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

The DOI matches its Crossref record; 65 of the 67 references it cites were found in a registry or an index; 9 of the 40 sources its claims cite were checked for support.

Chemical Fingerprints of Emotional Body Odor

Open access

Abstract

Chemical communication is common among animals. In humans, the chemical basis of social communication has remained a black box, despite psychological and neural research showing distinctive physiological, behavioral, and neural consequences of body odors emitted during emotional states like fear and happiness. We used a multidisciplinary approach to examine whether molecular cues could be associated with an emotional state in the emitter. Our research revealed that the volatile molecules transmitting different emotions to perceivers also have objectively different chemical properties. Chemical analysis of underarm sweat collected from the same donors in fearful, happy, and emotionally neutral states was conducted using untargeted two-dimensional (GC×GC) coupled with time of flight (ToF) MS-based profiling. Based on the multivariate statistical analyses, we find that the pattern of chemical volatiles (N = 1655 peaks) associated with fearful state is clearly different from that associated with (pleasant) neutral state. Happy sweat is also significantly different from the other states, chemically, but shows a bipolar pattern of overlap with fearful as well as neutral state. Candidate chemical classes associated with emotional and neutral sweat have been identified, specifically, linear aldehydes, ketones, esters, and cyclic molecules (5 rings). This research constitutes a first step toward identifying the chemical fingerprints of emotion.

Overview

CoChat Summary

TL;DR

This study used untargeted GC×GC-TOFMS profiling to test whether underarm sweat from fear, happy, and neutral states carries distinct chemical patterns. It found significant emotion-associated volatile signatures, with fear and neutral most clearly separated and happy sweat showing a bipolar pattern overlapping either fear or neutral.

Key findings

  • Across 24 male donors and 144 sweat pads, 1655 volatile peaks were analyzed; all three PLS-DA models were significant, with Neutral vs. Fear strongest (Q2 = 0.85), followed by Neutral vs. Happy (Q2 = 0.6) and Fear vs. Happy (Q2 = 0.33).
  • PCA on 94 discriminative peaks showed fear and neutral as distinct classes, while happy sweat split into subcluster HF (n = 12) overlapping fear and subcluster HN (n = 8) overlapping neutral.
  • Fear/HF sweat was high in linear aldehydes and ketones, whereas neutral/HN sweat was high in esters and cyclic molecules (5 rings); 84 of 94 peaks showed clear condition differences, and most changes were not explained by total sweat production.
  • Self-reported happiness did not differ between HF and HN subclusters (MHF = 3.79, SDHF = 0.42; MHN = 3.75, SDHN = 0.46; t(20) = 0.18, p = 0.856), so the chemical bipolarity was not explained by differences in experienced happiness.

Methods

  • Within-subjects design: 24 non-smoking Caucasian men (Mage = 22.42, SD = 3.54; range 19–34) participated in three emotion-induction sessions—fear, happy, and pleasant neutral—separated by one week, with fixed order (fear first, neutral last); happy sessions were run with donors in groups of 2–4.
  • Underarm sweat was collected on textile pads (144 pads total), stored frozen, and analyzed using headspace sorptive extraction with thermal desorption followed by GC×GC-TOFMS; ChromaTOF alignment produced 3796 peaks, cleaned to 1655 peaks and 135 pads (45 per condition).
  • Main analyses included PCA, PLS-DA with Pareto scaling, 7-fold cross-validation, permutation testing, ROC curves, box plots, and cross-validated regularized logistic regression on self-reported emotional states to explore happy subclusters.

Limitations

  • Stated: emotion-induction order was fixed and not counterbalanced; the happy condition was induced with 2–4 donors together rather than alone, so sweat volatiles may have reflected social processes rather than happiness; receiver validation was not possible because all collected sweat was used for chemical analysis; the conservative untargeted approach reveals only the strongest discriminators.
  • Stated: self-reported arousal may not have captured the proposed systematic relationship, and physiological measures such as heart rate or skin conductance would have been more appropriate.
  • Evident from the text: the sample was limited to young, non-smoking Caucasian men, so generalizability to other populations is unknown; movement was not measured, leaving its possible relation to volatile patterns unresolved.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3390/metabo10030084

Cites

65 of 65 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure A1 Receiver operating characteristic (ROC) curves of peak 2209 (left panel) and 234 (right panel) for the classifications Happy–Neutral (blue), Fear–Neutral (red), and Fear–Happy (green). For peak 2209 Neutral was the positive level for Happy–Neutral and Fear–Neutral, and Happy was the positive level for Fear–Happy. For 234 fear was the positive level for all three classifications.
Figure A1 Receiver operating characteristic (ROC) curves of peak 2209 (left panel) and 234 (right panel) for the classifications Happy–Neutral (blue), Fear–Neutral (red), and Fear–Happy (green). For peak 2209 Neutral was the positive level for Happy–Neutral and Fear–Neutral, and Happy was the positive level for Fear–Happy. For 234 fear was the positive level for all three classifications.CC BY · PubMed Central
Figure A2 Box plot of three condition classes based on the sum of 1655 peak areas per sweat sample.
Figure A2 Box plot of three condition classes based on the sum of 1655 peak areas per sweat sample.CC BY · PubMed Central
Figure A3 Example of a typical comprehensive GC×GC TOF MS contour plot of a test person, with x-axis denoting first dimension rt time in min (on non-polar column in Table A1), and y-axis denoting second dimension rt time is seconds (mid polar column in Table A1).
Figure A3 Example of a typical comprehensive GC×GC TOF MS contour plot of a test person, with x-axis denoting first dimension rt time in min (on non-polar column in Table A1), and y-axis denoting second dimension rt time is seconds (mid polar column in Table A1).CC BY · PubMed Central
Figure A4 PCA performed on 3796 RT/MS peaks and 169 pads using Pareto scaling. The scores plot (upper panel) displays the variation of the pads classified as clean (from clean pads), reference (from pads deposited in the room), and the condition samples (Neutral, Happy, Fear from the subjects). The loading plot (lower panel) displays the variation of the analytes with tentative assignments.
Figure A4 PCA performed on 3796 RT/MS peaks and 169 pads using Pareto scaling. The scores plot (upper panel) displays the variation of the pads classified as clean (from clean pads), reference (from pads deposited in the room), and the condition samples (Neutral, Happy, Fear from the subjects). The loading plot (lower panel) displays the variation of the analytes with tentative assignments.CC BY · PubMed Central
Figure 1 Effects of emotion induction (fear, neutral, happy) on self-reported emotional state: (A) Spider plot showing participants’ emotion profiles on affective circumplex [46]; (B) experienced core affect (valence, arousal) on affect grid [47]; (C) experienced discrete emotions. Error bars: ± 1 standard error of the mean (SEM).
Figure 1 Effects of emotion induction (fear, neutral, happy) on self-reported emotional state: (A) Spider plot showing participants’ emotion profiles on affective circumplex [46]; (B) experienced core affect (valence, arousal) on affect grid [47]; (C) experienced discrete emotions. Error bars: ± 1 standard error of the mean (SEM).CC BY · PubMed Central
Figure 2 Partial least squares discriminant analysis (PLS-DA) score plots (left panels) and loading plots (right panels) based on 1655 peak areas and on two condition classes at a time: Neutral versus Fear (upper panels), Neutral versus Happy (middle panels), Fear versus Happy (lower panel). The first two principal components (PCs) t[1] and t[2] are displayed. The loading plots show weighted coefficients of the principal components. The data were not normalized and Pareto-scaled prior to analysis. Q2 values > 0.2 were statistically significant.
Figure 2 Partial least squares discriminant analysis (PLS-DA) score plots (left panels) and loading plots (right panels) based on 1655 peak areas and on two condition classes at a time: Neutral versus Fear (upper panels), Neutral versus Happy (middle panels), Fear versus Happy (lower panel). The first two principal components (PCs) t[1] and t[2] are displayed. The loading plots show weighted coefficients of the principal components. The data were not normalized and Pareto-scaled prior to analysis. Q2 values > 0.2 were statistically significant.CC BY · PubMed Central
Figure 3 Principal component analysis (PCA) score plot based on 135 sweat samples and 94 peak areas that were significantly discriminating the condition groups in the PLS-DA models. The principal components t[1], t[2] are displayed, explaining R2X[1]= 0.437 and R2X[2] = 0.141 of the variances, respectively. The data were not normalized and Pareto scaling was applied.
Figure 3 Principal component analysis (PCA) score plot based on 135 sweat samples and 94 peak areas that were significantly discriminating the condition groups in the PLS-DA models. The principal components t[1], t[2] are displayed, explaining R2X[1]= 0.437 and R2X[2] = 0.141 of the variances, respectively. The data were not normalized and Pareto scaling was applied.CC BY · PubMed Central
Figure 4 Box plots showing individual peaks per condition class: Panels (A–E) show the boxplots of the peaks 146, 1850, 2578, 2019, and 1842, respectively, per condition for each of the 24 panelists as observed in underarm sweat both left and right (so 48 datapoints for each of the three conditions in each graph).
Figure 4 Box plots showing individual peaks per condition class: Panels (A–E) show the boxplots of the peaks 146, 1850, 2578, 2019, and 1842, respectively, per condition for each of the 24 panelists as observed in underarm sweat both left and right (so 48 datapoints for each of the three conditions in each graph).CC BY · PubMed Central

Verification

Claims the paper makes

7 Valid · 3 Not valid · 15 Not checked · 59 Own results and interpretations

35 more claims

Create a free account to see all 84 claims and to verify the 2 references still open.

See all claims — free account
  • It may state two separate claims as one. 27
  • Probably not a substantial claim. 23

Create a free account to see these claims.

Statistics recomputed

Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.

  • Consistentt(20) = 0.18reported p = .856 · recomputed p = .859

    “t(20) = 0.18, p = .856”

1 of 1

Self-check

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

Each contradiction between two passages listed here was confirmed by an independent check.

5 of 5

Citations

Cites

Show all 67 references

Research works printed but not linked: 1

  1. Preti, 2019, Analytical identification of stress odors from human breath, HDIAC J., 6, 10

Cited by

See all 60

Metadata

Authors
Smeets, M. A. M., Rosing, E. A. E., Jacobs, D. M., van Velzen, E., Koek, J. H., Blonk, C., Gortemaker, I., Eidhof, M. B., Markovitch, B., de Groot, J., & Semin, G. R.
Affiliations
  • Utrecht University (NL)
  • Unilever Research and Development Vlaardingen (NL)
  • Unilever R&D Vlaardingen, Olivier van Noortlaan 120, 3133 AT Vlaardingen, The Netherlands
  • Faculty of Social and Behavioural Sciences, Utrecht University, Heidelberglaan 1, 3584 CS Utrecht, The Netherlands
Corresponding author
Smeets, Monique A.M.
Venue
Metabolites
Publisher
MDPI AG
Published
2020-02-28
Volume
10
Issue
3
Pages
84
ISSN
2218-1989 (ISSN-L)
OpenAlex
W3010236843
Cited by
59
References
67
Citations by year
2026: 4 · 2025: 6 · 2024: 7 · 2023: 15 · 2022: 8 · 2021: 13 · 2020: 6
FWCI
3.30
Type
journal article
Language
en
Open access
Locations
  • Metabolites (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • Utrecht University Repository (Utrecht University) (submittedVersion, other-oa)
  • Minerva Access (University of Melbourne) (submittedVersion, cc-by)
  • Portuguese National Funding Agency for Science, Research and Technology (RCAAP Project by FCT) (submittedVersion, cc-by-nc-nd)
  • Radboud Repository (Radboud University) (submittedVersion)
  • Radboud Repository (Radboud University) (submittedVersion, other-oa)
Funders
Unilever R&D Netherlands (N/A)
Topics
  • Physical Sciences › Engineering › Biomedical Engineering › Advanced Chemical Sensor Technologies
  • Life Sciences › Neuroscience › Sensory Systems › Olfactory and Sensory Function Studies
  • Physical Sciences › Computer Science › Artificial Intelligence › Emotion and Mood Recognition
Keywords
chemical communication, fear, happiness, volatile organic compounds, emotional state, linear aldehydes, ketones, esters, cyclic molecules

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.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (qwen3.8-omni-flash and gpt-6-sol).

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: Chemical Fingerprints of Emotional Body Odor. https://app.cochat.ai/papers/10.3390/metabo10030084
The paper
Smeets, M. A. M., Rosing, E. A. E., Jacobs, D. M., van Velzen, E., Koek, J. H., Blonk, C., Gortemaker, I., Eidhof, M. B., Markovitch, B., de Groot, J., & Semin, G. R. (2020). Chemical Fingerprints of Emotional Body Odor. Metabolites. https://doi.org/10.3390/metabo10030084

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.