Psychiatry: Model of Language

Model of Language

In the past two decades, neurolinguistic research has been greatly influenced by the application of cognitive theories and computa-tional models to the study of language processes. Combined with functional neuroimaging techniques, these methodologies have enabled the analyses of much smaller units of language process-ing. The evidence has led to the development of the currentinformation-processing or PDP model for language that has sup-plemented the traditional Wernicke–Geschwind model as well as earlier cognitive models of language.

The traditional Wernicke–Geschwind model was more anatomically based and reflected a view of language functioning in terms of interconnections and disconnections between hier-archically arranged regions of the brain. According to this view, language was processed through a serial flow of information in the interconnecting pathways and cortical regions. For example, in a simple repetition task, the information flows through the following pathways: auditory input received by the auditory ap-paratus passes through the medial geniculate nucleus, primary auditory cortex (Brodmann’s area 41), and higher order auditory cortex (area 42) to the angular gyrus (area 39) and Wernicke’s area (area 22). Thereupon, the arcuate fasciculus transfers the in-formation to Broca’s area, where syntactic processing and articu-latory programming for production take place.

The application of cognitive theories to neuropsychologi-cal processes led to the development of information-processing models for language (as for other mental activities). However, the earlier models also tended to posit that information traversed across regions and across levels of processing (phonological, syntactic, semantic) in a serial fashion.

Within the framework of the current PDP model, linguistic operations are considered to be performed in more complex ways. One feature that characterizes this model is that any given task is thought to involve many operations that are localized within the various components of language (e.g., orthographical, semantic, phonetic, syntactic). These components, in turn, are subserved by different regions in the brain; that is, they are distributed (Paulesu et al., 1993). Investigations in this area therefore entail two stages: first, an attempt is made to locate the operations within the com-ponents or levels of processing; secondly, attempts are made to map these components onto specific brain areas. A second feature of this model is that it posits a view that the various linguistic op-erations that constitute performance are carried out interactively through various routes in a parallel, concurrent and simultaneous fashion. At least, in normal or near-normal performance this is the case. In defective language processing, the flow of informa-tion through the expected routes is disturbed. As suggested ear-lier, the concept of parallel processing emerged from a realization that the notion of a serial transfer of information did not account for the rapidity with which mental processes are carried out.

Given the complexity of linguistic phenomena and the technical and methodological limitations involved in applying a heuristic computational model to the empirical analyses of lan-guage processes, this field of inquiry is still in its infancy.