A meta-analysis on the correlations between statistical learning, language, and reading outcomes.

Abstract

The purpose of this meta-analytic review is to investigate the relation between statistical learning (SL) and language-related outcomes, and between SL and reading-related outcomes. A comprehensive search of peerreviewed published research resulted in 42 articles with 53 independent samples and 201 reported effect sizes (Pearson's r). Results of our robust variance estimation correlated effects model revealed a significant, moderate relation between SL and language-related outcomes, r = .236, p , .001, and a significant, moderate relation between SL and reading-related outcomes, r = .239, p , .001. Moreover, age, the writing system of the language, and SL paradigm moderate the strength of the association between SL and reading. Age is the only significant moderator on the strength of the association between SL and language. The findings from this meta-analysis shed light on the contribution of multiple factors that impact how SL relates to language and reading outcomes, with important implications for developing effective instructional practices that emphasize statistical regularities of oral and written materials in the classroom. Theoretical implications of these findings for language and reading development are discussed. Public Significance StatementThis study demonstrated a moderate correlation between SL and language and reading outcomes, which is relevant to public health. Our findings can inform policy and be used to promote children's well-being and success in school. Educators may cultivate early SL skills and embed SL into instruction as a pathway to promote language and reading development.

A meta-analysis of 97 studies reveals that statistical learning and language ability are only weakly correlated

“…While there appears to be an overall positive correlation between SL and language across the lifespan, our results indicate that its strength is not significantly affected by the age of the participants. This finding is in contrast with the results of Ren et al (2023), who found age to be the only significant moderator of the association between SL and language. Note that it made no difference whether we tested age as a continuous predictor, or as a categorical variable similar to the approach by Ren et al (2023).…”

“…This finding is in contrast with the results of Ren et al (2023), who found age to be the only significant moderator of the association between SL and language. Note that it made no difference whether we tested age as a continuous predictor, or as a categorical variable similar to the approach by Ren et al (2023). More work on the developmental trajectory of different aspects of SL in relation to language is necessary.…”

“…In line with our expectations (prediction 1) and with the results of two other recent meta-analysis (Olivera et al, 2023;Ren et al, 2023) 3 , we found that the estimated relationship of SL abilities and language skill is small (r = .16) yet significant. This result provides support for the hypothesis that SL and language are, to some degree, related.…”

“…We also chose to omit studies that did not use an independent test of SL and language abilities (e.g., Brice et al, 2021;. While the initial submission of this meta-analysis was under review, two related meta-analyses were published (Oliveira et al, 2023;Ren et al, 2023). We chose to merge the studies included in those meta-analyses that were missing in our search Figure 2 PRISMA Flowchart…”

Contribution of statistical learning in learning to read across languages

“…In the handful of studies that have examined how SL relates to reading in alphabetic writing systems, some studies examined domain-specific SL effects in alphabetic languages (e.g., [15][16][17][18][19]), and some other studies focused on domain-general SL effects. Among these studies, some revealed positive relationships between visual SL (VSL) and word reading, and between auditory SL (ASL) and word reading (e.g., [20][21][22][23]). Arciuli and Simpson [20] utilized the classic triplet paradigm to test non-linguistic VSL, and the reading subtest of the Wide Range Achievement Test (WRAT-4; [24]) to assess children's ability to read aloud individually presented English orthographic strings.…”