Effects of bimodal divided attention on cortical representations of linguistic context during continuous speech perception in noise

Xie, Zilong · 2026 · Crossref

DOI: 10.64898/2026.04.28.721419

archive: archived pipeline: cataloged

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study investigates how bimodal divided attention—specifically, splitting cognitive resources between a demanding visual task and auditory speech perception in noise—modulates cortical representations of linguistic context. While previous research established that linguistic context at sublexical, word-form, and sentence levels facilitates speech processing in quiet conditions, it remained unclear how background noise and crossmodal attentional demands interact to influence these hierarchical processing stages. The authors hypothesized that sentence-level context processing, which requires integrating information over longer timescales, would be more susceptible to attentional disruption than lower-level processing. The experiment involved 24 young adults who performed a dual-task paradigm while EEG was recorded. Participants listened to audiobook excerpts of *Alice’s Adventures in Wonderland* mixed with multi-talker babble noise (SNR 5 dB) as a secondary task. Concurrently, they performed a primary visuospatial n-back task, which imposed either low cognitive load (0-back) or high cognitive load (3-back). Behavioral data showed that speech comprehension accuracy was significantly lower under high-load conditions (mean 49.6%) compared to low-load conditions (mean 77.0%). To assess neural processing, the authors used multivariate temporal response function (mTRF) models to predict EEG responses from information-theoretic predictors (entropy and surprisal) derived from sublexical, word-form, and sentence-level linguistic models, as well as acoustic and visual features. Results indicated that neutral tracking of linguistic context was significant at the word-form and sentence levels, but not at the sublexical level. Critically, the strength of neural tracking for sentence-level linguistic representations was significantly reduced under high cognitive load compared to low load, with these effects emerging at latencies beyond 200 ms. In contrast, neural tracking of word-form-level representations and acoustic features of the speech signal remained unaffected by the dual-task load. These findings suggest that bimodal divided attention selectively disrupts the cortical processing of higher-level, sentence-based linguistic context, while preserving lower-level acoustic and word-form processing. This selective impairment in higher-level linguistic integration likely contributes to the observed deficits in speech comprehension when individuals attempt to multitask in noisy environments.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.8-27b-gittensor on 2026-08-23; verification: pending re-verification.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• auditory