CoChat

The DOI matches its Crossref record; 69 of the 69 references it cites were found in a registry or an index; none of the 67 sources its claims cite is checked for support yet.

Cortical Tracking of Continuous Speech Under Bimodal Divided Attention

Open access
Xie, Zilong, Brodbeck, Christian, Chandrasekaran, Bharath · 2023 · Neurobiology of Language

Published version of a preprint

Abstract

Speech processing often occurs amid competing inputs from other modalities, for example, listening to the radio while driving. We examined the extent to which dividing attention between auditory and visual modalities (bimodal divided attention) impacts neural processing of natural continuous speech from acoustic to linguistic levels of representation. We recorded electroencephalographic (EEG) responses when human participants performed a challenging primary visual task, imposing low or high cognitive load while listening to audiobook stories as a secondary task. The two dual-task conditions were contrasted with an auditory single-task condition in which participants attended to stories while ignoring visual stimuli. Behaviorally, the high load dual-task condition was associated with lower speech comprehension accuracy relative to the other two conditions. We fitted multivariate temporal response function encoding models to predict EEG responses from acoustic and linguistic speech features at different representation levels, including auditory spectrograms and information-theoretic models of sublexical-, word-form-, and sentence-level representations. Neural tracking of most acoustic and linguistic features remained unchanged with increasing dual-task load, despite unambiguous behavioral and neural evidence of the high load dual-task condition being more demanding. Compared to the auditory single-task condition, dual-task conditions selectively reduced neural tracking of only some acoustic and linguistic features, mainly at latencies >200 ms, while earlier latencies were surprisingly unaffected. These findings indicate that behavioral effects of bimodal divided attention on continuous speech processing occur not because of impaired early sensory representations but likely at later cognitive processing stages. Crossmodal attention-related mechanisms may not be uniform across different speech processing levels.

Overview

CoChat Summary

TL;DR

Bimodal divided attention (visual n-back while listening to audiobook stories) reduced behavioral speech comprehension at high visual load and selectively reduced late neural tracking of acoustic onsets and linguistic predictive features, but most acoustic/linguistic tracking did not scale with dual-task load. The results suggest divided-attention costs on continuous speech arise at later cognitive stages rather than early sensory representation.

Key findings

  • High load dual-task lowered story comprehension accuracy (65.63%) relative to low load (84.58%) and auditory single-task (88.96%); auditory single-task and low load did not differ significantly.
  • Visual neural tracking increased with task relevance/load: mean Δz was 0.075 in high load, 0.053 in low load, and 0.020 in auditory single-task.
  • Compared with auditory single-task, dual-task reduced neural tracking of onset spectrogram (F(2, 30) = 4.93, p = .033) and combined linguistic predictors (F(1.41, 21.15) = 6.66, p = .029), mainly at latencies >200 ms; gammatone spectrogram and word onset showed no significant task effect.
  • Increasing dual-task load did not further reduce speech tracking; in dual-task conditions only, phoneme surprisal tracking was numerically higher at high vs. low load (t(15) = 2.20, uncorrected p = .044), but this was not robust after correction.

Methods

  • EEG study with 16 participants (18–23 years; 11 female, 5 male) after exclusions; native American English speakers without significant formal music training.
  • Participants performed a primary visuospatial n-back task (3-back high load, 0-back low load) while listening to ~60-s *Alice’s Adventures in Wonderland* audiobook segments as a secondary task; contrasted with an auditory single-task condition; comprehension questions followed each trial.
  • Neural tracking was assessed with multivariate temporal response function encoding models predicting EEG from visual, acoustic (gammatone and onset spectrograms), and linguistic predictors (sublexical, word-form, sentence surprisal/entropy; word/phoneme onsets), using fivefold cross-validation and Δz prediction accuracy; cluster-based permutation and repeated-measures ANOVAs tested effects.

Limitations

  • Paper states dual-task conditions required button-press responses (~6 per trial) whereas auditory single-task did not, a potential confound; authors argue it is unlikely because reductions were feature-specific rather than uniform.
  • Paper states the audiobook was a well-known text, so prior familiarity could affect performance; exit questionnaire found no significant main effects/interactions of reading the book or watching a movie.
  • Text makes evident: small sample (N = 16), young adult native English speakers, single story; some effects (e.g., surprisal load effect) were not significant after FDR correction or only emerged when excluding single-task; paper also notes mTRF response functions may include components confounded with correlated speech features.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1162/nol_a_00100

Cites

69 of 69 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1. Trial design. In the two dual-task conditions, the primary task was to respond to the visual stimuli and the secondary task was to attend to auditory stimuli (story segments of about 60 s). (A) In the high load condition, participants responded only when the current blue square matched the one 3 positions back (examples highlighted in red squares). (B) In the low load condition, participants responded only when the current blue square matched the first square in each trial (highlighted in the red square). (C) In the auditory single-task condition, participants were instructed to attend to the auditory stimulus and ignore the visual stimuli. At the end of each trial, participants responded to two multiple-choice comprehension questions for the story segments. ISI: interstimulus interval.
Figure 1. Trial design. In the two dual-task conditions, the primary task was to respond to the visual stimuli and the secondary task was to attend to auditory stimuli (story segments of about 60 s). (A) In the high load condition, participants responded only when the current blue square matched the one 3 positions back (examples highlighted in red squares). (B) In the low load condition, participants responded only when the current blue square matched the first square in each trial (highlighted in the red square). (C) In the auditory single-task condition, participants were instructed to attend to the auditory stimulus and ignore the visual stimuli. At the end of each trial, participants responded to two multiple-choice comprehension questions for the story segments. ISI: interstimulus interval.CC BY · PubMed Central
Figure 2. An excerpt of raw electroencephalography (EEG) responses from all 64 electrodes (top row) and the predictor variables (subsequent rows) used to model the EEG responses. Note that visual predictors consist of a separate one-dimensional array with impulses for onsets and offsets of the blue squares. They are combined into a single predictor in this example for illustration purposes.
Figure 2. An excerpt of raw electroencephalography (EEG) responses from all 64 electrodes (top row) and the predictor variables (subsequent rows) used to model the EEG responses. Note that visual predictors consist of a separate one-dimensional array with impulses for onsets and offsets of the blue squares. They are combined into a single predictor in this example for illustration purposes.CC BY · PubMed Central
Figure 3. Behavioral performance on visual and auditory tasks. (A) Accuracy on the low load (0-back) and high load (3-back) visual tasks, which was calculated as the difference in hit rates (i.e., correctly responding to a target) and false alarm rates (i.e., identifying a non-target as being a target). (B) Reaction time on the low load (0-back) and high load (3-back) visual tasks, which was calculated for hits only. (C) Accuracy on the auditory task, which was calculated as the percentage of correctly answered story questions. Individual lines in A–C denote individual participants (n = 16). (D) Correlation between the change in auditory accuracy (i.e., (low load − high load)/low load) and the change in visual RT (i.e., (high load − low load)/low load). The gray line is the linear regression line. n.s. (not significant): p > 0.05, *** p < 0.001.
Figure 3. Behavioral performance on visual and auditory tasks. (A) Accuracy on the low load (0-back) and high load (3-back) visual tasks, which was calculated as the difference in hit rates (i.e., correctly responding to a target) and false alarm rates (i.e., identifying a non-target as being a target). (B) Reaction time on the low load (0-back) and high load (3-back) visual tasks, which was calculated for hits only. (C) Accuracy on the auditory task, which was calculated as the percentage of correctly answered story questions. Individual lines in A–C denote individual participants (n = 16). (D) Correlation between the change in auditory accuracy (i.e., (low load − high load)/low load) and the change in visual RT (i.e., (high load − low load)/low load). The gray line is the linear regression line. n.s. (not significant): p > 0.05, *** p < 0.001.CC BY · PubMed Central
Figure 4. Neural tracking of visual stimuli across task conditions. Visual stimuli were associated with a robust response, which further increased with task relevance and load. (A) Topography showing the increase in prediction accuracy (Δz) due to visual predictors, which was significantly above zero in a single cluster encompassing the highlighted (yellow) electrodes. (B) Prediction accuracy across task conditions. Blue lines denote individual participants: Thicker lines indicate higher prediction accuracy for the high vs. low load condition, and thinner lines indicate lower accuracy for the high versus low load condition. Red asterisks denote p values for comparison between conditions. Error bars denote the 95% within-subject confidence interval (Loftus & Masson, 1994). (C) Global field power of the visual temporal response functions (TRFs). Visual stimuli lasted from 0 to 500 ms. Shaded areas denote within-subject standard errors around the mean (for color labels see panel B). Horizontal lines denote time windows in which TRFs were significantly different between conditions. (D) Topographies of selected times in panel C (grey vertical lines). A-ST: auditory single-task, Lo-DT: low load dual-task, Hi-DT: high load dual-task. ** p < 0.01, *** p < 0.001.
Figure 4. Neural tracking of visual stimuli across task conditions. Visual stimuli were associated with a robust response, which further increased with task relevance and load. (A) Topography showing the increase in prediction accuracy (Δz) due to visual predictors, which was significantly above zero in a single cluster encompassing the highlighted (yellow) electrodes. (B) Prediction accuracy across task conditions. Blue lines denote individual participants: Thicker lines indicate higher prediction accuracy for the high vs. low load condition, and thinner lines indicate lower accuracy for the high versus low load condition. Red asterisks denote p values for comparison between conditions. Error bars denote the 95% within-subject confidence interval (Loftus & Masson, 1994). (C) Global field power of the visual temporal response functions (TRFs). Visual stimuli lasted from 0 to 500 ms. Shaded areas denote within-subject standard errors around the mean (for color labels see panel B). Horizontal lines denote time windows in which TRFs were significantly different between conditions. (D) Topographies of selected times in panel C (grey vertical lines). A-ST: auditory single-task, Lo-DT: low load dual-task, Hi-DT: high load dual-task. ** p < 0.01, *** p < 0.001.CC BY · PubMed Central
Figure 5. Neural tracking of acoustic information across task conditions. (A) Increase in prediction accuracy (Δz) due to acoustic predictors of speech (gammatone and onset spectrogram), which was significantly above zero in a cluster encompassing the highlighted (yellow) electrodes. Blue dots denote individual participants. (B) Prediction accuracy across task conditions for acoustic predictors, i.e., combined gammatone and onset spectrogram. (C) and (D) Prediction accuracy across task conditions for gammatone spectrogram and onset spectrogram separately. Topographies highlight electrodes (yellow) with prediction accuracy that was significantly above zero. Black asterisks denote p values for testing against (above) zero at individual conditions. In B–D, blue lines denote individual participants: Thicker lines indicate lower accuracy for high vs. low load condition, and thinner lines indicate higher accuracy for high vs. low load condition. Red asterisks denote p values for comparison between conditions. Error bars denote 95% confidence interval. (E) and (F) GFP (top) of mTRFs and related topographies (bottom) for gammatone and onset spectrogram. The mTRFs were averaged across the eight frequency bands. Shaded areas denote within-subject standard errors around the mean. Horizontal lines denote time windows in which mTRFs were significantly different between conditions. Topographies are shown for selected times indicated in GFPs (grey vertical lines). * p < 0.05, ** p < 0.01, *** p < 0.001.
Figure 5. Neural tracking of acoustic information across task conditions. (A) Increase in prediction accuracy (Δz) due to acoustic predictors of speech (gammatone and onset spectrogram), which was significantly above zero in a cluster encompassing the highlighted (yellow) electrodes. Blue dots denote individual participants. (B) Prediction accuracy across task conditions for acoustic predictors, i.e., combined gammatone and onset spectrogram. (C) and (D) Prediction accuracy across task conditions for gammatone spectrogram and onset spectrogram separately. Topographies highlight electrodes (yellow) with prediction accuracy that was significantly above zero. Black asterisks denote p values for testing against (above) zero at individual conditions. In B–D, blue lines denote individual participants: Thicker lines indicate lower accuracy for high vs. low load condition, and thinner lines indicate higher accuracy for high vs. low load condition. Red asterisks denote p values for comparison between conditions. Error bars denote 95% confidence interval. (E) and (F) GFP (top) of mTRFs and related topographies (bottom) for gammatone and onset spectrogram. The mTRFs were averaged across the eight frequency bands. Shaded areas denote within-subject standard errors around the mean. Horizontal lines denote time windows in which mTRFs were significantly different between conditions. Topographies are shown for selected times indicated in GFPs (grey vertical lines). * p < 0.05, ** p < 0.01, *** p < 0.001.CC BY · PubMed Central
Figure 6. Neural tracking of linguistic information across task conditions. (A) Increase in prediction accuracy (Δz) due to linguistic predictors of speech (word onsets, phoneme onsets, sublexical surprisal and entropy, word-form surprisal and entropy, and sentence surprisal and entropy), which was significantly above zero across highlighted (yellow) electrodes in the topography. Blue dots denote individual participants. (B) Prediction accuracy for combined linguistic predictors across conditions. Blue lines denote individual participants: Thicker lines indicate lower accuracy for high versus low load condition, and thinner lines indicate higher accuracy for high versus low load condition. Red asterisks denote p values for comparison between conditions. (C) Prediction accuracy for three context levels (sublexical, word-form, and sentence) across conditions. Each level includes entropy and surprisal predictors. (D) Global field power (top) of mTRFs and related topographies (bottom) for each context level. The mTRF GFPs were averaged across entropy and surprisal. (E) Prediction accuracy for entropy and surprisal. Each predictor includes the three context levels. (F) GFP of mTRFs and related topographies for entropy and surprisal. In B, C, and E, error bars denote 95% confidence interval. In D and F, shaded areas denote standard errors around the mean. Horizontal lines denote time windows in which the mTRFs were significantly different between task conditions. Topographies are shown for selected times indicated in GFPs (grey vertical lines). * p < 0.05, *** p < 0.001.
Figure 6. Neural tracking of linguistic information across task conditions. (A) Increase in prediction accuracy (Δz) due to linguistic predictors of speech (word onsets, phoneme onsets, sublexical surprisal and entropy, word-form surprisal and entropy, and sentence surprisal and entropy), which was significantly above zero across highlighted (yellow) electrodes in the topography. Blue dots denote individual participants. (B) Prediction accuracy for combined linguistic predictors across conditions. Blue lines denote individual participants: Thicker lines indicate lower accuracy for high versus low load condition, and thinner lines indicate higher accuracy for high versus low load condition. Red asterisks denote p values for comparison between conditions. (C) Prediction accuracy for three context levels (sublexical, word-form, and sentence) across conditions. Each level includes entropy and surprisal predictors. (D) Global field power (top) of mTRFs and related topographies (bottom) for each context level. The mTRF GFPs were averaged across entropy and surprisal. (E) Prediction accuracy for entropy and surprisal. Each predictor includes the three context levels. (F) GFP of mTRFs and related topographies for entropy and surprisal. In B, C, and E, error bars denote 95% confidence interval. In D and F, shaded areas denote standard errors around the mean. Horizontal lines denote time windows in which the mTRFs were significantly different between task conditions. Topographies are shown for selected times indicated in GFPs (grey vertical lines). * p < 0.05, *** p < 0.001.CC BY · PubMed Central
Figure 7. Neural tracking of word onsets across task conditions. (A) Topography showing the increase in prediction accuracy (Δz) due to word onsets, which was significantly above zero across highlighted (yellow) electrodes. (B) Prediction accuracy across conditions. Blue lines denote individual participants: Thicker lines indicate higher accuracy for the high vs. low load condition, and thinner lines indicate lower accuracy for the high vs low load condition. Error bars denote 95% within-subject confidence interval (Loftus & Masson, 1994). (C) Global field power of mTRFs. Shaded areas denote within-subject standard error around the mean. (D) Topographies of selected times in panel C (grey vertical lines).
Figure 7. Neural tracking of word onsets across task conditions. (A) Topography showing the increase in prediction accuracy (Δz) due to word onsets, which was significantly above zero across highlighted (yellow) electrodes. (B) Prediction accuracy across conditions. Blue lines denote individual participants: Thicker lines indicate higher accuracy for the high vs. low load condition, and thinner lines indicate lower accuracy for the high vs low load condition. Error bars denote 95% within-subject confidence interval (Loftus & Masson, 1994). (C) Global field power of mTRFs. Shaded areas denote within-subject standard error around the mean. (D) Topographies of selected times in panel C (grey vertical lines).CC BY · PubMed Central

Verification

Claims the paper makes

0 Valid · 1 Not valid · 34 Not checked · 134 Own results and interpretations

94 more claims

Create a free account to see all 169 claims and to verify the 64 references still open.

See all claims — free account
  • It may state two separate claims as one. 39
  • Probably not a substantial claim. 34

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

4 more claims about this paper

Create a free account to see every claim other papers make about this one.

See all claims — free account

Statistics recomputed

Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.

  • Not confirmedF(1.41, 21.15) = 6.66reported p = .029 · recomputed p = .011

    “F(1.41, 21.15) = 6.66, p = .029”

  • Not confirmedF(2.55, 38.18) = 1.19reported p = .4 · recomputed p = .323

    “F(2.55, 38.18) = 1.19, p = .40”

  • Not confirmedF(2, 30) = 0.32reported p = .77 · recomputed p = .729

    “F(2, 30) = 0.32, p = .77”

  • Not confirmedF(1.32, 19.86) = 1.29reported p = .4 · recomputed p = .283

    “F(1.32, 19.86) = 1.29, p = .40”

  • Not confirmedF(1, 15) = 0.31reported p = .65 · recomputed p = .586

    “F(1, 15) = 0.31, p = .65”

5 of 41 · the rest aren’t loaded

36 more recomputed tests

Create a free account to see every test the Stats check recomputed.

See all tests — free account

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

2 of 2

Citations

Cites

Show all 69 references

Research works printed but not linked: 7

  1. Brain Products, 2023, ActiCAP
  2. Brain Products, 2023, ActiCHAmp and Pycoder
  3. Easycap, 2023, Easycap
  4. Etymotic, 2023, ER-3
  5. Interacoustics, 2023, Equinox 2.0
  6. Schneider, 2002, E-Prime: User’s guide. Reference guide. Getting started guide
  7. VPixx Technologies, 2022, VIEWPixx

Cited by

See all 21

Metadata

Authors
Xie, Z., Brodbeck, C., & Chandrasekaran, B.
Affiliations
  • Florida State University (US)
  • University of Connecticut (US)
  • University of Pittsburgh (US)
Corresponding author
  • Xie, Zilong
  • Chandrasekaran, Bharath
Venue
Neurobiology of Language
Publisher
MIT Press
Published
2023
Volume
4
Issue
2
Pages
318-343
ISSN
2641-4368 (ISSN-L)
OpenAlex
W4317434061
Cited by
21
References
69
Citations by year
2026: 10 · 2025: 7 · 2024: 1 · 2023: 3
FWCI
2.20
Type
journal article
Language
en
Open access
Locations
  • Neurobiology of Language (publishedVersion, cc-by)
  • bioRxiv (Cold Spring Harbor Laboratory) (submittedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • OSF Preprints (OSF Preprints) (submittedVersion)
Funders
  • National Institute on Deafness and Other Communication Disorders (R01DC013315)
  • National Science Foundation (BCS-1754284)
Topics
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Neurobiology of Language and Bilingualism
  • Social Sciences › Psychology › Experimental and Cognitive Psychology › Multisensory perception and integration
  • Social Sciences › Psychology › Experimental and Cognitive Psychology › Phonetics and Phonology Research
Keywords
cross-modal attention, speech understanding, audiobook listening, cognitive load

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-11-19, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-11-19 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Cortical Tracking of Continuous Speech Under Bimodal Divided Attention. https://app.cochat.ai/papers/10.1162/nol_a_00100
The paper
Xie, Z., Brodbeck, C., & Chandrasekaran, B. (2023). Cortical Tracking of Continuous Speech Under Bimodal Divided Attention. Neurobiology of Language. https://doi.org/10.1162/nol_a_00100

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.