The combination of volatile compounds in underarm sweat differed after watching frightening films and calm footage. After enjoyable films, however, the compounds did not follow a single pattern: some people’s samples resembled those from the fear condition, while others resembled those from the calm condition.
The 2020 study investigated the chemical characteristics of body odor associated with emotion. Rather than looking only for a single compound, it examined patterns related to emotional states by considering increases and decreases across many compounds together.
Collecting sweat in three moods from the same 24 people
The participants were 24 nonsmoking White men aged 19–34. They watched approximately 30 minutes of footage intended to induce fear, happiness, or a calm, mildly pleasant neutral state. Sessions were a week apart, and everyone participated in the order fear, happiness, then neutral. Only in the happiness condition did participants sit together in groups of two to four, to help elicit a happy mood.
Diet, fragranced products, and other potential sources of contamination were restricted for two days before collection. Absorbent pads were attached to the left and right underarms of cleaned T-shirts. After collection, the pads were frozen, and the compounds released from them were captured and examined using two-dimensional gas chromatography and mass spectrometry. Questionnaire responses also confirmed the mood induction: fear was higher in the fear condition and happiness higher in the happiness condition.
The analysis retained 135 pads and 1,655 analytical peaks. A peak is a chemical feature detected by the instrument; this does not mean that the identity of that many different molecules was conclusively established. The study first looked for combinations that distinguished the conditions, then grouped candidate compounds by chemical type.
The clearest distinction was between fear and calm
In statistical models using combinations of compounds, the distinction between fear and neutral was clearest. Happiness versus neutral and happiness versus fear also differed statistically, but discrimination became weaker in that order. Cross-validation, which divides the data to check performance within the model, was also conducted.
A closer examination of the 94 peaks involved in distinguishing conditions showed that many increased or decreased in both fear and happiness relative to neutral. Some changes were specific to fear or to happiness, so the pattern was not simply one compound appearing for each emotion.
The paper reported a Q² model statistic of 0.85 for fear versus neutral, but this does not mean that the model identified people’s moods with 85% accuracy. This was an exploratory analysis that treated left and right underarms from the same participant as separate samples. Predictive performance has not yet been tested in a new group of people.
A pattern higher in aldehydes and another higher in esters
Organizing the chemical candidates revealed a tendency for signals from linear aldehydes, ketones, and related compounds to be higher during fear, and for esters and some cyclic compounds to be higher in the neutral condition. Specific candidates included hexanal and octanal in the former group, and ethyl acetate and propyl acetate in the latter.
Here, higher refers to a comparison of instrument signals. The study did not directly measure perceived odor intensity or a compound’s ability to convey emotion. Fatty acids associated with ordinary underarm odor were also detected, but were not among the 94 peaks that distinguished the conditions. Compounds that make body odor noticeable and candidate compounds that reflect differences in state do not necessarily overlap.
Why happy sweat split into two patterns remains unclear
Samples from the happiness condition included a fear-like pattern with higher aldehyde and ketone signals, and a neutral-like pattern with higher signals from esters and related compounds. The researchers considered whether this might reflect excited versus calm happiness, and examined associations with questionnaire responses.
However, reported happiness was nearly identical in the two patterns, and self-reported arousal and other emotions did not adequately explain the split. Future work will need to include physiological measures such as heart rate and physical movement. The subgroup counts also differ between the paper’s text and table, so it is more appropriate to focus on the two reported tendencies than on detailed comparisons of their sizes.
How can emotion-related changes be tested further?
Condition order was fixed, and only the happiness condition included other people sitting together. The effects of becoming familiar with the experiment or being with others therefore cannot be fully separated from emotion itself. Participants were also limited to young men. Replication will require studies that vary the order and match the social conditions.
All collected samples were used for chemical analysis; the researchers did not test them by having other people smell them. Combining candidate compounds and testing whether they actually change a recipient’s responses could connect the chemical patterns with emotional communication. The paper also discusses sweat production, but the measure compared there was the sum of analytical peak areas. This should be distinguished from weighing the sweat itself.
Summary
Underarm sweat collected from the same men showed a chemical pattern with relatively higher aldehyde and ketone signals during fear, and higher ester and some cyclic-compound signals in the calm neutral condition. Happiness split into two patterns resembling one or the other, a difference that self-reported happiness and arousal did not fully explain.
The findings provide clues for investigating emotion-related body odor as combinations of compounds. They point toward further experiments to establish which molecules provide information to other people.
The research was funded by Unilever R&D. The paper states that authors employed by the company at the time participated in chemical analysis and related statistical analyses, while Utrecht University authors led study design, sweat collection, and analysis of self-report data. The authors declared no potential conflicts of interest.
References
Smeets, M. A. M., et al. (2020). Chemical Fingerprints of Emotional Body Odor. Metabolites, 10(3), 84. https://doi.org/10.3390/metabo10030084 (Open-access full text)
