Multiscale neural signatures of long-lag repetition learning: a simultaneous fNIRS-EEG study.

SIGNIFICANCE: Repetition-based learning supports word recognition and reading fluency, yet its neural mechanisms remain poorly understood, particularly where deficits in consolidating repeated linguistic input may underlie reading and language difficulties. Establishing reliable neural markers of repetition-based learning is a critical first step toward characterizing the phenotypes of these difficulties in children.

AIM: We developed a whole-brain simultaneous functional near-infrared spectroscopy-electroencephalography (fNIRS-EEG) approach to capture fast temporal electrophysiological and spatially accurate hemodynamic signals during long-lag repetition learning.

APPROACH: Thirty participants performed a semantic judgment task with words repeated across four runs with intervening items, modeling repetition demands central to reading instruction. A multimodal decoding framework integrating electrophysiological, hemodynamic, and cross-modal fNIRS-EEG correlation features was used to classify repetition-based learning.

RESULTS

Repeated words showed response time reductions. EEG captured a frontal N400-range repetition-related negativity and attenuation of alpha-beta suppression, reflecting changes in recognition-related and post-recognition processing. fNIRS revealed left inferior frontal and superior temporal efficiency increases alongside right supramarginal activation, suggesting integrative processing. Multimodal classifiers outperformed unimodal classifiers, indicating that each modality captured complementary aspects of learning.

CONCLUSIONS

Combining fNIRS and EEG with multimodal machine learning provides a richer characterization of repetition-based learning than either modality alone, establishing a framework for phenotyping language learning capacity and identifying neural markers of reading difficulties in pediatric populations.