InterJournal Complex Systems, 412
Status: Submitted
Manuscript Number: [412]
Submission Date: 608
Revised On: 718
Complexity, Nonlinearity, and Human Cognition
Author(s): Richard Heath

Subject(s): CX.04, CX.14, CX.41

Category: Brief Article


Despite the extensive application of computational models, such as multilayer neural networks, to explain the fundamental features of human cognitive processes, some important questions, such as why the capacity of short-term memory is so small and why memory retention is limited because of the stability-plasticity tradeoff, remain unanswered. Such questions are important because they impose constraints on the types of mathematical model that could possibly explain the functioning of complex cognitive systems. They are rarely raised in cognitive theory in psychology where the emphasis is usually on devising alternative models to explain a set of constrained empirical phenomena. This paper presents a theoretical framework that considers cognitive processes as adaptive and evolutionary, providing a necessary interactive link between mind/body and the environment. The fundamental principles are embodied in the types of models developed for cellular automata by Kauffman (1993), Langton (1990) and others. These models exhibit qualitative changes in their dynamics that depend on the rate at which information flows through the network, the rule that transforms information from one time period to the next and, in particular, the number of units to which each network unit is connected. Evidence suggests that optimal information processing, as indicated by spontaneously self-organized network behavior, occurs for relatively sparse connectivity during a phase transition at the edge-of-chaos. A review of a number of cognitive processes, including memory recognition, memory constraints on cognitive development, motor skill acquisition and decision making suggests that optimal cognitive performance occurs when the system has a limited memory capacity. This conclusion is confirmed when an adaptive model for human cognition provides good fits to recognition memory data and incorporates both the memory capacity limitations and the stability-plasticity tradeoff characteristic of the automata models (Heath, 1998). More precisely, the model predicts the U shaped serial position curves in a recognition memory task that varies the depth of information processing required from participants. A modular extension of the basic model with a chaotic tonic state was developed. This model contains at least two modules, including one that allows the cognitive system to interact with the environment. Environmental influences, governed usually by physical stimulation, alter the internal dynamics of the cognitive system by means of control of chaos. This leads to some interesting predictions, such as an optimal stochastic decision model, based on an Ornstein-Uhlenbeck process (Heath, 2000b); the representation of memory storage and retrieval as a change in qualitative dynamics rather than simple placement into a distributed or fixed-address memory system; and the prediction of motor control dynamics, such as handwriting, when two or more coupled processing units interact nonlinearly. The paper concludes with a general discussion of the important role played by nonlinear dynamical systems in human cognition (Heath, 2000a). Both the optimal "edge-of-chaos" representation, and one based on the coupled interaction between cognitive and environmental modules provide the necessary dynamical processes to justify this nonlinear system emphasis.

Retrieve Manuscript
Retrieve Previous Revision's Abstract
Submit referee report/comment

Public Comments: