Part of the charm of search and rescue is the mix of low and high technology it employs. A search that uses a latest-generation helicopter with thermal imaging may also use our oldest, simplest technologies: travel by foot; search by eyes, ears, and nose. Human SAR responders can use their noses -- on my first search, we confirmed the dog's alerts when we got close enough to smell smoke. But dogs' noses are far superior for the task: Lilly had been smelling that fire, and the search subject who lit it, all along.
Although the use of dogs by humans is a very old technology, it is not a simple one. Understanding how our canine SAR partners work requires us to understand their sense of smell. Scent research hasn't uncovered anything that would overturn most good-faith training and operational practices of SAR dog teams. However, science sometimes surprises us. We can also be doing the right things for the wrong reasons. Understanding why proper training and operations are proper is, I think, an important part of the increasing professionalism of SAR dog handlers.
Most dog handlers' first exposure to the science of scent came from either Scent and the Scenting Dog by Syrotuck (1972) or Scent by Pearsall and Verbruggen (1982). While excellent books for their time, both were written before research findings of the 1980s and 1990s exploded scientific knowledge of scent and the detection and characterization of odors. In this article, I'd like to review what scientists have learned about the sense of smell and what practical lessons these findings tell us about the process our dogs use to find lost people by their scents.
Scent detection: The key fits the lock.
In 1991, a team of researchers at Columbia University reported how they had isolated the genes that tell the body how to make olfactory receptors. These receptors are proteins that sit on the surface of specialized nerve cells in the nasal membranes and detect odorants -- specific chemicals that carry odor.
Detection of an odor begins when an odorant molecule sticks to a pocket in an olfactory receptor protein, which in turn signals its nerve cell to fire a message to the brain's olfactory lobe. Each olfactory receptor's "binding site" fits only a small number of odorants of similar molecular shape and chemical properties -- much as a lock fits only keys of a certain shape. (See figure 1.) Many types of receptors with different binding sites gives animals the ability to detect a wide variety of odorants.
Fig. 1: An olfactory receptor will only bind to an odorant that has a complementary shape and chemistry (symbolized here by color). When it does (bottom), it fires a signal that in turn causes its nerve cell to fire a signal to a specific part of the brain's olfactory lobe.
Interestingly, both humans and dogs appear to have about the same number of olfactory receptor genes: roughly 1,000. But as many as 75 percent of human olfactory receptor genes are "pseudogenes," which are false genes that don't produce any working receptors. Dogs, which appear to have far fewer pseudogenes than us, may be able to smell a large number of odors that we can't detect at all, though that question is a matter of heated debate among scent researchers at the moment.
Olfactory nerve cells with the same receptor type send their signals to the same part of the olfactory lobe; because of this, different odorant molecules cause different patterns of nerve-cell activity in the brain. Dogs' olfactory lobes are larger than ours. We don't fully understand what advantage this gives them. It may contribute to their superior odor-processing, allowing them to pick out faint target odors against a confusing background better than we can.
The next step in the detection of smell is when the brain takes the olfactory lobe activity caused by mixtures of odors and somehow translates it into a unified scent or smell. The same odor can be a part of very different smells. For example, a type of carboxylic acid called L-lactic acid is the primary reason for the sour smell of milk that's gone bad. Accompanied by a different set of other odors, it's a component of the smell of humans. In smaller amounts, it forms part of the pleasant sharpness of some cheeses. Clearly, the way we experience lactic acid depends on both the amount of it present and the other odors we experience at the same time.
An example of this may be found in SAR dog work. Many handlers have observed and described an interesting phenomenon in which a cadaver-trained dog will sniff all around but not approach an especially strong cadaver-scent source. Some handlers wonder if the dog is being overwhelmed by the strength of the source. My guess is that the dogs are being overwhelmed, but in a chemical rather than a behavioral way. When present at high concentrations, an odorant may begin to bind different receptors than it does at lower concentrations. So the intense scent at the center of a real cadaver may smell different to the dog than the unavoidably weaker samples she trained on. Fortunately, a properly trained dog will lead the handler quite close if not all the way to the cadaver.
Keynotes or combination locks: What is human scent?
One of the big unanswered questions about how the brain (dog or human) interprets smells is how it combines individual odors into a recognizable, unique scent. In one theory for how this happens, the "keynote" hypothesis, the brain ignores many or even most of the odors coming from a substance, focusing only on certain keynote odors that are unique to that substance. (See figure 2.)
Figure 2a: keynote hypothesis
Figure 2b: combination lock hypothesis
Fig 2: Two ways that the brain may put odors together to create complex scents. 2a, left: keynote hypothesis. The brain picks out only one or a few key odors that are unique to that scent: A = human; C = deer. 2b, right: combination lock hypothesis. The brain compares the relative amounts of many non-unique odors: 9A 6B 0C 2D 3E 5F = human; 0A 2B 10C 7D 4E 6F = deer.
For instance, a human may smell different from a deer because human body scent contains large amounts of carboxylic acids (from bacterial action on human skin secretions). Our skin also tends to emit traces of non-natural compounds such as hydrocarbon derivatives, because we travel in and refuel vehicles that burn hydrocarbons. Since neither of these are present to any extent in deer scent, all are possible human keynotes.
Another possibility, which I call the "combination lock" hypothesis, is that there are no individual keynote odors that are unique to human scent. Instead, the brain takes a look at all the odorants and their relative amounts. The dog identifies human scent when the right proportions of odors exist, although a different combination of the very same odors may say "deer." Finally, dogs may use both keynotes and an overall combination to interpret smells.
These hypotheses are important when we consider the use of artificial scent products for training SAR dogs. Say, for instance, we wanted to produce "human scent in a can." Such a product would be useful because in training, we could place it in locations that are hazardous or impossible to place a human being. If the keynote hypothesis is correct, "human in a can" may be more easily achievable. The only trick would be including all the necessary keynotes without inadvertently putting in false ones. But will the keynotes detected by one dog necessarily be the same keynotes another dog uses? If combinations are the answer, the task becomes trickier. Not only may it be necessary to pack more individual substances in your can, the proportions of these substances would have to be "right," even after storage. If both keynote and combination are important, would a keynote-style artificial scent be "good enough" to pass small-scale tests but fail in a significant percentage of real-world problems or with certain dogs?
I don't mean to criticize any specific artificial scent product that may be on the market. In practice, we don't really need to know the composition of an artificial scent product; but we do need to know how often dogs trained with the artificial human scent alert on scent that isn't human (false positives) and how often they fail to alert on the real thing (false negatives). This is, I think, the most important question I'd ask before I felt that I understood what I was using.
Human scent: Some specific research that bears on SAR dogs
Having sketched out the modern understanding of scent, I'd like to wrap up by talking about a few recent studies that cover indirectly some topics of importance in SAR dog use. A warning label comes attached to these studies, though: much of this research draws on experiments with animals that are not dogs. Important differences could exist between dogs and other animals; but so far, scientists have discovered surprisingly few variations in the sense of smell in most mammals.
In a 1991 paper, researchers from the Savannah River Ecology Laboratory in South Carolina and from Harvard University showed that dogs trained to discriminate between individual humans could not reliably match smells from one part of a person's body to another. For instance, when trained on scent from a person's hand, the dog could not distinguish between scent from that person's elbow crook and from another person's hand. The authors of the study suggested that there might be no such thing as a single individual human scent.
It's important to understand what this experiment doesn't show. It doesn't "prove" that our SAR dogs can't trail individual humans. It does, however, suggest that when they do trail individuals, they're doing something far more clever than we realize. When given a hat as a scent article, they may need to pick out the scent from that person's head from those of the rest of his body in order to trail him. Interestingly enough, many experienced discrimination dog handlers anticipated this scientific finding long ago. For quite some time, it's been common for dog handlers to make a point of using a variety of objects containing human scent while training. The main reason is to ensure the dog learns to pick up scent from any possible scent object -- but it may also train the dogs to pick out scents from different parts of the body and follow them.
A paper published in April, 2000, by researchers at the Monell Chemical Senses Center in Philadelphia and the University of Pennsylvania reviewed discoveries of the 1980s and 1990s about human-detectable human body smell. This paper offers interesting clues as to what human scent might be and what our dogs may be doing with it.
The parts of the human body most responsible for what we can smell are the armpits and the genital areas -- in fact, human infants can identify their mothers via armpit scent alone. Both these areas contain especially rich populations of eccrine sweat glands, which produce moisture; sebaceous glands, which produce oils that bacteria turn into carboxylic acids; and apocrine glands, which produce steroid-like odor molecules similar to those that have been implicated in forming the individual scents of animals. Of the two body areas, the armpits are probably more important, because with normal hygiene the genital area is considerably dryer; and as dog handlers know, moisture is crucial for the bacterial growth that produces much of human scent.
Researchers found that carboxylic acids and odiferous steroids are among the most important parts of human scent. This may be important, because most parts of the body -- lacking apocrine glands especially -- don't produce nearly as much of the scent that comes from bacterial action on apocrine/sebaceous/eccrine secretions (which I'll call BASE scent) as the armpits.
You'll recall the 1991 paper showing that dogs couldn't match scent from the hand with the same person's elbow crook. But if BASE scent plays a major role in individual human scents, the dogs may have been trying to match one trace scent with another -- a very difficult task, unless they were trained specifically to do it. When we train our dogs to scent discriminate via a pocketknife touched by the subject, they may not be learning to detect "hand odor." They may be learning to pick up trace amounts of BASE scent on the hands. For this reason, scent articles that contain larger amounts of BASE scent, such as underwear or shirts that have touched the armpits, may be more resistant to contamination than other objects (although this hasn't been proved).
An important facet of these findings is that pre-pubescent children, who lack apocrine glands, can't have full BASE scent. This obviously doesn't prevent dogs from finding them; but dog handlers whose dogs haven't trained on children sometimes see unusual behavior in their dogs (for example, failure to give a bark signal) when they first encounter children on a search task. It's an argument for using pre-pubescent subjects in training, so that we understand and can work around any quirks in our dogs' reactions to incomplete BASE scent.
The final article I'd like to discuss was published in February, 2000. A team from the University of Florida and the U. S. Department of Agriculture's research service used a new method of collecting human skin emanations to create what may be the most complete and exhaustive catalog to date of the volatile molecules that come off human skin. The researchers identified at least 277 chemical compounds that could be potential human odorants -- and they were careful to say that there are probably more.
Interestingly, some of these compounds were present in consistent amounts from person to person; some varied between people; and some varied on a single person from day to day. I'm especially interested in seeing if these compounds could be screened to see which of them, or which combinations, search dogs will alert on. These researchers weren't interested in dogs -- they were trying to figure out what mosquitoes use to find humans (it turns out that the little buggers air scent). Now, mosquitoes don't favor human scent over animal to anywhere near the extent that we expect from our dogs. But this research may still provide important clues as to what dogs sense.
Interestingly, mosquitoes home in on carboxylic acids -- in particular L-lactic acid, the human-body-smell component we discussed earlier. They also home in on carbon dioxide, which we exhale. But they are attracted even more to a combination of lactic acid and carbon dioxide. The researchers suspect that, in the real world, mosquitoes home in on a combined profile of odors.
Of course, the fact that a compound comes off human skin doesn't necessarily mean it's part of what dogs are smelling; and dogs may very well not be detecting the same compounds as mosquitoes. But this research does show that an extremely simple organism like a mosquito can use a rather sophisticated combination of signals to home in on hosts. Dogs, which are far more complex, may be capable of far more sophisticated analyses of scent profiles. The research casts a conditional vote for the combination-lock hypothesis, and a cautionary note about whatever artificial scents are likely to be developed in the future.
In this article, I've tried to give an update on human scent as dogs detect it. I haven't touched on how (or whether) scent discrimination between individual humans is different than detecting generic human scent. I haven't even mentioned the vomeronasal (Jacobson's) organ, a "sixth sense" scent organ we're just beginning to understand. And we haven't discussed the role of environmental factors, such as diet, soaps, and so forth. But these are tales for another day.
I'd like to thank Dr. Ulrich Bernier of the U. S. Department of Agriculture and the members of Allegheny Mountain Rescue Group and Search Dogs Northeast for comments and criticism on this article.
*An earlier version of this article appeared in Advanced Rescue Technology, October/November 2000, pages 24 to 30. Copyright 2000 by Summer Communications, Inc. Reproduced with the permission of the publisher.
General information on scent Axel, R., "The molecular logic of smell," Scientific American 10:154-9, 1995.
Marples, M. J., "Life on the human skin," Scientific American 220:108-15, 1969.
Pearsall, M. D., and Verbruggen, H., Scent, Alpine Publications, 1982.
Syrotuck, W. G., Scent and the scenting dog, Arner Publications, 1972.
Scientific articles on scent
Bernier, U. R. et al., "Analysis of human skin emanations by gas chromatography/mass spectrometry. 2. Identification of volatile components that are candidate attractants for the Yellow Fever mosquito (Aedes aegypti)," Analytical Chemistry 72(4):747-56, 2000.
Brisbin, I. L., and Austad, S. N., "Testing the individual odour theory of canine olfaction," Animal Behavior 42:63-69, 1991.
Wysocki, C. J., and Preti, G., "Human body odors and their perception," Japan Journal of Taste and Smell Research 7(1):19-42, 2000.