Lecture 29: Selecting your Approach to reduced reinforcement
The topic of reducing reinforcers is complicated, because your path is dependent on many variables. In the first part of this class we began the process of reducing reinforcers - let's look more deeply at this issue at this time. We'll start by considering several of the questions I ask myself when deciding on a plan to reduce reinforcers.
Factors to be considered when reducing reinforcers:
How fluent is the targeted behavior? In the learning phase, we use classic reinforcers (concrete things like food, toys, balls, etc.) to mark correct behavior - we condition our dogs to understand that when they do something we like, we'll give them something they like. In the training phase, this is important, but over the long run it's a problem because in competition, there will be no classic reinforcer no matter how clever the dog is. The dog must accept a reduced reinforcement schedule even when the behaviors are excellent in their quality.
As a result, we separate these issues. Here are some questions to ask yourself:
In the learning phase, 100% of the time we reward correct work; you cannot reduce reinforcers until you are convinced that the dog knows what is the correct response to a given cue. NEVER reduce the reinforcement rate under 100% until you are very confident that your dog knows what is correct under a given circumstance - if you do not follow this rule you will either weaken your dog's known behaviors or create a situation where your dog offers random behaviors, whether or not you want them.
How mature is your dog?
Older dogs have much better frustration tolerance than puppies. I tend to wait until my dogs are reasonably mature before I start the process of reducing the reward schedule, or at least I factor this in when deciding how much to reduce the rewards. The more mature the dog, the more I can ask in terms of time between classic reinforcements.
How much working drive does your dog have?
Note that I didn't say "love of reinforcers". These are not the same thing. Working drive is the dog's interest in doing activities and interacting with work in a training session. There is a big difference between the two. A dog can have a very high working drive and relatively low drives for classic reinforcers (though this is somewhat uncommon). More typical is a dog with very high drive for classic reinforcers but a low working drive. This is demonstrated by the dog's attitude when you have nothing for the dog - do they still love to perform even though they know there will be no rewards? If yes, then the dog has strong working drive. The more the dog enjoys work for the sake of the activity, the stronger the dog's working drive.
What environment are you working in?
The more distractions, worries, temperature challenges, and unpleasant working conditions that you are competing with, the less you can reduce classic reinforcement. Familiar, quiet and nonthreatening environments on a cool day are much easier than novel, busy and scary places in the middle of summer!
What is your dog working for?
Different motivators bring out different qualities in the dog. I find that food gives the poorest endurance - I prefer toys and interpersonal games (that may or may not include food or tug) to get more sustained performance. It's ok if your dog's primary interest is in food but to get endurance then it helps if your dog's interest in food is very high.
The more your dog values the reinforcer which is available, the longer you'll be able to go between reinforcers.
What kind of relationship based motivators do you have? How much does your dog value them?
Relationship based motivators are things that affect your dog's desire to be with you in general - they are rarely used to actually train behaviors but they are excellent for maintaining behavior - because the dog likes doing things with you. For example, play, petting, praise, belly rubs, etc, - these things are relationship based because their value is largely a function of how the dog feels about the person offering them. Some dogs LOVE to do a few tricks for a new person - the fact that the person is novel matters! But most dogs will only offer sustained work and a positive attitude if they are working for a person that they have a strong relationship with. The dog is not working for any specific thing. They are working because they either care how you feel about them (could be positive or negative), or because being with you in a close way satisfies a need in the dog (for example, some dogs heel because they are afraid of the world so they want to be very close to their human - it makes them feel safe. Other dogs heel because they love the joyful interaction of working with precision with their human. Very different reason for working but both powerful in their own ways).
The more your dog values relationship based motivators, the easier it is to reduce classic motivators. That is one of the main reasons for engagement training - to work on developing the dog's love of strong interpersonal interaction - which can go into the competition ring.
How much effort is your dog putting forth?
As we look over the above list, we can see that these factors combine to create an "effort" score for your dog. A high effort score would be reflected in a dog who is immature and working in a difficult environment vs. a low "effort" score for a dog that is mature and working at home. I tend to think in terms of a unit of reward for a unit of effort - pay your dog for how hard they are trying, rather than for what they do. Effort isn't simply the work you are requesting; it is a function of all of the above variables (maturity of the dog, environment, challenges, choice of reinforcers, etc.)
HOW to reduce reinforcers:
Now that we've looked at what factors to consider, let's also look at the process itself:
First of all, recognize that the process of reducing reinforcers goes on for months and years, not weeks, and it is not a linear process. Each time you train, you need to re-evaluate the circumstances that you are working under, and then pick a reward schedule that makes sense. Be prepared to change it if you see that your dog needs some help!
For most dogs, you'll never need (or want) to completely reduce the reinforcers in training - it's more a matter of frequently substituting ring acceptable motivators as much as possible, and making decisions about what you are trying to accomplish. If you are going for long stretches of flow so that you can do well in competition - you'll need many fewer classic rewards. If you are working to regain some specific precision (for example, your fronts), then you will bring back all of your classic motivators, because they are so good at "marking" what you want.
Just think about how you will differentiate these (flow from skill) for your dog. My dogs know the difference between flow training (moving out and doing stuff!) and skill based training. In skill training, we do one thing over and over and the classic rewards are generous. In flow training, I tend to ignore most errors or I use specific handling moves to fix errors (a conversation about that is beyond the scope of this class but is the basis of most of my advanced heeling class for those of you who have taken it). For our purposes, just think about developing a pattern of flow training that uses many fewer classic rewards and that focuses on movement and interaction above what the dog is actually "getting."
As you're reducing reinforcers, try to change only one factor at a time - that is easier said than done, since we are not computers who can compute all factors at one time, but do your best. The factors that are generally most obvious and most relevant include changes in the environment (temperature, location, equipment, surfaces, etc.), changes in the complexity of the behavior (behavior chains instead of discrete behaviors), changes in the location of the reinforcers (on your body? on a table? In the house?), or changes in the challenge level of the behaviors themselves (raising criteria).
The more your dog will accept personal praise and play as a reinforcer, and the more your dog will accept higher value work as a reward for lower value work (most relevant for dogs with stronger working drive), the better off you will be.
Substitution vs. back chaining (intervals)
This brings us to the topic of "Substitution vs. back chaining" for changing a reinforcement schedule.
In this class, we have talked about "substitution" of higher value work for classic rewards. We also talked about using games and play when you had the urge to offer a classic reinforcer. But...what if your dog doesn't care enough about play or praise? What if your dog has NO preferred exercises? What if your dog has no real working drive, or the handler has not yet developed the ability to make the above-mentioned relationship based factors reinforcing? If we exhaust all of our efforts on substitutions and get nowhere, then we consider "interval" training. In short - back chaining the dog's work to a reward found outside the ring.
Backchaining training and substitution of praise/play for work have a similar base; both require that you consider all of the relevant factors before making a plan. Both are systematic in their approach (to the ability that we humans can be systematic while training a dog at the same time). And both share the same end goal - getting into the ring. A dog with a strong foundation in substitution training will pick up interval training much more quickly, since they have already experienced a reduction in rewards and the concept of reinforcement coming from off of the body.
The difference is that interval training is much less focused on finding alternatives and much more focused on convincing the dog that hard work for a long time is what pays with whatever classic reinforcer they really really want. It is not a relationship based approach. Once the dog understands it, in theory anyone could handle the dog and the dog would perform because it is not about the person handling or the relationship - it's a business transaction.
If you go with the interval training route, you must be very slow and systematic with your progression. You must wait to apply it until your dog is extremely fluent on the exercises that you wish to chain, and has a clear "no work means no reward" attitude. An example would be a dog working on the OTCH who used to qualify regularly and is no longer able to do so. It is NOT appropriate if there is even a shred of possibility that the dog is either unable to work due to stress or confusion over expectations.
Intervals are most appropriate for dogs that are temperamentally not very interested in working unless a classic reinforcer is available. They are appropriate for dogs that are very thoroughly trained and clear on the required exercises. They are appropriate for dogs that show no appreciation for alternatives like personal play or approval and for dogs that have a low basic working drive but a high drive for classic motivators.
The basis of interval training is back chaining. Showing the dog what he is working for, leaving it outside the ring and training your dog to understand that only very correct work will earn that reinforcement. To use this method you must be extremely systematic, determined, and willing to withhold reinforcement altogether. You must be absolutely committed to the path - only correct work at whatever level of criteria you have set and trained for, for incremental improvements in time, will lead to reinforcement.
Let's say you want to back chain your dog's work and you will be entering Novice B. Your steps would go something like this:1. Show your dog the cookies that you are placing outside of the ring. Place your dog in front position and ask for a finish. Immediately after the finish, tell your dog the rewards are coming! Be excited about it! Put the leash on and bring the dog to the location of the cookies - and let him have a ton of them! Make him excited about this!
Continue this finish until your dog LOVES to finish.
Now we add the recall - the behavior immediately before the finish.
2. Ideally have someone hold your dog back for a recall, or send them around an object to start the recall, or...leave the dog on a sit stay and then perform the recall. When your dog gets to front, wait quietly for a few seconds and then finish your dog. Now - start the whole excited process of leaving the ring one more time! Soon, your dog will LOVE to do that recall and finish. If your dog anticipates the finish in their enthusiasm, you start the whole chain over - including the recall before the finish. Be cheerful about it but do start over. Do not proceed until your dog is nailing the combination of sit, stay, recall, finish and wait for leash. You can proof those behaviors by adding some challenges (for example, the proofing work we did in this class), and be sure that you feel they are ring ready before adding the next exercise.
Since this is novice B the exercise before the recall is the off leash heeling. But since we are back chaining, only do a few steps of heeling before moving on to that recall. Slowly increase the amount of heeling before the recall and finish.
When you have added in a full heeling pattern you would add in the stand for exam - the exercise before the recall.
The tricky question is what to do if your dog makes an error, and there is no definitive answer. What you cannot do is go get cookies and make it better. What you can do is start over. If the chain is quite long and you feel that starting over will deflate your dog, AND if you believe that the error is something to address outside of a chain, then ignore it and continue on with the chain. But if you think the error was a result of the dog rushing to the end of the chain for the classic reinforcement, then repeat the prior exercise, or even go back deeper towards the beginning of the chain. Because of the risk of demoralizing the dog, I would rarely recommend going back more than a minute or so.
So in my above example with Novice B- let's say that your dog did the entire Novice B routine -and then got so excited at the finish after the recall that the dog anticipated and finished with a cue. I would not go back further than the off leash heeling because I do not believe that dogs can really remember all of the things that they did in that chain. It's enough to go back two exercises. In this case, off leash heeling and a recall/finish.
I'm not a huge fan of back chaining - I much prefer the relationship based approach because I do not train dogs to perform as a business transaction. I train my dogs because I enjoy what we do together. But I can recognize that some dogs have relatively low interest in me or the work itself so if that is the case then you have a business transaction - and that might mean having your dog repeat several behaviors to "get it right" without a ton of sympathy on your part.
Some dogs do extremely well with intervals but others fall apart and become demoralized when they make an error and do not "win" reinforcement. That leads to careful and methodical work - not something I strive for. Take a look at your dog. The more your dog cares about your relationship the more they will suffer in interval training and the less value it has because you can substitute relationship based reinforcers for classic ones. The less your dog cares about the final reward the less effective intervals are. Your dog may decide they simply don't care enough about the final reward to put out more than a set amount of effort - lets say 90 second max. How are they handling the program that you have set out for them? If your dog gets more excited but stays clear headed (thinking) as you progress, then you're probably doing a good job. If your dog gets TOO excited (hectic), then maybe you need to reduce the value of the reinforcement outside the ring. Or possibly you need to be a bit slower and more systematic in your training. Done well you shouldn't see the dog building stress, though you might see the dog building excitement as they believe they are getting closer to the end of the chain. Can you tell these apart? If not, err on the side of assuming stress and reconsider your next training plan. Slow down - build your variable more slowly. Hectic dogs are stressed dogs - and it is not something we want to build into your dog's CER (Conditioned Emotional Response).