arxiv: 2604.14694 · v1 · submitted 2026-04-16 · 🧬 q-bio.NC

Recognition: unknown

Robust Evaluation of Neural Encoding Models via ground-truth approximation

Giovanni M. Di Liberto

Authors on Pith no claims yet

Pith reviewed 2026-05-10 09:09 UTC · model grok-4.3

classification 🧬 q-bio.NC

keywords neural encoding modelsMEEGground-truth approximationcanonical correlation analysisparticipant averagingmodel evaluationbrain signal analysisstimulus-related activity

0 comments

The pith

Encoding models are evaluated more accurately by comparing predictions to a ground-truth approximation of neural activity obtained via canonical correlation analysis and participant averaging.

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

Encoding models aim to describe how brains represent sensory inputs from MEEG measurements. The true underlying neural activity remains unknown, so standard evaluation metrics compare predictions against noisy data dominated by unrelated variance. This work creates an approximation of the relevant neural signal by aligning the measurements to the predictions with canonical correlation analysis and averaging the aligned signals across participants. The resulting CPA-PA metric delivers evaluations for individual participants that exceed conventional scores by large margins on both synthetic and real data. A sympathetic reader would care because this makes it easier to trust what the models say about brain function and to test scientific ideas with more power.

Core claim

The paper establishes that aligning MEEG signals with model predictions using canonical correlation analysis and participant averaging produces a ground-truth approximation that supports the CPA-PA metric. This metric yields single-participant evaluations outperforming conventional scores by 300-1000% on synthetic EEG data and 250% on 34 real MEEG datasets with 818 datapoints. The gains come from higher sensitivity to stimulus-relevant neural activity and reduced dependence on signal-to-noise ratio, which together establish ground-truth approximation as a robust framework for evaluating encoding models.

What carries the argument

The ground-truth approximation of stimulus-related neural activity, created by aligning MEEG signals to model predictions with canonical correlation analysis and then averaging across participants to form the CPA-PA metric.

Load-bearing premise

Aligning MEEG signals with model predictions using canonical correlation analysis and participant averaging produces a valid approximation of the underlying stimulus-related neural activity without introducing alignment-induced biases or circular dependence on the predictions themselves.

What would settle it

Applying the CPA-PA metric to synthetic EEG data with explicitly known ground-truth neural signals and finding that it does not outperform conventional scores by the claimed margins would indicate the approximation does not capture the relevant activity as intended.

read the original abstract

Encoding models enable measurement of how our brains represent sensory inputs using electro-and magneto-encephalography (MEEG). Evaluating how closely encoding models reflect the underlying brain functions is a crucial premise for model interpretation and hypothesis testing. However, the ground-truth neural activity is unknown, preventing model evaluation with respect to the target neural signal. Existing evaluation metrics must therefore relate model's predictions to noisy MEEG measurements, where most variance is stimulus-unrelated. Here, I introduce an evaluation framework where model predictions are compared to a ground-truth approximation, obtained by aligning MEEG signals with predictions using canonical correlation analysis and via participant averaging. The resulting metric (CPA-PA) yields single-participant evaluations outperforming conventional scores by 300-1000% on synthetic EEG data and 250% on 34 real MEEG datasets (818 datapoints). These gains reflect increased sensitivity to stimulus-relevant neural activity and reduced dependence on SNR, establishing ground-truth approximation as a robust framework for evaluating encoding models.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The CPA-PA metric claims big gains in evaluating neural encoding models but builds its ground-truth proxy by aligning MEEG data directly to the model's predictions via CCA, which creates a circularity problem.

read the letter

The main takeaway is that this paper introduces CPA-PA, a framework that approximates ground truth for model evaluation by aligning MEEG signals to predictions with canonical correlation analysis and then averaging across participants. It reports large improvements over standard metrics on both synthetic EEG and 34 real datasets. The idea targets a genuine issue in the field where single-participant MEEG data is dominated by stimulus-unrelated noise, making direct comparisons unreliable for model interpretation. That part of the motivation is solid and practical. The combination of CCA alignment plus averaging as a specific proxy is new enough in this context, and the empirical deltas on real data give something concrete to examine. Participant averaging itself is a straightforward way to boost signal, and the goal of reducing SNR dependence makes sense for cognitive neuroscience work. The soft spot is the circularity built into the construction. CCA maximizes correlation between the signals and the predictions by design, so the resulting approximation is partly shaped by the model under test. Any later score comparing predictions back to that approximation will include a component that rewards how well the model drives the alignment step rather than how faithfully it matches independent neural activity. The abstract gives no sign that alignment uses held-out predictions, cross-validation folds, or an external reference to break this dependence. On synthetic data where true ground truth is available, this could have been checked directly, but without that the reported 300-1000% gains are hard to trust at face value. The paper is aimed at researchers who build and compare encoding models against MEEG recordings and need better evaluation tools. A reader working on stimulus-driven neural responses would find the framework worth trying, especially if they can add their own controls for the alignment bias. It shows clear engagement with the evaluation problem and supplies data, so it deserves a serious referee to clarify the procedure and test whether the gains survive proper separation of the alignment from the scoring. I would send it to peer review rather than desk reject.

Referee Report

3 major / 2 minor

Summary. The manuscript introduces an evaluation framework for neural encoding models that approximates ground-truth stimulus-related neural activity by aligning MEEG signals with model predictions via canonical correlation analysis (CCA) followed by participant averaging, yielding the CPA-PA metric. It claims this produces single-participant evaluations that outperform conventional scores by 300-1000% on synthetic EEG data and 250% on 34 real MEEG datasets (818 datapoints), with gains attributed to increased sensitivity to stimulus-relevant activity and reduced SNR dependence.

Significance. If the CPA-PA approximation can be shown to be non-circular and closer to independent ground truth, the framework would offer a valuable tool for more reliable assessment of encoding models in MEEG research, potentially improving model interpretation and hypothesis testing. The scale of the empirical evaluation (synthetic plus 34 real datasets) is a strength that could support broader adoption if the core validity concerns are resolved.

major comments (3)

[Methods (CPA-PA definition and algorithm)] Methods section on CPA-PA construction: The ground-truth approximation is formed by CCA alignment of MEEG signals directly to the model's predictions (which maximizes their correlation by design) before participant averaging. This makes the reference signal a function of the predictions being scored. The manuscript does not state whether alignment uses held-out predictions, cross-validated folds, or an external fixed reference; without such separation the subsequent correlation metric contains a self-referential component that can inflate scores even for null models.
[Results (synthetic EEG experiments)] Results on synthetic data: The reported 300-1000% gains are presented without an ablation replacing the encoding model with random or null predictions while keeping the CCA alignment step. Such a control is required to demonstrate that performance improvements arise from better capture of stimulus-driven variance rather than from the alignment process itself optimizing the reference to any input.
[Results or Methods (validation subsection)] Validation of approximation quality: No quantitative check is provided (even on synthetic data where true stimulus-related activity is known) showing that the CPA-PA reference correlates more strongly with the actual ground-truth signal than raw or conventionally processed MEEG. This direct comparison is load-bearing for the central claim that CPA-PA constitutes a valid ground-truth approximation.

minor comments (2)

[Abstract] Abstract: The dramatic percentage gains are stated without a mathematical definition or equation for CPA-PA; a concise definition or pointer to the methods equation should be added.
[Throughout] Notation: Ensure consistent definition of acronyms (MEEG, CCA, CPA-PA) at first use and clarify whether participant averaging occurs before or after CCA in the pipeline.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive and detailed comments, which have helped us identify areas where the manuscript can be strengthened. We address each major comment below and outline the revisions we will make.

read point-by-point responses

Referee: Methods section on CPA-PA construction: The ground-truth approximation is formed by CCA alignment of MEEG signals directly to the model's predictions (which maximizes their correlation by design) before participant averaging. This makes the reference signal a function of the predictions being scored. The manuscript does not state whether alignment uses held-out predictions, cross-validated folds, or an external fixed reference; without such separation the subsequent correlation metric contains a self-referential component that can inflate scores even for null models.

Authors: We agree that explicit separation between the data used for CCA alignment and the evaluation data is necessary to avoid any self-referential inflation. The current implementation fits the CCA on training folds and applies the learned transformations to held-out test data when computing the CPA-PA metric for each participant. However, the manuscript does not describe this cross-validation procedure in sufficient detail. We will revise the Methods section to include a clear description of the cross-validation scheme, along with pseudocode illustrating the full pipeline. This will confirm that the alignment step does not introduce circularity into the reported scores. revision: yes
Referee: Results on synthetic data: The reported 300-1000% gains are presented without an ablation replacing the encoding model with random or null predictions while keeping the CCA alignment step. Such a control is required to demonstrate that performance improvements arise from better capture of stimulus-driven variance rather than from the alignment process itself optimizing the reference to any input.

Authors: We concur that an ablation using null or random predictions is required to isolate the contribution of the encoding model from the alignment procedure. We will add this control experiment to the synthetic EEG results section. Specifically, we will recompute the CPA-PA metric after replacing model predictions with random Gaussian noise and with temporally shuffled predictions, while keeping the CCA alignment and participant averaging steps unchanged. The results will be reported alongside the original gains to demonstrate that the improvements are attributable to stimulus-related variance captured by the model. revision: yes
Referee: Validation of approximation quality: No quantitative check is provided (even on synthetic data where true stimulus-related activity is known) showing that the CPA-PA reference correlates more strongly with the actual ground-truth signal than raw or conventionally processed MEEG. This direct comparison is load-bearing for the central claim that CPA-PA constitutes a valid ground-truth approximation.

Authors: We acknowledge that a direct quantitative validation of the approximation quality against known ground truth is important for supporting the central claim. On the synthetic EEG data, where the true stimulus-related neural activity is available by construction, we will add a comparison of the correlation between the CPA-PA reference and the true ground-truth signal versus the correlations obtained from raw MEEG signals and from standard preprocessing pipelines. These results will be included in a new or expanded validation subsection to provide the requested evidence. revision: yes

Circularity Check

1 steps flagged

CPA-PA constructs its 'ground-truth' approximation by CCA-aligning MEEG signals directly to the evaluated model's predictions

specific steps

fitted input called prediction [Abstract]
"a ground-truth approximation, obtained by aligning MEEG signals with predictions using canonical correlation analysis and via participant averaging. The resulting metric (CPA-PA)"

CCA is applied to maximize the correlation between MEEG signals and the model's predictions; the resulting aligned signal is then treated as the independent ground-truth reference against which those same predictions are scored. The performance metric therefore contains a component that rewards the model's ability to drive the alignment rather than its match to stimulus-related activity measured independently of the model.

full rationale

The paper's core evaluation metric is defined by first using CCA to align the noisy MEEG recordings to the model's own predictions (maximizing their correlation by construction) and then averaging across participants to form the reference. The model is subsequently scored by its correlation to this reference. Because the alignment step is performed on the same predictions being evaluated, any reported gains (300-1000% on synthetic data, 250% on real data) necessarily include the contribution of the CCA optimization itself rather than measuring fidelity to an independent neural target. The abstract provides no indication that alignment uses held-out predictions, cross-validation, or an external reference, confirming the reduction is by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 3 axioms · 0 invented entities

Review is based on abstract only; full details of parameters and assumptions are unavailable. The method implicitly relies on standard MEEG analysis assumptions but introduces a new combination for approximation.

axioms (3)

domain assumption MEEG recordings contain a mixture of stimulus-related neural activity and unrelated noise
Stated in the problem setup of the abstract.
domain assumption CCA alignment can isolate stimulus-relevant components when matching predictions to data
Central mechanism for obtaining the ground-truth approximation.
domain assumption Averaging across participants reduces stimulus-unrelated variance
Used to improve the approximation quality.

pith-pipeline@v0.9.0 · 5460 in / 1517 out tokens · 64639 ms · 2026-05-10T09:09:31.499788+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

81 extracted references · 49 canonical work pages

[1]

Holdgraf, C.R., Rieger, J.W., Micheli, C., Martin, S., Knight, R.T., and Theunissen, F.E. (2017). Encoding and Decoding Models in Cognitive Electrophysiology. Front Syst Neurosci 11, 61. 10.3389/fnsys.2017.00061

work page doi:10.3389/fnsys.2017.00061 2017
[2]

Crosse, M.J., Zuk, N.J., Di Liberto, G.M., Nidiffer, A.R., Molholm, S., and Lalor, E.C. (2021). Linear Modeling of Neurophysiological Responses to Speech and Other Continuous Stimuli: Methodological Considerations for Applied Research. Frontiers in neuros cience 15, 705621 -705621. 10.3389/fnins.2021.705621

work page doi:10.3389/fnins.2021.705621 2021
[3]

-R., and Gramfort, A

King, J. -R., and Gramfort, A. (2018). Encoding and decoding neuronal dynamics: Methodological framework to uncover the algorithms of cognition

2018
[4]

Broderick, M.P., Anderson, A.J., Di Liberto, G.M., Crosse, M.J., and Lalor, E.C. (2018). Electrophysiological Correlates of Semantic Dissimilarity Reflect the Comprehension of Natural, Narrative Speech. Current Biology. 10.1016/j.cub.2018.01.080

work page doi:10.1016/j.cub.2018.01.080 2018
[5]

Brodbeck, C., Hong, L.E., and Simon, J.Z. (2018). Rapid Transformation from Auditory to Linguistic Representations of Continuous Speech. Current Biology 28, 3976-3983.e3975

2018
[6]

Leonard, M.K., Gwilliams, L., Sellers, K.K., Chung, J.E., Xu, D., Mischler, G., Mesgarani, N., Welkenhuysen, M., Dutta, B., and Chang, E.F. (2024). Large -scale single -neuron speech sound encoding across the depth of human cortex. Nature 626, 593-602. 10.1038/s41586-023-06839-2

work page doi:10.1038/s41586-023-06839-2 2024
[7]

de Cheveigné, A., and Simon, J.Z. (2008). Denoising based on spatial filtering. Journal of Neuroscience Methods 171, 331-339. 10.1016/j.jneumeth.2008.03.015

work page doi:10.1016/j.jneumeth.2008.03.015 2008
[8]

Crosse, M.J., Di Liberto, G.M., Bednar, A., and Lalor, E.C. (2016). The multivariate temporal response function (mTRF) toolbox: A MATLAB toolbox for relating neural signals to continuous stimuli. Frontiers in Human Neuroscience 10. 10.3389/fnhum.2016.00604

work page doi:10.3389/fnhum.2016.00604 2016
[9]

Lalor, E.C., Power, A.J., Reilly, R.B., and Foxe, J.J. (2009). Resolving Precise Temporal Processing Properties of the Auditory System Using Continuous Stimuli. Journal of Neurophysiology 102, 349-

2009
[10]

10.1152/jn.90896.2008

work page doi:10.1152/jn.90896.2008 2008
[11]

Di Liberto, G.M., Barsotti, M., Vecchiato, G., Ambeck -Madsen, J., Del Vecchio, M., Avanzini, P., and Ascari, L. (2021). Robust anticipation of continuous steering actions from electroencephalographic data during simulated driving. Scientific Reports 11, 23383. 10.1038/s41598-021-02750-w

work page doi:10.1038/s41598-021-02750-w 2021
[12]

Nidiffer, A.R., Cao, C.Z., O'Sullivan, A., and Lalor, E.C. (2023). A representation of abstract linguistic categories in the visual system underlies successful lipreading. NeuroImage 282, 120391. https://doi.org/10.1016/j.neuroimage.2023.120391

work page doi:10.1016/j.neuroimage.2023.120391 2023
[13]

Rogachev, A., and Sysoeva, O. (2024). Neural tracking of natural speech in children in relation to their receptive speech abilities. Cognitive Systems Research 86, 101236

2024
[14]

Kalashnikova, M., Peter, V., Di Liberto, G.M., Lalor, E.C., and Burnham, D. (2018). Infant -directed speech facilitates seven -month-old infants’ cortical tracking of speech. Scientific Reports 8. 10.1038/s41598-018-32150-6

work page doi:10.1038/s41598-018-32150-6 2018
[15]

Di Liberto, G.M., Pelofi, C., Bianco, R., Patel, P., Mehta, A.D., Herrero, J.L., De Cheveigné, A., Shamma, S., and Mesgarani, N. (2020). Cortical encoding of melodic expectations in human temporal cortex. Elife 9, e51784

2020
[16]

Lalor, E.C., Pearlmutter, B.A., Reilly, R.B., McDarby, G., and Foxe, J.J. (2006). The VESPA: a method for the rapid estimation of a visual evoked potential. NeuroImage 32, 1549-1561

2006
[17]

Brodbeck, C., Presacco, A., and Simon, J.Z. (2018). Neural source dynamics of brain responses to continuous stimuli: Speech processing from acoustics to comprehension. NeuroImage 172, 162-174. https://doi.org/10.1016/j.neuroimage.2018.01.042

work page doi:10.1016/j.neuroimage.2018.01.042 2018
[18]

Thornton, M., Mandic, D., and Reichenbach, T. (2022). Robust decoding of the speech envelope from EEG recordings through deep neural networks. Journal of neural engineering 19, 046007

2022
[19]

Broderick, M., Anderson, A., Di Liberto, G., Crosse, M., and Lalor, E. (2018). Data from: electrophysiological correlates of semantic dissimilarity reflect the comprehension of natural, narrative speech. Dryad Digital Repository. Published online February 23, 2018. 21

2018
[20]

Di Liberto, G.M., O'Sullivan, J.A., and Lalor, E.C. (2015). Low-frequency cortical entrainment to speech reflects phoneme-level processing. Current Biology 25. 10.1016/j.cub.2015.08.030

work page doi:10.1016/j.cub.2015.08.030 2015
[21]

O'Sullivan, J.A., Power, A.J., Mesgarani, N., Rajaram, S., Foxe, J.J., Shinn -Cunningham, B.G., Slaney, M., Shamma, S.A., and Lalor, E.C. (2014). Attentional Selection in a Cocktail Party Environment Can Be Decoded from Single-Trial EEG. Cerebral Cortex, bht355-bht355

2014
[22]

Vanthornhout, J., Decruy, L., Wouters, J., Simon, J.Z., and Francart, T. (2018). Speech Intelligibility Predicted from Neural Entrainment of the Speech Envelope. Journal of the Association for Research in Otolaryngology 19, 181-191. 10.1007/s10162-018-0654-z

work page doi:10.1007/s10162-018-0654-z 2018
[23]

Ding, N., Chatterjee, M., and Simon, J.Z. (2014). Robust cortical entrainment to the speech envelope relies on the spectro-temporal fine structure. NeuroImage 88, 41-46

2014
[24]

de Cheveigné, A., Wong, D.E., Di Liberto, G.M., Hjortkjær, J., Slaney, M., and Lalor, E. (2018). Decoding the auditory brain with canonical component analysis. NeuroImage 172, 206 -216. 10.1016/j.neuroimage.2018.01.033

work page doi:10.1016/j.neuroimage.2018.01.033 2018
[25]

Di Liberto, G.M., Wong, D., Melnik, G.A., and de Cheveigne, A. (2019). Low -frequency cortical responses to natural speech reflect probabilistic phonotactics. NeuroImage 196, 237 -247. 10.1016/j.neuroimage.2019.04.037

work page doi:10.1016/j.neuroimage.2019.04.037 2019
[26]

Geirnaert, S., Vandecappelle, S., Alickovic, E., de Cheveigné, A., Lalor, E., Meyer, B.T., Miran, S., Francart, T., and Bertrand, A. (2021). Neuro -Steered Hearing Devices: Decoding Auditory Attention From the Brain

2021
[27]

Hannah, J., and Di Liberto, G.M. (2026). Trust Modulates Speech Entrainment: Enhanced Cortical Tracking for Low Trust Speakers. bioRxiv, 2026.2003. 2011.711118

work page arXiv 2026
[28]

Ip, E.Y., Akkaya, A., Winchester, M.M., Bishop, S.J., Cowan, B.R., and Di Liberto, G.M. (2025). Exploring the impact of social relevance on the cortical tracking of speech: viability and temporal response characterisation. bioRxiv, 2025.2009. 2023.674728

work page arXiv 2025
[29]

Chalehchaleh, A., Winchester, M.M., and Di Liberto, G. (2024). Robust assessment of the cortical encoding of word -level expectations using the temporal response function. Journal of Neural Engineering, 2024-2004. https://doi.org/10.1088/1741-2552/ada30a

work page doi:10.1088/1741-2552/ada30a 2024
[30]

Jessen, S., Obleser, J., and Tune, S. (2021). Neural tracking in infants – An analytical tool for multisensory social processing in development. Developmental Cognitive Neuroscience 52, 101034. https://doi.org/10.1016/j.dcn.2021.101034

work page doi:10.1016/j.dcn.2021.101034 2021
[31]

Herbst, S.K., Fiedler, L., and Obleser, J. (2018). Tracking Temporal Hazard in the Human Electroencephalogram Using a Forward Encoding Model. eneuro 5, ENEURO.0017 -0018.2018. 10.1523/ENEURO.0017-18.2018

work page doi:10.1523/eneuro.0017-18.2018 2018
[32]

Di Liberto, G.M., Peter, V., Kalashnikova, M., Goswami, U., Burnham, D., and Lalor, E.C. (2018). Atypical cortical entrainment to speech in the right hemisphere underpins phonemic deficits in dyslexia. NeuroImage NIMG-17-29, 70-79. 10.1016/J.NEUROIMAGE.2018.03.072

work page doi:10.1016/j.neuroimage.2018.03.072 2018
[33]

De Cheveigné, A., Wong, D.D.E., Di Liberto, G.M., Hjortkjær, J., Slaney, M., and Lalor, E. (2018). Decoding the auditory brain with canonical component analysis. NeuroImage 172, 206-216

2018
[34]

Jessen, S., Fiedler, L., Münte, T.F., and Obleser, J. (2019). Quantifying the individual auditory and visual brain response in 7 -month-old infants watching a brief cartoon movie. NeuroImage 202, 116060-116060. 10.1016/j.neuroimage.2019.116060

work page doi:10.1016/j.neuroimage.2019.116060 2019
[35]

Di Liberto, G.M., Attaheri, A., Cantisani, G., Reilly, R.B., Ní Choisdealbha, Á., Rocha, S., Brusini, P., and Goswami, U. (2023). Emergence of the cortical encoding of phonetic features in the first year of life. Nature Communications 14, 7789

2023
[36]

Attaheri, A., Choisdealbha, Á.N., Di Liberto, G.M., Rocha, S., Brusini, P., Mead, N., Olawole-Scott, H., Boutris, P., Gibbon, S., and Williams, I. (2022). Delta -and theta -band cortical tracking and phase - amplitude coupling to sung speech by infants. NeuroImage 247, 118698

2022
[37]

Um…, it’s really difficult to… um… speak fluently

Agmon, G., Jaeger, M., Tsarfaty, R., Bleichner, M.G., and Zion Golumbic, E. (2023). “Um…, it’s really difficult to… um… speak fluently”: Neural tracking of spontaneous speech. Neurobiology of Language 4, 435-454

2023
[38]

Solanki, V.J. (2017). Brains in dialogue: investigating accommodation in live conversational speech for both speech and EEG data. (University of Glasgow). 22

2017
[39]

Van de Ryck, I., Heintz, N., Rotaru, I., Geirnaert, S., Bertrand, A., and Francart, T. (2026). EEG -based Decoding of Auditory Attention to Conversations with Turn -taking Speakers. Hearing Research, 109539

2026
[40]

Orf, M., Tune, S., Hannemann, R., and Obleser, J. (2025). Speech, gait, and brain dynamics during natural conversation in motion. bioRxiv, 2025.2012.2023.696153. 10.64898/2025.12.23.696153

work page doi:10.64898/2025.12.23.696153 2025
[41]

King, J.-R., Charton, F., Lopez-Paz, D., and Oquab, M. (2020). Back-to-back regression: Disentangling the influence of correlated factors from multivariate observations. NeuroImage 220, 117028

2020
[42]

de Cheveigné, A., Di Liberto, G.M., Arzounian, D., Wong, D., Hjortkjaer, J., Fuglsang, S.A., and Parra, L.C. (2018). Multiway Canonical Correlation Analysis of Brain Signals. bioRxiv, 344960 -344960. 10.1101/344960

work page doi:10.1101/344960 2018
[43]

Accou, B., Vanthornhout, J., hamme, H.V., and Francart, T. (2023). Decoding of the speech envelope from EEG using the VLAAI deep neural network. Scientific Reports 13, 812. 10.1038/s41598 -022- 27332-2

work page doi:10.1038/s41598 2023
[44]

Broderick, M., Di Liberto, G., Anderson, A., Rofes, A., and Lalor, E. (2020). Dissociable electrophysiological measures of natural language processing reveal differences in speech comprehension strategy in healthy ageing. bioRxiv, 2020.2004.2017.046201 - 042020.046204.046217.046201. 10.1101/2020.04.17.046201

work page doi:10.1101/2020.04.17.046201 2020
[45]

Mischler, G., Li, Y.A., Bickel, S., Mehta, A.D., and Mesgarani, N. (2024). Contextual Feature Extraction Hierarchies Converge in Large Language Models and the Brain. arXiv preprint arXiv:2401.17671

work page arXiv 2024
[46]

Broderick, M.P., Di Liberto, G.M., Anderson, A.J., Rofes, A., and Lalor, E.C. (2021). Dissociable electrophysiological measures of natural language processing reveal differences in speech comprehension strategy in healthy ageing. Scientific Reports 11, 4963. 10.1038/s41598-021-84597- 9

work page doi:10.1038/s41598-021-84597- 2021
[47]

Di Liberto, G.M., Nie, J., Yeaton, J., Khalighinejad, B., Shamma, S.A., and Mesgarani, N. (2021). Neural representation of linguistic feature hierarchy reflects second-language proficiency. NeuroImage 227, 117586-117586. 10.1016/j.neuroimage.2020.117586

work page doi:10.1016/j.neuroimage.2020.117586 2021
[48]

Di Liberto, G.M., Pelofi, C., Shamma, S., and de Cheveigné, A. (2020). Musical expertise enhances the cortical tracking of the acoustic envelope during naturalistic music listening. Acoustical Science and Technology 41

2020
[49]

Cohen, S.S., and Parra, L.C. (2016). Memorable Audiovisual Narratives Synchronize Sensory and Supramodal Neural Responses. eneuro 3, ENEURO.0203-0216.2016. 10.1523/eneuro.0203-16.2016

work page doi:10.1523/eneuro.0203-16.2016 2016
[50]

de Cheveigné, A., Di Liberto, G.M., Arzounian, D., Wong, D.D.E., Hjortkjær, J., Fuglsang, S., and Parra, L.C. (2019). Multiway canonical correlation analysis of brain data. NeuroImage 186, 728 -740. 10.1016/J.NEUROIMAGE.2018.11.026

work page doi:10.1016/j.neuroimage.2018.11.026 2019
[51]

Obleser, J., and Kayser, C. (2019). Neural Entrainment and Attentional Selection in the Listening Brain. Trends in Cognitive Sciences. Elsevier Ltd

2019
[52]

Klimovich-Gray, A., Di Liberto, G., Amoruso, L., Barrena, A., Agirre, E., and Molinaro, N. (2023). Increased top -down semantic processing in natural speech linked to better reading in dyslexia. NeuroImage 273, 120072. https://doi.org/10.1016/j.neuroimage.2023.120072

work page doi:10.1016/j.neuroimage.2023.120072 2023
[53]

Attaheri, A., Ní Choisdealbha Á, Di Liberto, G.M., Rocha, S., Brusini, P., Mead, N., Olawole -Scott, H., Boutris, P., Gibbon, S., Williams, I., et al. (2022). Delta - and theta-band cortical tracking and phase - amplitude coupling to sung speech by infants. N euroimage 247, 118698. 10.1016/j.neuroimage.2021.118698

work page doi:10.1016/j.neuroimage.2021.118698 2022
[54]

Verschueren, E., Somers, B., and Francart, T. (2019). Neural envelope tracking as a measure of speech understanding in cochlear implant users. Hearing Research 373, 23 -31. https://doi.org/10.1016/j.heares.2018.12.004

work page doi:10.1016/j.heares.2018.12.004 2019
[55]

Somers, B., Verschueren, E., and Francart, T. (2018). Neural tracking of the speech envelope in cochlear implant users. Journal of Neural Engineering 16, 16003-16003. 10.1088/1741-2552/aae6b9

work page doi:10.1088/1741-2552/aae6b9 2018
[56]

Chen, Y.-P., Schmidt, F., Keitel, A., Rösch, S., Hauswald, A., and Weisz, N. (2023). Speech intelligibility changes the temporal evolution of neural speech tracking. NeuroImage 268, 119894. https://doi.org/10.1016/j.neuroimage.2023.119894. 23

work page doi:10.1016/j.neuroimage.2023.119894 2023
[57]

Van Hirtum, T., Somers, B., Dieudonné, B., Verschueren, E., Wouters, J., and Francart, T. (2023). Neural envelope tracking predicts speech intelligibility and hearing aid benefit in children with hearing loss. Hearing Research 439, 108893. https://doi.org/10.1016/j.heares.2023.108893

work page doi:10.1016/j.heares.2023.108893 2023
[58]

Van Hirtum, T., Somers, B., Verschueren, E., Dieudonné, B., and Francart, T. (2023). Delta-band neural envelope tracking predicts speech intelligibility in noise in preschoolers. Hearing Research 434, 108785. https://doi.org/10.1016/j.heares.2023.108785

work page doi:10.1016/j.heares.2023.108785 2023
[59]

Carta, S., Aličković, E., Zaar, J., Valdés, A.L., and Di Liberto, G.M. (2025). Simultaneous cortical tracking of competing speech streams during attention switching. bioRxiv, 2025.2007. 2002.662762

work page arXiv 2025
[60]

-D., Blankertz, B., and Bießmann, F

Haufe, S., Meinecke, F., Görgen, K., Dähne, S., Haynes, J. -D., Blankertz, B., and Bießmann, F. (2014). On the interpretation of weight vectors of linear models in multivariate neuroimaging. NeuroImage 87, 96-110. https://doi.org/10.1016/j.neuroimage.2013.10.067

work page doi:10.1016/j.neuroimage.2013.10.067 2014
[61]

Di Liberto, G.M., Marion, G., and Shamma, S.A. (2021). The Music of Silence: Part II: Music Listening Induces Imagery Responses. The Journal of Neuroscience 41, 7449. 10.1523/JNEUROSCI.0184 - 21.2021

work page doi:10.1523/jneurosci.0184 2021
[62]

Marion, G., Di Liberto, G.M., and Shamma, S.A. (2021). The Music of Silence. Part I: Responses to Musical Imagery Accurately Encode Melodic Expectations and Acoustics. Journal of Neuroscience

2021
[63]

Varma, S., and Simon, R. (2006). Bias in error estimation when using cross -validation for model selection. BMC Bioinformatics 7, 91. 10.1186/1471-2105-7-91

work page doi:10.1186/1471-2105-7-91 2006
[64]

Wong, D.D.E., Fuglsang, S.A., Hjortkjaer, J., Ceolini, E., Slaney, M., and De Cheveigne, A. (2018). A Comparison of Regularization Methods in Forward and Backward Models for Auditory Attention Decoding. Frontiers in Neuroscience 12, 531-531. 10.3389/FNINS.2018.00531

work page doi:10.3389/fnins.2018.00531 2018
[65]

Di Liberto, G.M., and Lalor, E.C. (2017). Indexing cortical entrainment to natural speech at the phonemic level: Methodological considerations for applied research. Hearing Research 348, 70-77. 10.1016/j.heares.2017.02.015

work page doi:10.1016/j.heares.2017.02.015 2017
[66]

Brookshire, G., Lu, J., Nusbaum, H.C., Goldin -Meadow, S., and Casasanto, D. (2017). Visual cortex entrains to sign language. Proceedings of the National Academy of Sciences of the United States of America 114, 6352-6357. 10.1073/pnas.1620350114

work page doi:10.1073/pnas.1620350114 2017
[67]

Ding, N., and Simon, J.Z. (2012). Neural coding of continuous speech in auditory cortex during monaural and dichotic listening. Journal of Neurophysiology 107, 78-89. 10.1152/jn.00297.2011

work page doi:10.1152/jn.00297.2011 2012
[68]

Wilroth, J., Silva, N.S., Tafakkor, A., de Avo Mesquita, B., Ip, E.Y.J., Lau, B., Hannah, J., and Di Liberto, G.M. (2026). Investigating neural speech processing with functional near infrared spectroscopy: considerations for temporal response functions. b ioRxiv, 2026.2003.2020.713212. 10.64898/2026.03.20.713212

work page doi:10.64898/2026.03.20.713212 2026
[69]

Di Liberto, G.M., Nidiffer, A., Crosse, M.J., Zuk, N., Haro, S., Cantisani, G., Winchester, M.M., Igoe, A., McCrann, R., and Chandra, S. (2024). A standardised open science framework for sharing and re - analysing neural data acquired to continuous stimuli. Neurons, Behavior, Data analysis, and Theory, 1-25

2024
[70]

Delorme, A., and Makeig, S. (2004). EEGLAB: an open source toolbox for analysis of single -trial EEG dynamics including independent component analysis. J Neurosci Methods 134, 9 -21. 10.1016/j.jneumeth.2003.10.009

work page doi:10.1016/j.jneumeth.2003.10.009 2004
[71]

Alice in Wonderland

Brennan, J.R. (2018). EEG Datasets for Naturalistic Listening to "Alice in Wonderland" (Version 1)

2018
[72]

Gwilliams, L., Flick, G., Marantz, A., Pylkkänen, L., Poeppel, D., and King, J. -R. (2023). Introducing MEG-MASC a high -quality magneto -encephalography dataset for evaluating natural speech processing. Scientific Data 10, 862. 10.1038/s41597-023-02752-5

work page doi:10.1038/s41597-023-02752-5 2023
[73]

Piazza, G., Carta, S., Ip, E., Pérez -Navarro, J., Kalashnikova, M., Martin, C.D., and Di Liberto, G.M. (2024). Are you talking to me? How the choice of speech register impacts listeners' hierarchical encoding of speech. bioRxiv, 2024-2009

2024
[74]

Piazza, G., Carta, S., Ip, E.Y.J., Pérez-Navarro, J., Kalashnikova, M., Martin, C.D., and Di Liberto, G.M. (2025). Are you talking to me? How the choice of speech register impacts listeners’ hierarchical encoding of speech

2025
[75]

Bollens, L., Accou, B., Van hamme, H., and Francart, T. (2023). SparrKULee: A Speech-evoked Auditory Response Repository of the KU Leuven, containing EEG of 85 participants. V3 ed. KU Leuven RDR. 24

2023
[76]

De Palma, I.C., Lopez, L.S., and Lopez-Valdes, A. (2023). Effects of spectral and temporal modulation degradation on intelligibility and cortical tracking of speech signals. pp. 5192-5196

2023
[77]

Emergence of the cortical encoding of phonetic features in the first year of life

Di Liberto, G.M., Goswami, U., Attaheri, A., Choisdealbha Á, N., Rocha, S., Mead, N., Olawole -Scott, H., and Grey, C. (2023). Data and code from “Emergence of the cortical encoding of phonetic features in the first year of life". OSF https://osf.io/mdnwg

2023
[78]

Das, N., Francart, T., and Bertrand, A. (2019). Auditory Attention Detection Dataset KULeuven (Version 2.0). Zenodo

2019
[79]

Cantisani, G., Shamma, S., and Di Liberto, G.M. (2024). Neural signatures of musical and linguistic interactions during natural song listening

2024
[80]

Di Liberto, G.M., Pelofi, C., Bianco, R., Patel, P., Mehta, A.D., Herrero, J.L., De Cheveigné, A., Shamma, S.A., and Mesgarani, N. (2021). Cortical encoding of melodic expectations in human temporal cortex

2021

Showing first 80 references.