Probability of a choice
Dogs make choices. Training is really about increasing the probability that your dog makes the choice that you prefer over all other choices, and that’s why it's important to understand how animals make choices. The good news is that the probability of a behaviour is actually quite predictable—which is not so comforting if you discover that the behaviour you want has a low probability! But the other piece of good news is that there are ways to increase probability, and the study of “choice behaviour” explains how.
You get what you reward
Dogs will usually seek to maximise reinforcement; they will make choices about what they do based on what is most likely to maximise their pay-off. Don’t assume that dogs think too much about this process, as some responses can become automatic. While humans are capable of very deep decision-making processes, most of the time we don't need to think too deeply about our choices. We just do what we have learned in the past. Similarly, dogs probably don't stop to think things out unless they really have to.
Many experiments have been conducted to investigate how animals (including humans) make choices. In these experiments the animals are usually given two choices—a pigeon might be given two keys to peck, for example. Each key is assigned a different schedule of reinforcement, or each key produces a different number of food pellets, or there is a delay on the reward from one key, or one key requires a harder peck to activate the food hopper. Using these experiments, behavioural scientists have been able to answer questions like:
How closely does rate of response match rate of reinforcement?
How many small reinforcements are required to influence choice when a larger reinforcement is also available?
Do animals tend to choose a small, immediate reward over a larger, delayed reward?
How do we get an animal to choose a more difficult task over an easier task?
Data from these experiments has also helped to answer many common questions about choice behaviour in dogs.
If a dog has a choice between chasing a rabbit and coming when called, how do we influence that choice so that the dog comes when called?
Given choices, what do dogs prefer, as demonstrated by their response? What is more likely to bring a higher probability of a trained response? Studies about choice behaviour reveal that responses that are more likely to earn reinforcement are more likely to occur. Responses that produce a more immediate reward are more likely to occur. Responses that produce a bigger reward, or a better reward, are more likely to occur. So, a high rate of reinforcement is going to make a behaviour more probable. Delivering a reward immediately after clicking is going to make a behaviour more probable. A big reward is going to make a behaviour more probable.
Dogs form expectations about reward. If a reward is significantly below a dog's expectation, then it probably won't be a reinforcer in that situation. Be pragmatic in deciding what to use for a reward. One approach is to teach a behaviour using a convenient reinforcer, one that doesn't take too much time to administer and that is readily available (such as small pieces of food). This is economical, and can result in lots of reinforced repetitions. But later, when the criteria are set higher, use less convenient, but higher value rewards. Of course, the dog gets to decide what a higher value reward is! Using higher value rewards adds more value per repetition for the behaviour you are seeking to train, i.e. the choice behaviour you prefer.
As most readers of this article already use a clicker, there is no need to explain why it's a good idea to use one! One of the best things about a clicker is that it allows a bridge between the response and the reinforcer. Even with this bridge, remember that it is preferable to reinforce as soon as possible after clicking. Animals prefer a more immediate reward to one that is delayed.
Tipping the scales in favour of your choice
How do we put what we have learned about choice behaviour to practical use? Positive reinforcement can definitely help teach a dog to come when called instead of chasing a rabbit or playing with another dog. A responsive and reliable recall is a great training goal; here are some tips for achieving that goal as well as other training goals.
To encourage the choice you prefer, manage access to other reinforcers, particularly early on. Begin training in a low distraction environment. Train somewhere where there aren't too many choices. In an environment like that, it’s more likely that your dog will choose to pay attention to you and offer responses that you are likely to reinforce.
Later, use fenced areas or leashes to limit access to other choices. Long lines are a great tool when teaching recalls, and proper tracking leashes are less likely to injure you or your dog. Harnesses are another great tool, as they keep the line out of your dog's way and limit any damage if the dog is pulled up abruptly for some reason.
Train often. There really are no shortcuts here. The more you train, the more recalls you reinforce, the more likely your dog will choose to come when called. It's a pretty simple formula. As your dog becomes fluent at simple recalls, start adding difficulty through distance and distractions. This should be done at a rate at which your dog is always succeeding so that there is always something to reinforce. Don't be in a hurry to make things too difficult too soon—while your dog may do well one day, no one knows what tomorrow will bring. A good foundation makes a strong house.
Distractions should be added in controlled environments in order to manage access to those reinforcers. For example, ask a helper to hold out a handful of food as your dog recalls past. If your dog stops to investigate, your helper should close the hand around the food so that this response does not pay off. The novelty will wear off and your dog will soon recall straight past that distraction. Your dog may check in every now and then, but this is a clear example of the probabilities being shifted in your favour. You can see it happen. While you can't dictate which choices your dog makes, you can tip the probabilities strongly in your favour.
The best recall?
At some point, start reinforcing the best recalls selectively and let others extinguish. But, what is the “best” recall? As dogs are expected to recall over a variety of distances, long and short, don’t weed out short recalls—they are still desirable! Nor do we expect our dogs to always run as fast as possible. It isn't always practical, especially in the house.
The best recalls are those where your dog responds immediately, doesn't dawdle, and doesn't stop to smell the roses or make small detours along the way. It's a good idea to work on how quickly your dog responds (known as “latency”) early on, and then maintain that response throughout the rest of training. The less time the dog has to think about the recall, the less time he will have to notice the squirrel in the tree above him!
The best reward?
Food can be the perfect reinforcer. For some dogs, it will be all you ever need to use. Others might have a preference for balls, or tugs, or bite sleeves, or dummies, or access to a pen of sheep (just to name a few examples of potential reinforcers!). Until you have worked with a natural retriever, you cannot imagine how pointless it is to offer a piece of cheese while you are standing at the water's edge holding a retrieving dummy. All the cues are there for the dog’s favourite activity, but instead you're offering... cheese?
However, teaching everything with a retrieve as a reinforcer can become a bit slow and tedious! Similarly, trying to teach a working German Shepherd everything with a tug or a bite sleeve could make you tired and sore very quickly. One option is to shape the basic mechanics of a behaviour with food, and then when working with larger chunks of behaviour make the switch to what the dog prefers in that situation.
Good odds for success!
Big rewards for good recalls in increasingly difficult scenarios will strongly shift the probabilities in your favour. Just be careful not to ask too much. When you fail, your dog is rewarded for doing something else, and that lowers the probability of coming when called.
Dogs always have choices. Nothing is certain; no conditioned behaviour has a probability of 100%. The very fact that other behaviours have some probability rules this out. But by applying what we have learned from the study of choice behaviour, in earnest, we can give ourselves and our dogs the best chances in life.
Seeing training as shifting probabilities
This is a great article. I think it's important that people understand that the dog, ultimately, always has a choice, but that good training with appropriate rewards shifts this choice in our favour. Controlling "self-rewarding" is another important aspect, and probably the hardest one to implement. Check out Leslie McDevitt for ideas on how to use your dog's favourite activity (such as chasing rabbits) in your favour!
Probability of a Choice
Very helpful article. Excellent suggestions are easy to implement.
how best to proceed?
My adopted 18 month old has an extremely high exercise need so the several acre fenced dog park is a wonderful resource. I've been using a long lead line in order to maintain better control. I understand the premise of 'not asking too much,' but those situations seem to be precisely the ones where I need to call her back to me. She is large and sometimes smaller dogs seem intimidated by her. It is just those times that I want her to return to me. My question is that since we've worked hard on recall with no/low interest distractions with good results- how do I work on those high interest distractions without exposing her to them?
Mom of Two Great Danes
Can you work with someone else that has a distraction the dog likes - for example a toy, or small dog in the other person's arms. Practice calling your dog and if he goes to the other distraction, have the person turn their back, hide the distraction, or walk away until your dog goes to you for a reward. That way there is a distinctive non-reward for the distraction, and reward for the recall.
asking too much
Great article. I often summarize the "secret" of training as making the desired behavior desireable to the dog , or bird, or...
Even though my dog has a great recall, I have practiced the last piece of advice - don't ask too much -
when I don't think that she'll come. Such as she is too far away from me and chasing something.
Besides rewarding the "wrong" behavior, I don't want to get to a place where she won't come
on the first call.
Post new comment