Dissociations Between Regularities and Irregularities in Language Processing: Computational Demonstrations Without Separable Processing Components

Two models are presented that compute a quasi-regular mapping. One was based on localist representations of items in the quasi-regular domain, the other was based on distributed representations. In each model, a control parameter termed input gain was modulated over the one and only level of representation that mapped inputs to outputs. Input gain caused both models to shift between regularitybased and item-based modes of processing. Performance on irregular items was selectively impaired in the regularitybased modes, whereas performance on novel items was selectively impaired in the item-based modes. Thus, each model exhibited a double dissociation without separable processing components. These results are discussed in the context of analogous dissociations found in language domains such as word reading and inflectional morphology.