Can computational models of reading aloud account for how individuals who are d/Deaf and hard of hearing read?

Abstract

Purpose: This paper proposes modifications to two current computational models of reading aloud - the Dual Route Cascaded (DRC) Model and the Connectionist Dual Process (CDP++) Model, that are specific to individuals who are d/Deaf and hard-of-hearing (d/Deaf/HoH). Methods: A brief summary of the of the DRC and the CDP++ models are first provided. Differences in how individuals who are d/Deaf/HoH read based on past literature are then reviewed in the context of the two models. Finally, adaptations to the existing models are proposed and new d/Deaf/HoH-specific versions of both models are discussed. Results: Several different strategies of reading instruction for individuals who are d/Deaf/HoH are identified from the proposed model modifications. These involve decreasing emphasis on auditory phonological processing instruction, adding multi-sensory phonological processing instruction, and increasing semantic knowledge instruction. Conclusions: Current computational models based on normal hearing data do not accurately account for how individuals who are d/Deaf/HoH read. Although the proposed modifications to the computational models should be tested, they hold potential to guide new reading instruction strategies that would help individuals who are d/Deaf/HoH achieve literacy levels comparable to their normal hearing peers. Keywords: Dual-route Theory, Deaf, hard-of-hearing, computational models, reading