Multivariate Decoding and Drift-Diffusion Modeling Reveal Adaptive Control in Trilingual Comprehension

TL;DR

Trilinguals adjust their brain control strategies during comprehension depending on the language context, with proactive control being key in high-conflict situations.

Contribution

This study reveals context-dependent proactive control in trilingual comprehension, contrasting with production and showing no reactive control.

Findings

Drift-diffusion modeling showed distinct processing profiles across language contexts, with L1–L2 having the lowest comprehension efficiency.

Proactive control was evident in the L1–L3 context through larger P300 and smaller N400 for L1-to-L3 switches.

MVPA identified distinct spatiotemporal neural patterns for different language contexts and switching directions.

Abstract

Background/Objectives: The Adaptive Control Hypothesis posits varying control demands across language contexts in production, but its role in comprehension is underexplored. We investigated if trilinguals, who manage three dual-language contexts (L1–L2, L2–L3, L1–L3), exhibit differential proactive and reactive control demands during comprehension across these contexts. Methods: Thirty-six Uyghur–Chinese–English trilinguals completed an auditory word-picture matching task across three dual-language contexts during EEG recording. We employed behavioral analysis, drift-diffusion modeling, event-related potential (ERP) analysis, and multivariate pattern analysis (MVPA) to examine comprehension efficiency, evidence accumulation, and neural mechanisms. The design crossed context (L1–L2, L2–L3, L1–L3) with trial type (switch vs. repetition) and switching direction (to dominant vs. non-dominant language). Results: Despite comparable behavioral performance, drift-diffusion modeling revealed distinct processing profiles across contexts, with the L1–L2 context showing the lowest comprehension efficiency due to slower evidence accumulation. In the L1–L3 context, comprehension-specific proactive control was indexed by a larger P300 and smaller N400 for L1-to-L3 switches. Notably, no reactive control (switch costs) was observed across any dual-language context. MVPA successfully classified contexts and switching directions, revealing distinct spatiotemporal neural patterns. Conclusions: Trilingual comprehension switching mechanisms differ from production. Reactive control is not essential, while proactive control is context-dependent, emerging only in the high-conflict L1–L3 context. This proactive strategy involves allocating more bottom-up attention to the weaker L3, which, unlike in production, enhances rather than hinders overall efficiency.

Figures