| Abstract: |
We report on the significant beneficial outcomes of the contrasted experimental rehabilitation techniques for grammaticality judgements operating on numbers in written word forms (verbal numbers). The information processing approach provided the theoretical framework for designing training programmes. The verbal primitives were classified into lexical categories (units, teens, decades, and hundreds). A set of lexico-syntactical rules was stated in terms of well-formed or illegal sequences of categories. Grammaticality judgement was simulated by a left-to-right parser performing the lexical categorization of verbal number primitives and operating with the relevance acceptance/rejection syntactical rules. In Therapy I, neither rules nor lexical categorization were taught to the patient, who was just presented with many verbal numbers in a grammaticality judgement task on a video screen. Automatic feedback to the patient qualified only the nature of his responses (correct or errors). In Therapy II, the patients was taught the rules and the lexical categories they operated upon. The two rehabilitation methods were successively administered and each induced different but significant improvement in grammaticality judgements not only on trained items but also on comparable undrilled items. Intratask leaning transfer was enhanced by training with a large variety of items sharing the same underlying lexicosyntactical structure, which proved more efficient than more repetition of exercises on the very same items. Intertask learning transfers were obtained in other number processing situations like writing verbal numbers from Arabic forms (or the reverse), magnitude comparisons, and numerical size estimation of verbal numbers |
