CoChat

Applying deep learning to single-trial EEG data provides evidence for complementary theories on action control

Open access

The DOI matches its Crossref record; 70 of the 70 references it cites were found in a registry or an index.

Abstract

Efficient action control is indispensable for goal-directed behaviour. Different theories have stressed the importance of either attention or response selection sub-processes for action control. Yet, it is unclear to what extent these processes can be identified in the dynamics of neurophysiological (EEG) processes at the single-trial level and be used to predict the presence of conflicts in a given moment. Applying deep learning, which was blind to cognitive theory, on single-trial EEG data allowed to predict the presence of conflict in ~95% of subjects ~33% above chance level. Neurophysiological features related to attentional and motor response selection processes in the occipital cortex and the superior frontal gyrus contributed most to prediction accuracy. Importantly, deep learning was able to identify predictive neurophysiological processes in single-trial neural dynamics. Hence, mathematical (artificial intelligence) approaches may be used to foster the validation and development of links between cognitive theory and neurophysiology of human behavior.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/s42003-020-0846-z

Cites

70 of 70 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Classification accuracy for this 4-class problem is shown for each individual subject. On the x-axis the individual subjects are shown. The y-axis denotes the classification accuracy for the 4-class problem in each individual. The dotted lines denotes the chance level calculated for the individual participant using the method by Combrisson and Jerbi53.
Fig. 1 Classification accuracy for this 4-class problem is shown for each individual subject. On the x-axis the individual subjects are shown. The y-axis denotes the classification accuracy for the 4-class problem in each individual. The dotted lines denotes the chance level calculated for the individual participant using the method by Combrisson and Jerbi53.CC BY · PubMed Central
Fig. 2 Confusion matrix showing the classification results for the different conditions. Colour shadings and numbers in the matrix denote the frequency at which the real data (“true”) label was classified into one of the four possible predicted classes.
Fig. 2 Confusion matrix showing the classification results for the different conditions. Colour shadings and numbers in the matrix denote the frequency at which the real data (“true”) label was classified into one of the four possible predicted classes.CC BY · PubMed Central
Fig. 3 Training and validation loss for two subjects. One with high accuracy (80%) and one with low accuracy (20%).
Fig. 3 Training and validation loss for two subjects. One with high accuracy (80%) and one with low accuracy (20%).CC BY · PubMed Central
Fig. 4 Visualization of the deep learning analysis. a Visualization maps showing the relevance of all timepoints and electrodes for classification between the four different classes of trials. Values close to 1 indicate that the specific feature at the specific timepoints contributes most to classification accuracy. The x-axis denotes the time in ms after target stimulus presentation. The y-axis denotes the different electrode sites. b Event-related potential at the electrode sites contributing most to classification accuracy in the deep learning model. The x-axis denotes the time in ms after target stimulus presentation. The y-axis denotes the voltage (note that the scaling of the y-axis differs between the plots). The gray-shadings denote the time interval that was found to contribute strongly to classification performance in the deep learning network. c The scalp topography plots (top) are shown denoting the distribution of amplitudes across the scalp in the time interval that contributed most to classification performance. For the sensor space images the ‘top-view’ is presented. Hence, electrodes appearing on the left in the figure are also place at the left of the scalp. Red colours denote positive amplitudes, blue colours negative amplitudes. At the bottom, the corresponding source localization results are shown for each of the different conditions. Only significant activations are shown (p < .05) corrected for multiple comparisons using voxel-wise randomization tests with 2000 permutations and statistical nonparametric mapping procedures (SnPM).
Fig. 4 Visualization of the deep learning analysis. a Visualization maps showing the relevance of all timepoints and electrodes for classification between the four different classes of trials. Values close to 1 indicate that the specific feature at the specific timepoints contributes most to classification accuracy. The x-axis denotes the time in ms after target stimulus presentation. The y-axis denotes the different electrode sites. b Event-related potential at the electrode sites contributing most to classification accuracy in the deep learning model. The x-axis denotes the time in ms after target stimulus presentation. The y-axis denotes the voltage (note that the scaling of the y-axis differs between the plots). The gray-shadings denote the time interval that was found to contribute strongly to classification performance in the deep learning network. c The scalp topography plots (top) are shown denoting the distribution of amplitudes across the scalp in the time interval that contributed most to classification performance. For the sensor space images the ‘top-view’ is presented. Hence, electrodes appearing on the left in the figure are also place at the left of the scalp. Red colours denote positive amplitudes, blue colours negative amplitudes. At the bottom, the corresponding source localization results are shown for each of the different conditions. Only significant activations are shown (p < .05) corrected for multiple comparisons using voxel-wise randomization tests with 2000 permutations and statistical nonparametric mapping procedures (SnPM).CC BY · PubMed Central
Fig. 5 The target stimuli (letters) could be located in either of the boxes on the left or the right of the fixation cross. Letter A required a reaction of the left hand (irrespective of the spatial position of the letter) while letter B required a reaction of the right hand (irrespective of the spatial position of the letter).
Fig. 5 The target stimuli (letters) could be located in either of the boxes on the left or the right of the fixation cross. Letter A required a reaction of the left hand (irrespective of the spatial position of the letter) while letter B required a reaction of the right hand (irrespective of the spatial position of the letter).CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

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

Self-check

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

The self-check needs the paper’s full text.

Citations

Cites

1 retracted

Show all 70 references

Research works printed but not linked: 1

  1. Shenhav, A., Botvinick, M. M. & Cohen, J. D. The expected value of control: an integrative theory of anterior cingulate cortex function. Neuron 79, 217–240 (2013). · DOI · Paper page

Cited by

See all 95

Metadata

Authors
Vahid, A., Mückschel, M., Stober, S., Stock, A.-K., & Beste, C.
Affiliations
  • Technische Universität Dresden (DE)
  • Otto-von-Guericke-Universität Magdeburg (DE)
Corresponding author
Beste, Christian
Venue
Communications Biology
Published
2020-03-09
Volume
3
Issue
1
Pages
112-112
ISSN
2399-3642 (ISSN-L)
OpenAlex
W3009188842
Cited by
87
References
68
Citations by year
2026: 3 · 2025: 11 · 2024: 15 · 2023: 19 · 2022: 17 · 2021: 20 · 2020: 2
FWCI
7.79
Type
journal article
Language
en
Open access
Licence
Locations
  • Communications Biology (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
Funders
Deutsche Forschungsgemeinschaft (SFB 940 project B8, SFB TRR 265)
Topics
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Neural and Behavioral Psychology Studies
  • Life Sciences › Neuroscience › Cognitive Neuroscience › EEG and Brain-Computer Interfaces
  • Social Sciences › Arts and Humanities › Philosophy › Free Will and Agency
Keywords
action control, single-trial EEG, conflict prediction, attention, response selection, cognitive control, neural dynamics, occipital cortex, superior frontal gyrus, deep learning
MeSH terms
Deep Learning, Adolescent, Adult, Attention, Brain, Cognition, Conflict, Psychological, Electroencephalography, Female, Goals, Humans, Male, Motor Activity, Neuropsychological Tests, Predictive Value of Tests, Reaction Time, Signal Processing, Computer-Assisted, Young Adult, Brain Waves, Support Vector Machine

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.

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: Applying deep learning to single-trial EEG data provides evidence for complementary theories on action control. https://app.cochat.ai/papers/10.1038/s42003-020-0846-z
The paper
Vahid, A., Mückschel, M., Stober, S., Stock, A.-K., & Beste, C. (2020). Applying deep learning to single-trial EEG data provides evidence for complementary theories on action control. Communications Biology. https://doi.org/10.1038/s42003-020-0846-z

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.