Language Processing: A Training Program
Graduate Training in Psycholinguistics
at the University of Illinois in Urbana-Champaign






Faculty Research Interests

Kathryn Bock

Fields of professional interest: Psycholinguistics, cognitive science

Kay Bock's work focuses on how people turn thoughts into speech. There are three questions behind much of the research. One is how the features of ideas affect language forms. A prospective graduate student might say either "The application took forever to fill out" or "It took forever to fill out the application". What determines which kind of sentence is used? Another question has to do with the cognitive processes that control how words are arranged. Speakers begin with ideas or scenes whose components may be present to the eye -- or the mind's eye -- all at once. But the words to communicate those ideas have to be spoken one at a time. If a picture is worth a thousand words, the question here is: In what order? How does the speaker manage the flow of information? The last question, closely related to the second, involves what goes wrong when speakers make errors in selecting or arranging words. If you say "I have a room in my phone" when you mean "I have a phone in my room," what slipped? And why?

The answers to these questions draw on theories about language and about cognitive processes, and develop from new experimental techniques for examining language production. These methods include measurements of the time needed to produce individual words, assessments of changes in sentence structure in response to changes in the linguistic context, induction of speech errors, and tracking the eye movements that precede and accompany the description of perceived events. In addition to normal adult speakers of English, the research extends to brain-damaged speakers and normal speakers of other languages.

Gary Dell

Fields of professional interest: Psycholinguistics, cognitive science; connectionist models.

Gary Dell's work deals with how people produce and understand sentences, and how these processes can be modelled using neural networks. For example, his research on language production attempts to understand production errors or "slips of the tongue." He has developed a neural net model that makes predictions about the qualitative and quantitative properties of speech errors. These predictions are tested using experimental procedures in which subjects produce words and sentences under controlled conditions. A particularly interesting aspect of the model is that it can be used to understand patterns of behavior resulting from brain damage. By changing the processing characteristics of the model, one can produce speech error patterns that are characteristic of certain types of aphasic patients.

Kara Federmeier

Human brains store vast amounts of knowledge about people, places, objects, events, and the words that describe them. Information is retrieved from this knowledge base, known as semantic memory, often in response to a spoken or written word. Remarkably, in only hundreds of milliseconds the brain of a language user can make sense of a complex, inherently meaningless signal using information from the ongoing discourse and from the environment. Federmeier's research program aims to uncover the mechanisms that support such complex, yet seemingly effortless, meaning processing. In particular, current and recent research has emphasized (1) how language context information and the structure of semantic memory interact to facilitate word processing in real time, (2) how information derived from different modalities comes to be represented in and accessed from long-term semantic memory, and (3) how the two cerebral hemispheres coordinate the task of building meaning representations.

Language contexts play a crucial role in shaping memory access, as, during comprehension, semantic and structural constraints make some words more likely to occur than others. These context effects have a profound influence on the speed and accuracy of word processing, as manifested both behaviorally and neurophysiologically. Federmeier's electrophysiological studies have revealed that a crucial part of a sentence context's influence on word processing is mediated through long-term memory structure: the physical and functional similarity between two semantic category members impacts processing, even when such relationships do not alter a word's subjective plausibility in its context. Moreover, this impact is actually more pronounced when contextual constraint is strong, suggesting that the brain uses context information to predict the semantic and possibly perceptual features of upcoming items (Federmeier & Kutas, 1999). Current work is examining different types of semantic relationships and various levels and types of contextual constraint to uncover more detail about the type of feature information used during comprehension and the mechanisms underlying predictive processing.

Research Federmeier has conducted in collaboration with a number of undergraduate students and visiting graduate students suggests further that people's use of world knowledge varies as a function of mood (Federmeier, Kirson, Moreno, & Kutas, 2001) and the availability of cognitive resources (Federmeier, McLennan, De Ochoa, & Kutas, 2002), and also as a function of the representation type used to cue that information. Different semantic information is accessed from words and pictures, suggesting that semantic processing takes place in a distributed system in which input modality information is not completely lost (Federmeier & Kutas, 2001). Different electrophysiological responses to words from different classes (e.g., nouns vs. verbs) further supports the notion of a structured yet shared semantic system, one sensitive to the nature of the information associated with a word as well as to the roles it plays in various contexts (Federmeier, Segal, Lombrozo, & Kutas, 2000). Current explorations of these issues are aimed at delineating when, where, and how brain areas involved in the processing of different types of information come together on-line to create the unified multimodal concepts we experience as meaning during language processing.

Finally, combining electrophysiological measures with visual-half-field techniques (used to preferentially stimulate one cerebral hemisphere), Federmeier has shown that the two hemispheres differ in how they use word and picture information to access world knowledge (Federmeier & Kutas, 1999; 2002). Whereas left hemisphere-initiated processing seems oriented toward prediction and the use of top-down cues, right hemisphere-initiated processing seems biased toward the veridical maintenance of information and integration with working memory. Current studies in the lab are testing the predictions of this framework, looking at hemispheric differences in language processing in conjunction with explorations of hemispheric differences in factors such as recognition memory and working memory.

 

Cynthia Fisher

Fields of professional interest: Language acquisition, psycholinguistics.

Cynthia Fisher's work explores questions about how children rapidly acquire the grammar and lexicon of their native languages. One area of her research explores what information children use to figure out the meanings of new verbs. Experimental evidence suggests that very young word learners use verbs' syntactic contexts to narrow down hypotheses about their meanings. Experiments explore how this works by introducing unfamiliar verbs to toddlers and preschoolers in different grammatical contexts, and using a variety of techniques to assess what the children take these new words to mean. These tasks include, for the youngest children, measuring visual fixations to two simultaneously presented pictures or videos as children listen to a soundtrack.

Theories of the role of sentence structure in early interpretation, and other fundamental issues in language acquisition, all depend on assumptions about what children take in of the linguistic input that surrounds them. Recent studies in Fisher's lab have begun to examine effects of experience on word recognition and the representation of sentences by very young children. To identify spoken words, children must build representations in long-term memory of the sound patterns of words, and use them to identify new instances of those words in ongoing speech. Many perceptual problems must be solved along the way, including the segmentation of words from continuous speech and compensation for sources of variability in speech. Fisher has adapted an auditory priming paradigm to examine mechanisms for change in children's representations of spoken words. This work is designed to explore the mechanisms by which language processing systems adapt to experience, and the representation and organization of linguistic input during the process of language acquisition.

 
 

Susan Garnsey

Fields of professional interest: Psycholinguistics, cognitive science, cognitive neuroscience.

Susan Garnsey's work investigates how people understand language, and especially how they do it so quickly - essentially as soon as they see or hear it. She uses a variety of techniques to tap into comprehension as it proceeds over time, including reaction times, eyetracking, and event-related brain potentials. She has also recently begun to use functional magnetic resonance imaging to address questions about the contributions of various brain regions to different aspects of comprehension. Most of her work has focused on how people combine specific knowledge about individual words with more general knowledge about sentence structures in coming to understand sentences. She has also investigated what people remember from one sentence to the next, with the goal of characterizing how they combine what they remember from previous sentences with the words that are currently coming in. Most recently, she is collaborating with Professor Hiroko Yamashita on studies comparing comprehension in native speakers of English and Japanese, in an attempt to determine which aspects of language comprehension are common to all languages, and which specific to particular languages. Work such as this will lead to a better understanding of the inherent flexibility in the human language capacity and how it is shaped by properties of the particular languages that are learned.

 
 

George McConkie

Fields of professional interest :Cognitive psychology, visual perception, psychology of reading.

George McConkie's research investigates how people perceive and acquire information from pictures, diagrams and other visual displays. In his laboratory, viewers' eye movements are monitored as they examine computer-controlled displays. In this environment, the display is dynamically changed in response to the viewer's own eye movements. For example, one can detect the eye movement that is taking the eyes to a preselected part of the display and, while the eyes are still en route, change the object that occupies that location. The display is then different when the eyes pause at the end of their movement than it was prior to the movement. This technique is called "eye movement contingent display control." It allows precise control to be exerted over the nature of the stimulus display that is present from moment to moment in response to the viewer's own behavior.

The eye movement contingent research technique makes it possible to study, in real time, the perceptual and cognitive processes of viewers as they examine pictures. McConkie is particularly interested in the dynamic interplay between the visual input, the viewer's prior knowledge, and the constraints imposed by the viewer's task, and how this interplay determines the deployment of visual attention and the control of eye movements. He is also concerned with the process of constructing a mental representation from the pictured information.

 
 

Kevin Miller

Fields of professional interest :Cognitive development, linguistic and cultural influences on development, relations between implicit and explicit knowledge.

Kevin Miller's research concerns the way in which language-based symbol systems affect the development of children's thinking. Symbolic systems are fundamental to the way we understand some of the most basic domains of human experience, including time, number, and language. Because symbols play such a fundamental role in the way we think about the domains they represent, research on the effects of symbols on cognitive development must look at how variations in symbol systems are associated with variations in learning.

Kevin Miller's research group has been studying how differences in symbol systems in two very different languages-Chinese and English-affect the course of cognitive development in Chinese and American children. For example, Chinese number names form a clear base-ten system in a way the English names do not. Compared to Chinese children, American preschoolers have trouble a) learning their language's set of number names; b) realizing that number names are formed around a base of ten; and c) mapping spoken names onto Arabic numerals. Differences in calendar systems and in the relation between spoken and written language in Chinese and English also are reflected in differences in children's acquisition of symbolic systems for these domains.

These cross-language studies provide a powerful method of determining the role that cognitive tools, such as number naming systems, can play in children's early understanding of the domains over which those tools apply, and of distinguishing universal problems in cognitive development from those that are contingent upon a particular system of symbolic representation.

 

Edward Shoben

Fields of professional interests :Memory, categorization, world knowledge, and comprehension.

Ed Shoben's primary research interests lie in two related areas. First, he is studying how people store and use magnitude information. For example, how do we know that sheep are larger than coyotes, that George Bush is taller than Jimmy Carter, or that leaving the tip comes after eating the meal. Our current work suggests two important principles: First, categorization into magnitude categories is an important and obligatory part of the decision. Second, information about relative magnitude does not flow in discrete jumps as most models assume, but instead flows continuously.

His second interest is generally in categorization and particularly in conceptual combinations. Specifically, how do we know that a wood stove is one that uses wood, but a honey bee is one that makes honey? One hypothesis that we are currently exploring is that people know what kinds of relations concepts can enter into and thus common ones are more readily comprehended than rare ones.

In addition to these interests, he is also interested in the differences (or lack thereof) between concrete and abstract materials and in the utility of various distinctions about memory.


return to Learning Processing home page