CoChat

The DOI matches its Crossref record; 40 of the 40 references it cites were found in a registry or an index; none of the 37 sources its claims cite is checked for support yet.

Neural Signatures of Hierarchical Linguistic Structures in Second Language Listening Comprehension

Open access
Lu, Lingxi, Deng, Yating, Xiao, Zhe, Jiang, Rong, Gao, Jia-Hong · 2023 · eneuro

Abstract

Native speakers excel at parsing continuous speech into smaller elements and entraining their neural activities to the linguistic hierarchy at different levels (e.g., syllables, phrases, and sentences) to achieve speech comprehension. However, how a nonnative brain tracks hierarchical linguistic structures in second language (L2) speech comprehension and whether it relates to top-down attention and language proficiency remains elusive. Here, we applied a frequency-tagging paradigm in human adults and investigated the neural tracking responses to hierarchically organized linguistic structures (i.e., the syllabic rate of 4 Hz, the phrasal rate of 2 Hz, and the sentential rate of 1 Hz) in both first language (L1) and L2 listeners when they attended to a speech stream or ignored it. We revealed disrupted neural responses to higher-order linguistic structures (i.e., phrases and sentences) for L2 listeners in which the phrasal-level tracking was functionally related to an L2 subject’s language proficiency. We also observed less efficient top-down modulation of attention in L2 speech comprehension than in L1 speech comprehension. Our results indicate that the reduced δ-band neuronal oscillations that subserve the internal construction of higher-order linguistic structures may compromise listening comprehension in a nonnative language.

Overview

CoChat Summary

TL;DR

L2 listeners showed disrupted neural tracking of higher-order linguistic structures (phrases and sentences) compared with L1 listeners, while syllabic tracking remained robust. Phrase-level tracking correlated with L2 proficiency, and L2 listeners showed less attentional gain than L1 listeners.

Key findings

  • L1 listeners tracked all tagged rates in active and passive listening: active 1 Hz t(23)=4.073, corrected p<0.001; 2 Hz t(23)=4.073, corrected p=0.002; 4 Hz t(23)=7.869, corrected p<0.001; passive 1 Hz t(23)=2.667, corrected p=0.027; 2 Hz t(23)=4.483, corrected p<0.001; 4 Hz t(23)=6.706, corrected p<0.001.
  • L2 listeners showed robust syllabic tracking at 4 Hz in active (t(23)=6.464, corrected p<0.001) and passive (t(23)=6.361, corrected p<0.001) listening, and a significant phrase-level 2 Hz response only in active listening (t(23)=2.926, corrected p=0.015); sentence-level 1 Hz tracking was disrupted.
  • L2 neural tracking was weaker than L1 in active listening (F(1,46)=13.743, p=0.001) but not passive listening (p=0.477); attentional gain was lower in L2 than L1 (F(1,46)=7.839, p=0.007).
  • Phrase-level 2 Hz ITPCz correlated with L2 proficiency in active listening (r=0.452, p=0.026) and marginally in passive listening (r=0.403, p=0.051), but syllable-level 4 Hz did not (active r=0.335, p=0.109; passive r=−0.206, p=0.335).

Methods

  • EEG frequency-tagging study with 24 native Mandarin L1 listeners and 24 Chinese-as-second-language L2 learners; L2 learners had varied native languages and 1–6 years of Mandarin study (mean 3.4 ± 1.6 years), with L2 proficiency indexed by a fixed-ratio cloze test (mean 20.2 ± 5.9 on 0–30).
  • Stimuli were 60 sentences (NP + VP; each phrase two 250-ms syllables) assembled into 48 isochronous 10-s sequences, tagging sentence, phrase, and syllable rates at 1, 2, and 4 Hz; each participant completed active-listening and passive-listening blocks (48 trials each).
  • EEG was recorded with 62 channels; intertrial phase coherence (ITPC) was z-normalized and analyzed at tagged frequencies with paired t tests against neighboring bins and ANOVAs on language, attention, and linguistic structure.

Limitations

  • The paper states that L1 background of L2 subjects was not strictly controlled, and subgroup analyses by L1 (English n=7, French n=5, Samoan n=4, Nepali n=4) had small sample sizes.
  • The paper states that the concurrent frequency-tagging paradigm risks contamination from harmonics, and EEG’s limited spatial resolution prevented localizing generators of delta tracking.
  • The paper states that watching a silent movie may not have been sufficiently challenging and attentional state was not behaviorally monitored; evident from the text, the absence of an attentional effect at 4 Hz in L2 and the marginal/absent sentence-level effects should be interpreted cautiously.

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

Integrity signals

Registry

Crossref · 10.1523/eneuro.0346-22.2023

Cites

40 of 40 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1. Schematic illustration of the experimental materials and procedures. A, Speech stimuli were sentence sequences. Each sentence consisted of a noun phrase (NP) and a verb phrase (VP), and each phrase contained two 250-ms syllables (Syl). Using EEG, we tested the neural activities at the tagged frequencies of 1, 2, and 4 Hz corresponding to the rhythm of sentences, phrases, and syllables, respectively. B, The spectrum for sound intensity showed a significant peak at 4 Hz (p < 0.001), corresponding to the syllable-level fluctuations of the speech. The shaded area represents the standard error (SE) across the 48 speech sequences. ***p < 0.001. C, Experimental protocol. In the active-listening block, participants listened carefully to the speech stimuli and responded to a comprehension question. In the passive-listening block, they concentrated on a silent movie while ignoring the sound. Details of language background of L2 participants are provided in Extended Data Figure 1-1.
Figure 1. Schematic illustration of the experimental materials and procedures. A, Speech stimuli were sentence sequences. Each sentence consisted of a noun phrase (NP) and a verb phrase (VP), and each phrase contained two 250-ms syllables (Syl). Using EEG, we tested the neural activities at the tagged frequencies of 1, 2, and 4 Hz corresponding to the rhythm of sentences, phrases, and syllables, respectively. B, The spectrum for sound intensity showed a significant peak at 4 Hz (p < 0.001), corresponding to the syllable-level fluctuations of the speech. The shaded area represents the standard error (SE) across the 48 speech sequences. ***p < 0.001. C, Experimental protocol. In the active-listening block, participants listened carefully to the speech stimuli and responded to a comprehension question. In the passive-listening block, they concentrated on a silent movie while ignoring the sound. Details of language background of L2 participants are provided in Extended Data Figure 1-1.CC BY · PubMed Central
Figure 2. Neural tracking of linguistic structures at syllabic, phrasal, and sentential rates. A, Significant peaks were observed at 1, 2, and 4 Hz in the L1 group, reflecting neural tracking of sentential, phrasal, and syllabic structures in both active-listening and passive-listening conditions. B, In the L2 group, there was robust tracking of the lower-level syllabic-rate fluctuations at 4 Hz and a significant response at the phrasal rate of 2 Hz in the active-listening condition. Shaded areas represent SE; *p < 0.05, **p < 0.01, ***p < 0.001. C, Topographic plots of EEG peak response at tagged frequencies showed whole-brain speech cortical tracking with a general central-frontal distribution. The EEG peak response relative to the average of their four neighboring bins (two at each side) is displayed. There was no difference in hemispheric lateralization between the left-handed and right-handed L2 learners (Extended Data Fig. 2-1).
Figure 2. Neural tracking of linguistic structures at syllabic, phrasal, and sentential rates. A, Significant peaks were observed at 1, 2, and 4 Hz in the L1 group, reflecting neural tracking of sentential, phrasal, and syllabic structures in both active-listening and passive-listening conditions. B, In the L2 group, there was robust tracking of the lower-level syllabic-rate fluctuations at 4 Hz and a significant response at the phrasal rate of 2 Hz in the active-listening condition. Shaded areas represent SE; *p < 0.05, **p < 0.01, ***p < 0.001. C, Topographic plots of EEG peak response at tagged frequencies showed whole-brain speech cortical tracking with a general central-frontal distribution. The EEG peak response relative to the average of their four neighboring bins (two at each side) is displayed. There was no difference in hemispheric lateralization between the left-handed and right-handed L2 learners (Extended Data Fig. 2-1).CC BY · PubMed Central
Figure 3. Comparisons of the frequency-tagged neural tracking response between groups. A, The L1 listener’s neural tracking response was stronger in active listening than in passive listening at all tagged frequencies, while the L2 learner’s tracking response was enhanced by attention at a phrasal rate of 2 Hz. B, Neural tracking responses in L2 were weaker than those in L1, especially when attentional resources were allocated to the speech stream. Error bars represent SE; *p < 0.05, **p < 0.01.
Figure 3. Comparisons of the frequency-tagged neural tracking response between groups. A, The L1 listener’s neural tracking response was stronger in active listening than in passive listening at all tagged frequencies, while the L2 learner’s tracking response was enhanced by attention at a phrasal rate of 2 Hz. B, Neural tracking responses in L2 were weaker than those in L1, especially when attentional resources were allocated to the speech stream. Error bars represent SE; *p < 0.05, **p < 0.01.CC BY · PubMed Central
Figure 4. Attentional modulation of speech cortical tracking in L1 and L2. Reduced neural tracking of linguistic structures in L2 relative to L1 was observed in the active-listening condition but not in the passive-listening condition. After normalizing the attentional effect at tagged frequencies, we found less attentional gain in L2 than L1. Data are displayed as box-whisker plots (box, 25/75% percentiles; whisker, 10/90% percentiles; line, median; square dot, mean). Circles indicate values outside the 10th–90th percentile range; * p < 0.05.
Figure 4. Attentional modulation of speech cortical tracking in L1 and L2. Reduced neural tracking of linguistic structures in L2 relative to L1 was observed in the active-listening condition but not in the passive-listening condition. After normalizing the attentional effect at tagged frequencies, we found less attentional gain in L2 than L1. Data are displayed as box-whisker plots (box, 25/75% percentiles; whisker, 10/90% percentiles; line, median; square dot, mean). Circles indicate values outside the 10th–90th percentile range; * p < 0.05.CC BY · PubMed Central
Figure 5. Correlation between language proficiency and cortical tracking of L2 speech. Pearson correlation analysis revealed that the oscillatory neural tracking at the phrase-level frequency (2 Hz) but not at the syllable-level frequency (4 Hz) was correlated with L2 language proficiency. *p < 0.05, # marginally significant; n.s., not significant.
Figure 5. Correlation between language proficiency and cortical tracking of L2 speech. Pearson correlation analysis revealed that the oscillatory neural tracking at the phrase-level frequency (2 Hz) but not at the syllable-level frequency (4 Hz) was correlated with L2 language proficiency. *p < 0.05, # marginally significant; n.s., not significant.CC BY · PubMed Central
Figure 6. Tracking responses in L2 subgroups whose L1 was English (n = 7), French (n = 5), Samoan (n = 4), or Nepali (n = 4). For all subgroups, a paired-sample t test showed no significant attentional effect at either 2 or 4 Hz (all p > 0.05). The shaded areas and error bars represent the SE. A connecting line between two dots indicates data from the same individual. act: active; pas: passive.
Figure 6. Tracking responses in L2 subgroups whose L1 was English (n = 7), French (n = 5), Samoan (n = 4), or Nepali (n = 4). For all subgroups, a paired-sample t test showed no significant attentional effect at either 2 or 4 Hz (all p > 0.05). The shaded areas and error bars represent the SE. A connecting line between two dots indicates data from the same individual. act: active; pas: passive.CC BY · PubMed Central

Verification

Claims the paper makes

0 Valid · 0 Not valid · 25 Not checked · 77 Own results and interpretations

60 more claims

Create a free account to see all 102 claims and to verify the 27 references still open.

See all claims — free account
  • It may state two separate claims as one. 24
  • Probably not a substantial claim. 20
  • The quote only partly states this claim. 2

Create a free account to see these claims.

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.

  • Consistentt(46) = 7.844reported p < .001 · recomputed p < .001

    “t(46) = 7.844, p < .001”

  • Consistentt(24) = 3.498reported p = .002 · recomputed p = .002

    “t(24) = 3.498, p = .002”

  • Consistentt(24) = 5.874reported p < .001 · recomputed p < .001

    “t(24) = 5.874, p < .001”

  • ConsistentF(1, 46) = 5.984reported p = .018 · recomputed p = .018

    “F(1,46) = 5.984, p = .018”

  • ConsistentF(1, 46) = 9.203reported p = .004 · recomputed p = .004

    “F(1,46) = 9.203, p = .004”

5 of 23 · the rest aren’t loaded

18 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 40 references

Research works printed but not linked: 1

  1. 2020, An analysis of proficiency test for CSL (Chinese as second language) based on fixed-ratio cloze questions., Appl Linguist, 3, 69

Cited by

See all 19

Metadata

Authors
Lu, L., Deng, Y., Xiao, Z., Jiang, R., & Gao, J.-H.
Affiliations
  • Beijing Language and Culture University (CN)
  • Peking University (CN)
Corresponding author
Lu, Lingxi
Venue
eneuro
Published
2023-06
Volume
10
Issue
6
Pages
ENEURO.0346-22.2023
ISSN
2373-2822 (ISSN-L)
OpenAlex
W4381046901
Cited by
19
References
40
Citations by year
2026: 9 · 2025: 3 · 2024: 5 · 2023: 2
FWCI
2.10
Type
journal article
Language
en
Open access
Locations
  • eNeuro (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
Funders
Beijing Social Science Foundation (21YYC010)
Topics
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Neurobiology of Language and Bilingualism
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Neuroscience and Music Perception
  • Social Sciences › Psychology › Experimental and Cognitive Psychology › Phonetics and Phonology Research
Keywords
second language listening comprehension, frequency tagging, neural entrainment, top-down attention, language proficiency, non-native speech perception
MeSH terms
Adult, Auditory Perception, Humans, Language, Linguistics, Speech Perception, Comprehension

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: Neural Signatures of Hierarchical Linguistic Structures in Second Language Listening Comprehension. https://app.cochat.ai/papers/10.1523/eneuro.0346-22.2023
The paper
Lu, L., Deng, Y., Xiao, Z., Jiang, R., & Gao, J.-H. (2023). Neural Signatures of Hierarchical Linguistic Structures in Second Language Listening Comprehension. eneuro. https://doi.org/10.1523/eneuro.0346-22.2023

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.