A cognitive architecture is a comprehensive, unified theory specifying the fixed structures, mechanisms, and processes that underlie human cognition. Unlike theories of specific cognitive phenomena (e.g., working memory, attention), architectures attempt to explain all of cognition within a single integrated framework. They specify how information is represented (symbolic, subsymbolic, or hybrid), how it is processed (serial, parallel, or both), how memory is organized (modular, distributed, or both), and how learning occurs.
Major Architectures
ACT-R (Adaptive Control of Thought-Rational), developed by John Anderson, combines symbolic production rules with subsymbolic activation processes. It includes modules for declarative memory, procedural memory, visual perception, motor control, and a central production system that coordinates them. SOAR (State, Operator, And Result), developed by Newell, Laird, and Rosenbloom, models cognition as problem solving in problem spaces using production rules, with learning through chunking. EPIC (Executive Process-Interactive Control) emphasizes parallel perceptual-motor processing. Global Workspace Theory architectures implement Baars's theory of consciousness.
ACT-R is the most widely used cognitive architecture. It specifies that declarative knowledge is stored as chunks with base-level activation that reflects recency and frequency of use (explaining memory effects like the power law of forgetting). Procedural knowledge is stored as production rules. A central bottleneck limits production firing to one rule per cycle (~50ms). The architecture makes detailed quantitative predictions about reaction times, error rates, and brain activation patterns that can be tested against human data.
Evaluation
Cognitive architectures have successfully modeled performance across hundreds of tasks, from simple reaction time to complex problem solving, text comprehension, and driving. However, no architecture has achieved the generality of human cognition. Current architectures struggle with creative thinking, analogical reasoning, and flexible adaptation to novel situations. The integration of connectionist learning mechanisms with symbolic architectures remains an active research challenge.