Starting from the premise that working memory is a system for providing access to representations for complex cognition, six requirements for a working memory system are delineated: (1) maintaining structural representations by
dynamic bindings, (2) manipulating structural representations, (3) flexible reconfiguration, (4) partial decoupling from long-term memory, (5) controlled
retrieval from long-term memory, and (6) encoding of new structures into longterm memory. The chapter proposes an architecture for a system that meets these requirements. The working memory system consists of a declarative and
a procedural part, each of which has three embedded components: the activated part of long-term memory, a component for creating new structural representations by dynamic bindings (the ‘‘region of direct access’’ for declarative working memory, and the ‘‘bridge’’ for procedural working memory), and a mechanism for selecting a single element (‘‘focus of attention’’ for declarative
working memory, and ‘‘response focus’’ for procedural working memory).
The architecture affords two modes of information processing, an analytical and an associative mode. This distinction provides a theoretically founded formulation of a dual-process theory of reasoning.