Is there a bilingual advantage on interference-control tasks? A multiverse meta-analysis of global reaction time and interference cost

A contentious issue in contemporary psycholinguistics is whether bilingualism enhances executive functions. Here, we report a meta-analysis of 80 studies (253 effect sizes) comparing performance of monolinguals and bilinguals on non-verbal interference-control tasks, while examining potential moderators of effects on two dependent variables (DVs): global reaction time (RT) and interference cost. We used a multiverse approach to determine how robust conclusions were to several dataset construction and analysis decisions. In our "preferred" analysis, using a broad definition of bilinguals and standard versions of interference-control tasks, there was a very small but significant bilingual advantage for global RT (g =.13), which became non-significant once corrected for publication bias. For interference cost, there was a very small but significant bilingual advantage (g =.11). Effects were not significantly moderated by task or participant age, but were moderated by an interaction between age of second language acquisition (AoA) and the DV. Unexpectedly, larger effect sizes for interference cost were observed for studies involving bilinguals with late as opposed to early AoA. The multiverse analysis produced results largely consistent with the preferred analysis, confirming our conclusion that evidence for a bilingual advantage on interference-control tasks is weak.