Beyond numbers: working memory deficits in children with mathematical difficulties — A longitudinal meta-analysis

Abstract

Working memory (WM) is central to mathematical development, yet its role in children with mathematical difficulties (MD) remains debated. Using a multilevel approach, this longitudinal meta -analysis synthesized 36 studies (38 samples; N = 6,015) comparing WM in children with MD and typically developing peers. Across 652 effect sizes, children with MD showed a robust, moderate deficit (Hedges’ g = –0.57), evident across all WM components (central executive, phonological loop, visuospatial sketchpad) and domains (verbal, numerical, visuospatial). The largest impairment appeared in the central executive, consistent with accounts highlighting attentional control and dual-task coordination. By integrating longitudinal designs, this study extends prior cross-sectional syntheses, demonstrating that WM deficits persist across development rather than reflecting transient delays. Diagnostic thresholds systematically shaped outcomes, with stricter criteria yielding larger effects, while demographic factors did not. Measurement approaches also influenced estimates, underscoring the methodological sensitivity of reported deficits. Overall, findings indicate that WM weaknesses in MD are broad and persistent, with the magnitude of observed deficits varying systematically as a function of diagnostic criteria. This is the first meta -analysis to integrate longitudinal evidence in mathematical difficulties, clarifying both the developmental stability of WM deficits and the methodological contingencies that shape reported outcomes. These results advance developmental theory by positioning WM as an enduring vulnerability and highlight the need for early identification and instructional approaches that are sensitive to working memory constraints and that carefully calibrate cognitive demands within mathematics instruction, in order to prevent cascading disadvantages in mathematics learning and promote more equitable learning trajectories.