Mark Gluck

Gluck works at the interface between cognition, animal learning, behavioral neuroscience, and computational modeling. He combines his training in human and animal psychology with neural-network analyses to understand the mechanisms of learning and memory.

His earlier work in cognitive psychology demonstrated how behavioral properties of animal learning can be related to higher-order forms of learning in humans. Currently, his lab works to to identify, model, and empirical evaluate fundamental components of both animal and human memory systems, with a special emphasis on the functional role of the hippocampus in learning.

1996 - National Science Foundation (NSF) Presidential Early Career Award for Scientists and Engineers (Social, Behavioral, and Economics Sciences Directorate).

"For outstanding contributions to understanding the cognitive neuroscience of human learning by evaluating computational models of neural networks that relate brain mechanisms to emergent behaviors and integrating behavioral and psychobiological approaches to animal and human learning."

1996 - American Psychological Association (APA) Distinguished Scientific Award for an Early Career Contribution to Psychology (in Behavioral Neuroscience/Animal Learning and Behavior).

"For his many contributions at the interface between computational modeling, associative learning, and behavioral neuroscience. His original and broadly conceived research on historically important processes and issues in learning and memory has demonstrated quantitative sophistication and an ability to appreciate the value of several levels of analysis. In the process, he has shown how connectionist modeling can benefit from being constrained by both behavioral and physiological data. Allready, his work is having a major impact in the exciting new area that is emerging at the intersection of the cognitive and neurosciences."

1992 - Office of Naval Research (ONR) Young Investigator Award (Cognitive and Neural Sciences Division).

Mark spent much of his early life surrounded by mathematics; his father, mother, and aunt are all mathematicians or mathematics teachers. His early exposure to mathematics lead to an interest in computers when his junior high school acquired an early teletype terminal. Soon Mark began staying after school and working between (and during) classes on the computer. Frequently late for class because of his work on some new computer project, Mark could often be seen running down the school hallway with the pink paper tape storing his latest program unravelling in a trail behind him. Mark's first exposure to the potential of computer modelling as a scientific tool came in high school when his father took him to a conference on mathematical biology. After seeing his father derive an analytic solution to a problem in evolutionary theory, Mark found that he could arrive at an identical solution through computer simulation. Entering Harvard as an undergraduate in the fall of 1977, Mark was torn between studying mathematics, computer science, or perhaps something entirely different. Growing up in academia, it had been common for Mark to hear discussions of intelligence and creativity, and he always wondered what we really knew about these concepts. This led him to the Psychology Department and a seminar given by a new Assistant Professor, Stephen Kosslyn. For his final paper, Mark proposed several experiments on perceptual access of semantic memory. Kosslyn invited Mark to return the next semester to run these experiments as an independent research project. Later, Mark took a graduate seminar in Mathematical Psychology with W. K. Estes. As the editor of Psychological Review, Estes was publishing many of the early connectionist modelling papers by Grossberg, Anderson, McClelland and Rumelhart which he passed on to Mark to read. Excited by this new brain-style approach to cognitive modelling, Mark began to work under Estes' supervision on connectionist models of basic levels in category hierarchies, seeking to explain some of the data he had collected earlier with Kosslyn.

While at Harvard, Mark also profitted greatly by interacting with many of the Psychology graduate students including Robert Nosofsky, Martha Farah, Jamshed Bharucha, Laura Pettito, Beth Adelson, Pierre Jolicoeur, and Steven Pinker (Mark's sophmore tutorial leader). Mark was also active in the college mountaineering club, ski team, and worked as a DJ at the college radio station. In addition to his school activities, Mark began programming on the side, including a summer working on the London Metals Exchange, and collaborating with the owner of an Irish pub in Cambridge on a software system for bar and restaurant management. Midway through college, Mark took a leave of absence for nine months, during which he worked on his research, his computer consulting, and then spent several months traveling and hiking in Nepal and India. After college, Mark worked for half a year at the Institute for Defense Analyses (IDA) in Princeton and then spent several months travelling up the Amazon River in South America, going from Brazil to the Andes Mountains in Peru.

When Mark decided to go to graduate school in cognitive psychology, his advisors were unanimous in their recommendation that he work with Gordon Bower at Stanford, Kosslyn's own former Ph.D. advisor and Estes' former colleague. Arriving at Stanford in the fall of 1982, Mark began working with Roger Shepard and Amos Tversky while Gordon Bower was away on a sabbatical. When Bower returned, he and Mark began by working on experimental studies of category learning and probabilistic reasoning. But Mark soon grew frustrated with his lack of progress and depressed by his lack of direction; for a while he considered dropping out of graduate school to go into business. Then he heard Richard Thompson give a talk on the neural bases of classical conditioning. Learning of Mark's interest and prior work in network modelling, Thompson suggested the two of them work together on modelling the cerebellar substrates of rabbit eyeblink conditioning.

Having trained up to then as a cognitive psychologist, Mark knew relatively little about animal behavior; the prevailing zeitgeist in the early 1980s was that the study of higher human cognitive processes had little to gain from looking at elementary conditioning behaviors seen in lower animals. Working with Thompson, Mark began to study the history of of animal learning theory and was especially impressed by the simplicity and power of Rescorla and Wagner's 1972 model of classical conditioning. Recognizing that the Rescorla-Wagner model addressed many issues currently of interest to cognitive psychologists, Mark proposed to Bower that they use it to model human probabilistic category learning. Building upon the Rescorla-Wagner model's similiarity to connectionist learning rules, Mark turned his attention back to cognitive psychology with a new interest in using network models to seek a closer rapprochement between theories of animal and human learning. This became the core of Mark's Ph.D. dissertation with Bower, and included several experimental studies of human learning that verified suprising predictions of the Gluck and Bower model.

While still at Stanford, Mark continued his outside business activities. With a local computer magazine publisher, Mark organized and taught two-and three-day courses on the Theory and Applications of Neural Networks for engineers and programmers working in industry. With Bernard Widrow, a professor in the Engineering department and an early pioneer in neural networks, Mark organized an industry-government conference on Neural Networks for Defense Applications.

After receiving his Ph.D in 1987, Mark remained at Stanford for several years with support from the NSF, ONR, and Sloan Foundation while he continued to collaborate with Bower and Thompson. For recreation Mark took up kayaking. In the winter of 1988, Mark and a friend took a collapsable two-man kayak to Central America where they kayaked the length of the Belize River, from the Guatemalan border to the Caribbean coast. Mark had initially assumed that after leaving Stanford he would look for a job in a traditional Psychology department. But an intriguing alternative arose when Paula Tallal and Ian Creese took Mark on a tour of the construction site for the future Center for Molecular and Behavioral Neuroscience, at Rutgers University. As the three of them walked in hard hats among the girders and concrete, Tallal and Creese described their vision for a new research and training center that would emphasize integrative approaches to the study of brain and behavior. In 1991, Mark accepted their offer of a faculty position and moved to Rutgers-Newark. There he began to work on a problem that had interested him for some time: What is the functional role of the hippocampus in learning and memory? From a theoretical perspective, Mark was intrigued to note that many behaviors that are not explainable by the Rescorla-Wagner model are the very same behaviors that are eliminated by lesions to the hippocampal region. In collaboration with a new postdoctoral fellow, Catherine Myers, Mark began to work on the puzzle of hippocampal function. In 1992 he and Myers proposed a theory that argued that a wide range of superficially disparate conditioning behaviors that depend on an intact hippocampal region can be understood as being those that require adaptive changes in the underlying representation of stimulus events. The framework for this theory built drew on several sources from various fields including recent work in connectionist network theory, early models in animal learning theory, as well as the mathematical models of similarity and psychological space developed by Roger Shepard.

In 1992 Mark was awarded the Young Investigator Award from the Office of Naval Research (ONR) in the area of Cognitive and Neural Sciences. With matching funds from this program, Mark began work with Robert Kolesar of the Naval Ocean System' Command (NOSC) in San Diego to apply the hippocampal model to engineering problems in pattern recognition. Exploiting the ability of the hippocampal model to recognize novel patterns, Gluck and his Navy collaborators developed an automatic mechanical fault detection system for helicopter transmission gear-boxes. During preliminary testing, their system was able to confirm a significant gear problem in an active-duty helicopter that the Navy had incorrectly thought to be in fine working order. Mark's contributions in this area were recently cited by the Office of Naval Research, in a report to the President's Office on Science and Technology, as one of the ONR's three most exciting accomplishments of the last year. The Navy has now built a prototype hardware system incorporating this hippocampal model in a chip, and this system is currently undergoing test flights.

Mark's more recent efforts in computational neuroscience have moved in two distinct, but inter-related, directions. One line of modelling has shown how the Gluck and Myers model of hippocampal function can apply to a wider range of memory phenomena, including real-time temporal processing and probabilistic category learning in humans. Another line of modeling has sought closer contact with the underlying anatomy and physiology, mapping computational interpretations of hippocampal-region function onto more physiologically realistic models of the entorhinal cortex, the hippocampal formation, and the septo-hippocampal cholinergic pathways involved in learning and memory.

To test predictions of these models, and to garner additional data to constrain and inform future theory development, Mark has recently expanded the empirical component of his research. With support from the Johnson & Johnson Foundation and the Hoechst-Celanese drug company, he has built an animal laboratory at Rutgers to do lesion and drug studies of rabbit eyelid conditioning.

In a related line of research, Gluck and colleagues are currently studying learning and memory in hippocampal-impaired human amnesics. Working in collaboration with several different groups of memory researchers including Larry Squire, John Gabrieli, Barbara Knowlton, Mona Hopkins, Ray Kesner, and John Disterhoft, Mark has been testing amnesic patients using three classes of increasingly complex learning paradigms: motor-reflex (eyeblink) conditioning , simple cognitive analogs of conditioning paradigms, and the probabilistic category learning tasks previously studied and modelled by Gluck and Bower. These neuropsychological studies have yeilded some preliminary confirmation for predictions of the Gluck and Myers theory as applied to human learning. In related work with colleagues at the NYU Aging and Dementia center (and with support from the National Institute on Aging and the Alzheimers Association), Mark's lab has shown that a battery of tasks predicted by the theory to be especially hippocampal-sensitive, are highly correlated with structural imaging assesments of hippocampal atrophy in non-demented elderly at risk for Alzheimers Disease.

In December of 1996, Mark was honored at the White House where he was presented with the National Science Foundation's Presidential Early Career Award for Scientists and Engineers in the area of "Social, Behavioral and Economic Sciences" in recognition of his outstanding contributions as a researcher and educator. In the award citation, President Clinton praised Gluck for his "outstanding contributions to understanding the cognitive neuroscience of human learning by evaluating computational models of neural networks that relate brain mechanisms to emergent behaviors and integrating behavioral and psychobiological approaches to animal and human learning." According to Assistant to the President for Science and Technology John H. Gibbons, this award "is the highest honor bestowed by the U.S. government on outstanding scientists and engineers beginning their independent careers.

In addition to his research, Mark has also been active as a teacher and mentor. His lab is run as a team in which everyone works inter-dependently on a core set of research issues. When not in the lab, Mark and his students oftengo as a group to the mountains for lab rafting, skiing, and hiking trips. Back in the lab, their research continues to concentrate on seeking to understand fundamental principles and mechanisms of learning and memory through the integration of behavioral, biological and computational approaches.