Professor, Department of Psychology Director, Center for Mind, Brain and Computation Stanford University 344 Jordan Hall, Bldg 420 450 Serra Mall Stanford, CA 94305 +1-650-736-4278 jlm@psych.stanford.edu Curriculum Vitae and Career Highlights. Publications and software. Interviews and news articles.

Recent Book

Research Interests

One line of research in my lab was inspired by the striking pattern of spared and impaired memory in patients with damage to the hippocampal region, suggesting that different parts of the brain play different, specialized roles in memory. In a paper with a colleague and a former student (McClelland, McNaughton, and O'Reilly, 1995), we suggested that the hippocampus and neocortex may play complementary roles in learning and memory. The neocortex uses a very gradual learning procedure that allows it to exploit the structure in ensembles of inputs. The hippocampus is needed to complement the neocortex, providing a mechanism for rapid learning of the specific arbitrary aspects of particular items. The results of this rapid learning are gradually integrated into the neocortical system, accounting for the pattern of retrograde amnesia seen in many amnesic patients. We are now considering how the neocortex may learn and represent semantic knowledge, addressing children's acquisition of knowledge of living things and the deterioration of this knowledge in dementia. Developmental data suggest a progressive differentiation of concepts: We distinguish animals from plants before we distinguish birds from fish, or canaries from robins. In semantic dementia this trend reverses, so that the finer distinctions are lost before more general ones. These findings coexist with the fact that in some tasks, there appears to be a priority for accessing and naming concepts at an intermediate ("basic") level of specificity (e.g., bird is more accessible than animal or robin). Current work with a former student (McClelland and Rogers, 2003; Rogers and McClelland, 2004) addresses all these phenomena in a single model, where they reflect the interplay of frequency and concept similarity in determining the semantic representations of concepts and the ease with which distinct names may be assigned to them. We also apply the approach to show that many findings that have been taken as supporting the existance of innate naive domain theories can arise from the kind of learning process that takes place in connectionist networks.

Another project considers why the neocortical learning system in adults, clearly capable of learning in many cases, nevertheless shows some important failures. We focus on the failure of Japanese adults to learn the distinction between /r/ and /l/ as an example. Some approaches suggest this failure reflects a simple switching off of learning about speech sounds as a function of age or puberty. We suggest that it may reflect, at least in part, a characteristic of the neural mechanism that underlies learning, which may be based on Hebbian synaptic modification: when one neuron participates in firing another, the connection between them is strengthened. This form of learning can tend to strengthen a network's tendency to keep doing what it is already doing. Thus, Japanese adults' tendency to perceive /r/ and /l/ as the same may simply be strengthened each time they hear either sound. A model implemented by a student in my lab shows how these ideas, in conjunction with some other biologically motivated assumptions, can lead to failure to learn new phonemic distinctions in adulthood. The model suggests training methods that may allow adults to learn such distinctions. Experiments with several collaborators (McCandliss et al, 2002) support the predictions of the Hebbian account, but also indicate that feedback plays a substantial role. Ongoing modeling work considers the incorporation of feedback into biologically plausible synaptic learning rules.

Among several other ongoing projects, I remain quite interested in the implications of the idea that knowledge of linguistic rules may best be viewed as residing in the strengths of connections among neurons, as proposed in Rumelhart and McClelland (1986); for recent developments, see McClelland and Patterson (2002a) and Bird et al. (in press). Currently we are exploring the interplay of phonological constraints, a pressure to keep frequent messages short, and a natural pressure in connectionist networks to prefer systematic representations; we suspect that the interplay of these forces may be responsible for many of the exceptions found among very high frequency English verbs (e.g., had, said, did, made, kept; see McClelland and Patterson, 2003b).

Publications

Over his career, McClelland has contributed to both the experimental and theoretical literatures in a number of areas, most notably in the application of connectionist/parallel distributed processing models to problems in perception, cognitive development, language learning, and the neurobiology of memory. He was a co-founder with David E. Rumelhart of the Parallel Distributed Processing research group, and together with Rumelhart he led the effort leading to the publication in 1986 of the two-volume book, Parallel Distributed Processing, in which the parallel distributed processing framework was laid out and applied to a wide range of topics in cognitive psychology and cognitive neuroscience. McClelland and Rumelhart jointly received the 1993 Howard Crosby Warren Medal from the Society of Experimental Psychologists, the 1996 Distinguished Scientific Contribution Award (see citation) from the American Psychological Association, the 2001 Grawemeyer Prize in Psychology, and the 2002 IEEE Neural Networks Pioneer Award for their pioneering work in this area. McClelland has served as Senior Editor of Cognitive Science, as President of the Cognitive Science Society, and as a member of the National Advisory Mental Health Council. He is a member of the National Academy of Sciences, and he has received the APS William James Fellow Award for lifetime contributions to the basic science of psychology.

McClelland currently teaches cognitive neuroscience and conducts research on learning, memory, conceptual development, spoken language, decision making, and semantic cognition.