A similarity-based memory retrieval model to orthographic processing

• Based on the memory-based view of automaticity, an orthographic processing model operating through a similarity-based memory retrieval mechanism is proposed. • Orthographic processing is item-based and involves the use of specific orthographic information: lexical and sublexical. • Upon the onset of literacy, within-category orthographic similarity is facilitatory, while between-category orthographic similarity is inhibitory. • Prior to the onset of literacy, a deficit in perceiving speech sounds within and between categories could impede sound–character mapping and play a causal role in dyslexia. • The universality of the automaticity experience could qualify the model to explain orthographic processing with a unified mechanism across different orthographies. Orthographic processing abilities are vital for the efficient execution of literacy-related tasks. Drawing from automaticity research, which addresses how skilled performance develops with practice, this article proposes a new model in which orthographic processing operates through a similarity-based memory retrieval mechanism that utilizes both lexical and sublexical orthographic knowledge. The model captures the learner’s experience of developing literacy through perceptual categorization and how it could be facilitated—and sometimes inhibited—by the orthographic similarity effect within and between categories. The model further predicts a situation in which a deficit in perceiving speech sounds within and between categories could impede children’s earlier attempts at sound–character mapping, leading them to deviate from the normal course of literacy development and become diagnosed with dyslexia. Critical issues to the model—such as the conceptualization of similarity and the relationship between perceptual categorization and categorical perception and how they both relate to the automaticity phenomenon—are discussed throughout the article. The universality of the automaticity experience could qualify the model to extend its predictions across different orthographies that appear to be processed differently yet underpinned by the same universal cognitive mechanism.