The argument is intuitive: a lifetime of holding two languages apart should train the machinery that holds anything apart. Decades of work later there is still no consensus about whether the advantage exists. So a team pooled 891 effect sizes from 152 adult studies across six domains, added unpublished data, and then did the step most reviews skip: corrected for . What survived the correction?
Before correction there was a very small advantage, and after correction there was none. The uncorrected estimates showed bilinguals slightly ahead on , shifting and working memory, with nothing for monitoring or attention. Once the analyses were corrected for the publication bias they observed, no evidence of a bilingual advantage remained. showed a small disadvantage, plausibly from less exposure to each language.
Managing two languages for a lifetime should, on the face of it, train general cognitive control. Bilinguals have therefore been argued to develop superior executive functioning compared with monolinguals, and the claim has been investigated extensively without producing a consensus.
This review went at the whole literature. The authors synthesized 891 effect sizes from 152 studies of adults, covering six executive domains, and they included unpublished data as well as published. They also considered a range of study, task and participant variables as potential moderators.
Before correcting for bias, the picture was a very small bilingual advantage for inhibition, shifting and working memory, and nothing for monitoring or attention. That is already weaker than the popular version of the claim. Then the correction for observed publication bias was applied, and no evidence for a bilingual advantage remained.
Two further details sharpen the conclusion. Verbal fluency showed a small bilingual disadvantage, possibly reflecting less exposure to each language when both are used in a balanced manner. And moderator analyses did not support the theoretical presuppositions about when the advantage should appear, which is what you would expect if there is nothing there to moderate. The authors conclude that the available evidence gives no systematic support for the widely held notion. Note the boundary: this is adults, and a bias correction is an estimate with assumptions of its own.
wat. The bilingual edge on executive tasks was small before anyone corrected for missing null results, and gone afterward.
wat wat
It is a demonstration of what publication bias does rather than an argument about language. A very small effect that disappears under correction is the signature of a literature where null results stayed in drawers.
How do we know?
Lehtonen and colleagues synthesized 891 effect sizes from 152 studies of adults across six executive domains, included unpublished data, tested study and participant moderators, and corrected for observed publication bias.
What is still unknown
This covers adults, so it says nothing directly about children or older adults, and bias-correction methods are estimates with assumptions of their own rather than a definitive accounting.
Learn something?
Where to next
How fast do adults actually read?meta-analysis
What does your brain already know about a word that escaped you?language
Phone, laptop and television at once. Does that really make you easier to distract?meta-analysis