| Sumario: | This review discusses the plausibility of computational analysis of brain-language relations hypotheses. Proposals from linguistics and neurosciences are examined, as well as the origin and relevance of the neurocomputational models. Development: computational models inspired on brain dynamics have triggered an intense debate within cognitive sciences about various psychological issues. One of the most thoroughly examined issues has been language and the cognitive processes involved. Linguistic evidence points to serial processing within a modular framework and based on a generative system which uses two kinds of elements: lexical items stored in long-term memory as unitary blocks retrieved to be used as building blocks for sentence construction, and rules that are applied to build morphological or syntactic structures in accordance to the requirements of a particular grammar and stored somewhere else. However, neuroscientific studies on language generation suggest that both components work in parallel, providing a fuzzy modular view of language with parallel processing considering overlapped representations and reciprocal inhibitory mechanisms, pointing to a connectionist view. Several neurocomputational models dealing with the implementation of language processing on brain-like systems are exposed to demonstrate that certain aspects of modular language processing theory have to be revisited. Finally, it is proposed that a theory of language processing can no longer continue to ignore the properties of the system on which it is implemented, as there is a growing number of characteristics of brain-like systems which seem to restrict the operations they perform and even determine.
|
|---|