No “far transfer” – chess, memory training and music just make you better at chess, memory training and music

Learning to ride a BMX obviously helps you handle a racing bike. How about a motorbike? A unicycle? A helicopter? The question of how far learning generalises beyond the original context has continued to vex psychologists. The answer has real-life implications for education and health. For instance, it bears on whether, by undertaking activities like brain training or learning chess, we can expect to boost our overall memory or intelligence – what's known as "far transfer". In a new review in Current Directions in Psychological Science, Giovanni Sala and Fernand Gobet of the University of Liverpool conclude that in fact the evidence for far-transfer is very weak.

Proponents of "far transfer" point to highly-cited studies suggesting it can happen, such as a paper from 2008 that claimed the kind of "working memory" exercise that's found in many brain training programmes led to improvements in problem-solving abilities or what's known as "fluid intelligence". However, it can be risky to read too much into single studies, so Sala and Gobet conducted three meta-analyses (which combine the data from multiple previous studies), involving three activities that are strong candidates for far transfer: chess, music and working memory training. The research was all focused on children, because you would expect any far-reaching benefits to be greater in those whose cognitive ability is still very much in development.

The results suggested that instruction in chess or music, or working memory training, all led to small-to-moderate gains in broader abilities, like memory, general intelligence, and academic attainment. On the face of it, this is evidence for far transfer. But Sala and Gobet picked apart the studies within the analyses to find something dispiriting: "the size of the effects was inversely related to the quality of the experimental design."

Limiting the analysis to the best-designed studies, they found little or no evidence of far-transfer. The only exception was a robust effect of working memory training on other memory tasks, which is arguably "near transfer" rather than far transfer. This tallies with an in-depth evaluation of brain training published last year that concluded such training improves performance at the specific skills being practiced, but that claims about its broader benefits have little support once you discount the less stringently designed studies.

So far transfer remains a weakly supported concept. Theoretically, the evidence lines up with the "common elements theory" of learning, proposed by Thorndike and Woodworth more than one hundred years ago: learning will have the most relevance for the domains precisely being practiced, some relevance for domains with lots in common, and little for those more removed. The practical implications for education are straightforward: if you want to acquire a skill, train that skill, or at least something closely related.