Bill Gates admitted recently that he felt “pretty stupid” for not knowing any languages other than English. To be fair to Gates, most native English speakers in the US (and the UK) are monolingual (despite the fact that most experts believe that bilingualism— or even multilingualism—is the norm). While nobody seriously thinks that Bill Gates is dumb, his remarks touched on a fascinating controversy in the scientific literature: are bilingual people smarter than monolinguals?

As Judith Kroll notes, until relatively recently, it was generally assumed—albeit without much evidence—that “young children will be confused by exposure to more than one language and harmed by delays in development.” This view went hand-in-hand with the idea that monolingualism is somehow the norm, and that bilinguals constitute “a special group of language users, much like brain-damaged patients, children with language disorders, or deaf individuals who use a signed language to communicate.”

Then along came Ellen Bialystok of York University in Toronto. In a series of studies, Bialystok and colleagues found that, although bilinguals perform worse than monolinguals on very simple tasks (e.g., naming pictures or generating words that begin with a particular letter), they actually show better performance on cognitive control tasks—those that measure participants’ ability to exert deliberate control their mental processes. One example of a cognitive control task: a participant might be asked to listen to one set of words presented over headphones and see another set of words presented on a computer screen, before being asked to recall the words that they had heard, but not those that they had seen (or vice versa).

Why do bilinguals show an advantage over monolinguals on cognitive control tasks? Bialystok argues that bilinguals are better at suppressing irrelevant or interfering information, because this is exactly what they have to do every day. For example, when speaking Spanish, a bilingual Spanish-English speaker must access the relevant Spanish words for the objects and ideas s/he has in mind, while suppressing the corresponding English words.

This suggests that bilingual children, struggling to cope with two languages, are in fact giving their brain a workout, and end up smarter than their monolingual classmates.

Or so it seemed. This month, a group of psychologists based at the University of Edinburgh published a paper showing apparent evidence of publication bias in the bilingual-advantage literature. Before a study is fully completed and written up, many researchers choose to present their preliminary findings at a conference. In order to secure a speaking slot (which can be quite competitive), researchers must submit a short abstract summarizing the research. The abstracts for accepted talks are then published in the conference program. The Edinburgh team got hold of as many conference abstracts as possible on the topic of the bilingual advantage, and then looked to see how many of these were followed up by a bona-fide peer-reviewed journal article reporting the final detailed results of the research. What they found was that the conference abstracts reporting a bilingualism advantage generally made it into fully-fledged-article form, while those that failed to find such an advantage often sunk without trace. (If you’re worried that this method sounds a bit informal, you can rest assured; a funnel-plot—the standard statistical analysis used in such cases—also found evidence of publication bias).

Now, when this JSTOR Daily post originally appeared last week, I concluded, on the basis of this Edinburgh review, that the claim of a bilingual advantage is “probably not true.” However, as two readers—Dan Hirschman and Martino Comelli—pointed out on Twitter, I wasn’t justified in leaping to such a strong conclusion. Since the overall bilingualism advantage across the published papers (0.3 on the scale explained in this paper) seems rather small compared to the fact that “63% of the studies supporting the bilingual advantage were published, compared with only 36% of the studies that challenged it,” I assumed that the unpublished studies—were they actually to be published—would wipe out this difference. For what it’s worth, I still think this is the most plausible interpretation of these data. But it was rather gung-ho of me to baldly present this conclusion without even explaining my reasoning, let alone backing it up with a mathematical analysis [1]. Indeed, as the authors point out, another possible interpretation is that, despite the existence of publication bias, “the cognitive bilingual advantage is genuine, albeit smaller and less stable than often presented in the literature.”

So is there a bilingual advantage? Currently we just don’t know. But two things are clear. The first is that researchers—and, perhaps more importantly, the journal editors who decide which papers to publish and which to reject—are people too, and people just love a good story. Now, “Bilingualism is good for your brain” is a great story. But it is for precisely this reason that we must be extra careful to ensure that findings that challenge this story are as likely to be published as those that support it. The second—and the other side of the same coin—is that those of us who set ourselves up as myth busters need to be just as cautious—and just as explicit when showing our working—as those who make the claims to begin with.

[1] There is actually a way to calculate the number of null-effect studies that would be needed to wipe out this effect—the “fail-safe N.” But (a) it requires assumptions about the unpublished studies that we just don’t have and (b) calculating it is way beyond my expertise. Still, if any meta-analysis experts reading this would like to try, I’d be fascinated to know the result.