Extending Empirical Benchmarks of Working Memory to Children: Insights from an Adaptive Learning Environment

In this study, we explored whether the key benchmarks of working memory processing identified in adults by Oberauer and colleagues (2018) also apply to children, using data from a large adaptive learning environment. Over nine thousand children from Dutch primary schools (age between 6 and 12) played two serial recall tasks (verbal domain and visuospatial domain), providing a means for studying working memory processing in students' regular educational environment. Using Bayesian multilevel modeling, we found that the difficulty of the over 2000 items was affected by characteristics related to response facilitation, spatial grouping, and set size. Set size and spatial grouping also affected the accuracy of students' responses. Furthermore, we investigated primacy and recency effects and found that, as expected, the effect of serial position varies across set size. This result is also in line with previous findings on developmental changes in working memory processing where primacy and recency effects change as children grow older. Finally, key benchmark findings on error categorization were replicated, revealing that children were more prone to omission and intrusion errors than transposition errors. However, as children mature, the proportion of transposition errors increased. Additionally, we found limited evidence for an in-fill effect in transpositions in the verbal working memory tasks and substantial evidence for locality constraints on transpositions in both tasks. Our findings provide an understanding of the development of working memory processing in children and highlight the robustness of classical working memory findings in online educational data.