Human beings possessed a remarkable ability to unconsciously detect and internalize statistical patterns of sensory input from the environment, such as frequency, variability, and co-occurrence probability. This ability, known as statistical learning which is believed to be related to language acquisition. Researchers have investigated the link between statistical learning and developmental dyslexia (DD), a difficulty in learning to read even with normal intelligence and schooling, however, the results were mixed.

Lee and colleagues conducted a meta-analysis compared the differences in statistical learning ability between people with developmental dyslexia (DD) and their age-matched typically development (TD) counterpart and investigate various factors which may affect the extent of differences. One of the factors was the statistical learning task paradigm which included:

• Serial reaction time (SRT) task: to measure the reaction time to randomly and sequentially ordered stimuli. This is an implicit learning that participant becomes familiar with a repeated sequence from mere exposure.
• Artificial grammar learning (AGL) task: participants are presented with strings of letters, numbers, or symbols arranged by unknown rules (or grammar). The AGL tasks assess whether participants can implicitly learn the rules of the sequence.
• Other paradigms, for example, segmentation paradigm, Hebb learning paradigm.

Those various task paradigms test different types of statistical structures, that each task has different demands on cognitive faculties. The research identified 59 studies in 49 papers met the inclusive criteria, using robust variance estimation, results of the review indicated that

• Overall, DD group had poorer statistical learning than TD counterpart (ES = +0.47), included children (age < 18) and adults (age = 18 or above).
• The effect size of weaker statistical learning was 0.49 for children (k=34) and 0.46 for adult (k=23). The effects were not significant different between two groups.
• No significant difference was found across different types of paradigm task of statistical learning, types of modality (visual, auditory, and audiovisual), nor stimulus types (linguistic and nonlinguistic).
• However, verbal intelligence was significantly accounted for statistical learning difference between DD and TD, but not age nor other cognitive markers (e.g., nonverbal IQ, phonological skills). Specifically, individual with DD had higher verbal intelligence had lower statistical learning (β = -0.39).

The findings indicated that the statistical learning difference between DD and TD had no difference between adults and children. Moreover, the absence of effect of task paradigms, modality, and stimulus type, suggested that statistical learning weakness in dyslexia is domain general. The authors proposed a new statistical learning and reading model for further testing.

Note: k = no. of studies.

Source: Lee, S. M.-K., Cui, Y., & Tong, S. X. (2022). Toward a model of statistical learning and reading: Evidence from a meta-analysis. Review of Educational Research, 92(4), 651–691. https://doi.org/10.3102/00346543211073188