In March 2015, I attended the 7th annual Art and Science of Animal Training Conference in Denton, Texas. This post is part of my notes from the conference.
For the rest of my notes from the conference, please visit this page.
I loved Ken Ramirez’s talk at this year’s Art and Science of Animal Training Conference. Ken is probably best known for the many years he spent as head of the training department at the Shedd aquarium in Chicago. However, Ken has also spent hundreds of hours training dogs. At this year’s conference, Ken talked about a fascinating topic – teaching advanced concepts and conceptual thinking skills to dogs.
Ken has done a lot of this kind of work. Interestingly, however, he’s also encountered a lot of resistance when he presents his results. Other professionals tell Ken that what he’s doing just can’t be done. So, it was really fascinating to hear Ken talk in-depth about these training projects and discuss his training process (with plenty of awesome videos, of course!).
What is concept training?
When learning a concept, the dog learns a set of rules that can be applied to new circumstances, rather than just a specific pattern of behaviors. For example, in guide dog training, a beginning dog might learn not to take his handler under a certain low-hanging sign that the handler might bump her head on. However, by the time this same dog gets to the end of the training program, he will have learned the general concept of “low-hanging obstacles.” He’ll be able to spot overhead obstacles that he would be able to pass under, but that his handler would not, and he’ll be able to steer her around these so that she doesn’t bonk her head. Most importantly, the dog will be able to do this even in completely new settings and with completely new types of obstacles.
Concept training takes the dog and trainer a step beyond normal training. Rather than a simple relation between a cue, a behavior, and a consequence (such as going around a certain sign on a certain street corner), the dog learns a broader rule or set of rules that can be applied to new situations. At one point, Ken Ramirez described this as creativity within a set of rules and within a specific framework, which I thought was an interesting description.
Getting started with concept training
Now, concept training is pretty fun stuff. However, Ken Ramirez cautioned that before beginning to train concepts, it’s really important for the dog to have a solid foundation of basic skills. Ken didn’t go into a whole lot of detail here, but he did mention that it is important for the dog to understand a marker signal, be fluent with cues, and be focused and eager to work. The dog also should be desensitized to lots of new things and should have practiced lots of generalization with known behaviors.
If your dog has experience with shaping new behaviors, this will also make it easier for your dog to learn new concepts. Ken specifically mentioned Kay Laurence’s Learning Games book, which I highly recommend checking out if you are not familiar with it. (Although this book is geared toward dogs, I think a thoughtful trainer working with other species would also be able to get a lot of ideas from this book.)
What concepts can dogs learn?
Lots of them, in fact! The bulk of Ken Ramirez’s presentation centered around sharing some of the concepts that he has trained to dogs. Here are a few of the concepts that Ken discussed in his presentation:
Mimicry: Several years ago, Ken and his staff at the Shedd aquarium embarked on a huge project to teach several dogs a “copy” cue. The dogs learned to watch a model dog do a behavior and then repeat that behavior when the trainer gave the “copy” cue. Interestingly, Claudia Fugazza was doing her “do as I do” research at about this same time, but neither knew about the other one’s work until later on.
During the training, the dog first learns to watch a model dog and then copy behaviors that he already knows how to do. Much later on, the model dog will do behaviors that are not in the other dog’s repertoire. Sometimes the dogs were able to copy a new behavior very accurately, other times the performance wasn’t quite as accurate, and this often depended on how complex the behavior was. The trainers found they could almost always get an approximation of the behavior, but not always the fine detail.
Ken has had to content with many naysayers who claim that his results are impossible because dogs are not “supposed to” be able to learn by watching other dogs. However, his experimental set up involved multiple trainers and an elaborate system of barriers to guard against accidental cuing or a clever Hans effect. Interestingly, Ken also discussed that one of the first systematic uses of mimicry in animal training was done by the Navy. They used mimicry to reduce the training time in their dolphin training programs.
Modifier cues: Ken Ramirez also discussed some training projects that he has done with what he calls “modifier cues.” Modifier cues allow the trainer to give more specific instructions or to alter a behavior. This includes concepts such as right / left, up / down, large / small, hard / soft, and more. For example, if you’ve taught your dog to retrieve toys, you could also teach a modifier cue so that you could ask the dog to examine a row of toys and bring back the toy on the right side or to bring back the largest toy.
In his presentation, Ken went into some detail about teaching the concept of left and right, starting with teaching the dog to touch either a target on the right side or a target on the left side. Once the dog could touch the correct target from only a verbal cue, Ken taught left and right with a second behavior, going into one of two kennels. Ken discussed that from his experience, the modifier cue usually has to be specifically trained with 3-5 behaviors that the dog knows well. After this, Ken’s dogs are often able to apply the modifier cue to other behaviors, even if they haven’t specifically practiced the modifier cue with these behaviors.
Counting: One of Ken Ramirez’s most recent projects has focused on teaching a dog to count. Or, more specifically, teaching the dog to distinguish how many objects were presented. Previous research has shown that dogs can distinguish between one to three objects and a few studies have gone up to five, but there’s no real research or documentation with larger numbers. Over the years, Ken has taught a number of dogs (and other animals) to count, but always as a fun training project. This time, he planned to have the systematic controls and tests in place to demonstrate that the dog was actually counting the number of objects presented.
Ken’s set up involved placing several objects in a tray. The dog had to look into the tray, determine how many objects were present and then go touch a target on the back wall to indicate how many objects were in the tray. At the beginning of the project, the targets were different shapes with numbers written on them. Later on in the project, Ken changed the set up so that instead of targets, he had boards with dots on them. The dog had to match the number of items in the tray to the board with the equivalent number of dots. In the final stages of the project, the trainers would change the patterns of dots on the boards and randomize the order of the boards. They also conducted double blind trials. During these tests, Ken would ask the dog to do the behavior. However, Ken would not be able to see how many objects were in the tray and would not know if the dog made a correct selection. This project is still ongoing and the dog can currently count accurately up to ten objects.
For those of you who might want to try this at home, Ken also discussed a fun variation that he has used in the past when training a dog to distinguish up to three or four numbers. You can choose a different behavior to represent each number and teach the dog to do the behavior when he sees that number of objects. For example, the dog might wave his paw if two objects are present and spin in a circle if four objects are present. The dog learns to do a certain behavior to “tell” you how many objects are present.
What is possible? What is impossible?
For many of these projects, Ken’s colleagues told him not to even bother, because dogs wouldn’t be able to figure out how to do it. He would just be wasting his time. And even after completing some of these projects, Ken still has naysayers telling him his results just can’t be true.
Ken believes that we often limit ourselves and our animals by making assumptions regarding what is and is not possible. These expectations then constrain what we are able to train our animals to do. However, dogs (and other creatures) are capable of far more than we sometimes give them credit for!
For all of these projects, it took Ken quite a bit of experimenting at the beginning to figure out exactly how to do the project. (It wasn’t that he got “lucky” or has exceptionally “smart” animals.) He followed a systematic training process and paid attention to the tiny details, so that he was certain the animals were really learning the correct concepts. Importantly, he also started with teaching a solid foundation and the necessary pre-requisite skills, so that the dogs were able to learn these tasks with minimal errors.
So, just remember, when training concepts (or anything, really) don’t limit yourself and your animal by assuming your animal won’t be able to learn certain behaviors or concepts.
Have you done any concept training with your animals?