Two Languages in One Head

Ask a Russian-English bilingual to pick up the marker, in an English-only session, and track their eyes. They flick briefly towards a stamp, because the Russian word for stamp sounds like the English word marker. There was no reason for Russian to be active. It was active anyway.

That is co-activation, and it is the fact everything else in this chapter rests on. Both languages of a bilingual remain active during comprehension and production, always, which means a bilingual is continuously selecting one and suppressing the other. It explains the wrong word intruding, it explains the slightly slower retrieval, and it generated the hypothesis that has dominated the field for twenty years.

The hypothesis was that all this practice at selecting and suppressing should transfer: bilinguals ought to have better executive control generally (better at ignoring distraction, at switching tasks, at holding conflicting information). Early studies found exactly that. The finding was appealing, it was widely reported, and it entered educational policy arguments.

It has largely not survived. As sample sizes grew and pre-registration spread, the effect shrank towards zero. Analyses of conference abstracts found that studies supporting the advantage were far more likely to reach publication than studies finding nothing, which is the mechanism made visible: the literature looked strong because the null results were invisible. It is now one of the standard teaching examples of publication bias, alongside several others from the same decade.

The claim in the opposite direction never had support at all and persists anyway, usually delivered as advice to parents. Raising a child bilingually does not delay language development. Bilingual children hit milestones on the normal schedule when both languages are counted together.

What has held up is narrower and mechanically sensible. Vocabulary in each individual language is somewhat smaller, while the combined total matches monolingual peers, dividing exposure across two systems means less of each. Word retrieval is slightly slower, because two candidates are active and one must be suppressed. Neither has any practical consequence.

Except in one setting, where it has caused real harm. A bilingual child assessed on a vocabulary test administered in one language will score below age norms, because the test measures half their vocabulary. Children have been placed in remedial provision on exactly this basis. The result is not meaningless (it accurately measures the language it was given in, which may matter for schooling in that language) but it cannot support a conclusion about the child's language development overall, and it has repeatedly been used to do so.

Finally, code-switching deserves rehabilitating. Switching languages mid-conversation or mid-sentence is frequently described as sloppiness, or as evidence of not knowing either language properly. It is neither. Fluent bilinguals switch at grammatically licensed boundaries and almost never at points that would break either grammar, which requires simultaneous command of both systems and is demonstrably rule-governed. The right summary of bilingualism generally is the one Grosjean gave: a bilingual is not two monolinguals in one person. It is a different arrangement, rather than a better or a worse one.