The Contribution of Attention and Working Memory to Statistical Learning in Young Children

Abstract

Purpose: Statistical learning is a mechanism that allows rapid learning without conscious effort (Plante & Gómez, 2018). Attention and working memory may underlie this mechanism (Erickson & Thiessen, 2015). Yet, empirical evidence for this is mixed, and there is an age-related gap in the literature in this area for young children. The goals of this study were (a) to develop a statistical learning task on which young children could demonstrate learning and (b) to use that task to examine the contribution of attention and working memory to statistical learning outcomes in that population. Addressing these goals could provide insight into how we might individualize statistical learning-based language interventions for children with language impairments by considering their unique cognitive profiles. Method: Fifty 5- and 6-year-old children with typical development were recruited. They completed computerized, in-person attention, working memory, and statistical learning tasks. The primary analysis was a multiple linear regression with age as a covariate and outcome measures on the attention and working memory tasks as predictors. The dependent variable was children’s accuracy on the statistical learning task. Retesting was completed for a subset of children to calculate the measures’ test-retest reliability. Results: Test-retest analyses showed that the attention and working memory tasks were moderately to strongly reliable. Descriptive statistics showed that most children were able to successfully demonstrate learning on the statistical learning task, with 30 of 50 achieving a score above chance. The regression analysis showed that neither attention nor working memory scores significantly predicted children’s scores on the statistical learning task. Conclusion: The statistical learning task was feasible for young children, but this was largely at the group level. Attention and working memory did not significantly predict children’s statistical learning outcomes. This might be explained by the mitigating influence of facilitative principles known to support statistical learning. Alternatively, this result might be explained by methodological aspects of the study. Though the findings cannot speak directly to ways we might individualize statistical learning-based interventions just yet, they have inspired questions to be addressed in future studies which would add depth to our understanding of statistical learning in children and the components that underlie it.