s work “a garden of horrors,” either. “I turned in plenty of papers to Donca, and I never got a comment like that,” allows Flemming. “She can be very charitable toward modest papers if they at least have interesting ideas.”
While Steriade proudly identifies herself as a student of Halle, today they take opposite sides of a prominent linguistics debate over a concept called Optimality Theory, which differs from Halle’s account about the mechanisms we use to turn words into sounds . Why do we pronounce the final “c” in “electric” with a “k” sound, but then give the same letter an “s” sound when we say the word “electricity”? Why does the “s” in “consign” sound like, well, “s,” while in “resign” the same letter now has a “z” sound?
Halle believes that as we learn to speak a language from others, we acquire a kind of pronunciation checklist with an order of operations; this allows us to make correct alterations even as we learn new rules . “This is the most interesting puzzle in the sound structure of language,” says Halle. “The answer, I believe — and I taught this to Donca, whether she agrees or not — is that there are computational steps. So one step is: Change ‘s’ to ‘z’ [as a sound] if it comes between vowels. Another rule says: Change ‘k’ to ‘s’ [as a sound] before ‘ity,’ as in ‘electricity.’ A small computer program in your head does that.”
In the early 1990s, some phonologists began arguing that we do not alter sounds through these steps, but instead use a set of “conflicting constraints” arranged in a hierarchy. “Not all of these conflicting preferences can be satisfied at once,” explains Steriade, “so you have to prioritize among them.” A preference that we keep the “k” sound in “electric” might be overruled by a preference for the “s” sound when possible. Optimality Theory’s advocates still seek universal components of language; however, they think learning a new language entails acquiring a new ranking of globally similar preferences.
Steriade diplomatically claims that in one way this debate “is a tiny difference that has been magnified in contemporary phonology.” We may indeed run though a sequence of computations while turning underlying words into sounds, she suggests, so in this regard, while optimality theorists “hoped they were going to eliminate the view Morris has, it’s become obvious that’s not possible.”
On the other hand, Steriade believes there is still a “fundamental conceptual difference” between the views. While Halle describes words becoming sounds through a more arbitrary, ad-hoc series of conventions that evolve in a given language, Optimality Theory asserts that the conflicting preferences that apply to pronunciation are not arbitrary at all.
In theory, Halle should be amenable to whatever the balance of evidence suggests. In 1974, Halle delivered an address to the Linguistic Society of America in which he declared, “the linguist must be prepared to lose as well as to win.” How does that idea strike him today? “I was younger then,” quips Halle, adding: “If you believe something, put your money where your mouth is, and 10 years later you will know if you were right or wrong. Nobody else will need to tell you when you’re whipped.”