Lecture by James McClelland, laureate of the 2014 C.L. de Carvalho-Heineken Prize for Cognitive Science, on interactive processes in perception and language
Abstract KNAW Heineken lecture 'Interactive processes in perception and language'
Perception and language processing require the simultaneous use of multiple sources of information, allowing the human mind to approach optimality in interpreting sensory information at many levels of representation in real time as information arrives. I present the Interactive Activation hypothesis—the idea that the mechanism used in perception and comprehension to achieve these feats exploits an interactive activation process implemented through the bidirectional propagation of activation among simple neuron-like processing units, and review interactive activation models of visual and spoken word perception my colleagues and I have developed. I examine evidence motivating and supporting the approach from psychological and neuroscience research, and consider empirical and theoretical controversies surrounding the idea of interactive processing, including a controversy that swirls around the relationship between interactive computation and optimal Bayesian inference. Current versions of interactive models overcome the limitations of earlier versions. I will review a recent version that explicitly links interactive activation and optimal probabilistic computation. There remain important challenges ahead for interactive models, but I argue that contemporary versions of such models provide a basis for efforts to achieve a fuller understanding of the processes of perception and comprehension, as well as several other aspects of cognition.