In this book, well known researchers in philosophy, cognitive psychology linguistics, and neuroscience address two questions about connectionism as an alternative to traditional symbolic cognitive models. Are connectionist models computationally sufficient to account for complex cognitive functions? Can connectionism account for such functions in a neurobiologically realistic way?The chapters in part I explore the computational adequacy and cognitive plausibility of connectionist or parallel distributed processing (PDP) models. Those in part II focus on the constraints these models impose on the PDP theorist who wants to remain true to known brain architecture.The contributors attempt to derive facts about brain architecture that might be crucial to connectionist formalisms the processes that one should expect to uncover if notions like connection strength, the delta rule, activation value, and back propagation are to find neurological concomitants.The authors include Valentino Braitenberg, Patricia Smith Churchland, Terrence J. Sejnowski, Gary Lynch, Richard Granger, John Larson, Michel Baudry, Leif H. Finkel, George N. Reeke, Jr, Gerald M. Edelman, Roger N. Shepard, Paul Smolensky, Jerome A. Feldman, Bruce McNaughton, and John O'Keefe. The editors contribute an introductory chapter.Lynn Nadel and Lynn Cooper are professors of psychology and research cognitive scientists at the University of Arizona; Peter Culicover is professor of linguistics at Ohio State University; R. M. Harnish is professor of philosophy at the University of Arizona. "Neural Connections, Mental Computation" is included in the series Computational Models of Cognition and Perception edited by Jerome A Feldman,Patrick J. Hayes, and David Rumelhart. A Bradford Book.