CAN NEUROPSYCHOLOGICAL EVIDENCE INFORM CONNECTIONIST MODELING - ANALYSES OF SPELLING

The symbolic information-processing paradigm in cognitive psychology has seen a growing challenge from neural network models over the past decade or so. While neuropsychological evidence has been of great utility in informing information-processing modelling, the emergence of less rigidly modular connectionist models raises the possibility that constraints from the behaviour of a damaged system may give relatively little information about these more complex structures. We believe that this will not prove to be the case, however, and discuss connectionist models of two sub-components of the spelling process which, internally, blur modular boundaries, and which explain, rather than describe, the relevant neuropsychological evidence. The models operate serially, and thus fall within a domain which has been a stronghold of symbolic modelling. The strong neuropsychological support which emerges for such models is therefore of particular interest.