Date of Completion


Embargo Period



reflective learning, reflexive learning, artificial orthography learning, individual differences

Major Advisor

Jay G. Rueckl

Associate Advisor

Kenneth Pugh

Associate Advisor

James Magnuson

Associate Advisor

Nicole Landi

Field of Study



Doctor of Philosophy

Open Access

Open Access


How people learn to read is an interesting question which has been investigated by many studies with various approaches. Some recent studies have related learning to read with domain-general abilities and have found a positive relationship between statistical learning and learning to read, as well as between procedural learning and learning to read. However, evidence on these relationships is still inconsistent, which probably because reading, statistical learning and procedural learning are componential capabilities. The current study provided another approach to explore how people learn to read, especially how to learn the orthography-phonology (O-P) and orthography-semantics (O-S) correspondences, with multiple learning systems: reflective learning which mostly underlies rule-based learning, and reflexive learning which mostly underlies information-integration. An artificial orthography learning paradigm (AOL) was used as the measure of learning to read with statistical regularities built in O-P and O-S correspondences. In Experiment 1, different manipulations were used on AOL tasks to disrupt either reflective learning or reflexive learning. Disrupting reflective learning significantly impaired performance on AOL tasks, and the O-S learning was more impaired than O-P learning. However, disrupting reflexive learning did not affect overall learning. Experiment 2 further examined the relationship between reflective/reflexive learning and the individual differences in learning to read, this time reflective and reflexive learning were directly measured. Reflective learning was a significant and robust predictor for AOL performance, but reflexive learning was only a predictor to AOL training but not categorization. A trend of competition between the two learning types was also shown by the interaction between them. In addition, reflexive learning but not reflective learning predicted visual statistical leaning, and working memory was found to be positively correlated with both types of learning. Taken together, this study showed that reflective learning was engaged in learning to read with the AOL tasks. The engagement of reflexive learning was also possible, but probably was diminished by the competition between the two learnings and the paradigm of tasks. Although we should be cautious when generalize the findings to a broader question of learning to read, this study provides insights in understanding reading acquisition and education.