Toward an analog neural substrate for production systems

Symbolic, rule-based systems seem essential for modeling high-level cognition. Subsymbolic dynamical systems, in con- trast, seem essential for modeling low-level perception and action, and can be mapped more easily onto the brain. Here we review existing work showing that critical features of symbolic production systems can be implemented in a subsymbolic, dynamical systems substrate, and that optimal tuning of connections between that substrate’s analog circuit elements accounts for fundamental laws of behavior in psychology. We then show that emergent properties of these elements are reflected in behavioral and electrophysiological data, lending support to a theory about the physical substructure of productions. The theory states that: 1) productions are defined by connection strengths between circuit elements; 2) conflict resolution among competing productions is equivalent to optimal hypothesis testing; 3) sequential process timing is parallel and distributed; 4) memory allocation and representational binding are controlled by competing relaxation oscillators.