Understanding the training and transfer effects in N-back training

Abstract

Current research looking at the effect of working memory training on constructs such as fluid intelligence has generated mixed findings. Some researchers have found that training participants on working memory tasks leads to an improvement on fluid intelligence scores, others have failed to find this effect. To reconcile these different findings, there is a need to understand the underlying mechanism of the transfer effect. In this study, a modified N-Back task was decomposed into its component processes, namely updating, focus switching one-step retrieval, and focus switching requiring search; the effect of training on each of the components was examined. Since updating has been found to be associated with both working memory (e.g., Miyake et al., 2000) and fluid intelligence (e.g., Friedman et al., 2006), the study specifically looked at the role of the updating component in eliciting transfer to other cognitive control processes (task switching and inhibition) as well as measures of fluid intelligence. The study employed two groups of participants—experimental and active control, which were trained for 10 hours over a period of two weeks and assessed on the transfer measures before and after training. The experimental group was trained on all the components of the modified N-Back task, whereas the active control group was trained on all components, except updating. If updating were the crucial link between training and transfer effects, the two groups should have shown differential effects on the transfer measures. However, this hypothesis was not supported. Training in the updating aspect of the N-Back task did not generalize to other cognitive control processes implicated in working memory nor did it lead to transfer to measures of fluid intelligence. However, we did find that training effects are different on the different working memory components. The updating component is more malleable than the focus switching requiring search and focus switching direct retrieval. Thus working memory training protocols targeting the updating component might be more effective than the ones which don’t include it.