Weak Links in the Canine Scent Detection Behavior Chain
Canine scent detection is a fun, cheap, and convenient way to provide dogs with exercise and enrichment, build confidence, and keep them busy while sheltering in place during COVID-19 lockdowns. Ironically, since humans primarily use vision, it can be challenging to understand how dogs detect invisible scents. Learning how canines navigate the invisible world of scent is both fascinating and necessary to achieve reliable performance in sniffer dogs. For example, sniffer dogs need to learn to ignore food and toy distractions in the search environment to find the target odor. While we want to maintain motivation to search, sniffer dogs eventually need to learn not to steal all foods they smell. To earn their reward, they need to show the handler where the target odor is, without damaging, biting or digging at objects, which are valuable (like vehicles being searched) and may even be harmful (such as narcotics or explosives).
In this article, I’ll break down the scent detection behavior chain, demonstrating indication training, including videos of what to reward (and not to reward). From the beginning, it’s crucially important not to weld undesirable behaviors into the behavior chain that you’ll wish would disappear later. Deciding do’s and don’ts before you train will help you to forge chains without faulty links that cause errors, confusion, and frustration. We know more about olfaction today than we ever did before, and by applying the latest scientific evidence to detection protocols, you can increase your odds of success. Preventing problems is far easier than retraining after bad habits are firmly established.
Backchaining scent detection training
To train reliable working detection dogs, we rely on backchaining. As shown in this video, we teach the dog what to do when they find odor first (the last step in the chain), and then work backward from there. The behavior chain is detailed below. To summarize, we start by building motivation for the reward (usually a favorite ball or food-stuffed toy). Once the value is established, the dog learns the indication behavior first (such as sitting while frozen like a statue and staring, until released to chase the reward). The dog learns to show their handler where that toy is located with a clear indication. This is the stage when we teach the dog what to do and what not to do when they find the toy. When natural but undesirable behaviors occur (such as digging at the toy), we withhold rewards. When the indication behavior is trained properly, the rate of reinfocement is very high and it becomes the dog’s favorite game.
Once we are happy with the dog’s indication, we progress to finding and indicating smaller pieces of the toy (such that the scent is less intense). One advantage of this method is that if the trainer pushes the dog too hard too fast and the dog has a negative experience, they can begin anew with a new toy, without damaging the dog’s association with target odor. The next step is to add a target odor (anything from bed bugs to human remains) to the toy the dog is finding and indicating. The dog learns to find the toy and the target scent, separately or together. Finally, we put all the components together into the final performance.
Our informal indication survey
Students have many questions about this training process. At Hunter’s Heart free public webinar, “Desmystifying Indications” on Nov. 16, 2020, there were 157 registrants, hailing from countries from Ecuador to Australia, with some IAABC members, professional detection teams, and inexperienced handlers. I conducted an informal, anonymous survey, by asking participants, “Have you trained a dog to indicate for scent detection?” Approximately 60% had trained a formal indication, 30% had never trained a formal indication, and 10% of participants chose “Other (e.g., I don’t know enough to respond with confidence).” That suggests that indications can be complex and confusing, but there is a fairly widespread interest in learning more.
Components of succesful scent detection
Let’s begin by breaking down scent detection into its components. Trioisi et al.’s 2019 review, “Behavioral and Cognitive Factors That Affect the Success of Scent Detection Dogs,” cited four components:
- “Search an area, often indicated by its handler
- Locate the target odor
- Follow the target odor to its source, and
- Reliably alert at the source of the odor without alerting to [aka indicating] nontarget odors”.1
Think of scent detection as a complex behavior chain, demonstrated in this video. In the words of Ken Ramirez, a chain is “a series of behaviors linked together in a continuous sequence by cues, delivered by the handler or from the environment, and maintained by a reinforcer at the end of the chain.”2
When it comes to behavior chains, you may be more familiar with retrieving (depicted in Fig. 2). The chain is initiated by the cue to “fetch”. The dog moves to the object (e.g. bird), picks it up, holds it while returning to deliver it to the handler and receive their reward. To succeed, the dog needs to know where to go (to the bird’s current location) and what to do when they get there (deliver it to the handler undamaged).
Many teams have problems when some links in the behavior chain are satisfactory, but others aren’t, for example, the dog chews the bird before delivering. The more the dog practices that undesirable sequence, the more hardwired it becomes. Ideally, the dog should not be rewarded if some links are faulty. So, the instant the dog chews, it’s better to interrupt the sequence and remove the opportunity for reinforcement. Isolate link number three and reward the dog only for holding gently. (For example, the dog can learn to gently hold an obedience dumbbell, before they are asked to retrieve more tasty birds.) When the dog is proficient at gently holding by themself, then combine it with other links in the behavior chain.
Note that chewing a bird is a natural dog behavior, but it doesn’t serve the handler’s best interests. If a feral dog is hunting to survive, chewing the bird is desirable. But if the goal is for the dog to deliver the bird to their handler, chewing is problematic.
Scent detection is similar, in the respect that the sniffer dog performs a behavior chain where chewing is undesirable. In fact, the dog must ignore intentional food and toy distractions in the search environment to find scent, as shown in Fig. 3.
The chain is initiated with cues, e.g., “find it.” The sniffer dog searches for target odor (i.e., a scent that they’ve been trained to detect). Upon encountering scent, the sniffer dog compares the intensity of scent in their right versus left nostrils to navigate the cloud of scent (aka plume).3 They pinpoint source, which is the highest concentration of target odor. Next, they communicate its location to the handler, by performing an indication (aka alert/trained final response), until released to their reward. To summarize, sniffer dogs need to know where to go (the source of target odor) and what to do when they get there (indicate).
Let’s take a closer look at the indication (shown in Fig. 4). According to Canine Olfaction Science and Law, “A well-trained dog should exhibit an alert that any observer can clearly conclude is overt, decisive, and highly repeatable. While conditions occasionally exist preventing an obvious alert such as limited physical space, the majority of alert problems involve training, handling, or other human involvement such as alert selection, alert reinforcement, odor threshold training, cueing, lead control, blocking the dog, etc.”4
An indication is a consistent, trained sequence of behaviors. Fig. 5 shows my dog indicating, including all of the following behaviors until released to the reward:
- Touching source with his nose
- Stop searching or sniffing (a dramatic difference you can hear)
- Stare at source
- Sit or down, and
- Freeze like a statue..
This obvious, repeatable indication clearly communicates the exact location of source. And when detecting bed bugs, that location information helps me to verify the presence of bugs, e.g., bugs, feces and shed skins. Without my dog, it would take far longer to strip all beds and disassemble furniture, resulting in a more invasive, expensive visual inspection.
Given the ideal behavior chain shown above, Fig. 6 summarizes the do’s and don’ts I use for indication training criteria.
|Fig. 6. Sample Indication Criteria – Do’s and Don’ts|
|Desirable Behaviors – Reward||Undesirable Behaviors – Withhold Reward|
Here’s video of a training session where I apply those criteria when my dog indicates a piece of his handler-scented toy. The first repetition is successful: He lies down and freezes and stares until released. The second repetition begins with sit and down and stare, but then he breaks his freeze and moves. I withhold the reward and remove the opportunity for reinforcement. In the final two repetitions, my dog succeeds and is rewarded for correctly performing all of the indication behaviors, and none of the don’ts.
Invisible olfactory cues
In fetching, it’s easier to evaluate success visually. But in scent detection, the cue to indicate is olfactory: when the dog finds source. The cue should not be a word the handler says, or a gesture the handler makes. This minimizes the Clever Hans Effect. Many beginners think that lying down is the most important component. They frequently use body language (like a tight leash) to bring the dog to source. Then they tell the dog to lie down and reward them for looking at the handler. This is problematic, because the dog is following the handler’s visual and tactile cues. When the handler no longer knows where scent is located (in deployments/competitions/tests), they can no longer help. Then the dog may fail, since they think the cue to indicate is the handler’s body language and words (incorrect) instead of learning that the source is the olfactory cue to indicate (correct). Teaching and practicing an indication based on an olfactory cue is essential for the dog to have an independent, reliable performance.
Consider Fig. 7, where my dog lies down between two containers, looking at the handler. Lying down does not communicate whether the right box contains odor or the left. This may not matter if your dog is finding food for fun or enrichment, but leads to failures in deployments. In nosework competitions (where pet dogs find essential oils), if scent is behind the top left corner of a license plate and the handler says its the bottom left corner, the team may be faulted or failed. If the dog constantly looks at the handler, it’s hard to tell if they gave up or they’re at source.
From the time that the handler gives the cue to search until the indication, the sniffer dog is experiencing pertinent olfactory cues, which are invisible and imperceptible to the handler. The handler can’t detect the odor plume or know for certain where source is. Unfortunately, the handler might believe their dog is wrong when actually they are correct. For example:
- There might be one bug egg in a bed frame, which the handler fails to detect upon visual inspection, so they do not reward the dog for indicating.
- The handler may handle scent improperly, so scent is present where the handler believes it is absent.
To address this conundrum, when imprinting puppies, we secure scent next to a hole in a box, which leaves an explicit, visual criterion for the handler. When the dog’s nose reaches the hole, they’re at source, and the handler can mark and reward in a timely fashion.
This method has succeeded with thousands of dogs, regardless of age, breed, and temperament. Most dogs sniff from the day they were born. Provided that the dog is healthy enough to participate and enjoys a food or toy reward, they quickly form a positive association with the target odor.
Don’t weld links into the behavior chain you’ll wish would disappear later
Failure to provide timely feedback may lead to rehearsals of undesirable behaviors. It’s easy to weld faulty links of the behavior chain, and they’re difficult to remove once firmly established through practice. The more you rehearse undesirable behaviors (like barking in frustration, as shown in Fig. 8), the stronger the welds become, and the more resistant they are to disappearing.
Think of tying your shoes. Many people would be hard pressed to explain the steps they go through to tie their shoes, but they usually follow the same sequence of behaviors, without conscious thought. How hard would it be for you to tie your shoes using only your non-dominant hand? (When I tried it yesterday, the results were entertaining.) If it’s important to tie exclusively with your non-dominant hand, it’s easier to learn that hand from the very beginning. Then every time you tie your shoes, you’ll be rehearsing success. The heuristic “get it perfect before you practice” is true in scent detection too.
If you only remember one thing from this article, I hope you’ll decide in advance what your scent detection behavior chain will include, write it down (similar to Fig. 6), and post it on your fridge. Decide what you want and what you don’t want, and be consistent. You get what you reward. If you plan to rely solely on trusting your dog and seeing what happens, you’ll get a little bit of everything. It’s easy to practice undesirable behaviors that are components in the normal predation sequence.
Hunting dogs’ normal behaviors
According to the book Canine Olfaction Science and Law, Advances in Forensic Science, Medicine, Conservation, and Environmental Remediation, “Hunting for prey desired by man is at the heart of canine olfactory demonstration and forms the basis of the varied disciplines in which a dog is used today.” Dogs frequently perform the following behaviors in the predation sequence: search, smell, see, stalk (see Fig. 9), chase, grab, eat. While those are normal behaviors, unfortunately aggressive alerts (where the dog indicates by biting, mouthing, and disturbing the source) cause failures in most disciplines. Aggressive alerts don’t serve human interests, since they may cause destruction of property and may be unsafe, e.g., when an explosives-detection dog digs at a bomb, or a narcotics dog digs at drugs and suffers an overdose. Watch this video for more tips on hiding food for dogs to find in ways that avoid aggressive alerts.
I vividly recall stepping in at the last moment to handle a dog at a nosework competition who barked before the search, and after sniffing every object. In that case, barking occurred so frequently that it was difficult for me to determine the exact location of source. Whether from excitement or frustration, barking is a normal dog behavior. But it’s far more obvious when the detection dog remains silent while searching and sourcing, and only barks to indicate they found source.
If you’re an experienced handler, you may have sound reasons for selecting different do’s and don’ts. For example, some human remains detection dogs are trained to bark to indicate when they find source at a distance, where the handler cannot accompany or even see them, e.g., in piles of rubble. But, if you’re new to training scent detection, the general public does not like barking, and clients will likely complain. The important thing is to commit in advance, so when your dog offers barking, you have already decided if you should reward it or not.
If the dog is successful at recognizing the target odor, you may be wondering whether any indication thereafter should be believed (and, therefore, rewarded). It depends. Trusting your dog is important, since they’re the one with the nose. But trust before training the dog is insufficient, and does not set the dog up for success. While it usually takes our clients and students seven days to train odor recognition, getting the dog comfortable with different search environments and generalizing their scent detection skills takes far longer than most teams predict or want. Further, no amount of trust removes the requirement for ongoing maintenance training. Teams get better results when they train, test/verify the skills in realistic field scenarios, and then trust their dog in similar scenarios. Dogs and handlers inevitably make mistakes, which highlight holes in their training. Mistakes are helpful information, and serve as an opportunity to go back to training isolated weak areas in the behavior chain.
To train the dog, the handler must know where the source of scent is located. Blind searches (where the handler doesn’t know where scent is located) do not train the dog! When the search is blind, the handler doesn’t know if the dog is right or wrong. They don’t know if the dog should be rewarded or not. In real life, this means that the handler is no longer able to give timely feedback to the dog, and may even physically pull the dog away from scent. In blind searches, many inexperienced handlers don’t recognize when the dog is finding source, and without handler support, the dog leaves source, often several times, which is an error. The solution is to reserve blind searches to test the skills of the handler (and the dog), until the dog searches, finds, and indicates, independently of the handler, with obvious, reproducible indication behaviors. Train, test, and then trust your dog in an iterative cycle.
Relying solely on “trusting your dog” is not an effective strategy when your dog offers normal dog behaviors like barking, digging, or chewing the source (which lead to failures in most disciplines). To increase your chances of success, write down the behavior chain you want. For every behavior listed in this article, write down if you will reward it or not. Have a mistake protocol in place, so you aren’t thinking about what to do while your dog is working, instead of providing timely feedback. Beware: The more you rehearse undesirable behaviors, the more difficult it becomes to remove those faulty links later. Just because your dog lies down doesn’t mean they succeeded.
After your training sessions, note all the behaviors your dog performed between the start of the search and the reward, then compare them with your ideal behavior chain. Video can help keep you honest, and show you the minutiae you might not notice while handling. Once you’ve decided what you want your indication to include, only reward behaviors you want. When your dog performs an undesirable behavior (such as pawing at source), interrupt, withhold the reward, and retry. If the dog fails twice, you probably made your criteria too hard, too fast. Lastly, remember you don’t need to reinvent the wheel. Learn more in our 2021 IAABC scent detection course, featuring protocols for training your sniffer dog. Good luck and happy hunting!
From bed bugs to birds, from truffles to nosework, Carla Simon, BSc, MD, MBA, has trained thousands of scent detection teams around the world. She’s been breeding Brittanys for Hunter’s Heart since 1999 and is based in Calgary, Canada.
- Troisi, C. A., Mills, D. S., Wilkinson, A. and Zulch, H. E. (2019) Behavioral and Cognitive Factors That Affect the Success of Scent Detection Dogs. Comparative Cognition & Behavior Reviews 14, 51-76.
- Ramirez, K. (Feb. 26, 2020). Behavior Chains – Untangling the Confusion. Retrieved from Clickertraining.com
- Craven, B.A., Paterson E.G., and Settles, G.S. (2009) The fluid dynamics of canine olfaction: unique nasal airflow patterns as an explanation of macrosomia. Journal of the Royal Society Interface 7, 933-943.
- Minhinnick, S. (2016) Statistical reliability confounders and improvement in advanced dog training: Patterns, routines, targets, alerts, distractors, reinforcement, and other issues. In: Canine Olfaction Science and Law. Ed. J. Tadeusz, et al. Pg. 201.