Please consider downloading the latest version of Internet Explorer
to experience this site as intended.
Tools Search Main Menu

Professor recognized for transforming understanding of human language

September 4, 2018

Michael K. Tanenhaus, a longtime professor of brain and cognitive sciences, is being recognized for work that has “transformed our understanding of human language and its relation to perception, action, and communication” by the premier academic society in his field.

portrait of Michael Tanenhaus.

University of Rochester Brain and Cognitive Sciences professor Michael K. Tanenhaus. (University of Rochester photo / J. Adam Fenster)

At the annual meeting of the Cognitive Sciences Society this summer, Tanenhaus was formally awarded the David E. Rumelhart Prize from the Cognitive Science Society and the Robert J. Glushko and Pamela Samuelson Foundation. The prize is the highest honor given by the Cognitive Science Society to recognize a “significant contribution to the theoretical foundations of human cognition.”

Tanenhaus, the Beverly Petterson Bishop and Charles W. Bishop Professor of Brain and Cognitive Sciences, says he’s humbled to join a list of honorees that include “giants in cognitive science from multiple disciplines, including computer science, linguistics, and psychology.”

Over the course of his 40-year career, Tanenhaus has focused his research on the mechanisms underlying language comprehension. He is best known as the creator of the Visual World Paradigm, which uses eye movements to study the mechanisms behind speech and language comprehension. According to the society, the paradigm has been widely adopted for studying language development and disorders.

Tanenhaus joined the Department of Brain and Cognitive Sciences in 1983. He was a long-time director of the Center for Language Sciences, an umbrella organization that brings together faculty, postdocs, and graduate students who conduct research on human language. The interdisciplinary center’s goal is to foster research and activities that reach across a broad group of disciplines, “allowing researchers to ask how language processing change across development or breaks down in developmental disorders.”

Tanenhaus earned his PhD in cognitive psychology from Columbia University.

What were your thoughts when you heard you had won the Rumelhart Prize?

To use the English expression, I was gobsmacked. The previous winners have been giants in cognitive science from multiple disciplines, including computer science, linguistics, and psychology, so it’s humbling to be considered worthy. Moreover, nearly all of the previous winners have been known for their theoretical contributions, typically for formal mathematical work or computational contributions. My work is theoretically driven, but I’m primarily an experimentalist.

What are some of the projects you’ve worked on that have “transformed your field,” as the awards committee indicated?

 All of my influential work was conducted in collaboration with a remarkable group of colleagues and graduate students and post-docs who flowed through the lab. Consider the development of the Visual World Paradigm. This paradigm allowed us to ask questions about spoken language processing that ranged from speech perception to interactive conversation. Students led these projects. Their projects formed the initial foundation for their research programs and pioneered application of the paradigm to new areas of inquiry within language processing.

What are some specific examples of projects that use the Visual World Paradigm?

One example project investigated how listeners recognize words in continuous speech, using eye-movements to trace comprehension as words unfolded over time. We asked subjects to follow instructions to click on and move clip art images, for example, “Click on the beaker,” in a display that included four pictures: a beetle, beaker, speaker, and carriage. The initial sounds of “beaker” would be consistent with either beetle and beaker, which are sometimes referred to as “cohorts,” and that was reflected in an equal number of fixations to the beetle and the beaker.

Another project asked people to pick up an object—for example, a cube—and “Put it in the can” when they were presented with two cans. Definite reference—that is, using “the”—requires that there be a unique referent: in this example, a single can. If the cube fit in both cans, people were confused. If, however, the cube fit in only one of the cans, then eye movements showed no delays, not even a temporary confusion. This experiment, and ones that followed, demonstrated that uniqueness takes into account the particular actions and objects in the language, seamlessly integrating linguistic processing, visual perception, and action. This result is about language processing, not just problem solving. If the instruction used the word “a,” suggesting that there was more than one relevant can, then eye movements revealed that people were temporarily confused.

What made you most interested in studying how humans comprehend language?

I grew up in a family that emphasized literature and language. While in high school in Iowa some graduate students from the Speech and Hearing Science Center gave a presentation about speech and language disorders that I found interesting. In college I was a research assistant for a professor who had studied with one of the giants of behaviorism. I happened to read a few chapters of a collection of essays and reprinted papers on applying behaviorism to language, both by proponents who viewed language as the final frontier for behaviorism and critics. Some of the critics were pursuing a new field called psycholinguistics.  I asked my advisor what he thought about language (in the context of behaviorism). He said, “I avoid thinking about it as much as I can.” I immediately thought, “Then this must be worth thinking about.”

What are some of the overarching questions you seek to answer with your research?

One of the main puzzles is: How do humans so rapidly construct the contexts that allow us to develop rich interpretations so quickly? As I mentioned in the cube example, definite noun phrases assume a unique identifiable referent. The definite noun phrase “the wine bottle” in “Pass the wine” assumes that there is a single identifiable bottle. However, speakers use definite noun phrases in situations where there are multiple bottles of wine—think of a restaurant with tables close to one another, where you are sitting at a table with multiple bottles of wine. You interpret “the wine” to be the bottle that is most relevant in the context. But how do we know what the relevant context is and compute it rapidly enough to constrain our interpretations?

What are some of your current research projects?

These are a few projects I’m most excited about:

  1. In interactive conversation, it’s important to take into account who is the better “source” of a piece of information. Sometimes this is easy—I’m a better source of how many children I have than anyone I might speak with. But what about other situations, like where someone says, “I think that’s a chickadee.” We relate objective certainty/uncertainty with how people express uncertainty.
  2. Most people in the field believe that listeners map the sound of the word onto a name that is “activated” by the pictured referent. In my lab, we’re exploring another hypothesis. We believe that as we hear a word, perceptual representations become active and we use these to map the word onto potential referents. If we use potential target pictures that have a dominant and a subordinate name (e.g., a picture that could be called a couch or a sofa), people are as likely to look at that picture if the target word is “soda” as they are if it is “cow” (the display contains a picture of a cow, soda, the couch/sofa, and an unrelated word). However, if people have to generate a name to remember the picture, then we see fixations to the synonym picture only for “cow.” This result is a first step toward understanding how different types of representations are coordinated to achieve efficient communication.
  3. In casual conversational speech, many of the most frequently occurring words are highly coarticulated. For example, the acoustic signal can be identical—depending on the context—for “saw,” “saw a,” and “saw uh,” where the “uh” is a filler. By manipulating the surrounding speech rate we can change whether the same acoustic sequence will be heard as “saw a raccoon swimming” or “saw raccoons swimming.” Eye movements let us determine that listeners generate probabilistic alternative possibilities, which influence and are influenced by down-stream information.

The Rumelhart Prize has been presented annually since 2001. Previous recipients include faculty members at Northwestern University, the University of California at Berkeley, Tufts University, Stanford University, Indiana University, University College London, and other institutions.

A workshop titled “Real-world language: Future directions in the science of communication and the communication of science,” funded in part by the National Science Foundation, was held to honor Tanenhaus following the Cognitive Science Society meeting this summer.

Tags: , , , ,

Category: University News