What Should a Connectionist Philosophy of Science Look Like?

This is not the place to present an introduction to connectionism, an approach to modeling cognitive phenomena that was first developed in the 1950s and once again gained prominence in the 1980s, in part with the publication of Rumelhart, McClelland, and the PDP Research Group (1986). Both Paul and Patricia Churchland have presented introductions to connectionism in various of their writings. See also Bechtel and Abrahamsen (1991) for an introduction. The understanding of connectionism varies significantly among authors. For both of the Churchlands the importance of connectionism seems to be that the parallels between connectionist networks and real neural networks allows our emerging understanding of brain function to inform cognitive modeling. For many others the importance lies not so much in the similarity of connectionist networks to neural architecture as in the fact that connectionist models seem to exhibit features of cognition lacking in other approaches to cognitive modeling.