Modelling individual differences in reading using an optimised MikeNet simulator: the impact of reading instruction

Area of Science:

• Cognitive Science
• Computational Linguistics
• Educational Psychology

Background:

• Reading acquisition is crucial for knowledge.
• Phonology-focused interventions often outperform meaning-focused ones for children with reading disabilities.
• Individual differences in reading skills are influenced by instruction, exposure, and oral language.

Purpose of the Study:

• To investigate how different reading instructional methods impact reading strategies and individual differences.
• To develop and utilize a computational model to simulate reading acquisition under varied training conditions.

Main Methods:

• Developed a computational reading model using an optimized MikeNet simulator.
• Trained the model with three distinct methods: phonology-focused, meaning-focused, and balanced phonology-meaning.
• Utilized semantic reliance (SR) as a metric for individual differences, reflecting reliance on print-to-sound vs. print-to-meaning mappings.

Main Results:

• Simulation results revealed a direct correlation between semantic reliance (SR) levels and the type of reading instruction received.
• Higher SR scores predicted poorer performance in reading-aloud tasks, indicating increased phonological errors and reduced phonological activation.
• Findings align with behavioral and neuroimaging data, supporting the model's validity.

Conclusions:

• Reading instruction significantly impacts reading behaviors and individual differences.
• Integrating phonological (OP) and semantic (OS) instruction is crucial for beginning readers.
• Semantic reliance (SR) serves as a valuable indicator of reading strategy and potential difficulties.