| Abstract: | Four related sciences, cognitive psychology, artificial intelligence, linguistics and neurobiology, are compared in a cross-cognitive way with respect to their approaches in the study of discourse comprehension, in particular its final product, semantic post-representations. The nature and structure of these, as they are built in a human mind after processing a short piece of discourse (one or a few sentences), seem to be best described in the framework of activation models, a family in which the basic processes of comprehension are considered to be activation of semantic units from long-term memory, predication and construction of higher-level propositional constituents. The notion of “activation level”, applied to such representational units in working memory, is particularly fruitful in this framework. Besides, a satisfying neural interpretation of this psychological type of model can be proposed. The paper shortly presents a series of experiments, involving a semantic probing technique and three main categories of factors, with results that support the semantic post-representation view, in addition to others. A critical comparison of this analysis with neurofunctional imagery data confirms the necessity of cross-cognitive exchanges. |
