The asymmetry of working memory training transfer: A systematic review and meta-analysis

Working memory (WM) training is widely used to improve individuals’ cognitive abilities, yet its transfer to untrained tasks remains controversial. Although traditional WM transfer theory emphasizes task similarity as the primary determinant of training transfer, emerging evidence challenges this view by proposing asymmetric transfers between tasks. Two primary hypotheses have been posited: resourced-based transfer advantage suggests training on tasks with higher cognitive demands transfer better to lower demanding tasks, whereas ability-based transfer advantage proposes training broader abilities transfer better to narrower ones. To systematically investigate transfer asymmetry and its underlying mechanisms, we conducted a three-level meta-analysis comparing transfer effects between two typical WM paradigms: span tasks and updating tasks. This comparison builds on the established evidence that updating tasks impose higher cognitive demands but engage narrower abilities than span tasks. Our analysis included 55 studies (208 effect sizes) involving 3,492 participants. Results first confirmed significant transfer effects between updating and span tasks (Hedges’ g = 0.146, p < .001), with substantial between-studies variance (τ² = 0.023, p < .001). Crucially, transfer direction modulator analysis revealed significant transfer effect from updating to span tasks (g = 0.176, p < .001) but not conversely (g = 0.048, p = .453), supporting the resource-based transfer advantage hypothesis. Furthermore, secondary moderator analysis demonstrated this asymmetry was amplified under lower training doses, in child populations and with verbal training stimuli. Together, the study advances understanding of WM training transfer beyond the task similarity and inspires WM training practices. notReviewed other