James L. McClelland

My research is based on the premise that it is useful to view human cognition as emerging from the parallel, distributed processing activity of neural populations. Over the years my colleagues and I have applied these ideas to a wide range of topics in human cognition, perception, language processing, and memory, with additional applications to topics in perceptual, cognitive, and linguistic development. In work on context effects in perception we developed an explicit model in which perception reflects an interactive activation process, involving synergistic, joint use of bottom-up or sensory information together with top-down or contextual information. In work on the human ability to apply familiar and regular patterns to novel examples (e.g., to infer that the past tense of a new verb "glub" is "glubbed") we argued that such abilities can arise in networks of simple neuron-like processing units without the need to formulate explicit rules to capture the pattern. Other work explores the implications of our parallel distributed processing approach for the organization of learning and memory in the brain, for basic aspects of human information processing and attention, and for the causes of the patterns of change seen over time during cognitive development.