Student or educator? Sign up with your school email and get 6 months of Pro free.

The DOI matches its Crossref record; 72 of the 74 references it cites were found in a registry or an index.

A distributed fMRI-based signature for the subjective experience of fear

Open access

Abstract

The specific neural systems underlying the subjective feeling of fear are debated in affective neuroscience. Here, we combine functional MRI with machine learning to identify and evaluate a sensitive and generalizable neural signature predictive of the momentary self-reported subjective fear experience across discovery (n = 67), validation (n = 20) and generalization (n = 31) cohorts. We systematically demonstrate that accurate fear prediction crucially requires distributed brain systems, with important contributions from cortical (e.g., prefrontal, midcingulate and insular cortices) and subcortical (e.g., thalamus, periaqueductal gray, basal forebrain and amygdala) regions. We further demonstrate that the neural representation of subjective fear is distinguishable from the representation of conditioned threat and general negative affect. Overall, our findings suggest that subjective fear, which exhibits distinct neural representation with some other aversive states, is encoded in distributed systems rather than isolated ‘fear centers’.

Overview

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

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

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

Integrity signals

Registry

Crossref · 10.1038/s41467-021-26977-3

Cites

72 of 72 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Experimental paradigms and analysis stages. Discovery (a) and validation (b) cohorts underwent two similar but not identical fear induction and rating paradigms during fMRI acquisition. Of note, examples of the fear-evoking photos are pictures only for display purposes and not included in the original stimulus set. The pictures have been obtained from pixabay.com under the Pixabay License, and are free for commercial and noncommercial use across print and digital. c depicts the analytic stages and datasets used in the present study. Specifically, a whole-brain multivariate pattern predictive of the level of subjective experience of fear was trained on the discovery sample (n = 67) using support vector regression and further evaluated in discovery (cross-validated), validation (n = 20), and generalization (n = 31) cohorts. We next systematically applied univariate and multivariate analyses to determine the spatial scale and local contributions of specific regions to the momentary subjective fear representation. Finally, we tested whether subjective fear was encoded with a neural signature that was distinct from the representation of conditioned threat (CS+ versus CS− cues) and general negative affect. GLM general linear model, SVR support vector regression, VIFS visually induced fear signature developed in the current study, AFSS animal fear schema signature developed by Vincent Taschereau-Dumouchel and colleagues, TPS threat-predictive signature developed by Reddan and colleagues, PINES picture-induced negative emotion signature developed by Luke Chang and colleagues. See ‘Methods’ and ‘Results’ for the details of the datasets and brain signatures used in this study.
Fig. 1 Experimental paradigms and analysis stages. Discovery (a) and validation (b) cohorts underwent two similar but not identical fear induction and rating paradigms during fMRI acquisition. Of note, examples of the fear-evoking photos are pictures only for display purposes and not included in the original stimulus set. The pictures have been obtained from pixabay.com under the Pixabay License, and are free for commercial and noncommercial use across print and digital. c depicts the analytic stages and datasets used in the present study. Specifically, a whole-brain multivariate pattern predictive of the level of subjective experience of fear was trained on the discovery sample (n = 67) using support vector regression and further evaluated in discovery (cross-validated), validation (n = 20), and generalization (n = 31) cohorts. We next systematically applied univariate and multivariate analyses to determine the spatial scale and local contributions of specific regions to the momentary subjective fear representation. Finally, we tested whether subjective fear was encoded with a neural signature that was distinct from the representation of conditioned threat (CS+ versus CS− cues) and general negative affect. GLM general linear model, SVR support vector regression, VIFS visually induced fear signature developed in the current study, AFSS animal fear schema signature developed by Vincent Taschereau-Dumouchel and colleagues, TPS threat-predictive signature developed by Reddan and colleagues, PINES picture-induced negative emotion signature developed by Luke Chang and colleagues. See ‘Methods’ and ‘Results’ for the details of the datasets and brain signatures used in this study.CC BY · PubMed Central
Fig. 2 Visually induced fear signature (VIFS). a depicts the VIFS pattern thresholded using a 10,000-sample bootstrap procedure at q < 0.05, FDR corrected. Inserts show the spatial topography of the unthresholded patterns in the left ACC, right PAG, bilateral AMG, and bilateral insula. AMG denotes amygdala, THAL thalamus, vmPFC ventromedial prefrontal cortex, dmPFC dorsal medial prefrontal cortex, dACC dorsal anterior cingulate cortex, MCC middle cingulate cortex, ACC anterior cingulate cortex, and PAG periaqueductal gray. b, c depict the predicted fear experience (subjective ratings; mean ± SE) compared to the actual level of fear for the cross-validated discovery cohort and the independent validation cohort, respectively. Accuracies reflect forced-choice comparisons. Two-sided binomial tests were used to test whether the classification accuracies were higher than chance level. r indicates the Pearson correlation coefficient between predicted and true ratings. d, e depict an average peristimulus plot (mean ± SE) of the VIFS response to the cross-validated discovery cohort and the independent validation cohort. This reflects the average VIFS response at every repetition time (TR; 2 s) in the time series separated by the fear ratings. Of note, the VIFS reacts with a latency of approximate 4 s after stimulus onset which corresponds to the timing of the hemodynamic response function (HRF) following stimulus onset. f, g compare the fear prediction (mean ± SE) of AFSS with the VIFS on discover, validation, and generalization cohorts, respectively. * indicates P < 0.05, **P < 0.01 and ***P < 0.001 (uncorrected). Error bars and shaded regions indicate SEs. VIFS visually induced fear signature, AFSS animal fear schema signature, SE standard error of mean. Source data are provided as a Source Data file.
Fig. 2 Visually induced fear signature (VIFS). a depicts the VIFS pattern thresholded using a 10,000-sample bootstrap procedure at q < 0.05, FDR corrected. Inserts show the spatial topography of the unthresholded patterns in the left ACC, right PAG, bilateral AMG, and bilateral insula. AMG denotes amygdala, THAL thalamus, vmPFC ventromedial prefrontal cortex, dmPFC dorsal medial prefrontal cortex, dACC dorsal anterior cingulate cortex, MCC middle cingulate cortex, ACC anterior cingulate cortex, and PAG periaqueductal gray. b, c depict the predicted fear experience (subjective ratings; mean ± SE) compared to the actual level of fear for the cross-validated discovery cohort and the independent validation cohort, respectively. Accuracies reflect forced-choice comparisons. Two-sided binomial tests were used to test whether the classification accuracies were higher than chance level. r indicates the Pearson correlation coefficient between predicted and true ratings. d, e depict an average peristimulus plot (mean ± SE) of the VIFS response to the cross-validated discovery cohort and the independent validation cohort. This reflects the average VIFS response at every repetition time (TR; 2 s) in the time series separated by the fear ratings. Of note, the VIFS reacts with a latency of approximate 4 s after stimulus onset which corresponds to the timing of the hemodynamic response function (HRF) following stimulus onset. f, g compare the fear prediction (mean ± SE) of AFSS with the VIFS on discover, validation, and generalization cohorts, respectively. * indicates P < 0.05, **P < 0.01 and ***P < 0.001 (uncorrected). Error bars and shaded regions indicate SEs. VIFS visually induced fear signature, AFSS animal fear schema signature, SE standard error of mean. Source data are provided as a Source Data file.CC BY · PubMed Central
Fig. 3 Subjective experience of fear is associated with and predicted by distributed brain regions. a shows the thresholded VIFS. b depicts the threshholded transformed ‘activation pattern’ from the VIFS. c shows the overlap between VIFS and transformed ‘activation pattern’. All images are thresholded at q < 0.05, FDR corrected. Hot color indicates positive associations (a) or weights (b) whereas cold color indicates negative associations (a) or weights (b). VIFS visually induced fear signature.
Fig. 3 Subjective experience of fear is associated with and predicted by distributed brain regions. a shows the thresholded VIFS. b depicts the threshholded transformed ‘activation pattern’ from the VIFS. c shows the overlap between VIFS and transformed ‘activation pattern’. All images are thresholded at q < 0.05, FDR corrected. Hot color indicates positive associations (a) or weights (b) whereas cold color indicates negative associations (a) or weights (b). VIFS visually induced fear signature.CC BY · PubMed Central
Fig. 4 Local brain region and network predictions. a shows brain regions that can significantly (P < 0.001 uncorrected, equivalent to FDR q < 0.004; two-sided) predict subjective fear ratings revealed by searchlight analysis. Histograms: cross-validated predictions (correlations) from local searchlight analysis. Red line indicates the prediction-outcome correlation from VIFS. b depicts brain regions which can significantly (P < 0.001 uncorrected, equivalent to FDR q < 0.003; two-sided) predict subjective fear revealed by parcellation-based analysis. Histograms: cross-validated predictions (correlations) from parcellations. Red line indicates the prediction-outcome correlation from VIFS. c shows cross-validated predictions (mean ± SE) from amygdala-based prediction analysis. Error bar indicates standard error of the mean; r indicates overall (between- and within-subjects; i.e., n = 333 pairs) prediction-outcome Pearson correlation coefficient. d demonstrates that the information about subjective experience of fear is distributed across multiple systems. Model performance was evaluated as increasing numbers of voxels/features (x axis) were used to predict subjective fear in different regions of interest including the entire brain (black), consciousness network (red), subcortical regions (light purple) or large-scale resting-state networks. The y axis denotes the cross-validated prediction-outcome correlation. Colored dots indicate the mean correlation coefficients, solid lines indicate the mean parametric fit and shaded regions indicate standard deviation. Model performance is optimized when approximately 10,000 voxels are randomly sampled across the whole-brain. VIFS visually induced fear signature. Source data are provided as a Source Data file.
Fig. 4 Local brain region and network predictions. a shows brain regions that can significantly (P < 0.001 uncorrected, equivalent to FDR q < 0.004; two-sided) predict subjective fear ratings revealed by searchlight analysis. Histograms: cross-validated predictions (correlations) from local searchlight analysis. Red line indicates the prediction-outcome correlation from VIFS. b depicts brain regions which can significantly (P < 0.001 uncorrected, equivalent to FDR q < 0.003; two-sided) predict subjective fear revealed by parcellation-based analysis. Histograms: cross-validated predictions (correlations) from parcellations. Red line indicates the prediction-outcome correlation from VIFS. c shows cross-validated predictions (mean ± SE) from amygdala-based prediction analysis. Error bar indicates standard error of the mean; r indicates overall (between- and within-subjects; i.e., n = 333 pairs) prediction-outcome Pearson correlation coefficient. d demonstrates that the information about subjective experience of fear is distributed across multiple systems. Model performance was evaluated as increasing numbers of voxels/features (x axis) were used to predict subjective fear in different regions of interest including the entire brain (black), consciousness network (red), subcortical regions (light purple) or large-scale resting-state networks. The y axis denotes the cross-validated prediction-outcome correlation. Colored dots indicate the mean correlation coefficients, solid lines indicate the mean parametric fit and shaded regions indicate standard deviation. Model performance is optimized when approximately 10,000 voxels are randomly sampled across the whole-brain. VIFS visually induced fear signature. Source data are provided as a Source Data file.CC BY · PubMed Central
Fig. 5 Comparing fear- and threat-predictive signatures. a depicts that visually induced fear signature (VIFS) does not distinguish unreinforced CS+ versus CS−. b shows the histograms of prediction of threat-predictive signature (TPS) on fear data from nonparametric permutation test. Histograms show the distribution of null-hypothesis prediction-outcome correlations, and the red line shows the actual correlation coefficient. c demonstrates the scatter plot displaying normalized voxel weights for VIFS (y axis) and TPS (x axis). Bars on the right represent the sum of squared distances from the origin (0,0) for each Octant. Different colors are assigned to the eight Octants that reflect voxels of shared positive or shared negative weights (Octants 2 and 6, respectively), selectively positive weights for the VIFS (Octant 1) or for TPS (Octant 3), selectively negative weights for the VIFS (Octant 5) or TPS (Octant 7), and voxels with opposite weights for the two neural signatures (Octants 4 and 8). Source data are provided as a Source Data file.
Fig. 5 Comparing fear- and threat-predictive signatures. a depicts that visually induced fear signature (VIFS) does not distinguish unreinforced CS+ versus CS−. b shows the histograms of prediction of threat-predictive signature (TPS) on fear data from nonparametric permutation test. Histograms show the distribution of null-hypothesis prediction-outcome correlations, and the red line shows the actual correlation coefficient. c demonstrates the scatter plot displaying normalized voxel weights for VIFS (y axis) and TPS (x axis). Bars on the right represent the sum of squared distances from the origin (0,0) for each Octant. Different colors are assigned to the eight Octants that reflect voxels of shared positive or shared negative weights (Octants 2 and 6, respectively), selectively positive weights for the VIFS (Octant 1) or for TPS (Octant 3), selectively negative weights for the VIFS (Octant 5) or TPS (Octant 7), and voxels with opposite weights for the two neural signatures (Octants 4 and 8). Source data are provided as a Source Data file.CC BY · PubMed Central
Fig. 6 Comparing VIFS and PINES responses. a depicts that VIFS more accurately (shown as forced-choice classification accuracy and Cohen’s d) predicts high versus low subjective fear while PINES is more sensitive to distinguish high versus low general negative emotion. b shows the multilevel mediation analysis results showing that VIFS response mediates the PINES response—fear rating association in both discovery and validation cohorts. c shows that the PINES response does not mediate the VIFS response—fear rating association in the discovery cohort. Although the mediation effect is significant in the validation cohort, the effect size (Cohen’s d = 0.06) is very small. ** indicates P < 0.01, ***P < 0.001, NSP > 0.5 (bootstrap tests; two-sided; uncorrected). VIFS visually induced fear signature, PINES picture-induced negative emotion signature. Source data are provided as a Source Data file.
Fig. 6 Comparing VIFS and PINES responses. a depicts that VIFS more accurately (shown as forced-choice classification accuracy and Cohen’s d) predicts high versus low subjective fear while PINES is more sensitive to distinguish high versus low general negative emotion. b shows the multilevel mediation analysis results showing that VIFS response mediates the PINES response—fear rating association in both discovery and validation cohorts. c shows that the PINES response does not mediate the VIFS response—fear rating association in the discovery cohort. Although the mediation effect is significant in the validation cohort, the effect size (Cohen’s d = 0.06) is very small. ** indicates P < 0.01, ***P < 0.001, NSP > 0.5 (bootstrap tests; two-sided; uncorrected). VIFS visually induced fear signature, PINES picture-induced negative emotion signature. Source data are provided as a Source Data file.CC BY · PubMed Central

Verification

Claims the paper makes

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

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

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

Show all 74 references

Research works printed but not linked: 2

  1. Clore, G. L. & Ortony, A. Appraisal theories: How cognition shapes affect into emotion. In Handbook of emotions (eds Lewis, M., Haviland-Jones, J. M. & Barrett, L. F.) (pp. 628–642) (The Guilford Press, New York City, 2008).
  2. Zhou, F. et al. Empathic pain evoked by sensory and emotional-communicative cues share common and process-specific neural representations. Elife 9, e56929 (2020). · DOI

Cited by

See all 155

Metadata

Authors
Zhou, F., Zhao, W., Qi, Z., Geng, Y., Yao, S., Kendrick, K. M., Wager, T. D., & Becker, B.
Affiliations
  • Dartmouth College (US)
  • University of Electronic Science and Technology of China (CN)
Corresponding author
  • Wager, Tor D.
  • Becker, Benjamin
Venue
Nature Communications
Published
2021-11-17
Volume
12
Issue
1
Pages
6643-6643
ISSN
2041-1723 (ISSN-L)
OpenAlex
W3214010604
Cited by
155
References
74
Citations by year
2026: 28 · 2025: 37 · 2024: 33 · 2023: 30 · 2022: 27
FWCI
11.83
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Communications (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • China Postdoctoral Science Foundation (2018M643432)
  • National Natural Science Foundation of China (31530032)
  • U.S. Department of Health & Human Services | NIH | National Institute of Mental Health (R01MH076136, R01MH116026)
  • National Key Research and Development Program of China
Topics
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Memory and Neural Mechanisms
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Neural and Behavioral Psychology Studies
  • Social Sciences › Psychology › Experimental and Cognitive Psychology › Anxiety, Depression, Psychometrics, Treatment, Cognitive Processes
Keywords
subjective fear, neural signatures, amygdala, thalamus, periaqueductal gray, insula, prefrontal cortex, midcingulate cortex, negative affect, aversive states, affective neuroscience
MeSH terms
Brain, Brain Mapping, Conditioning, Classical, Emotions, Fear, Female, Humans, Magnetic Resonance Imaging, Male, Nerve Net, Photic Stimulation, Reproducibility of Results, Cohort Studies, Young Adult

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: A distributed fMRI-based signature for the subjective experience of fear. https://app.cochat.ai/papers/10.1038/s41467-021-26977-3
The paper
Zhou, F., Zhao, W., Qi, Z., Geng, Y., Yao, S., Kendrick, K. M., Wager, T. D., & Becker, B. (2021). A distributed fMRI-based signature for the subjective experience of fear. Nature Communications. https://doi.org/10.1038/s41467-021-26977-3

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.