The role of statistical learning in reading : evidence from meta-analysis and controlled-trial studies

Abstract

Humans can effortlessly sense and internalize the pattern of sets of information in the environment. This ability, known as statistical learning, might contribute to reading acquisition. However, many aspects of statistical learning were not fully understood. The present series of studies were conducted to better unearth statistical learning from both theoretical and practical perspectives for developing an ecological model. The first study investigated the relationship among statistical learning, dyslexia, and reading development by systematically evaluating the evidence from the literature. The second study examined the timescale and capacity of multidimensional statistical learning in 40 neuro-typical adults. The third study investigated the potential role of statistical learning in Chinese dyslexia intervention by a randomized controlled trial. Ninety-five children with dyslexia were randomly assigned to one of the following intervention groups: implicit, explicit, and no statistical learning of phonetic radicals that predict the pronunciations of Chinese characters.

The first study showed that individuals with dyslexia exhibited statistical learning weakness; which was significantly influenced by verbal IQ, marginally moderated by working memory, and marginally dependent on the individuals’ acquired reading skills for opaque orthographies. The second study showed that, across all exposure phases, adult participants could unintentionally respond faster to a multi-regularity sequence that they could simultaneously acquire and maintain some, but not all statistical information. The third study demonstrated that the statistical learning-based intervention (implicit or explicit phonetic radical learning) significantly improved Chinese word reading performance.

On the whole, these studies contributed to the development of a new theoretical model and an intervention approach that bridge statistical learning to reading development, suggesting that humans, even those with dyslexia, possess the power of learning to crack statistical codes and read print codes effortlessly and subconsciously.