Kenneth Stevens, the Clarence J. LeBel Professor Emeritus of Electrical Engineering and Computer Science at MIT, whose pioneering work at the intersection of engineering and linguistics helped earn him the National Medal of Science, died Aug. 19 in Clackamas, Ore., from complications of Alzheimer’s disease. He was 89.
Born in Toronto on March 23, 1924, to British-born parents, Stevens lived there until 1948, earning both bachelor’s and master’s degrees in engineering physics from the University of Toronto. Having lost an eye to cancer when he was four years old, Stevens was ineligible for military service during World War II, but after completing his master’s, he stayed at the university for three more years, teaching returning soldiers under the Veterans Rehabilitation Act, the Canadian equivalent of the G.I. Bill. One of these students was his older brother, Pete.
In 1948, Stevens came to MIT as a doctoral student in electrical engineering. He would spend the next 59 years at the Institute, joining the faculty in 1954 and retiring, in 2007, at age 83.
In an oral history recorded for the American Institute of Physics, Stevens recalled that, when he was finishing his master’s, “control theory was a big thing, so I did a master’s thesis on what we called servomechanisms.” He expected that he might continue that work at MIT.
But, as it happened, MIT professor of communications engineering Leo Beranek — also one of the founders of BBN Technologies, a research and development organization that is now a subsidiary of Raytheon Co. — needed a teaching assistant for a class on acoustics. “Beranek noticed that in my vita, I had once taken a course in acoustics, so he got in touch with me and asked if I would be a teaching assistant,” Stevens recalled. “So my transition to acoustics was very much by chance.”
Originator of the ‘quantal theory of speech’
That chance transition launched him on a career that culminated in his winning of the National Medal of Science in 1999. Stevens is best known for his “quantal theory of speech,” which explored why — despite the apparent diversity of sounds across different languages — human speech actually exploits only a small fraction of the sounds that the vocal tract can produce.
In 1952, while Stevens was completing his doctorate, the MIT linguist Morris Halle, together with colleagues Gunnar Fant and Roman Jakobson, proposed that all human speech sounds could be described as combinations of 20-odd “distinctive features,” such as the placement of the tip of the tongue, the shape of the tongue, whether the glottis (voice box) was opened or closed, the shape of the lips, and so on.
Stevens, who collaborated closely with all three men, observed that these distinctive features seemed to describe configurations of the vocal tract’s “articulators” — such as the tongue, glottis and lips — in which small deviations had little effect on the sounds produced. This is by no means true of all configurations: In most cases, small deviations would actually yield large sonic differences. But, Stevens argued, language users would naturally converge on the more stable configurations, which would lead to greater consistency in sound production.
Quantal theory was not, however, just a theory of speech production; it was also a theory of speech recognition. If humans had a limited repertory of sounds that they could produce reliably, then the auditory system may very well have evolved to key in on them. Stevens spent much of his career indefatigably investigating the implications of quantal theory, both experimentally and through mathematical modeling, frequently in collaboration with Halle and, later, with Samuel Jay Keyser, another MIT linguist.
A draw for graduate students
Along the way, he advised more than 60 graduate students — including four after his retirement, when he and his second wife, Sharon Manuel, moved to Oregon.
“When I was looking for a place to get my doctorate, all I knew was that I wanted to study with Ken Stevens,” says Victor Zue, the Delta Electronics Professor of Electrical Engineering and Computer Science at MIT. “I didn’t come to MIT and then find Ken Stevens. I wanted to study with Ken, and OK, he’s at MIT, so I guess that’s where I will apply.”
“I always remember Ken as an incredibly tolerant advisor,” Zue adds. “He always allowed people the latitude to explore areas that weren’t central to his own interests.” As Dennis Klatt — who developed the speech synthesizer used by Stephen Hawking, among many others, as a researcher in Stevens’ group — once put it, “As a leader, Ken is known for his devotion to students and his miraculous ability to run a busy laboratory while appearing to manage by a principle of benevolent anarchy.”
Joseph Perkell — who took a summer job in Stevens’ group in 1964, returned as a PhD student in 1969, and remained with the group as a research scientist until his own retirement last year — recalls a particularly telling instance of Stevens’ devotion his to students. When he began working with Stevens, Perkell was a dental student at Harvard University, and, he recalls, “I did some work that Ken thought would make a good research monograph.” After completing his dental degree, however, Perkell went on active duty for two years as a dental officer in the Army.
“While I was in the Army, Ken took the work that I had done and put it together and wrote it up and submitted it to MIT Press — with just my name on it,” Perkell recalls. “So my first publication was a hardbound research monograph, and I was the sole author, even though he was the one who pulled the writing together. That’s the kind of person he was.”
Beyond MIT, a love of the outdoors
Stevens married his first wife, Phyllis, in 1957; together they had four children, two of whom went on to careers in the sciences. Stevens’ son Michael, a research physicist at the Naval Research Laboratory in Washington, recalls that, outside the lab and the classroom, Stevens loved the outdoors — skiing, hiking, camping and canoeing.
Michael Stevens remembers one summer in particular: His parents decided to rent a van and drive from Cambridge to Yellowstone National Park and back, camping out along the way, with their own children and two of their children’s friends — “an enormous undertaking, by any measure,” he says. “I remember that, lying down in the tent every night, he would put his hands behind his head, and he would say, ‘This is the life.’”
When they arrived at Yellowstone, the first night was cold enough that several of the campers took refuge in the van. Michael Stevens recalls waking early to find his father etching a fake claw print with his fingernail in the frost on the van’s windshield. When the rest of the group woke up, he played along with his father’s tale of a nocturnal ursine visitation.
Stevens’ first marriage ended in divorce in 1981, and in 1994 he married Manuel, a linguist who had received her PhD from Yale University in 1987. In his mid-70s, Stevens became a father again when the couple adopted two infant girls from China. Each of the girls — Kendra and MacKenzie — has her father’s first name embedded in hers, along with their mother’s surname.
“He was much more active, physically, than I am, so he would do more running around with them,” Manuel says. “He was a modern dad. He at least had plans of doing 50 percent of the work around the house.” But, she adds, he had some difficulty realizing those plans as, well into his 70s, “he was still working 60, 70 hours a week.”
“His personality was unique,” Michael Stevens says. “He had no pretensions about him. He had no arrogance about him. He was humble until the very end.”
Stevens is survived by his wife, Sharon Manuel, of Clackamas, Ore., and their two children, Kendra and MacKenzie. He is survived by four grown children by his first wife, Phyllis Stevens: Rebecca Stevens of Silver Spring, Md., Andrea Stevens of Schwenksville, Pa., Michael Stevens of Clarksville, Md., and John Stevens of Kenmore, N.Y. Stevens is also survived by five grandchildren.
Stevens’ family asks that donations be made in his name to the Alzheimer’s Association.