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

The DOI matches its Crossref record; 68 of the 68 references it cites were found in a registry or an index.

Computations of uncertainty mediate acute stress responses in humans

Open access
de Berker, Archy O., Rutledge, Robb B., Mathys, Christoph, Marshall, Louise, Cross, Gemma F., Dolan, Raymond J., Bestmann, Sven · 2016 · Nature Communications

Abstract

The effects of stress are frequently studied, yet its proximal causes remain unclear. Here we demonstrate that subjective estimates of uncertainty predict the dynamics of subjective and physiological stress responses. Subjects learned a probabilistic mapping between visual stimuli and electric shocks. Salivary cortisol confirmed that our stressor elicited changes in endocrine activity. Using a hierarchical Bayesian learning model, we quantified the relationship between the different forms of subjective task uncertainty and acute stress responses. Subjective stress, pupil diameter and skin conductance all tracked the evolution of irreducible uncertainty. We observed a coupling between emotional and somatic state, with subjective and physiological tuning to uncertainty tightly correlated. Furthermore, the uncertainty tuning of subjective and physiological stress predicted individual task performance, consistent with an adaptive role for stress in learning under uncertain threat. Our finding that stress responses are tuned to environmental uncertainty provides new insight into their generation and likely adaptive function.

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/ncomms10996

Cites

68 of 68 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Task structure and stress measures. (a) Learning task. Visual stimuli (rocks) were probabilistically associated with outcomes (snake or no snake). Subjects made a prediction of the outcome on each trial. The appearance of a snake was accompanied by the delivery of a painful electric shock. (b) Example trial. Here the participant incorrectly predicts no snake. Timing was jittered; see Methods. (c) The probabilities governing stimulus–outcome relationships shifted unpredictably over time, producing fluctuations in uncertainty. (d) Subjective stress ratings were collected every four to six trials. Measures of skin conductance (n=45) and pupil dilatation (n=22) were collected in some subjects.
Figure 1 Task structure and stress measures. (a) Learning task. Visual stimuli (rocks) were probabilistically associated with outcomes (snake or no snake). Subjects made a prediction of the outcome on each trial. The appearance of a snake was accompanied by the delivery of a painful electric shock. (b) Example trial. Here the participant incorrectly predicts no snake. Timing was jittered; see Methods. (c) The probabilities governing stimulus–outcome relationships shifted unpredictably over time, producing fluctuations in uncertainty. (d) Subjective stress ratings were collected every four to six trials. Measures of skin conductance (n=45) and pupil dilatation (n=22) were collected in some subjects.CC BY · PubMed Central
Figure 2 Modelling of learning and stress. (a) Hierarchical Gaussian Filter model21. Beliefs are represented in probability distributions organized in a hierarchy, with the speed of updating at each level influenced by the estimate at the level above. This allows learning to occur more quickly in volatile environments. Each level is Gaussian, characterized by a mean (μ) and a variance (σ), which corresponds to uncertainty. These representations unfold over time, with the model furnishing an estimate at each level, for each trial. We take these dynamic representations of uncertainty from this model (b) and use them to predict stress responses (c) using linear modelling (d). The resultant regression coefficients (β1−n) quantify the influence of each form of uncertainty on that stress measure. Note that X1−n are the regressors in the linear model, which include but are not limited to uncertainty trajectories; we also include terms such as number of shocks, and nuisance variables such as gaze co-ordinates.
Figure 2 Modelling of learning and stress. (a) Hierarchical Gaussian Filter model21. Beliefs are represented in probability distributions organized in a hierarchy, with the speed of updating at each level influenced by the estimate at the level above. This allows learning to occur more quickly in volatile environments. Each level is Gaussian, characterized by a mean (μ) and a variance (σ), which corresponds to uncertainty. These representations unfold over time, with the model furnishing an estimate at each level, for each trial. We take these dynamic representations of uncertainty from this model (b) and use them to predict stress responses (c) using linear modelling (d). The resultant regression coefficients (β1−n) quantify the influence of each form of uncertainty on that stress measure. Note that X1−n are the regressors in the linear model, which include but are not limited to uncertainty trajectories; we also include terms such as number of shocks, and nuisance variables such as gaze co-ordinates.CC BY · PubMed Central
Figure 3 Assessing models of learning. (a) Random-Effects Bayesian Model Comparison confirmed that the Hierarchical Gaussian Filter (HGF) outperformed fixed-learning-rate models (Rescorla–Wagner) and variable-learning-rate non-Bayesian models (Sutton K1). (b) Life stress was assessed with a Perceived Stress Scale38. Life stress scores were correlated with the metavolatility parameter (ϑ) in the HGF, suggesting that more stressed individuals believe the world to be less stable (n=41; Spearman ρ=0.38, P=0.014).
Figure 3 Assessing models of learning. (a) Random-Effects Bayesian Model Comparison confirmed that the Hierarchical Gaussian Filter (HGF) outperformed fixed-learning-rate models (Rescorla–Wagner) and variable-learning-rate non-Bayesian models (Sutton K1). (b) Life stress was assessed with a Perceived Stress Scale38. Life stress scores were correlated with the metavolatility parameter (ϑ) in the HGF, suggesting that more stressed individuals believe the world to be less stable (n=41; Spearman ρ=0.38, P=0.014).CC BY · PubMed Central
Figure 4 Irreducible uncertainty predicts subjective stress. (a) Example subjective stress trajectory for one participant (blue) and irreducible uncertainty estimates for that individual (red). (b) Comparison of regression models of subjective stress. All models shared two components: the value of the previous rating and the number of shocks delivered since the last rating. The surprise model summed the surprise (|δ1|) for each outcome since the previous rating, while models σ1–3 included the estimated uncertainty at each level of the Hierarchical Gaussian Filter at the time of rating. Irreducible uncertainty () provided the best fit to our participants (n=45). (c) Irreducible uncertainty predicts prediction response times. A curve describing the variance of a Bernoulli distribution representing beliefs about probabilities, corresponding to irreducible uncertainty, predicts average response times (Pearson r=0.99, P<0.001). (d) The winning regression model predicting subjective stress responses (mean r2=0.25). Shocks and irreducible uncertainty both predicted subjective stress ratings (single-sample t-tests, P<0.001; P=0.0024). Error bars represent s.e.m.
Figure 4 Irreducible uncertainty predicts subjective stress. (a) Example subjective stress trajectory for one participant (blue) and irreducible uncertainty estimates for that individual (red). (b) Comparison of regression models of subjective stress. All models shared two components: the value of the previous rating and the number of shocks delivered since the last rating. The surprise model summed the surprise (|δ1|) for each outcome since the previous rating, while models σ1–3 included the estimated uncertainty at each level of the Hierarchical Gaussian Filter at the time of rating. Irreducible uncertainty () provided the best fit to our participants (n=45). (c) Irreducible uncertainty predicts prediction response times. A curve describing the variance of a Bernoulli distribution representing beliefs about probabilities, corresponding to irreducible uncertainty, predicts average response times (Pearson r=0.99, P<0.001). (d) The winning regression model predicting subjective stress responses (mean r2=0.25). Shocks and irreducible uncertainty both predicted subjective stress ratings (single-sample t-tests, P<0.001; P=0.0024). Error bars represent s.e.m.CC BY · PubMed Central
Figure 5 Physiological responses reflect uncertainty and surprise. (a) Median splits indicate that both irreducible uncertainty and surprise increase pupil diameter (both P<0.001). Shading is s.e.m. across participants. (b) Baseline pupil diameter and skin conductance on each trial displayed a clear inverted-U relationship with belief, as seen for subjective stress. A curve describing the variance of a Bernoulli distribution fit well (Pearson correlations: pupil diameter: r=0.96, P<0.001; skin conductance: r=0.84, P=0.002). Error bars represent s.e.m. (c) A multiple regression model demonstrated a role for uncertainty throughout the trial for both pupil diameter (robust regression β=0.11, single-sample t-test, t21=4.72, P<0.001) and skin conductance (robust regression β=0.044, single-sample t-test, t36=2.25, P=0.031). Error bars represent s.e.m. (d) Across subjects, the sensitivity of subjective stress to irreducible uncertainty correlated with the sensitivity observed in pupil diameter (Pearson correlation, n=22, r=0.52, P=0.013) and skin conductance (Pearson correlation, n=37, r=0.38, P=0.021) models. Each data point is one participant.
Figure 5 Physiological responses reflect uncertainty and surprise. (a) Median splits indicate that both irreducible uncertainty and surprise increase pupil diameter (both P<0.001). Shading is s.e.m. across participants. (b) Baseline pupil diameter and skin conductance on each trial displayed a clear inverted-U relationship with belief, as seen for subjective stress. A curve describing the variance of a Bernoulli distribution fit well (Pearson correlations: pupil diameter: r=0.96, P<0.001; skin conductance: r=0.84, P=0.002). Error bars represent s.e.m. (c) A multiple regression model demonstrated a role for uncertainty throughout the trial for both pupil diameter (robust regression β=0.11, single-sample t-test, t21=4.72, P<0.001) and skin conductance (robust regression β=0.044, single-sample t-test, t36=2.25, P=0.031). Error bars represent s.e.m. (d) Across subjects, the sensitivity of subjective stress to irreducible uncertainty correlated with the sensitivity observed in pupil diameter (Pearson correlation, n=22, r=0.52, P=0.013) and skin conductance (Pearson correlation, n=37, r=0.38, P=0.021) models. Each data point is one participant.CC BY · PubMed Central
Figure 6 Relationship between uncertainty sensitivity and task performance. (a) Subjective stress sensitivity (the regression coefficient for uncertainty in the subjective stress model) correlated with how frequently participants predicted the correct outcome (Pearson correlation, n=45, r=0.37, P=0.012). (b) Pupillary sensitivity to uncertainty also predicted performance (Pearson correlation, n=22, r=0.62, P=0.0023). Each data point is one participant.
Figure 6 Relationship between uncertainty sensitivity and task performance. (a) Subjective stress sensitivity (the regression coefficient for uncertainty in the subjective stress model) correlated with how frequently participants predicted the correct outcome (Pearson correlation, n=45, r=0.37, P=0.012). (b) Pupillary sensitivity to uncertainty also predicted performance (Pearson correlation, n=22, r=0.62, P=0.0023). Each data point is one participant.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

Show all 68 references

Cited by

See all 330

Metadata

Authors
de Berker, A. O., Rutledge, R. B., Mathys, C., Marshall, L., Cross, G. F., Dolan, R. J., & Bestmann, S.
Affiliations
  • Wellcome Centre for Human Neuroimaging (GB)
  • UCL Queen Square Institute of Neurology (GB)
  • King's College Hospital (GB)
Venue
Nature Communications
Published
2016-03-29
Volume
7
Issue
1
Pages
10996-10996
ISSN
2041-1723 (ISSN-L)
OpenAlex
W2314770590
Cited by
330
References
68
Citations by year
2026: 20 · 2025: 35 · 2024: 44 · 2023: 32 · 2022: 41 · 2021: 32 · 2020: 40 · 2019: 30 · 2018: 24 · 2017: 23 · 2016: 9
FWCI
18.31
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Communications (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • SDL (submittedVersion, cc-by)
  • RePEc: Research Papers in Economics (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • UCL Discovery (University College London) (submittedVersion)
  • MPG.PuRe (Max Planck Society) (submittedVersion)
Funders
  • Wellcome Trust (091593, 091593/Z/10/, 098362, 098362/Z/12/Z, 091593/Z/10/Z, 098362/Z/12/)
  • Max-Planck-Gesellschaft (098362/Z/12/Z)
  • Medical Research Council
  • FP7 Ideas: European Research Council (260424)
Topics
  • Life Sciences › Neuroscience › Behavioral Neuroscience › Stress Responses and Cortisol
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Memory and Neural Mechanisms
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Neural and Behavioral Psychology Studies
Keywords
acute stress response, uncertainty, subjective stress, salivary cortisol, pupil size, skin conductance, probabilistic learning, threat learning, Bayesian hierarchical model
MeSH terms
Bayes Theorem, Humans, Learning, Models, Biological, Stress, Psychological, Task Performance and Analysis, Uncertainty

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: Computations of uncertainty mediate acute stress responses in humans. https://app.cochat.ai/papers/10.1038/ncomms10996
The paper
de Berker, A. O., Rutledge, R. B., Mathys, C., Marshall, L., Cross, G. F., Dolan, R. J., & Bestmann, S. (2016). Computations of uncertainty mediate acute stress responses in humans. Nature Communications. https://doi.org/10.1038/ncomms10996

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.