Dogs shown to identify human stress condition odors

By Neha MathurReviewed by Aimee MolineuxSep 30 2022

In a recent study published in PLOS ONE, researchers used a controlled olfactory paradigm to assess whether dogs could discriminate between human odors in breath and sweat samples before and after experiencing experimentally induced negative psychological stress.

Background

A stress response is the principal physiological process associated with anxiety, panic attacks, and post-traumatic stress disorder (PTSD). Given domesticated dogs' remarkable sense of smell and their closeness with humans, it is possible that they could detect changes in human physiology through odor, for example, olfactory cues associated with acute negative stress.

So far, studies have assessed dogs’ detection of human psychological states primarily via emotional contagion, a process that mirrors the emotional states of individuals.

Sundman et al., for instance, reported that the long-term cortisol level of pet dogs mirrors that of their owners. However, they could not identify the mechanisms employed by dogs to detect their owner’s stress. Perhaps, dogs picked a combination of verbal, visual, and olfactory cues. Likewise, D’Aniello et al.’s study suggested that dogs could detect human psychological states primarily from olfactory cues. Another recent study by Reeve et al. found that trained Medical Alert Dogs were most responsive to stress.

Yet, a controlled olfactory study addresses whether dogs can discriminate between human odor samples taken when not under stress and when under stress.

About the study

In the present study, researchers used a bio-detection paradigm to show how trained dogs can discriminate between odors in different samples in a controlled setting. So far, these paradigms have been most commonly used concomitantly with non-human odors, e.g., amyl-acetate, and isoamyl-acetate. Integrating these types of paradigms into the field of dogs detecting human chemosignals could be interesting.

They collected a total of 13 participants’ data remotely. To this end, they delivered sample kits to each participant’s homes, and the experimenters conducted the stress induction protocol over Microsoft Teams or Zoom. They contacted participants via email with a meeting and an online survey link.

Participants self-confirmed via a survey questionnaire that they were non-smokers and had not consumed food or drink, other than water, or any mood-altering medication, for a minimum of one hour before the meeting. Participants answered demographic questions in the survey, including their age, gender, and ethnicity.

The researchers demonstrated how to make their baseline sample by wiping a piece of gauze on the back of their neck, placing it in the vial labeled D1, and then exhaling deeply into the vial three times before securing the lid. The team asked the participants to complete a Mental Arithmetic Task (MAT), where they counted backward from 9000 in units of 17 without using paper or a pen. The task continued for three minutes, regardless of the number of correct answers.

Once sample collection was complete, they instructed participants to complete the second self-report measure, which assessed their post-task level stress using a self-report measure termed a visual analog scale (VAS). A further 40 participants completed the protocol on campus and in person, with the addition of physiological measures for 25 of 36 samples.  All samples that passed the criteria of a two-point increase in self-report stress from the self-report VAS and an increase in the mean heart rate (HR) and mean arterial pressure (blood pressure: BP) were shown to dogs.

The team trained dogs on a two-phase, three-alternative forced-choice paradigm with increasing difficulty levels. They made them discriminate against odor between two people and then for the same person at two times of the day. Each dog carried out 20 discrimination trials within each session to assess its ability to discriminate between the samples.

Dogs with above chance performance of 80%  progressed to the testing stages. This sequential method allowed researchers to ensure that if a dog’s performance dropped to chance at the testing stage, it happened because the stress and baseline samples were indistinguishable from the dog.

The researchers imposed strict odor controls. For example, they collected samples from each participant in the same room at four minutes intervals, which reduced the possibility of dogs being able to inform their indication decisions by using background volatile organic compounds (VOCs) from the air in the room.

Study findings

The dogs’ performances were consistently above chance, ranging between 90% to 96.88% during discrimination tests, with an aggregate performance of 93.75% across sessions. Also, dogs were able to discriminate between them on first exposure. No dog showed signs of distress when encountering the human stress samples. On the contrary, dogs appeared excited when they came to the stress sample in anticipation of the clicker and food reward for a correct alert.

Each participant’s samples were distinct at baseline compared to after the stress induction. Dogs also recognized that the baseline sample (D1) was distinct from what they were previously rewarded for. The dogs successfully passed the first trial of each discrimination phase session and correctly alerted to the stress sample in 94.44% of first exposure trials. They incorrectly alerted on the baseline sample in their first exposure only twice. Dogs recognized that the new stress sample (T2) was the same odor profile used in the learning trials (T1).

Conclusions

The current study evidenced that dogs can detect an odor associated with acute stress in humans from breath and sweat alone. This finding laid a strong foundation for future investigations into emotional contagion because it confirmed that an odor component to acute negative stress could be detected by dogs in the absence of visual or vocal cues.

Notably, it was a proof of principle study, with study samples comprising only four dogs; however, this did not compromise the study findings. It demonstrated that some highly trained dogs could successfully discriminate between samples from different humans. Moreover, a small number of dogs could detect odor differences in baseline and stress samples, which suggested that an odor difference existed.