Can you train executive functions? A second meta-analysis says…

In 2015 I wrote this blogpost on a meta-analysis on the training of executive functions. The results were a bit depressing. Today I found a new meta-analysis via Jeroen Jansen and the results… are again not that good and it all has to do with one of the key concepts in education: transfer. The meta-analysis was based on 38 studies with 47 contrasts.

I’ll first share with you the abstract of the study:

In the present meta-analysis we examined the near- and far-transfer effects of training components of children’s executive functions skills: working memory, inhibitory control, and cognitive flexibility. We found a significant near-transfer effect (g⁺ = 0.44, k = 43, p < .001) showing that the interventions in the primary studies were successful in training the targeted components. However, we found no convincing evidence of far-transfer (g⁺ = 0.11, k = 17, p = .11). That is, training a component did not have a significant effect on the untrained components. By showing the absence of benefits that generalize beyond the trained components, we question the practical relevance of training specific executive function skills in isolation. Furthermore, the present results might explain the absence of far-transfer effects of working memory training on academic skills (Melby-Lervag & Hulme, 2013; Sala & Gobet, 2017).

So, this study is not that surprising as it states that near transfer can happen more easily, far transfer is far more difficult (pun intended). But the consequences of this finding can be pretty big:

The most important aim of the present meta-analysis was to gain a deeper understanding of whether and to what extent the different components of executive functions are trainable and whether train- ing a specific executive function has an ameliorating effect on other main executive function components. This is of crucial importance because in the long run the main aim of training executive functions skills is to improve children’s everyday functioning; for example, academic and social skills as well as emotion regulation. These complex skills are not supported by one sole executive function but generally rely on the interplay among most of them.

So to be clear the study doesn’t say that executive functions can’t be trained, or…

Overall, we found a significant, medium-sized near-transfer effect. However, no far-transfer effect appeared. More specifically, there were significant near-transfer effects on all three components: a moderate-sized effect on working memory and small-sized effects on inhibition and cognitive flexibility. In contrast, no far-transfer effects were found on working memory, inhibitory control, or flexibility. The finding that there was a significant near-transfer effect excludes the possibility that the interventions in the primary studies were not effective in training the components that they targeted. Instead, performance on the components that were trained did significantly improve, however, these gains did not transfer to the untrained components.

But what does this mean?

The lack of far-transfer effect found in the present meta-analysis even within the set of executive function skills makes it—though logically not impossible—still highly unlikely that training unique executive functions could have measurable ameliorating transfer effect on more distantly related and complex constructs, such as academic and social skills (Blair & Razza, 2007) that rely just as much on the trained executive function component as on the other untrained and largely unaffected components. The results of the present meta-analysis therefore provide a possible explanation for the previously found absence of far-transfer effects of working memory trainings on academic skills.

So what do we need to do?

The present meta-analysis shows that there are limited practical benefits— other than on the trained component— of training single executive function components in childhood. Thus, it might be more advisable, both in the educational and in the clinical practice, to use approaches that target multiple executive function components.

But maybe the most important question is in the end of the meta-analysis:

“The lack of causal evidence for significant relationships among the three core components might contradict accounts of executive functions as a single construct.”

So maybe you can’t train ‘executive functions‘, but you can train an executive function.