Sophia's Research [MannersMinder/Treat & Train]
One interesting aspect of scientific research is that you never know when a simple question will suddenly catapult you into a wild race for answers, followed by a lifetime’s worth of ideas and possibilities. In July of 2003, my research took such a turn when, out the blue, I received an e-mail message from The Sharper Image, a San Francisco–based gadget company. “We want to develop an ‘electronic mood translation device,’” they said. “One that really works.”
Because my research was on barking as communication in dogs, I knew what they meant. They wanted a bark translator. But I also knew that the cost of research and development for such a product would take many years and much more money to develop than any company would want to invest. Additionally, the accuracy of interpreting barks without any other contextual cues would be lower than they expected. Plus, it’s better to learn to translate your dog’s barks yourself. So when I got their message, I basically replied, “No that won’t work.” But, unable to resist the idea of working with a company that could probably make any animal training device I conjured up, I added, “But how about these ideas instead?”
Their product idea had to do with barking. Excessive barking is a huge problem for dog owners, and the current products dealing with excessive barking all focus on punishment and thus come with a number of pitfalls or unwanted behavioral side-effects. So I suggested, “How about making a device that addresses the barking issue by rewarding quiet behavior, and that’s backed up by research to prove that it works?” And I thought to myself, “A product like that would sure make my behavior consult work a lot easier.”
Well, 18 months later the product came out. It premiered on the front cover of The Sharper Image catalogue as the Treat & Train Dog Training System, designed to decrease barking, jumping, door dashing, and other unruly behaviors that dogs exhibit when guests come to the door. The techno-gadget part of the system is a remote-controlled kibble-dispensing machine that emits a tone and immediately releases a treat when your dog performs the correct behavior.
What’s so great about that, you ask? The key is the huge improvement in timing that a techno-gadget like this provides. With the click of a remote, owners could now time the reward to appear right as their dog performed the correct behavior, without having to fumble for the food or run up to Dozer to deliver the treat. As a result, the dog would know exactly what he was doing right. This type of device is especially useful when you’re training dogs at some distance from you, such as when you need to reward them for lying down or performing some behavior 20 feet away or in another room.
The Treat & Train unit is cool, and the first time I used it I finally understood why men like remote controls so much. The part of the Treat & Train that really makes it work, though, is the carefully developed protocol for training dogs to run to a rug and lie down quietly on cue, even in the face of major distractions such as repeated doorbell ringing or knocking, loud shouting, people running around, the front door wide open, guests walking around, or people trying to eat a peaceful dinner. The training protocol is based on two published research studies that I performed before developing the final product.
The Pilot Study
I designed and carried out the scientific studies in three stages. I started with pilot testing using 10 dogs, to look at different variations of techniques I had already used. This turned out to be more complicated than I initially thought. Training dogs myself is fairly straightforward, but my goal here was to design a program that the average dog-human team with no previous training experience could perform successfully with very few errors. Plus, they had to perform the steps correctly without a trainer present to coach them.
I quickly solicited suggestions from Bob Bailey, former general manager of Animal Behavior Enterprises (the largest animal training company based on scientific methods that has ever existed), Karen Pryor, author of Don’t Shoot the Dog, and Eduardo Fernandez, one of my research collaborators. Over a grueling several weeks, I came up with a training protocol—one that took me about six days to complete per dog but whose development taught me way more than I expected. Along the way, I learned the answer to many questions I had not previously pondered carefully, such as:
- Should you train a verbal and a visual cue simultaneously?
- If you train a visual signal first, when you start training the verbal cue should you say the word simultaneously with the visual signal or should you present it before or after you present the visual signal?
- Is it important where you dispense the food reward? Will the location affect the dog’s behavior?
- If you start with high-value treats, will you always have to use high-value treats to get the good behavior?
- Is positive reinforcement alone good enough to change a dog’s behavior in a real-life situation?
- Are verbal reprimands such as “ah” or “no” okay?
- When training dogs to ignore stimuli such as people knocking at the door, if you use food to distract them from the noise at the door will they ever learn to remain calm in the face of the knocking or other stimuli? Or are they just distracted only when the food is present but will never learn to actually behave?
Of course, learning a lot was fun and coming up with a training protocol was satisfying. But it was just the first step. Now I had to test how well it really worked.
The First Experiment: Examining the Protocol
In January of 2004, my assistants and I carried out the first research experiment at The Canine Connection Dog Training facility, owned by Dr. Sarah Richardson, in Chico, California. We took six unruly dogs whom owners had complained were misbehaving when people came to the door and we worked them through each step of the protocol. Each time we performed a repetition or trial, we recorded correct and incorrect responses. Then we followed strict criteria for moving on in the training: nine out of ten correct trials in a row and go to the next step; miss more than two out of ten and repeat the step; miss five or more out of ten and go back a step.
To my surprise, despite having dogs of different breeds and temperaments, all of the dogs made it through the protocol in eight days and the steps were easy enough that dogs performed each trial correctly more than 90% of the time. That means they made mistakes less than 10% of the time. So now we had a training protocol that we knew was easy to carry out.
The Second Experiment: Clinical Trial
The next step was the clinical trial to see how it worked in real homes. We sent out a call for the most poorly behaved door-greeting dogs we could find and made owners prove their dogs were unruly enough. Owners had to videotape their dogs for one minute during three guest visits so we could quantify the bad behaviors.
We got what we asked for: 15 dogs who barked on average 19.3 times per minute, jumped 8.2 times per minute and spent over 75% of their time crowding the door or crowding the guest. The worst in each category barked over 40 times, jumped over 20 times, gnawed on visitors’ arms, and one even had a history of lunging so hard to get out the door that he once dislocated his owner’s shoulder.
Some owners were skeptical about whether the program would work for their unruly dog, but all wanted their dog to be better behaved. So, armed with an instruction manual, a rough instructional video, and a prototype kibble-dispensing machine, the owners diligently worked through the program just as a regular person might.
That is, instead of practicing every day as they were directed, they skipped many days in a row, took long vacations, accidentally skipped steps, and performed steps incorrectly—which meant we had to check on them regularly to ensure they were staying on track. In spite of the setbacks, all dogs metamorphosed into polite door greeters within two to 16 weeks, with the average owner spending 20 to 30 minutes a day and taking about a month of consecutive training days to complete the study. By the end, dogs on average barked less than one time per minute, and since they all stayed on their rugs virtually the entire time, none jumped on or crowded the guests.
The Instructional DVD
At this point we had a training program that we knew worked even when owners did the training with their own dogs. Unfortunately, developing an effective training program is only half the work. Creating instructional materials that would compel owners to perform the steps correctly would be the key to success. This sounds simple, until you realize that animal training is a technical skill, a sport, just like tennis or golf. If your timing is off or you do something a little bit different, you don’t get the results you want. And just watching someone demonstrate the correct technique isn’t enough. Good instructors break the techniques into their component parts to reveal the important nuances.
With this in mind, I developed an instructional DVD using dogs in different stages of the learning process so that viewers could see how dogs look while they’re learning the exercises, as well as how they look once they know the exercises. Most steps are illustrated using several dogs, each step is illustrated several times, and the finer points are highlighted with close-ups and slow motion. Additionally, while the DVD features me demonstrating techniques correctly, it also features owners demonstrating the mistakes they have made and special “nerd alerts” that humorously illustrate additional errors you should avoid. We also show how to deal with most of the pitfalls owners might come across.
As you might guess, the DVD is quite extensive. It shows more than 30 dogs and it includes information on how your dog learns as well as ancillary exercises that will help you get your dog through the program quickly. And the DVD has even been tested with a focus group in which I asked viewers to perform the techniques they just observed so that I could evaluate their interpretation.
The DVD contains 13 chapters and two bonus chapters, and is meant to be watched in 10-minute segments. The total length is 3.5 hours.
Treat & Train, Now MannersMinder, Is Standing the Test of Time
The Treat & Train has been on the market for five years and is now called the MannersMinder, which is sold through Premier Pet. It’s recommended by trainers, veterinary behaviorists, and Ph.D. behaviorists for behaviors ranging from being unruly at the door, to separation anxiety, to aggression between household dogs where both dogs need to learn to run to a rug away from each other so they can chill out. It’s also used in canine agility for teaching go-outs, contacts, the pause table and right-left directions, and in obedience for the down-stay and down-stay with distractions.
Disabled clients and service dog organizations use it to train service dogs. Progressive shelters use it to help reward calm behavior in shelter dogs and as enrichment. Furthermore, I and others use it to train cats—both tricks and polite, calm behaviors such as sit.
Overall, while that fateful e-mail in 2003 has not made me rich in dollars, it has provided me with an invaluable product I can use with my own animals and to help others. And the process by which the training protocol was developed has provided me with a wealth of knowledge.
What I learned from my research
- Should you train a verbal and a visual cue simultaneously? For the protocol, my goal was to get through a given step in just 10 trials. When I attempted to train both the verbal and visual cue simultaneously, it clearly took well over 10 trials for the dog to learn the verbal cue. That is, they might learn the visual cue within 10 trials because the visual cue is more biologically relevant to them, but it might take three to five times as long to learn the verbal cue this way. For instance, if I was training the dog to touch a target (a rod with a ball on the end) with their nose by placing a dab of peanut butter on the ball, then presenting the target in front of their face and rewarding them immediately after they touched the ball, the presentation of the target would become the visual cue to touch the target. They learned this visual cue easily because it’s the obvious thing to do. If I said the word “target” at the same time I presented the visual cue, the dog didn’t need to pay attention to the verbal cue “target” because they got all the information they needed by seeing the target appear. In psychology, this phenomenon in which the more biologically noticeable cue blocks learning of the less noticeable cue is called overshadowing. A more efficient way to each the verbal cue is that once the dog knows the visual cue, say the verbal cue right before you present the visual cue. Then the verbal cue comes to predict that the visual cue (followed by the behavior) will appear.
- If you train a visual signal first, when you start training the verbal cue should you say the word simultaneously with the visual signal or should you present it before or after you present the visual signal? As I explained in the previous question, knowledge of the visual cue makes it difficult for the dog to learn the verbal cue. This makes sense. Why learn a verbal cue if you already know the visual cue and the visual cue always occurs with the verbal one? This psychological phenomenon has also been studied thoroughly; it’s called blocking. If the animal already knows one cue for a behavior and you present a new cue simultaneously with the cue he already knows, he will take much longer to learn the second cue. It’s best to present the new cue right before the cue he knows so that the new cue comes to predict that he will then see or hear the familiar cue.
- Is it important where you dispense the food reward? Will the location affect the dog’s behavior? In this study it was quickly apparent that reward location was important. For instance, when training dogs to run to a rug and lie down, they tended to lie down close to and facing the food reward. So if I used a method where the Treat & Train was located several feet away from the rug and the dog had to get off the rug to get his reward, the dog tended to have difficulty staying on the rug. He would run to the rug and tend to lie down with part of his body off the rug on the side closest to the machine. If, however, I placed the machine in a location where he had to stay on the rug and had to be lying down when he received the reward, he could learn much more quickly to stay in a down-stay, even with distractions. Incidentally, once he learned to do a down-stay with the machine nearby, the machine could just be removed and the dog rewarded at variable intervals with treats or some other reward that you deliver to him.
- If you start with high-value treats, will you always have to use high-value treats to get the good behavior? This was actually a huge concern with the first experiment. What we found was that once the dog knew the behavior well and had a strong history of reinforcement for the good behavior, you could start switching to lower-value rewards such as kibble or petting and praise. So, just as you would start by rewarding a dog frequently for staying in a down-stay position, and then systematically progress to fewer and fewer treats for the same behavior or switch to other rewards the dog likes, you could also start with higher-value treats and, once the behavior is learned well, go to lower-value rewards such as kibble.
- Is positive reinforcement alone good enough to change a dog’s behavior in a real-life situation? In the pilot study it actually seemed that positive reinforcement was good enough. But by experiment #1, it was clear that some dogs needed to learn that improper behaviors did not gain rewards. Why the discrepancy? It turns out that during the pilot study, out of habit, I required the dogs to follow house rules. That means they had to sit for all attention, petting and kibble, and to go out the door, and also could not pull on the leash—doing so meant they didn’t get to walk forward. It turns out that this subconscious training had an effect on the dogs’ behavior in the down-stay and distractions exercises. It taught the dogs to be more focused and calm, which allowed them to focus on the rewards for good behavior.In experiment #1, two of the seven dogs had to go through a few short learn-to-earn sessions before they could really focus on the training. In the clinical trial, the importance of the learn-to-earn program became really clear. Thirteen of the 15 dogs had to go through the learn-to-earn program before they could complete the very last step of training—practice with actual visitors at the door. Dogs had learned to do a down-stay and down-stay with door-type distractions (people running, shouting, door knocking and doorbell ringing), but these distractions were easy to control in terms of the level of distraction. Visitors coming to the door is less controllable. The dogs’ arousal level could be different, depending on time of day and type of visitor, too. So, for the dogs to learn that they should stay on the rug, first they had to learn that if they perform the wrong behavior, they will get no reward. So, for instance, if they got up from the rug the owner blocked them from getting to the door. Because they had gone through the learn-to-earn program, where they learned to sit to get to the item their owners was blocking them from, they would not try to get past the owners at the door. Instead, they’d wait for a few seconds and then realize they couldn’t get to the person, then choose to run back to the rug where they could earn rewards.
- Are verbal reprimands such as “ah” or “no” okay? At first I thought the verbal reprimand for dogs who broke their down-stay on the rug might work, but the dogs I used it on became confused or scared. If I reprimanded them when they got up from their down-stay and pointed for them to go back to their rug, they tended to get up again and submissively crawl over to me or just get up and freeze in place because they didn’t know what to do. It became clear that they didn’t really know what I wanted and that’s why they were getting up. The protocol needed to be adjusted so they were getting more frequent positive reinforcement for staying in the down-stay, rather than a reprimand for messing up.
- When training dogs to ignore stimuli such as people knocking at the door, if you use food to distract them from the noise at the door will they ever learn to remain calm in the face of the knocking or other stimuli? Or are they just distracted only when the food is present but will never learn to actually behave? During the distractions stage, the goal was to control the stimulus, such as knocking at the door, to occur while the dog was eating and at a level low enough so that the dog never took his head out of the Treat & Train bowl. Some people criticize that this is only a distraction and the dog cannot perceive the distracting stimulus so he won’t learn. But the studies clearly show that the dogs did learn. At first they would sometimes react to the sounds by twitching their ears or occasionally looking up at the distraction, but by the second or third one-minute trial, if the distractions were timed right they ignored the distractions. The next step was to time the food and the distractions more randomly so that now the dogs were getting rewarded for remaining in a down-stay in the face of the distraction. This transition was easy because, in fact, the food did not just serve as a distraction; it served as a way to change the dog’s emotional state from overly excited to happy and calm in a down-stay.