Limits of near transfer: content- and operation-specific effects of working memory training

Working memory (WM) training typically leads to large performance gains in the practiced tasks, but transfer of these gains to other contexts is elusive. One possible explanation for the inconsistent findings of past research is that transfer may only occur when cognitive strategies acquired during training can also be applied in the transfer tasks. Therefore, we systematically varied the content domains and WM operations assessed by training and transfer tasks and, thereby, the extent to which similar cognitive strategies could be applied. We randomly assigned 171 young adults to one of eight experimental groups who trained one of two WM operations (storage and processing or relational integration) with materials from one of four content domains (verbal, numerical, figural-icon, or figural-pattern) to an active or to a passive control group. Before and after 12 sessions of adaptive training within 2–3 weeks, performance was assessed in all eight WM tasks. Bayesian generalized linear mixed-effects models revealed improved performance in the trained tasks compared to the active control group. However, these improvements did not generalize to tasks measuring the same WM operation with different materials. Moreover, the comparison of the training groups with an active and a passive control group showed considerable differences, thus highlighting the importance of distinguishing between active and passive control groups. Overall, the findings revealed no evidence for transfer between tasks assumed to afford the same strategies. Therefore, the adoption of specific cognitive strategies alone is unlikely to be responsible for transfer of WM training gains between tasks.