Multilingual Stroop performance: Effects of trilingualism and proficiency on inhibitory control

Abstract

Previous research suggests that multilinguals' languages are constantly co-activated and that experience managing this co-activation changes inhibitory control function. The present study examined language interaction and inhibitory control using a colour-word Stroop task. Multilingual participants were tested in their three most proficient languages. The classic Stroop effect was detected in all three languages, with participants performing more accurately on congruent than on incongruent trials. Multilinguals were faster and more accurate in the within-language-competition condition than in the between-language-competition condition, indicating that additional processing costs are required when stimulus and response languages differ. Language proficiency influenced speed, accuracy and error patterns in multilingual Stroop task performance. These findings augment our understanding of language processing and inhibitory control in multilingual populations and suggest that experience using multiple languages changes demands on cognitive function.

Keywords:

• multilingualism
• Stroop task
• cognitive control
• proficiency
• within- and between-language interference

Acknowledgements

Preparation of this manuscript was supported in part by grant NICHD RO1HD059858 to the first author. We thank Maria Khristina S. Manueli and Vera Denisova for their contributions to data collection and the members of the Northwestern Bilingualism and Psycholinguistics Laboratory for comments on earlier drafts of this paper.

Notes

1. Response speed is reported across the three languages, but not for incongruent and congruent items separately. Because incongruent and congruent trials were presented in an intermixed fashion, with an equal number of responses in the two conditions, separate response speed measures for the two conditions could not be obtained.

2. For analyses on specific error types, only by-subjects analyses (F₁, t₁) are reported, because by-item analyses were not possible due to the small number of errors.

3. Analyses on match vs. mismatch of stimulus and response languages were conducted only on L1 and L2, because not enough participants were available who showed a match between stimulus language and response language for their L3 (i.e. L3 was English for three participants only), therefore this analysis was conducted with 23 participants.