I’ve been writing about the ABCs for a few weeks now. (That link takes you to a list of the posts. Unfortunately you need to start at the bottom to read in order. I’m working on fixing that.)
I touched on this subject in the preceding entry of this series, but it deserves a complete post: in addition to what kind of reward your dog receives, how often she receives it makes a big difference in how durable the behavior is too.
The frequency in which a reinforcer is delivered is called the reinforcement schedule. Reinforcement schedules can get pretty complicated, and completely covering them is worth a chapter in a pretty intense book, if not a book unto itself. I’m going to cover what we need for simple training and problem solving with pet dogs.
Before we cover schedules we need to cover extinction. When we talk about a behavior going extinct it is similar to what happened with dinosaurs and mini-disc players: the behavior went away. But extinction specifically refers to a behavior stopping because it was no longer reinforced. There are some other details involved, like the fact that it doesn’t really go away completely, and that extinction is usually preceded by a burst of activity (that can be very stressful for the dog) but I’ll cover that in a future post.
A continuous reinforcement schedule is when your dog receives a reinforcer every time she performs the behavior. We use this in the early stages of learning, and a good trainer will make sure you are never stingy when training a new behavior.
But a continuous schedule has a very huge pitfall: if your dog gets accustomed to receiving a reward every time, then performance can plummet or the behavior can even go extinct if a reward is missed. This is why it’s critical to move to a partial reinforcement schedule quickly.
A partial reinforcement schedule is exactly what is sounds like: not all behaviors are reinforced. You can base the schedule on the number of behaviors per reinforcement (fixed-ratio), the time between reinforcers (fixed-interval), or a random number of behaviors (variable-ratio). Each type of schedule has its advantages and disadvantages.
For pet dog training a variable-rato schedule is often the best solution because it creates a very high rate of response and is resistant to extinction because the rewards are so unpredictable.
This doesn’t just describe a training scenario though, does it? It describes real life too. If a dog finds a slice of pizza when jumping up on a kitchen counter one time in a hundred, will he check every time he’s in the kitchen? If he gets a friendly greeting and a scratch under the chin every fourth time he jumps up on a visitor, will he try the other three?
Back to training. Karen Pryor recommends moving to a variable schedule of reinforcement as soon as your dog reaches the level of performance you are currently looking for — in other words when your dog meets your current level of criteria.
Let’s look at an example of adjusting rewards to criteria. Each number below is a level of criteria. While training each level, your dog would get a click and treat (C/T) for each behavior. When each one is met, she no longer gets a C/T for that level.
- Teach your dog how to sit. Start in your kitchen with no one else around.
- Add another person in the kitchen.
- Have the other person make a peanut butter and banana sandwich while you are practicing sits.
- Move to the living room.
In this example we are raising criteria by adding distractions. Depending on our dog, the next level may be easier than the one before – moving to the living room may be easier than sitting in the face of a yummy peanut butter and banana sandwich. That’s all part of the game. Criteria setting is an art and sometimes the results are surprising.
What’s special here is we are creating a variable-ratio schedule while also improving performance! Technically this is a variation of the differential reinforcement I introduced in part three of this series, but I won’t tell your dog if you don’t. The end result is a dog that is working harder to get a reward, and exercising his mind trying to deal with distractions.
What do you think? Good? Bad? Indifferent? Leave a comment below.
Photo credit: jfinnirwin