CoChat

A Bayesian Model of Category-Specific Emotional Brain Responses

Open access

The DOI matches its Crossref record; 115 of the 117 references it cites were found in a registry or an index.

Abstract

Understanding emotion is critical for a science of healthy and disordered brain function, but the neurophysiological basis of emotional experience is still poorly understood. We analyzed human brain activity patterns from 148 studies of emotion categories (2159 total participants) using a novel hierarchical Bayesian model. The model allowed us to classify which of five categories--fear, anger, disgust, sadness, or happiness--is engaged by a study with 66% accuracy (43-86% across categories). Analyses of the activity patterns encoded in the model revealed that each emotion category is associated with unique, prototypical patterns of activity across multiple brain systems including the cortex, thalamus, amygdala, and other structures. The results indicate that emotion categories are not contained within any one region or system, but are represented as configurations across multiple brain networks. The model provides a precise summary of the prototypical patterns for each emotion category, and demonstrates that a sufficient characterization of emotion categories relies on (a) differential patterns of involvement in neocortical systems that differ between humans and other species, and (b) distinctive patterns of cortical-subcortical interactions. Thus, these findings are incompatible with several contemporary theories of emotion, including those that emphasize emotion-dedicated brain systems and those that propose emotion is localized primarily in subcortical activity. They are consistent with componential and constructionist views, which propose that emotions are differentiated by a combination of perceptual, mnemonic, prospective, and motivational elements. Such brain-based models of emotion provide a foundation for new translational and clinical approaches.

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 · 10.1371/journal.pcbi.1004066

Cites

115 of 115 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig 1 Classification of emotion category using the Bayesian Spatial Point Process model. A) A schematic of the method, which models the population density of activation across the brain with a sparse set of multivariate Gaussian distributions at two levels (study center and population center). The intensity function map summarizes the expected frequency of activation conditional on an emotion category. The model also represents the joint activation across multiple brain regions, which is not captured in the intensity map. The model can also be used for classification by calculating the conditional likelihood of each emotion category given a set of foci using Bayes’ rule. B) Confusion matrix for the 5-way classification of emotion category based on the model. Diagonal entries reflect classification accuracy. C) The intensity maps for each of the 5 emotion categories. Intensity maps are continuous over space, and their integral over any portion of the brain reflects the expected number of activation centers in that area for all studies with a particular emotion. The maps are thresholded for display at a voxel-wise intensity of 0.001 or above.
Fig 1 Classification of emotion category using the Bayesian Spatial Point Process model. A) A schematic of the method, which models the population density of activation across the brain with a sparse set of multivariate Gaussian distributions at two levels (study center and population center). The intensity function map summarizes the expected frequency of activation conditional on an emotion category. The model also represents the joint activation across multiple brain regions, which is not captured in the intensity map. The model can also be used for classification by calculating the conditional likelihood of each emotion category given a set of foci using Bayes’ rule. B) Confusion matrix for the 5-way classification of emotion category based on the model. Diagonal entries reflect classification accuracy. C) The intensity maps for each of the 5 emotion categories. Intensity maps are continuous over space, and their integral over any portion of the brain reflects the expected number of activation centers in that area for all studies with a particular emotion. The maps are thresholded for display at a voxel-wise intensity of 0.001 or above.CC BY · PubMed Central
Fig 2 Emotion-predictive patterns of activity across cortical networks and subcortical regions. A) Left: Seven resting-state connectivity networks from the Buckner Lab with cortical, basal ganglia, and cerebellar components. Colors reflect the network membership. Right: Published anatomical parcellations were used to supplement the resting-state networks to identify sub-regions in amygdala (131), hippocampus (131, 132), and thalamus (133). dAN: dorsal attention network; Def: default mode network; FPN: fronto-parietal network; Limbic: limbic network; SMN: somatomotor network; vAN: ventral attention network; Vis: visual network. B) The profile of activation intensity across the 7 cortical and basal ganglia resting-state networks, and anatomical amygdalar and thalamic regions. Colors indicate different emotion categories, as in Fig. 1. Red: anger; green: disgust; purple: fear; yellow: happiness; blue: sadness. Values farther toward the solid circle indicate greater average intensity in the network (i.e., more expected study centers). C) Two canonical patterns estimated using non-negative matrix factorization, and the distribution of intensity values for each emotion across the two canonical patterns. The colored area shows the 95% joint confidence interval (confidence ellipsoids) derived from the 10,000 Markov chain Monte Carlo samples in the Bayesian model. Non-overlapping confidence ellipsoids indicate significant differences across categories in the expression of each profile.
Fig 2 Emotion-predictive patterns of activity across cortical networks and subcortical regions. A) Left: Seven resting-state connectivity networks from the Buckner Lab with cortical, basal ganglia, and cerebellar components. Colors reflect the network membership. Right: Published anatomical parcellations were used to supplement the resting-state networks to identify sub-regions in amygdala (131), hippocampus (131, 132), and thalamus (133). dAN: dorsal attention network; Def: default mode network; FPN: fronto-parietal network; Limbic: limbic network; SMN: somatomotor network; vAN: ventral attention network; Vis: visual network. B) The profile of activation intensity across the 7 cortical and basal ganglia resting-state networks, and anatomical amygdalar and thalamic regions. Colors indicate different emotion categories, as in Fig. 1. Red: anger; green: disgust; purple: fear; yellow: happiness; blue: sadness. Values farther toward the solid circle indicate greater average intensity in the network (i.e., more expected study centers). C) Two canonical patterns estimated using non-negative matrix factorization, and the distribution of intensity values for each emotion across the two canonical patterns. The colored area shows the 95% joint confidence interval (confidence ellipsoids) derived from the 10,000 Markov chain Monte Carlo samples in the Bayesian model. Non-overlapping confidence ellipsoids indicate significant differences across categories in the expression of each profile.CC BY · PubMed Central
Fig 3 Co-activation graphs for each emotion category. A) Force-directed graphs for each emotion category, based on the Fruchterman-Reingold spring algorithm (134). The nodes (circles) are regions or networks, color-coded by anatomical system. The edges (lines) reflect co-activation between pairs of regions or networks, assessed based on the joint distribution of activation intensity in the Bayesian model (Pearson’s r across all MCMC iterations) and thresholded at P <. 05 corrected based on a permutation test. The size of each circle reflects its betweenness-centrality (48, 49), a measure of how strongly it connects disparate networks. (B) The same connections in the anatomical space of the brain. One location is depicted for each cortical network for visualization purposes, though the networks were distributed across regions (see Fig 3A). C) Global network efficiency (see refs. (135, 136)) within (diagonal elements) and between (off-diagonals) brain systems. Global efficiency (135, 136) is defined as the inverse of the average minimum path length between all members of each group of regions/nodes. Minimum path length is the minimum number of intervening nodes that must be traversed to reach one node from another, counting only paths with statistically significant associations and with distance values proportional to (2—Pearson’s r), rather than binary values, to better reflect the actual co-activation values. Higher efficiency reflects more direct relationships among the systems. Values of 0 indicates disjoint systems, with no significant co-activation paths connecting any pair of regions/networks, and values of 1 indicate the upper bound of efficiency, with a perfect association between each pair of regions. Co-activation is related to connectivity and network integration, though all fMRI-based connectivity measures only indirectly reflect actual neural connections. Efficiency is related to the average correlation among regions (r = 0.76) but not the average intensity (r = 0.02; see S5 Fig).
Fig 3 Co-activation graphs for each emotion category. A) Force-directed graphs for each emotion category, based on the Fruchterman-Reingold spring algorithm (134). The nodes (circles) are regions or networks, color-coded by anatomical system. The edges (lines) reflect co-activation between pairs of regions or networks, assessed based on the joint distribution of activation intensity in the Bayesian model (Pearson’s r across all MCMC iterations) and thresholded at P <. 05 corrected based on a permutation test. The size of each circle reflects its betweenness-centrality (48, 49), a measure of how strongly it connects disparate networks. (B) The same connections in the anatomical space of the brain. One location is depicted for each cortical network for visualization purposes, though the networks were distributed across regions (see Fig 3A). C) Global network efficiency (see refs. (135, 136)) within (diagonal elements) and between (off-diagonals) brain systems. Global efficiency (135, 136) is defined as the inverse of the average minimum path length between all members of each group of regions/nodes. Minimum path length is the minimum number of intervening nodes that must be traversed to reach one node from another, counting only paths with statistically significant associations and with distance values proportional to (2—Pearson’s r), rather than binary values, to better reflect the actual co-activation values. Higher efficiency reflects more direct relationships among the systems. Values of 0 indicates disjoint systems, with no significant co-activation paths connecting any pair of regions/networks, and values of 1 indicate the upper bound of efficiency, with a perfect association between each pair of regions. Co-activation is related to connectivity and network integration, though all fMRI-based connectivity measures only indirectly reflect actual neural connections. Efficiency is related to the average correlation among regions (r = 0.76) but not the average intensity (r = 0.02; see S5 Fig).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

Research works printed but not linked: 2

  1. J Price, 1999, Prefrontal cortical networks related to visceral function and mood, Annals of the New York Academy of Sciences
  2. DD Lee, 1999, Learning the parts of objects by non-negative matrix factorization, Nature, 401, 788, 10.1038/44565 · DOI · Paper page

Cited by

See all 265

Metadata

Authors
Wager, T. D., Kang, J., Johnson, T. D., Nichols, T. E., Satpute, A. B., & Barrett, L. F.
Affiliations
  • University of Colorado Boulder (US)
  • Emory University (US)
  • University of Michigan (US)
  • University of Warwick (GB)
  • Northeastern University (US)
Corresponding author
Wager, Tor D.
Venue
PLOS Computational Biology
Published
2015-04-08
Volume
11
Issue
4
Pages
e1004066
ISSN
  • 1553-734X (ISSN-L)
  • 1553-7358
OpenAlex
W2099078585
Cited by
262
References
117
Citations by year
2026: 11 · 2025: 20 · 2024: 21 · 2023: 30 · 2022: 31 · 2021: 25 · 2020: 27 · 2019: 21 · 2018: 18 · 2017: 34 · 2016: 17 · 2015: 6
FWCI
2.49
Type
journal article
Language
en
Open access
Locations
  • PLoS Computational Biology (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Digital Access to Scholarship at Harvard (DASH) (Harvard University) (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • Figshare (submittedVersion, cc-by)
Funders
  • Wellcome Trust (100309/A/12/Z, 100309)
  • National Institutes of Health (DP1OD003312)
  • National Institute of Mental Health (R01MH105561, R01MH076136)
  • National Institute of Neurological Disorders and Stroke (R01NS075066)
  • Army Research Institute for the Behavioral and Social Sciences (W5J9CQ-11-C-0046)
Topics
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Neural and Behavioral Psychology Studies
  • Physical Sciences › Computer Science › Artificial Intelligence › Emotion and Mood Recognition
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Memory and Neural Mechanisms
Keywords
emotion classification, Bayesian hierarchical model, fear, anger, disgust, sadness, happiness, amygdala, thalamus, cortical-subcortical interactions, brain networks
MeSH terms
Bayes Theorem, Brain, Computer Simulation, Emotions, Humans, Models, Neurological, Nerve Net, Mental Recall, Models, Statistical

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

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.