Cognitive science, the study of the nature of the human mind and cognition, is by definition interdisciplinary. This emerging field has generally been a cooperative effort of computer scientists, linguists, psychologists, philosophers, anthropologists and neuroscientists. UB's approach is even broader, including professors from geography, industrial engineering, biophysics, physiology, communication sciences and education.

"We're at a unique point in history," said Leonard Talmy, director of the center and associate professor of linguistics at UB. "After decades of science moving in ever more specialized, differentiated directions, the emergence of cognitive science signals a return in the other direction."

That directional change intensified in 1994, when UB hosted the First International Summer Institute in Cognitive Science, the first major international forum held exclusively for cognitive science. During the month of July, some 500 professors, researchers, students and industrial and government scientists from 32 countries attended intensive classes, workshops and lectures.

"In part because of that conference, Buffalo is now regarded as a major crossroads in the world for cognitive science," Talmy said.

Talmy cautioned that the number of universities with centers or groups approaching the study of the mind this way is still small. UB is one of only about 20 in the U.S., and only about 30 in the world, that has a formalized cognitive-science graduate track.

In recent trips to other universities that have an interest in studying cognition, Talmy has noticed how far ahead of other schools the UB effort has come. "At many schools I visit, the researchers ask very traditional kinds of questions," he said. "I don't hear echoes in them of all the additional understanding you can gain from disciplines outside your own."

The Spoken Language Group includes researchers from linguistics, communicative disorders, neurology and psychology. Operating with annual funding of $100,000 under a National Institutes of Health training grant, the group does research ranging from how infants recognize their names to what parts of the brain are responsible for producing and understanding speech.

"The problems we deal with are so large, no one individual could hope to tackle them," explained James Sawusch, associate professor of psychology and a member of the group.

Studies performed by group member Peter W. Jusczyk, UB psychologist, have shown that infants as young as six months become attuned to the characteristic sound patterns of words in their native languages.

Other projects of the Spoken Language Group are designed to explore the perceptual processes of adults.

"I deal with the sounds of language, not words," explained Sawusch. "We want to find out what in these sounds is most eye-catching, or rather, 'ear-catching.'

"If we can develop a normal model for how we perceive language, then we can begin to study disorders and the development of new treatments for them," he said.

The Cognitive Neurosciences/Neurolinguistics Research Group is using one of medicine's most advanced imaging technologies, the PET (positron emission tomography) scanner, combined with electrophysiological data, to pinpoint the parts of the brain involved in language processing.

One of a handful in the U.S. that has both a clinical and research capacity, the $12 million PET facility at UB and the Buffalo VA Medical Center is being used to study which areas of the brain are active while subjects perform language-processing tasks.

Project investigators are Robert D. Van Valin Jr., professor and chair of linguistics; Jeri Jaeger, assistant professor of linguistics; Alan Lockwood, professor of neurology and nuclear medicine and director of operations at the PET center, and doctoral candidate David Kemmerer. The research is funded by a UB multidisciplinary pilot project grant.

In these experiments, subjects are first asked to read aloud real verbs and nonsense words, such as "wug" or "spalk." Then they're asked to view lists of verbs with regular past-tense forms (such as "jump") and irregular past-tense forms (such as "ran"). For the nonsense verbs, the subjects convert them into past tenses, such as "wugged" or "spalked."

"When brain regions take on an additional workload, an increase in the amount of fuel is needed," explained Lockwood. "This leads to an increase in blood flow that we measure with PET."

While subjects perform these tasks, the PET scan records the rate of blood flow to various parts of the brain, compared to the rate of blood flow to the brain in the resting state. The result is a map of areas of the brain that are activated during specific language processes, such as forming the past tense.

"The past tense of English verbs is a hot topic in cognitive science," explained Van Valin.

The controversy has been over whether one or two systems allow the brain to process past tenses. Proponents of a single system say that people use either a single system of rules that allows them to produce past-tense forms of both regular and irregular verbs. A third group of researchers assumes that all forms, both present and past tense, are memorized and stored in the mental lexicon.

Linguists who believe in the dual-systems theory say that while past-tense forms of regular verbs are generated by the rule, irregular verbs are memorized and stored individually. The UB study's findings clearly bear this out, said Van Valin, demonstrating strong evidence for two distinct cognitive processes.

"It's not that past tenses are in themselves such an interesting phenomenon," said Jaeger. "The underlying argument here is over just what the brain is good at. How does the brain prefer to organize and process information?"

As linguists, Van Valin and Jaeger are interested in how humans process language in order to illuminate how the mind works.

But as a neurologist, Lockwood is interested in this research because of what it will reveal about disorders, such as strokes, that affect a patient's ability to process language.

"What allows some stroke patients to recover language processing and others not?" he asked. "What causes conditions where patients experience very subtle language impairments, such as in Parkinson's disease? The work that is being done here leads very naturally to these kinds of questions."

CASSIE, (Cognitive Agent of the SNePS System an Intelligent Entity) is a computerized cognitive agent they designed that communicates in natural language, the way humans do. The researchers key in questions and statements and CASSIE responds by flashing responses on the computer screen.

At a recent conference on artificial intelligence, Shapiro and CASSIE had a conversation that demonstrated CASSIE's ability to distinguish between talking about sentences and talking about beliefs. During the exchange, CASSIE seemed to understand that a person it was discussing believed one thing, but said another. SNePS has endowed CASSIE with an intelligence, an ability to reason about what it knows or does not know based on direct discourse.

"There are sentences and there are propositions, or beliefs," Shapiro said, "and we, as well as computers, need to talk about both of them."

A related focus of the SNePS Research Group is to discover how humans and cognitive agents can understand narrative.

"We are trying to understand how readers develop a 'story world' when they are reading narrative, how they keep track of what's happening and where," said Rapaport. "How do readers understand who is talking when they read an indirect quote?"

SNePS was recently incorporated into an intelligent electronic system that translates Chinese text into English and actually "understands" what it is translating.

The system utilizes SNePS as what's known as an interlingua, an intermediate, language-neutral representation.

"Because the system uses SNePS, it doesn't translate one sentence at a time," explained Rapaport. Instead, it translates all of the material into this interlingua, understands it and then translates it from the interlingua into English. "Since it understands everything it has translated, the system can be asked questions about it and it will respond," said Rapaport.

According to UB graduate student Min-Hung Liao, who developed the system, SNePS is much more powerful than other knowledge-representation systems.

"With SNePS, we can represent propositions or beliefs," said Liao. "Other systems cannot do that." Future applications for the SNePS-equipped translator might include a multilingual system that could instantly translate information on the World Wide Web into any one of several languages.

Additional projects being conducted by researchers with the UB Center for Cognitive Science include studies designed to: