Connectionist models and their properties (01 Jan 1982)
Although several researchers had worked with neurally-inspired computational models before 1982, this article was among the first to present a general and formal characterization of connectionist models. In fact, it introduced the term "connectionist model" to cognitive science. The authors provided formal results dealing with winner-take-all networks, coarse-coding, and the control of sequencing, and described the well known 100-time-step argument for the need for massively parallel computation. The specific approach taken by Feldman and Ballard is now represented by the many structured or localist connectionist models in the field.
Article URL: http://cognitrn.psych.indiana.edu/rgoldsto/cogsci/Feldman.pdf