06.23 The Smell of Stress

Highlights

Humans and dogs have worked together for thousands of years. In addition to their role as pets, dogs can be trained to support humans in a variety of tasks ranging from search and rescue to detecting diseases with their keen sense of smell.
These researchers demonstrated that dogs can smell the difference between someone in a relaxed state and someone who recently experienced stress.
The results of these experiments improve our understanding of canine smell detection. Next steps include further research to identify the biological mechanisms behind dogs’ sense of smell.

According to the American Veterinary Association, over one-third of American households include a pet dog. Research has shown multiple health benefits to dog ownership including improvements in mental health and heart health. In addition to the unconditional love and companionship dogs provide, they can also assist humans in their daily life, as they have done for thousands of years. Think of hunting dogs that help humans catch food, guard dogs that keep people and property safe, and sheep dogs that guard and protect the herd. In modern times, working dogs can choose from a variety of careers according to their skills and interests. For example, therapy dogs are trained to provide emotional support to groups of people who might be lonely or stressed, like elders in nursing homes or college students. Service dogs undergo intensive training to provide a specialized service to their owner, such as seeing eye dogs for people with vision impairments.

Additional work opportunities for dogs rely on their keen olfaction, or sense of smell. Research shows that dogs are able to smell vastly better than humans. Detection dogs can be trained to hone this sense of smell to identify specific substances like money or explosives. Ongoing research efforts are trying to train dogs to recognize disease states by smell such as cancer or COVID-19 infection. Better understanding of how exactly dogs smell and what they are sensing when they rely on smell to distinguish disease states could allow for the development of new diagnostic tools for these diseases. New research by Dr. Clara Wilson, Dr. Catherine Reeve, and colleagues at Queen's University Belfast in Northern Ireland, United Kingdom demonstrates that dogs can be trained to recognize the smell of human stress.

Dr. Clara Wilson grew up around service dogs. Her grandfather has been blind since the age of thirteen and has worked with eight seeing eye dogs over the course of his life. Watching the relationships her grandfather had with his service dogs sparked Dr. Wilson’s interest in animal behavior. In college she was exposed to the field of animal cognition, which led to her graduate work in dog behavior described here, conducted together with her advisor Dr. Catherine Reeve.

Previous researchers had developed a model for training dogs to detect disease states such as high blood sugar or low blood sugar, which can be dangerous for people with diabetes. Dr. Wilson and Dr. Reeve wanted to expand this work to new areas. They wondered whether dogs could detect emotional states, such as stress.

Stress is an important topic in health research because ongoing or chronic stress can be related to a variety of disease states. Stress can be thought of as a change in the internal or external environment that you need to respond to. The stress response is physiological and psychological and can affect many parts of the body including heart rate, blood pressure, breathing, appetite, and hormone levels.

Recently, it has been shown that dogs can detect a variety of human diseases from smell alone. However, whether dogs’ capabilities extend to detecting odors associated with psychological states has been explored far less. Dr. Reeve and Dr. Wilson set out to determine whether dogs could detect a difference between a person in a relaxed state and that same person while stressed.

For this study, the researchers used volunteer dogs from the community. They advertised for volunteer dogs who were high energy and might benefit from some extra training. Over twenty volunteer dogs and their owners came to the laboratory for an initial evaluation. Canine participants had to be motivated by food rewards for training, and they needed to be able to be left with the researchers, away from their owners, for the duration of the training session. In addition, the researchers only worked with dogs who demonstrated that they enjoyed their time training and did not show any behavioral signs of distress.

Each dog came to the laboratory for training for one hour per week over the course of several months. Dr. Wilson and Dr. Reeve trained these dogs using a model called the canine bio-detection paradigm, which has been used to train dogs to detect a variety of health conditions such as seizures, low blood sugar, different types of cancers, and COVID-19 infection. In this training paradigm, the dogs were first trained to recognize a specific odor, called the target odor. Next, they were exposed to a lineup of different odors, including the target odor, and asked to identify which was the target. Dr. Wilson was assisted by Kerry Campbell, a master’s student.

The training was divided into four stages. All stages utilized a specially designed apparatus with a port, or opening, for three vials.

In order to move from one stage to the next, each dog needed to complete a minimum of two sessions of training where they performed the task correctly at least 80% of the time. Dogs who could not successfully complete a task did not continue with the training protocol.

In stage one, a treat was left in one port while the other two ports were completely empty. The dog was asked to identify the port with the treat. Every dog should be able to find a piece of food – if they can’t, it probably isn’t because they can’t smell it but that they likely aren’t interested or motivated to do so. Dogs whose behavior indicated that they were not interested in the training paradigm were screened out for further training work.

In stage two, the task became much more challenging, though it was still something the researchers knew dogs could do. The researchers placed a glass vial containing medical gauze in each port. Two of the ports contained breath samples from two different people. The third port contained a blank, meaning that it included the same glass vial with gauze, but the gauze had no odor. Using blanks is a critical part of the training protocol because it allows the researchers to know that the dogs are behaving correctly due to their sense of smell and not due to differences they detect in the environment or equipment.

The most difficult part of the task in stage two is for the trainer to communicate instructions to the dogs. To do this, the first ten trials of the session used only the target breath sample in one port and two blanks. In this way, the dogs learned which scent they were looking for.

After the first ten trials were complete, the next twenty trials included two vials with breath samples from two different people, one the target odor and one as a distraction, and one blank. In order to successfully complete the task, the dog had to identify which of the three ports contained the target scent as taught in the first ten trials.

In stage three, the dogs were asked to tell the difference between the breath of the same person taken at different times of the day, once early in the morning and once later in the afternoon. This is more difficult for dogs because the samples come from the same person, so many of the background odors are the same. Still, there are some changes for the dogs to notice. For example, changes associated with the time of day, food consumed during the day, and hormone levels. This was the final stage before testing and was a closer approximation to the final task, but still no stress was involved. If the dogs were able to consistently smell these samples apart, they could progress to testing.

Of the initial twenty dog volunteers, four made it to the final stage of training to participate in the experiment: Soot (as seen above), Fingal, Winnie, and Treo. In the fourth and final stage, the dogs were asked to distinguish between breath and sweat samples taken from the same person while relaxed and four minutes later, immediately after the person completed the stressful Mental Arithmetic Task (MAT) described below.The goal was that nothing else would have changed in those four minutes, as the samples were taken in the same room and only minutes apart. The researchers expected the background odors would be the same, and any differences would be associated with the psychological stress response caused by the stressful task.

The samples were collected from 53 healthy volunteers. The researchers recruited volunteers across campus using email, social media, and word of mouth. A total of 40 participants came to the laboratory to provide samples. An additional 13 participants provided samples remotely due to COVID-19 restrictions. Smokers were not eligible to participate because the odor of the smoke can mask other scents. All participants were asked to refrain from eating and drinking any flavored beverages for one hour prior to sample collection, to avoid these smells overpowering any changes associated with stress.

In advertising the study, the researchers shared that participants would be asked to complete a mild stress-inducing task. They did not share the specific type of task in order to make sure volunteers weren’t self-selecting based on their enjoyment or fear of math. If volunteers knew the task involved math, the researchers thought that people who would be stressed by the task would not sign up (knowing that math is stressful) and people who enjoyed math (and wouldn’t be stressed out by the task) would. In order to collect the samples for the training protocol, they needed the volunteers to feel mild stress.

In the laboratory, participants completed a survey on the computer in the study room including a 10-point stress scale to indicate their level of stress at that moment.

After filling out the survey, the researchers collected physiological measurements of stress including heart rate and blood pressure. Once the measurements were taken, participants provided a sweat and breath sample. Participants were given medical grade gauze and asked to swab the back of their neck and place the swab in a test tube. They then blew into the test tube four times.

Next, the participants were asked to complete a Mental Arithmetic Task (MAT). In this task, they were asked to count backwards from 9,000 by 17 without using pen and paper or a calculator. This took place in front of the researchers, who would pressure participants by saying phrases such as “it is very important that you perform the task as quickly and efficiently as possible” and “you must keep going until the test is completed.” The task lasted for a total of three minutes. The second sweat and breath sample and the second set of physiological measurements were taken after the task had been completed. Participants were again asked to rate their level of stress on a 10-point scale.

The two breath and sweat samples were shown to a dog within three hours of collection. Samples had to meet specific criteria for use:

If a participant did not meet these criteria, their samples were not used, as the protocol had failed to induce stress in those individuals. About half of the samples collected from volunteers were used in the study.

Before beginning this experiment, the researchers did not know whether the dogs would successfully be able to complete the task. However, they trusted that the dogs were very well trained. By the final stage of testing, if the results decreased to what could be expected by chance, which would indicate that the dogs were guessing, the researchers could reasonably conclude the dogs could not tell the difference between the same person’s relaxed and stressed breath/sweat samples. Somewhat to the researcher’s surprise, the results clearly demonstrated the dogs were able to detect a difference between the relaxed and stressed samples.

“These are really exciting results,” Dr. Wilson concluded. “They show proof of concept for this type of training protocol using volunteer pet dogs and further suggest that dogs may have the capacity to smell stress in humans.” In future research, Dr. Wilson and Dr. Reeve would be curious to find out whether dogs can differentiate between human emotional states, for example between excitement and stress. Both of these emotions have a similar physiological response, such as increased blood pressure and heart rate. However, the hormonal response is different. “We don’t yet know what it is the dogs are picking up on when they can tell the difference between a relaxed and stressed state,” added Dr. Wilson, “but we would really like to know that.”

The researchers hypothesize that dogs can sense volatile organic compounds, a class of molecules that are found all around us and also emanate from the skin and other bodily secretions. Another area of future research for Dr. Wilson and colleagues is to better understand what the dogs are detecting to be able to develop diagnostic tools for clinical use.

Dr. Clara Wilson is now a postdoctoral research fellow at the Penn Vet Working Dog Center at the University of Pennsylvania. Her current research focuses on 1) training bio-detection dogs to help us learn more about the scent of human and animal diseases, and 2) investigating assessments of dog cognition, behavior, and physiology to learn more about these amazing animals and what makes them most suited for working roles such as search-and-rescue, medical, or explosives detection. When not conducting research, Dr. Wilson enjoys spending time working with pet dogs to overcome fear and reactivity issues using force-free training methods. Beyond dog training, Dr. Wilson enjoys trying new knitting patterns and exploring Philadelphia.

Dr. Catherine Reeve is an honorary lecturer at Queen’s University Belfast and a contract instructor at Dalhousie University in Halifax, Nova Scotia, Canada. She conducts research on the human-dog relationship, including topics such as olfactory detection of physiological changes and dogs’ responses to human spoken cues. Outside of academics, Dr. Reeve conducts behavior consulting for dog behavior problems, specializing in force-free behavior modification for reactivity. In her free time, Dr. Reeve enjoys hiking with her own dog and learning more about the psychology of learning.