Modeling the bilingual advantage : do results differ between analysis methods?

Abstract

Whether bilingualism results in improved executive function is controversial. According to some researchers, putative bilingual advantages can be explained by individual differences in unmeasured non-linguistic variables. Additionally, commonly used models containing exclusively fixed-effects do not account for the data structure inherent in multi-trial behavioral tasks. Mixed-effects models by contrast address both of these issues, allowing for a more valid test of the bilingual advantage hypothesis. The present study aimed to assess whether the choice of analysis method impacted on results when investigating the bilingual advantage in executive function. Simon task data from Mandarin-English speaking Chinese adolescents were analyzed using separate fixed and mixed-effects models. Comparisons between models revealed considerable differences in the pattern of results. Most notable was the observation that a number of previously significant main effects on task performance were no longer significant in a mixed model accounting for the inclusion of multiple trials for each participant. Implications for the bilingual advantage are discussed.