Neural signatures of second language proficiency in narrative processing

Making sense of speech in a second language relies on multiple abilities. Differences in brain activity related to proficiency in language tasks have often been attributed to processing demands. However, during naturalistic narrative comprehension, listeners at different proficiency levels may form different representations of the same speech. We hypothesized that the synchronization of these representations across people could thus be used to measure second-language proficiency. Using a searchlight shared response model, we found that highly proficient participants showed synchronization in regions similar to those of native speakers, including in the default mode network and in the lateral prefrontal cortex. In contrast, participants with low proficiency showed more synchronization in auditory cortex and word-level semantic processing areas in the temporal lobe. Moderate proficiency showed the greatest neural diversity, suggesting lower consistency in the source of this partial proficiency. Based on these synchronization differences, we were able to reliably classify the proficiency level or predict behavioral performance on an independent English test in held-out participants, suggesting the identified neural systems represented proficiency-sensitive information that was generalizable to other individuals. These findings suggest higher second-language proficiency leads to a more native-like neural processing of naturalistic language, including in systems beyond the cognitive control network or the core language network. Highlights Neural synchronization in second-language speech processing reflects proficiency. High-proficiency individuals neurally resemble native speakers. Low-proficiency individuals are synchronized in perceptual and word semantics areas. Proficiency level can be predicted using neural synchronization signatures.