2003-2004 William James Fellow Award
Carnegie Mellon University
Jay McClelland’s contributions to psychology constitute a revolution in how psychologists think about cognitive and perceptual processes. McClelland was a pioneer in developing the principles of connectionist modeling and applying them to a broad range of psychological phenomena. His research has linked computational models that use neuron-like processors to the functioning of neurons in biological systems. He has demonstrated the power of such models to produce seemingly rule-governed behavior at a global level through the interactions of local, independent processing units. This work has merited such honors as membership in the National Academy of Sciences, the Grawemeyer Award, and now, the William James Fellow Award.
McClelland’s career can be divided into three stages. Initially, he was a behavioral scientist who was among the first to document the effects of top-down processing from words on letter recognition. In the second phase, he modeled the perceptual phenomena he had been studying. He began by challenging the idea that sequentially organized processes acted in a strictly serial manner, each passing the finished product of its computation on to the next. He suggested instead that sequential processes could act in a cascade, such that intermediate products of computation were passed on before the computation completed. This led to interactive activation models in which elementary representations iteratively passed activation to one another – such models were capable of learning and handled top-down processing in a direct way.
During this phase of his research career, in collaboration with David Rumelhart, McClelland developed the revolutionary architecture called “parallel distributed processing,” in which the elements that passed activation had no particular identity of their own and representations were instantiated by activation across the system as a whole.
His research then moved on to its third states, in which he merged his modeling interests with their emerging context, the study of the brain itself. Through modeling, he has distinguished between learning in the hippocampus and the neocortex, a dichotomy that helps to explain how rapid assimilation of the here-and-now co-exists with slow learning of categories and rules.
McClelland currently serves as the founding director of the joint University of Pittsburgh/Carnegie Mellon Center for the Neural Basis of Cognition, which he has guided from its origins to a major force in bringing diverse disciplines to the study of neural structures underlying basic human cognitive abilities. He has united computer scientists, neuroscientists, psychologists, roboticists and even artists in his efforts toward this important goal. McClelland merits the William James Fellow Award for the ground-breaking trajectory of his career to date and the undoubted achievements of the future.