Pith. sign in

REVIEW 3 major objections 4 minor 80 references

Interpretable MEG Decoding of Perceived Speech: Cortical Sources and the Stimulus Features That Drive Retrieval

T0 review · 3 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A compact decoder whose front end is constrained by the physics of MEG matches black-box speech-retrieval accuracy while its weights map to cortical sources and paired occlusion reveals the stimulus features driving retrieval.

desk verdict A careful, well-controlled interpretable MEG decoder that matches retrieval performance, but the headline feature-use attribution is an upper bound that needs reworded or donor matching tightened. read the letter →

arxiv 2608.01481 v1 pith:OOYCBZG5 submitted 2026-08-02 cs.LG cs.SDq-bio.NC

Ilia Semenkov , Daria Kleeva , Ivan Dakhtin , Zarina Maksudova , Alex Ossadtchi This is my paper · ORCID
classification cs.LGcs.SDq-bio.NC
keywords MEGdecodingspeechperceptioninterpretabledeeplearningsourcelocalizationpairedocclusionanalysissphericalharmonicswav2vec2.0stimulusfeatureattribution
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper sets out to show that a speech-decoding network need not be a black box: if its front end is constrained by the physics of the MEG measurement and by the physiology of the sources, the trained weights can be read back as cortical topographies and dynamics while matching black-box retrieval accuracy. Its architecture replaces a planar spatial-attention layer with spherical harmonics on the three-dimensional helmet geometry, cuts the subject-specific representation to $K=25$ branches, adds a 150 ms temporal filter per branch, and removes ocular and cardiac components, so each branch acts as a matched spatial-temporal filter for a neuronal population. On MEG-MASC the model retrieves the correct 3 s audio segment among 1005 candidates at $39.75\pm0.34\%$ Top-1 accuracy with about 20-fold fewer decoder parameters, and its weights localize to the canonical speech-perception network, with left-lateralized branches carrying rhythmic components not seen on the right. Paired MEG occlusion—replacing feature-marked segments with matched donors from feature-present versus feature-absent intervals—shows that 15 of 19 stimulus features contribute to retrieval, led by silence, loudness, vowels, and acoustic onsets. The upshot is a demonstration that a compact, physically interpretable decoder closes the gap between an accuracy figure and a neuroscientific claim: what drives retrieval, and where in the brain it is computed, are both recoverable.

What carries the argument

The load-bearing mechanism is the factorized spatial–temporal front end. Each of the $K=25$ branches computes a spatially filtered signal via the product of spherical-harmonic attention, a shared unmixing layer, and a subject-specific projection, and then applies a trainable 150 ms depthwise temporal filter (15 samples at 100 Hz). Because the spatial and temporal filters adapt together, the paper uses the covariance-based recipe of Eqs. (5)–(7) to recover each branch's sensor-space pattern and temporal pattern, which are then localized to cortex by a minimum-norm inverse and a recursive subspace-correlation scan (RAP-MUSIC). Paired MEG occlusion is the second mechanism: replacing feature-mar

What would settle it

What would settle it: run the same paired-occlusion battery on a stimulus set in which the 19 features are deliberately decorrelated (e.g., resynthesized speech varying silence, intensity, onset strength, and vowel content independently), and check whether the fifteen positive rank contrasts persist. If, for instance, the silence contrast vanishes when donors are energy-matched, or the vowel contrast disappears when surrounding phonemes are held fixed, the feature-use claim collapses; a cheaper check is to regress the 19 rank contrasts on mask duration and show that they survive that covariate

Watch

Extended reading notes

Core claim

The central claim is that retrieval of perceived speech from non-invasive MEG can be performed accurately by a decoder that is interpretable at the level of classical electrophysiology, and that the same architecture doubles as a knowledge-discovery instrument. The authors demonstrate that spherical-harmonic attention over the helmet geometry, subject-specific projection into 25 branches, and branch-wise temporal filtering produce a decoder that reaches $39.75\pm0.34\%$ Top-1 accuracy among 1005 candidates across six trained solutions with roughly 20 times fewer parameters than the reference brain decoder. More importantly, the trained branch weights, mapped with covariance-based spatial/tem

Load-bearing premise

The paired-occlusion test assumes that a feature-absent donor interval matches the replaced interval in every way except the tested feature; because feature annotations overlap and covary (silence with low energy, stops with onsets, vowels with voicing), a positive rank contrast may reflect a correlated acoustic or contextual state rather than the feature itself, so the fifteen 'contributing' effects are an upper bound on feature-specific encoding.

Editorial extensions

If this is right

  • Retrieval accuracy becomes decomposable: the decoder's score can be attributed to identifiable stimulus features (silence, intensity, acoustic onsets, phonetic classes, surprisal) and to identifiable cortical regions, so a single accuracy number no longer has to be taken on faith.
  • The architecture offers a template for applied decoders (speech neuroprostheses, surgical language mapping) that need both high accuracy and a spatial/temporal readout of what the network is using.
  • The feature-space compression result implies that a dozen learned dimensions of the wav2vec target suffice for MEG alignment, while temporal trajectory must be preserved—so efficient MEG-to-audio interfaces can shrink the target side drastically without losing retrieval.
  • The opposing word-list result indicates that narrative structure, not merely word-level acoustics or semantics, supports decodable neural activity, and implies that retrieval systems may drop sharply in accuracy on non-narrative or scrambled material.
  • The ablation grid shows that subject-conditioned spatial mappings, 3D attention, and temporal filtering each contribute and that a compact branch space around $K=10$–25 lies on the performance plateau, bounding the task-relevant cortical subspace.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The fifteen positive occlusion effects should be read as an upper bound on feature-specific encoding: because feature annotations covary (silence with low energy, stops with onsets, vowels with voicing), a decorrelated stimulus corpus would be needed to test whether each feature has an independent causal contribution.
  • If the left-lateralized faster rhythm replicates with longer temporal filters (which give finer frequency resolution), it would corroborate the asymmetric-sampling-in-time account of speech lateralization; the current 15-tap filters limit frequency resolution to about 6.7 Hz.
  • The temporal-compression failure suggests that stimulus timing, not just identity, is the bottleneck for non-invasive retrieval; one testable extension is to compare retrieval of the same words at different speaking rates to see whether the timing requirement is absolute or relative.
  • A practical extension of the occlusion protocol would be to apply it to imagined or internally generated speech, where stimulus features cannot be annotated from the audio stream, to see which of the 19 features remain retrievable when the acoustic reference is absent.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper presents an interpretable MEG-to-audio retrieval architecture for naturalistic speech perception. Building on the prior framework of Petrosyan et al., the authors replace 2D Fourier spatial attention with spherical harmonics on the MEG helmet, reduce the subject-specific latent to K=25 branches, add depthwise temporal filters, and remove ocular/cardiac ICAs before training. On MEG-MASC they report 39.75±0.34% Top-1 accuracy among 1005 candidates across six seeds, with roughly 20× fewer decoder parameters than the reference model. They map learned branch weights to cortical sources via MNE and RAP-MUSIC, finding sources in auditory, frontal, and temporal regions. They also perform paired MEG occlusion with feature-present vs feature-absent donor substitutions for 19 stimulus features, reporting corrected positive effects for 15 features, including silence, loudness, vowels, and acoustic onsets. Additional experiments examine segment duration, feature-space and temporal compression, architectural ablations, and temporal-filter support.

Significance. If the results are robust, the paper is a valuable contribution: it demonstrates that a physically and physiologically constrained decoder can match black-box retrieval accuracy while yielding interpretable spatial and temporal components. Strengths include the public code and ICA components, careful cross-seed replication of the occlusion results (Appendix B), controlled internal ablations, and the clear separation between the f→0 vs f→f arms to control generic replacement damage. The finding that the wav2vec target can be compressed to ~12 learned feature dimensions, while temporal compression is harmful, is interesting and well supported. However, the central knowledge-discovery claim hinges on the feature-specific attribution in the occlusion analysis, and that attribution is currently confounded. The source-localization claims also rest on a common-template approximation that needs either validation or softer wording.

major comments (3)
  1. [§3.6, Eq. (18)–(20); §4.5; §6] The paired occlusion contrast is not matched on the other 18 stimulus features. Donor intervals are selected only for duration, participant/session, and audio file; no conditioning or covariate adjustment is described. Because annotated features covary (silence with low energy, vowels with voicing/formant structure, onsets with stops), a positive rank contrast r_f→0 − r_f→f may be driven by a correlated acoustic or contextual state rather than the tested feature. The f→f arm controls for generic replacement damage, not for covariate imbalance; averaging over five donor pairs reduces donor-selection variance only. The paper concedes this in §6, but the abstract and §4.5 state that “15 of 19 stimulus features contribute” without this qualification. Since the knowledge-discovery claim rests on feature-specific attribution, this needs either covariate-matched donors, a sensitivity analysis,
  2. [§4.9, §3.8] The comparisons that motivate the 3D spherical-harmonic attention and the other front-end choices are run with a single seed (seed 42). Figure 12 reports about one percentage point advantage of 3D over 2D attention. The text says this is larger than the 0.34 pp seed-to-seed variability, but that variability is measured for the main model only, not for the 2D-attention variant. A single-seed difference of ~1 pp is not sufficient to establish a reliable advantage. Please provide multi-seed estimates for the 2D-vs-3D comparison and for at least the key ablations, or soften the quantitative claim.
  3. [§3.4, §4.4, §6] Cortical source claims are mapped through a single fsaverage template with a common coregistration. The paper notes that individual surface reconstruction succeeded for only six participants and therefore all maps are common-template estimates. The abstract's claim that weights “map to source space, recovering generators consistent with the speech-perception network” is accordingly not validated against individual anatomy, and no quantitative bound on the spatial error is provided. The six individual-anatomy cases could be used as a validation, or the source-localization claims should be tempered to reflect template-based estimates.
minor comments (4)
  1. [Abstract and §4.5] The wording “15 of 19 stimulus features contribute” should be aligned with the limitation in §6: the evidence supports that the decoder uses MEG information that distinguishes feature-present from feature-absent states, not that each feature is an independent causal contributor. Please add a qualifier in the abstract and results.
  2. [Figure 12] The ablation bars are shown without error bars or seed counts. Since these are single-seed runs, adding a note in the caption or a small multi-seed panel for the main comparisons would help the reader calibrate the differences.
  3. [§4.5] The effect magnitudes are described as not comparable across features because masks differ in duration and eligible-window coverage. This caveat appears only in text and Figure 7; consider stating it more prominently in the abstract or figure caption to avoid over-reading.
  4. [Equation (17)] The weakest-view combination distance is explained in words, but a one-sentence justification of why the minimum is chosen rather than the sum or average would improve reproducibility for readers new to this clustering approach.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: retrieval, occlusion, and compression results are data-driven; source-mapping formulas are a parameter-free prior framework applied as a tool.

full rationale

The paper's central retrieval result is measured against a fixed held-out 1005-candidate bank and compared with published scores from Défossez et al. and Zhang et al.; it is not derived from its own inputs. The source-mapping pipeline (Eqs. 5–7) is imported from Petrosyan et al. [4], a parameter-free derivation with stated linearity and factorized space-time assumptions; although one current author is a co-author of [4], the framework does not assume the MEG-MASC results, so the citation is independent support under Rule 4 rather than load-bearing circularity. The paired occlusion analysis defines feature use operationally via rank contrasts between feature-absent and feature-present real-MEG substitutions; this is a new data-driven contrast, and the Section 6 caveat that annotations covary and the intervention does not isolate independent causal contributions is a validity limitation, not a circular step. The compression, duration, and ablation results are direct empirical measurements. No step was found in which a prediction reduces by construction to a fitted parameter or to a self-citation chain.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

The central quantitative claims rest on a handful of architectural choices (K, L, B, temporal-filter length), on the Petrosyan interpretation framework, and on a common-template forward model. No new physical entities are introduced; the branch decomposition is an interpretability device rather than a new biological mechanism.

free parameters (6)
  • K (number of interpretable branches) = 25
    Chosen by hand from a sweep (Figure 3); plateau at K=10-25. Central to the compact-representation claim.
  • L (spherical-harmonic degree) = 24
    Selected as the basis size for the 3D attention layer; not swept.
  • B (convolutional blocks) = 2
    Chosen from the ablation grid; 2-5 blocks perform similarly.
  • Temporal filter length = 15 samples (150 ms)
    Chosen as default; later sweep shows a plateau beyond 150 ms with the best result at 490 ms.
  • MNE regularization lambda^2 = 1/3
    Ad hoc choice for source localization; affects cortical map spread.
  • MNE depth weighting and orientation constraint = 0.5, 0.5
    Ad hoc choices in the source mapping pipeline.
assumptions (6)
  • standard math Linear superposition generative model of MEG (Eq. 1): measured signals are a linear mixture of task-related dipoles, noise sources, and sensor noise.
    Standard biophysical forward model assumed throughout Section 2.1.
  • domain assumption Branches of the factorized front end learn to match individual neuronal sources (Petrosyan et al. framework, Eq. 4-7).
    The entire source-mapping and temporal-pattern interpretation depends on this premise; the paper flags it as a limitation in Section 6.
  • ad hoc to paper The fsaverage template with shared coregistration approximates each participant's anatomy well enough for source localization.
    Only 6 of 27 participants had usable surfaces; all maps use the template (Section 5).
  • domain assumption wav2vec 2.0 frozen embeddings are a valid target for perceived-speech retrieval.
    Target representation inherited from prior work (Défossez et al.).
  • domain assumption Ocular and cardiac ICA component removal eliminates shortcut pathways without removing the cortical signal of interest.
    Key preprocessing choice argued in Section 5; not empirically validated within the paper.
  • domain assumption Feature-absent donor MEG intervals provide a valid counterfactual for the tested feature.
    Underlies the paired occlusion inference; conceded as imperfect in Section 6.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Interpretable MEG Decoding of Perceived Speech: Cortical Sources and the Stimulus Features That Drive Retrieval." pith.science (2026). https://pith.science/paper/OOYCBZG5

@misc{pith2026260801481,
  author       = {Pith},
  title        = {Pith review of: Interpretable MEG Decoding of Perceived Speech: Cortical Sources and the Stimulus Features That Drive Retrieval},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OOYCBZG5}},
  note         = {Machine review of arXiv:2608.01481}
}
read the original abstract

Short segments of perceived speech can be retrieved from non-invasive magnetoencephalographic (MEG) recordings by deep networks trained with a CLIP-style objective against wav2vec 2.0 audio embeddings. Yet their weights do not map onto electrophysiological quantities, and it remains unclear which speech properties drive retrieval. We build on a high-performing MEG-to-audio retrieval architecture but redesign both its front end and decoder. Its spatial attention operates on a flattened sensor layout; we replace it with spherical harmonics defined on the three-dimensional MEG helmet geometry. We reduce the subject-specific representation from 270 to 25 branches, add a temporal filter to each branch to match it to a neuronal source in space and time, and make the convolutional decoder shallower. Ocular and cardiac components are removed before training to reduce the risk of stimulus-locked shortcuts. On MEG-MASC, the model reaches 39.75 +/- 0.34% Top-1 accuracy among 1005 candidates across six trained solutions, with about 20 times fewer decoder parameters. Its weights map to source space, recovering generators consistent with the speech-perception network, while left-lateralized branches carry higher-frequency rhythmic components not evident on the right. Paired MEG occlusion shows that 15 of 19 stimulus features contribute, with the largest effects for silence, sound intensity, vowels, and acoustic onsets. Random word lists behave oppositely: substituting narrative MEG into them improves retrieval, indicating that activity without narrative structure carries less recoverable information than activity during coherent speech. The wav2vec target can be reduced to about twelve learned feature dimensions without loss of accuracy, whereas strong temporal compression causes a clear loss. Together, source mapping and input interventions reveal what drives retrieval.

Figures

Figures reproduced from arXiv: 2608.01481 by the authors.

Figure 1
Figure 1. Interpretable front-end made as a collection of branches, the [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Our network’s architecture for MEG-to-audio embedding alignment. A 3-second, 208-channel MEG segment [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Retrieval accuracy as a function of the number of interpretable branches [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: 3D spherical-harmonic attention learned by architectures with varying numbers of non-linear convolutional [PITH_FULL_IMAGE:figures/full_fig_p014_4.png]
Figure 5
Figure 5. Figure 5: Leading ten singular vectors of the across-subject spatial filter and spatial pattern matrices aggregated from [PITH_FULL_IMAGE:figures/full_fig_p015_5.png]
Figure 6
Figure 6. Figure 6: The 12 largest clusters for the main K = 25 model, computed using the Cable Spool Fort recordings from the first session. Each column shows the medoid of one cluster. Rows show, from top to bottom, the sensor-space spatial pattern, the temporal pattern computed using t…
Figure 7
Figure 7. Figure 7: Paired MEG occlusion effects for 19 stimulus features. For each participant, the plotted effect is the retrieval [PITH_FULL_IMAGE:figures/full_fig_p018_7.png]
Figure 8
Figure 8. Figure 8: Test retrieval accuracy as a function of paired MEG–audio segment duration for models with [PITH_FULL_IMAGE:figures/full_fig_p020_8.png]
Figure 9
Figure 9. Figure 9: Top-n test retrieval accuracy for the main two-block, K = 25 model across paired MEG–audio segment durations. Curves were calculated using the similarity rank of the correct audio segment among the same 991 candidates. The advantage of longer segments is present throug…
Figure 10
Figure 10. Figure 10: Feature-space compression of the target wav2vec representation. The feature dimension is reduced using [PITH_FULL_IMAGE:figures/full_fig_p021_10.png]
Figure 11
Figure 11. Figure 11: Temporal-resolution compression of the target wav2vec trajectory. Temporal PCA, trainable linear reduction, [PITH_FULL_IMAGE:figures/full_fig_p022_11.png]
Figure 12
Figure 12. Figure 12: Architectural ablations of the network’s front end. Bars show the change in Top-1 and Top-10 retrieval [PITH_FULL_IMAGE:figures/full_fig_p023_12.png]
Figure 13
Figure 13. Figure 13: Effect of temporal-filter support on retrieval accuracy. Points show changes in Top-1 and Top-10 accuracy [PITH_FULL_IMAGE:figures/full_fig_p024_13.png]
Figure 14
Figure 14. Figure 14: Robustness of paired MEG occlusion effects across six model initializations. Small grey points show the [PITH_FULL_IMAGE:figures/full_fig_p034_14.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

80 extracted references · 68 canonical work pages

  1. [1]

    Decoding speech perception from non-invasive brain recordings.Nature Machine Intelligence, 5(10):1097–1107, 2023

    Alexandre Défossez, Charlotte Caucheteux, Jérémy Rapin, Ori Kabeli, and Jean-Rémi King. Decoding speech perception from non-invasive brain recordings.Nature Machine Intelligence, 5(10):1097–1107, 2023

    2023

  2. [2]

    wav2vec 2.0: A framework for self- supervised learning of speech representations

    Alexei Baevski, Yuhao Zhou, Abdelrahman Mohamed, and Michael Auli. wav2vec 2.0: A framework for self- supervised learning of speech representations. InAdvances in Neural Information Processing Systems, volume 33, pages 12449–12460, 2020

    work page 2020

  3. [3]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. Learning transferable visual models from natural language supervision. In Marina Meila and Tong Zhang, editors,Proceedings of the 38th International Conference on Machine ...

    2021

  4. [4]

    Decoding and interpreting cortical signals with a compact convolutional neural network.Journal of Neural Engineering, 18(2):026019, 2021

    Artur Petrosyan, Mikhail Sinkin, Mikhail Lebedev, and Alexei Ossadtchi. Decoding and interpreting cortical signals with a compact convolutional neural network.Journal of Neural Engineering, 18(2):026019, 2021

    work page 2021

  5. [5]

    Artur Petrosyan, Alexey V oskoboinikov, Dmitrii Sukhinin, Anna Makarova, Anastasia Skalnaya, Nastasia Arkhipova, Mikhail Sinkin, and Alexei Ossadtchi. Speech decoding from a small set of spatially segregated minimally invasive intracranial EEG electrodes with a compact and interpretable neural network.Journal of Neural Engineering, 19(6):066016, 2022

    work page 2022

  6. [6]

    Spherical harmonics reveal standing EEG waves and long-range neural synchronization during non-REM sleep.Frontiers in Computational Neuroscience, 10:59, 2016

    Siddharth S Sivakumar, Amalia G Namath, and Roberto F Galán. Spherical harmonics reveal standing EEG waves and long-range neural synchronization during non-REM sleep.Frontiers in Computational Neuroscience, 10:59, 2016

    work page 2016

  7. [7]

    Introducing MEG-MASC: a high-quality magneto-encephalography dataset for evaluating natural speech processing.Scientific Data, 10(1):862, 2023

    Laura Gwilliams, Graham Flick, Alec Marantz, Liina Pylkkänen, David Poeppel, and Jean-Rémi King. Introducing MEG-MASC: a high-quality magneto-encephalography dataset for evaluating natural speech processing.Scientific Data, 10(1):862, 2023

    work page 2023

  8. [8]

    On the interpretation of weight vectors of linear models in multivariate neuroimaging.NeuroImage, 87:96–110, 2014

    Stefan Haufe, Frank Meinecke, Kai Görgen, Sven Dähne, John-Dylan Haynes, Benjamin Blankertz, and Felix Bießmann. On the interpretation of weight vectors of linear models in multivariate neuroimaging.NeuroImage, 87:96–110, 2014

    2014

Show all 80 references
  1. [9]

    Towards decoding individual words from non-invasive brain recordings.Nature Communications, 16:10521, 2025

    Stéphane d’Ascoli, Corentin Bel, Jérémy Rapin, Hubert Banville, Yohann Benchetrit, Christophe Pallier, and Jean-Rémi King. Towards decoding individual words from non-invasive brain recordings.Nature Communications, 16:10521, 2025. doi:10.1038/s41467-025-65499-0

    2025 doi

  2. [10]

    What are we actually decoding? source attribution for non-invasive brain-to-language retrieval.arXiv preprint arXiv:2605.24524, 2026

    Xinyu Zhang, Sichao Liu, Runhao Lu, Alexandra Woolgar, and Lihui Wang. What are we actually decoding? source attribution for non-invasive brain-to-language retrieval.arXiv preprint arXiv:2605.24524, 2026. 28 Interpretable MEG Decoding of Perceived Speech

    2026 arXiv

  3. [11]

    Magnetoencephalography: evidence of magnetic fields produced by alpha-rhythm currents.Science, 161(3843):784–786, 1968

    David Cohen. Magnetoencephalography: evidence of magnetic fields produced by alpha-rhythm currents.Science, 161(3843):784–786, 1968

    1968

  4. [12]

    Magnetoencephalography for brain electrophysiology and imaging.Nature Neuroscience, 20(3): 327–339, 2017

    Sylvain Baillet. Magnetoencephalography for brain electrophysiology and imaging.Nature Neuroscience, 20(3): 327–339, 2017. doi:10.1038/nn.4504

    2017 doi

  5. [13]

    Princeton University Press, Princeton, NJ, 1959

    Wilder Penfield and Lamar Roberts.Speech and Brain Mechanisms. Princeton University Press, Princeton, NJ, 1959

    1959

  6. [14]

    The organization of language and the brain.Science, 170(3961):940–944, 1970

    Norman Geschwind. The organization of language and the brain.Science, 170(3961):940–944, 1970. doi:10.1126/science.170.3961.940

    1970 doi

  7. [15]

    The brain basis of language processing: from structure to function.Physiological Reviews, 91(4):1357–1392, 2011

    Angela D Friederici. The brain basis of language processing: from structure to function.Physiological Reviews, 91(4):1357–1392, 2011. doi:10.1152/physrev.00006.2011

    2011

  8. [16]

    Speech listening specifically modulates the excitability of tongue muscles: a TMS study.European Journal of Neuroscience, 15(2):399–402, 2002

    Luciano Fadiga, Laila Craighero, Giovanni Buccino, and Giacomo Rizzolatti. Speech listening specifically modulates the excitability of tongue muscles: a TMS study.European Journal of Neuroscience, 15(2):399–402, 2002

    2002

  9. [17]

    Listening to speech activates motor areas involved in speech production.Nature Neuroscience, 7(7):701–702, 2004

    Stephen M Wilson, Ay¸ se Pinar Saygin, Martin I Sereno, and Marco Iacoboni. Listening to speech activates motor areas involved in speech production.Nature Neuroscience, 7(7):701–702, 2004. doi:10.1038/nn1263

    2004 doi

  10. [18]

    Neural circuits underlying language control and modality control in bilinguals: An fMRI study.Neuropsychologia, 178:108430, 2023

    Huanhuan Liu, Zibin Guo, Yishan Jiang, John W Schwieter, and Fenqi Wang. Neural circuits underlying language control and modality control in bilinguals: An fMRI study.Neuropsychologia, 178:108430, 2023

    2023

  11. [19]

    Neural substrates of vocalization feedback monitoring in primate auditory cortex.Nature, 453(7198):1102–1106, 2008

    Steven J Eliades and Xiaoqin Wang. Neural substrates of vocalization feedback monitoring in primate auditory cortex.Nature, 453(7198):1102–1106, 2008

    2008

  12. [20]

    An accurate and rapidly calibrating speech neuroprosthesis.New England Journal of Medicine, 391(7):609–618, 2024

    Nicholas S Card, Maitreyee Wairagkar, Carrina Iacobacci, Xianda Hou, Tyler Singer-Clark, Francis R Willett, Erin M Kunz, Chaofei Fan, Maryam Vahdati Nia, Darrel R Deo, et al. An accurate and rapidly calibrating speech neuroprosthesis.New England Journal of Medicine, 391(7):609...

    2024

  13. [21]

    Meenakshi Bisla and R.S. Anand. A comprehensive review on state-of-the-art imagined speech decoding techniques using electroencephalography. In M.A. Ansari, R.S. Anand, Pragati Tripathi, Rajat Mehrotra, and Md Belal Bin Heyat, editors,Artificial Intelligence in Biomedical and ...

    2025 doi

  14. [22]

    Swift, Adriana de Pesters, Christian S

    Amin Nourmohammadi, James R. Swift, Adriana de Pesters, Christian S. Guay, Matthew A. Adamo, John C. Dalfino, Anthony L. Ritaccio, Gerwin Schalk, and Peter Brunner. Passive functional mapping of receptive language cortex during general anesthesia using electrocorticography.Cli...

    2023 doi

  15. [23]

    Taplin, Adriana de Pesters, Peter Brunner, Dora Hermes, John C

    AmiLyn M. Taplin, Adriana de Pesters, Peter Brunner, Dora Hermes, John C. Dalfino, Matthew A. Adamo, Anthony L. Ritaccio, and Gerwin Schalk. Intraoperative mapping of expressive language cortex using passive real- time electrocorticography.Epilepsy & Behavior Case Reports, 5:4...

    2016 doi

  16. [24]

    Towards stimulation-free automatic electrocorticographic speech mapping in neurosurgery patients.Journal of Neural Engineering, 22(5):056002, sep 2025

    Alexei V oskoboynikov, Magomed Aliverdiev, Yulia Nekrasova, Ilia Semenkov, Anastasia Skalnaya, Mikhail Sinkin, and Alexei Ossadtchi. Towards stimulation-free automatic electrocorticographic speech mapping in neurosurgery patients.Journal of Neural Engineering, 22(5):056002, se...

    2025 doi

  17. [25]

    Mechanistic interpretability of EEG foundation models via sparse autoencoders

    William Lehn-Schiøler, Magnus Ruud Kjær, Rahul Thapa, Magnus Guldberg Pedersen, Anton Mosquera Storgaard, Nick Williams, Radu Gatej, Tue Lehn-Schiøler, Andreas Brink-Kjær, Sadasivan Puthusserypady, Sándor Beniczky, James Zou, and Lars Kai Hansen. Mechanistic interpretability o...

    2026 arXiv

  18. [26]

    Interpretable functional specialization emerges in deep convolutional networks trained on brain signals.Journal of Neural Engineering, 19(3):036006, may 2022

    J Hammer, R T Schirrmeister, K Hartmann, P Marusic, A Schulze-Bonhage, and T Ball. Interpretable functional specialization emerges in deep convolutional networks trained on brain signals.Journal of Neural Engineering, 19(3):036006, may 2022. doi:10.1088/1741-2552/ac6770

    2022 doi

  19. [27]

    Mechanistic interpretability of brain-to-speech models across speech modes.arXiv preprint arXiv:2602.01247, 2026

    Maryam Maghsoudi and Ayushi Mishra. Mechanistic interpretability of brain-to-speech models across speech modes.arXiv preprint arXiv:2602.01247, 2026

    2026

  20. [28]

    EEGNet: a compact convolutional neural network for EEG-based brain–computer interfaces.Journal of Neural Engineering, 15(5):056013, 2018

    Vernon J Lawhern, Amelia J Solon, Nicholas R Waytowich, Stephen M Gordon, Chou P Hung, and Brent J Lance. EEGNet: a compact convolutional neural network for EEG-based brain–computer interfaces.Journal of Neural Engineering, 15(5):056013, 2018

    2018

  21. [29]

    Deep learning with convolutional neural networks for EEG decoding and visualization.Human Brain Mapping, 38(11):5391– 5420, 2017

    Robin Tibor Schirrmeister, Jost Tobias Springenberg, Lukas Dominique Josef Fiederer, Martin Glasstetter, Katharina Eggensperger, Michael Tangermann, Frank Hutter, Wolfram Burgard, and Tonio Ball. Deep learning with convolutional neural networks for EEG decoding and visualizati...

    2017

  22. [30]

    Adaptive neural network classifier for de- coding MEG signals.NeuroImage, 197:425–434, 2019

    Ivan Zubarev, Rasmus Zetter, Hanna-Leena Halme, and Lauri Parkkonen. Adaptive neural network classifier for de- coding MEG signals.NeuroImage, 197:425–434, 2019. ISSN 1053-8119. doi:10.1016/j.neuroimage.2019.04.068

    2019 doi

  23. [31]

    Eye activity tracks task-relevant structures during speech and auditory sequence perception.Nature Communications, 9(1):5374, 2018

    Peiqing Jin, Jiajie Zou, Tao Zhou, and Nai Ding. Eye activity tracks task-relevant structures during speech and auditory sequence perception.Nature Communications, 9(1):5374, 2018

    2018

  24. [32]

    Eye movements track prioritized auditory features in selective attention to natural speech.Nature Communications, 15(1):3692, 2024

    Quirin Gehmacher, Juliane Schubert, Fabian Schmidt, Thomas Hartmann, Patrick Reisinger, Sebastian Rösch, Konrad Schwarz, Tzvetan Popov, Maria Chait, and Nathan Weisz. Eye movements track prioritized auditory features in selective attention to natural speech.Nature Communicatio...

    2024

  25. [33]

    Eye movements during auditory attention predict individual differences in dorsal attention network activity.Frontiers in Human Neuroscience, 10:164, 2016

    Rodrigo M Braga, Richard Z Fu, Barry M Seemungal, Richard JS Wise, and Robert Leech. Eye movements during auditory attention predict individual differences in dorsal attention network activity.Frontiers in Human Neuroscience, 10:164, 2016

    2016

  26. [34]

    Amygdala and heart rate variability responses from listening to emotionally intense parts of a story

    Mikkel Wallentin, Andreas Højlund Nielsen, Peter Vuust, Anders Dohn, Andreas Roepstorff, and Torben Ellegaard Lund. Amygdala and heart rate variability responses from listening to emotionally intense parts of a story. NeuroImage, 58(3):963–973, 2011. ISSN 1053-8119. doi:10.101...

    2011 doi

  27. [35]

    Parra, and Jacobo D

    Pauline Pérez, Jens Madsen, Leah Banellis, Ba¸ sak Türker, Federico Raimondo, Vincent Perlbarg, Melanie Valente, Marie-Cécile Niérat, Louis Puybasset, Lionel Naccache, Thomas Similowski, Damian Cruse, Lucas C. Parra, and Jacobo D. Sitt. Conscious processing of narrative stimul...

    2021

  28. [36]

    Contributions of principal neocortical neurons to magnetoen- cephalography and electroencephalography signals.The Journal of Physiology, 575(3):925–936, 2006

    Shingo Murakami and Yoshio Okada. Contributions of principal neocortical neurons to magnetoen- cephalography and electroencephalography signals.The Journal of Physiology, 575(3):925–936, 2006. doi:10.1113/jphysiol.2006.105379

    2006 arXiv

  29. [37]

    FreeSurfer.NeuroImage, 62(2):774–781, 2012

    Bruce Fischl. FreeSurfer.NeuroImage, 62(2):774–781, 2012

    2012

  30. [38]

    Prentice Hall, 1993

    Steven M Kay.Fundamentals of statistical signal processing: estimation theory. Prentice Hall, 1993

    1993

  31. [39]

    A novel method for reliable and fast extraction of neuronal EEG/MEG oscillations on the basis of spatio-spectral decomposition.NeuroImage, 55(4):1528–1535, 2011

    Vadim V Nikulin, Guido Nolte, and Gabriel Curio. A novel method for reliable and fast extraction of neuronal EEG/MEG oscillations on the basis of spatio-spectral decomposition.NeuroImage, 55(4):1528–1535, 2011

    2011

  32. [40]

    SPoC: a novel framework for relating the amplitude of neuronal oscillations to behaviorally relevant parameters.NeuroImage, 86:111–122, 2014

    Sven Dähne, Frank C Meinecke, Stefan Haufe, Johannes Höhne, Michael Tangermann, Klaus-Robert Müller, and Vadim V Nikulin. SPoC: a novel framework for relating the amplitude of neuronal oscillations to behaviorally relevant parameters.NeuroImage, 86:111–122, 2014

    2014

  33. [41]

    Representational dissimilarity component analysis (ReDisCA).NeuroImage, 301:120868, 2024

    Alexei Ossadtchi, Ilia Semenkov, Anna Zhuravleva, Vladimir Kozunov, Oleg Serikov, and Ekaterina V oloshina. Representational dissimilarity component analysis (ReDisCA).NeuroImage, 301:120868, 2024

    2024

  34. [42]

    Compact convolutional neural networks for classification of asynchronous steady-state visual evoked potentials.Journal of Neural Engineering, 15(6):066031, 2018

    Nicholas Waytowich, Vernon J Lawhern, Javier O Garcia, Jennifer Cummings, Josef Faller, Paul Sajda, and Jean M Vettel. Compact convolutional neural networks for classification of asynchronous steady-state visual evoked potentials.Journal of Neural Engineering, 15(6):066031, 2018

    2018

  35. [43]

    invertmeeg: A benchmark and unified Python library for EEG inverse solvers.bioRxiv, 2026

    Lukas Hecker. invertmeeg: A benchmark and unified Python library for EEG inverse solvers.bioRxiv, 2026. doi:10.64898/2026.03.06.710103

    2026 doi

  36. [44]

    M. S. Hämäläinen and R. J. Ilmoniemi. Interpreting magnetic fields of the brain: minimum norm estimates. Medical & Biological Engineering & Computing, 32:35–42, 1994

    1994

  37. [45]

    Mosher and R.M

    J.C. Mosher and R.M. Leahy. Source localization using recursively applied and projected (RAP) MUSIC.IEEE Transactions on Signal Processing, 47(2):332–340, 1999. doi:10.1109/78.740118

    1999 doi

  38. [46]

    Oxford University Press, 2006

    György Buzsáki.Rhythms of the Brain. Oxford University Press, 2006

    2006

  39. [47]

    MEG and EEG data analysis with MNE-Python.Frontiers in Neuroscience, 7:267, 2013

    Alexandre Gramfort, Martin Luessi, Eric Larson, Denis A Engemann, Daniel Strohmeier, Christian Brodbeck, Roman Goj, Mainak Jas, Teon Brooks, Lauri Parkkonen, and Matti Hämäläinen. MEG and EEG data analysis with MNE-Python.Frontiers in Neuroscience, 7:267, 2013

    2013

  40. [48]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. InInternational Conference on Learning Representations, 2019

    2019

  41. [49]

    Emergence of neural encoding of auditory objects while listening to competing speakers.Proceedings of the National Academy of Sciences, 109(29):11854–11859, 2012

    Nai Ding and Jonathan Z Simon. Emergence of neural encoding of auditory objects while listening to competing speakers.Proceedings of the National Academy of Sciences, 109(29):11854–11859, 2012

    2012

  42. [50]

    Speech rhythms and multiplexed oscillatory sensory coding in the human brain.PLoS Biology, 11(12): e1001752, 2013

    Joachim Gross, Nienke Hoogenboom, Gregor Thut, Philippe Schyns, Stefano Panzeri, Pascal Belin, and Simon Garrod. Speech rhythms and multiplexed oscillatory sensory coding in the human brain.PLoS Biology, 11(12): e1001752, 2013

    2013

  43. [51]

    Temporal envelope of time-compressed speech represented in the human auditory cortex.Journal of Neuroscience, 29(49):15564–15574, 2009

    Kirill V Nourski, Richard A Reale, Hiroyuki Oya, Hiroto Kawasaki, Christopher K Kovach, Haiming Chen, Matthew A Howard, and John F Brugge. Temporal envelope of time-compressed speech represented in the human auditory cortex.Journal of Neuroscience, 29(49):15564–15574, 2009. 30...

    2009

  44. [52]

    The tracking of speech envelope in the human cortex.PLoS One, 8(1):e53398, 2013

    Jan Kubanek, Peter Brunner, Aysegul Gunduz, David Poeppel, and Gerwin Schalk. The tracking of speech envelope in the human cortex.PLoS One, 8(1):e53398, 2013

    2013

  45. [53]

    Continuous speech processing.Current Opinion in Physiology, 18: 25–31, 2020

    Christian Brodbeck and Jonathan Z Simon. Continuous speech processing.Current Opinion in Physiology, 18: 25–31, 2020

    2020

  46. [54]

    Yulia Oganian and Edward F. Chang. A speech envelope landmark for syllable encoding in human superior temporal gyrus.Science Advances, 5(11):eaay6279, 2019. doi:10.1126/sciadv.aay6279

    2019 doi

  47. [55]

    The cortical organization of speech processing.Nature Reviews Neuroscience, 8(5):393–402, 2007

    Gregory Hickok and David Poeppel. The cortical organization of speech processing.Nature Reviews Neuroscience, 8(5):393–402, 2007

    2007

  48. [56]

    Cope, Ediz Sohoglu, Katie A

    Thomas E. Cope, Ediz Sohoglu, Katie A. Peterson, P. Simon Jones, Catarina Rua, Luca Passamonti, William Sedley, Brechtje Post, Jan Coebergh, Christopher R. Butler, Peter Garrard, Khaled Abdel-Aziz, Masud Husain, Timothy D. Griffiths, Karalyn Patterson, Matthew H. Davis, and Ja...

    2023

  49. [57]

    Temporal lobe networks supporting the comprehension of spoken words.Brain, 140(9):2370–2380, 2017

    Leonardo Bonilha, Argye E Hillis, Gregory Hickok, Dirk B den Ouden, Chris Rorden, and Julius Fridriksson. Temporal lobe networks supporting the comprehension of spoken words.Brain, 140(9):2370–2380, 2017

    2017

  50. [58]

    The analysis of speech in different temporal integration windows: cerebral lateralization as ‘asymmetric sampling in time’.Speech Communication, 41(1):245–255, 2003

    David Poeppel. The analysis of speech in different temporal integration windows: cerebral lateralization as ‘asymmetric sampling in time’.Speech Communication, 41(1):245–255, 2003. doi:10.1016/S0167-6393(02)00107- 3

    2003 doi

  51. [59]

    Hierarchical and asymmetric temporal sensitivity in human auditory cortices.Nature Neuroscience, 8(3):389–395, 2005

    Anthony Boemio, Stephen Fromm, Allen Braun, and David Poeppel. Hierarchical and asymmetric temporal sensitivity in human auditory cortices.Nature Neuroscience, 8(3):389–395, 2005. doi:10.1038/nn1409

    2005 doi

  52. [60]

    Cortical oscillations and speech processing: emerging computational principles and operations.Nature Neuroscience, 15(4):511–517, 2012

    Anne-Lise Giraud and David Poeppel. Cortical oscillations and speech processing: emerging computational principles and operations.Nature Neuroscience, 15(4):511–517, 2012

    2012

  53. [61]

    Naturalistic spoken language comprehension is supported by alpha and beta oscillations.Journal of Neuroscience, 43(20):3718–3732, 2023

    Ioanna Zioga, Hugo Weissbart, Ashley G Lewis, Saskia Haegens, and Andrea E Martin. Naturalistic spoken language comprehension is supported by alpha and beta oscillations.Journal of Neuroscience, 43(20):3718–3732, 2023

    2023

  54. [62]

    Heard or understood? neural tracking of language features in a comprehensible story, an incomprehensible story and a word list.eNeuro, 10(7):ENEURO.0075–23.2023,

    Marlies Gillis, Jonas Vanthornhout, and Tom Francart. Heard or understood? neural tracking of language features in a comprehensible story, an incomprehensible story and a word list.eNeuro, 10(7):ENEURO.0075–23.2023,

    2023

  55. [63]

    Reading senseless sentences: Brain potentials reflect semantic incongruity

    Marta Kutas and Steven A Hillyard. Reading senseless sentences: Brain potentials reflect semantic incongruity. Science, 207(4427):203–205, 1980

    1980

  56. [64]

    A cortical network for semantics: (de)constructing the N400

    Ellen F Lau, Colin Phillips, and David Poeppel. A cortical network for semantics: (de)constructing the N400. Nature Reviews Neuroscience, 9(12):920–933, 2008

    2008

  57. [65]

    Wichmann

    Robert Geirhos, Jörn-Henrik Jacobsen, Claudio Michaelis, Richard Zemel, Wieland Brendel, Matthias Bethge, and Felix A. Wichmann. Shortcut learning in deep neural networks.Nature Machine Intelligence, 2(11):665–673,

  58. [66]

    Mapping Whisper representations to human ECoG responses with interpretable time-resolved neural encoding.arXiv preprint arXiv:2606.02305, 2026

    Matteo Ciferri, Tommaso Boccato, Michal Olak, Matteo Ferrante, and Nicola Toschi. Mapping Whisper representations to human ECoG responses with interpretable time-resolved neural encoding.arXiv preprint arXiv:2606.02305, 2026

    2026 arXiv

  59. [67]

    Neural tracking of natural speech in children in relation to their receptive speech abilities.Cognitive Systems Research, 86:101236, 2024

    Anton Rogachev and Olga Sysoeva. Neural tracking of natural speech in children in relation to their receptive speech abilities.Cognitive Systems Research, 86:101236, 2024

    2024

  60. [68]

    Chang, and Srikantan S

    Yulia Oganian, Katsuaki Kojima, Assaf Breska, Chang Cai, Anne Findlay, Edward F. Chang, and Srikantan S. Nagarajan. Phase alignment of low-frequency neural activity to the amplitude envelope of speech reflects evoked responses to acoustic edges, not oscillatory entrainment.Jou...

    2023 doi

  61. [69]

    Vanheusden, Steven L

    Suwijak Deoisres, Yuhan Lu, Frederique J. Vanheusden, Steven L. Bell, and David M. Simpson. Continuous speech with pauses inserted between words increases cortical tracking of speech envelope.PLOS ONE, 18(7): e0289288, 07 2023. doi:10.1371/journal.pone.0289288

    2023 doi

  62. [70]

    A hierarchy of linguistic predictions during natural language comprehension.Proceedings of the National Academy of Sciences, 119(32):e2201968119, 2022

    Micha Heilbron, Kristijan Armeni, Jan-Mathijs Schoffelen, Peter Hagoort, and Floris P de Lange. A hierarchy of linguistic predictions during natural language comprehension.Proceedings of the National Academy of Sciences, 119(32):e2201968119, 2022

    2022

  63. [71]

    Evidence of a predictive coding hierarchy in the human brain listening to speech.Nature Human Behaviour, 7(3):430–441, 2023

    Charlotte Caucheteux, Alexandre Gramfort, and Jean-Rémi King. Evidence of a predictive coding hierarchy in the human brain listening to speech.Nature Human Behaviour, 7(3):430–441, 2023. 31 Interpretable MEG Decoding of Perceived Speech

    2023

  64. [72]

    Shared computational principles for language processing in humans and deep language models.Nature Neuroscience, 25(3):369–380, 2022

    Ariel Goldstein, Zaid Zada, Eliav Buchnik, Mariano Schain, Amy Price, Bobbi Aubrey, Samuel A Nastase, Amir Feder, Dotan Emanuel, Alon Cohen, et al. Shared computational principles for language processing in humans and deep language models.Nature Neuroscience, 25(3):369–380, 2022

    2022

  65. [73]

    The neural architecture of language: Integrative modeling converges on predictive processing.Proceedings of the National Academy of Sciences, 118(45):e2105646118, 2021

    Martin Schrimpf, Idan Asher Blank, Greta Tuckute, Carina Kauf, Eghbal A Hosseini, Nancy Kanwisher, Joshua B Tenenbaum, and Evelina Fedorenko. The neural architecture of language: Integrative modeling converges on predictive processing.Proceedings of the National Academy of Sci...

    2021

  66. [74]

    Toward a realistic model of speech processing in the brain with self-supervised learning

    Juliette Millet, Charlotte Caucheteux, Pierre Orhan, Yves Boubenec, Alexandre Gramfort, Ewan Dunbar, Christophe Pallier, and Jean-Rémi King. Toward a realistic model of speech processing in the brain with self-supervised learning. In S. Koyejo, S. Mohamed, A. Agarwal, D. Belgr...

  67. [75]

    Self-supervised models of audio effectively explain human cortical responses to speech

    Aditya R Vaidya, Shailee Jain, and Alexander Huth. Self-supervised models of audio effectively explain human cortical responses to speech. In Kamalika Chaudhuri, Stefanie Jegelka, Le Song, Csaba Szepesvari, Gang Niu, and Sivan Sabato, editors,Proceedings of the 39th Internatio...

    2022

  68. [76]

    Geodesic optimization for predictive shift adaptation on EEG data

    Apolline Mellot, Antoine Collas, Sylvain Chevallier, Alexandre Gramfort, and Denis A Engemann. Geodesic optimization for predictive shift adaptation on EEG data. InAdvances in Neural Information Processing Systems, volume 37, pages 32828–32855, 2024

    2024

  69. [77]

    Johns Hopkins University Press, 2013

    Gene H Golub and Charles F Van Loan.Matrix Computations. Johns Hopkins University Press, 2013. 32 Interpretable MEG Decoding of Perceived Speech Appendix A Mapping the signal subspace to cortical sources via recursive subspace correlation scan Algorithm 1Mapping the signal sub...

    2013

  70. [2020]

    doi:10.1038/s42256-020-00257-z

    doi

  71. [2022]

    doi:10.52202/068431-2422

    doi

  72. [2023]

    doi:10.1523/ENEURO.0075-23.2023

    2023 doi

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.