Working memory updating and binding training: Bayesian evidence supporting the absence of transfer

As working memory (WM) predicts a wide range of other abilities, it has become a popular target for training interventions. However, its effectiveness to elicit generalized cognitive benefits is still under debate. Previous research yielded inconsistent findings and focused only little on the mechanisms underlying transfer effects. To disentangle training effects on WM capacity and efficiency, we evaluated near transfer to untrained, structurally different WM tasks and far transfer to closely related abilities (i.e., reasoning, processing speed, task switching, and inhibitory control) in addition to process-specific effects on three WM mechanisms (i.e., focus switching, removal of WM contents, and interference resolution). We randomly assigned 197 young adults to one of two experimental groups (updating or item-to-context binding) or to an active control group practicing visual search tasks. Before and after five weeks of adaptive training, performance was assessed measuring each of the cognitive processes and abilities of interest with four tasks covering verbal-numerical and visual-spatial materials. Despite the relatively large sample size, large practice effects in the trained tasks, and at least moderate correlations between WM training tasks and transfer measures, we found consistent evidence for the absence of any training-induced improvements across all ranges of transfer and mechanisms. Instead, additional analyses of error patterns and self-reported strategy use indicated that WM training encouraged the development of stimulus-specific expertise and use of paradigm-specific strategies. Thus, the results suggest that the WM training interventions examined here enhanced neither WM capacity nor the WM mechanisms assumed to underlie transfer.