ABSTRACT
A framework for using neural network models to study high-level cognition needs to be established. This framework should encompass different paradigms for linking neural networks to cognition. In this paper I suggest four paradigms that should be included in such a framework: 1- Decomposing complex cognitive abilities into their component subsystems; 2- Examining cognitive deficits and normal cognition by exploring lesioned neural networks; 3- Exploring cognitive processing by manipulating the availability of information; and 4- Investigating which principles and properties of information processing give rise to certain cognitive phenomena.
