Working memory functioning is strongly correlated with fluid intelligence. So, does working memory training raise intelligence? In this study, the answer was no, even after the participants were paid to receive 40 hours of training across 2 years.
Working memory training did improve working memory functioning substantially. However, this did not lead to improvements in intelligence. In specific fluid intelligence tasks (which, in theory, should benefit the most from working memory training) improved d = -.12 to .11, with a latent variable improvement improvement of just d = .08 (p = .52). For crystallized intelligence tasks, the training group improved d = -.33 to d = .21, and the latent variable “improved” d = -.10 (p = .38).
In other words, the crystallized and fluid intelligence tasks showed a mix of improvement and worsening in the training group, but the underlying intelligence ability did not change after working memory training.
The authors were very clear about their results: “Thus, the training-induced improvements in WM were not accompanied by significant improvements in either of two prominent factors of intelligence” (p. 724). “. . . our findings showed reliable evidence for the lack of transfer from WM training to intelligence . . .” (p. 725). “Thus, given our results and the available meta-analytic evidence, we do not think that WM training in its current form allows to improve [sic] cognitive abilities” (p. 727).
Trying to raise intelligence is a worthy goal, but many studies like this one show that improving working memory training does not increase intelligence. If 2 years of working memory training can’t raise intelligence, then it is unlikely this will be a productive line of research to continue. Scientists should look elsewhere in their efforts to raise intelligence.
The 2-year timeframe and 40 hours of training make this study really definitive. If working memory training were going to transfer to intelligence, this should have been enough to show it. The fact that WM improved substantially but fluid and crystallized intelligence showed zero transfer confirms what meta-analyses have been suggesting: you get better at the specific tasks you practice, but that skill doesn’t generalize to broader cognitive abilities. The training group got better at n-back and similar WM exercises, but that’s all they got better at. This has huge implications for the brain training industry, which is basically selling false hope.
What’s important is they measured both manifest performance on individual tasks and latent factors underlying intelligence. Some individual tasks showed small improvements, but the underlying g factor didn’t budge. This means any task-level gains were just practice effects on those specific items, not real cognitive enhancement. The d = .08 and d = -.10 effects for fluid and crystallized intelligence are essentially zero, especially after 40 hours of training. This study basically kills the idea that you can raise intelligence through cognitive training exercises. If it worked, we’d see it here. The correlation between WM and intelligence exists because they share underlying neural architecture, not because one causes the other.
“Thus, the training-induced improvements in WM were not accompanied by significant improvements in either of two prominent factors of intelligence” (p. 724).