Ken Stevens
passed away last week. He was without question one of the giants in the field
of speech. His contributions are legion – he is best known for the quantal theory of speech, a theoretical
framework that ‘explains’ the mapping between acoustic and articulatory events
and the basis for the finite inventory of speech sounds found in natural
language. Ken proposed that there are regions in articulatory space which give
rise to small acoustic changes and other regions which give rise to large
acoustic changes. These areas of change define the finite inventory of possible
speech sounds. That there are regions of acoustic stability suggests that there
is acoustic invariance in speech.
Namely, there are acoustic properties corresponding to phonetic features that
remain stable across phonetic context, position, and speaker. Ken’s work
provided an empirical basis for the distinctive features proposed by Roman
Jakobson, Gunnar Fant, and Morris Halle in their 1951 monograph Preliminaries to Speech Analysis. Distinctive
features have been considered to be the representational units underlying the
phonological inventories of natural language. Throughout his work, Ken emphasized the
importance of acoustic landmarks which define natural boundaries used by the
perceptual and ultimately the linguistic system for characterizing the
phonetic/phonological properties of speech. Ken, in collaborative work with
Morris Halle in the 1960’s, also proposed a model of speech, analysis by synthesis. Here, the input
signal matches to internally generated representations and ultimately selects
the ‘best match’. This theory has modern day applications and implications for
current Bayesian and predictive coding approaches to perceptual
processing.
I had the
good fortune to work with Ken over a period of 18 years. I first approached Ken
in 1974 on the advice of Roman Jakobson to see if I could learn acoustics while
on leave from Brown. I spent that semester at MIT and Ken and I worked on our
first project examining the acoustic and perceptual properties of retroflex
consonants. That was the start of a collaboration that resulted in a series of
papers and chapters on acoustic invariance in speech and the acoustic
properties of a number of phonetic parameters.
Ken’s
working style was and is something to emulate. Low-key, focused, collegial, his
lab was a bevy of activity. He was a fantastic leader of the Speech Group at
MIT welcoming students and colleagues from all over the world. The only
requirement was an interest and a commitment to research in speech. No one
would have known he had won the National Medical of Science, was elected to the
National Academy of Sciences, or had been the President of the Acoustical
Society of America. He was just Ken, a great researcher, mentor, collaborator, colleague,
and friend!
Sheila Blumstein
Brown University
