Training with positive reinforcement is pretty simple: when you see a behavior you like, reward it with something your dog likes. As a result your dog will perform that behavior more often. For the most part that’s all you really need to know.
But under the surface there are actually quite a few things going on at once, and it’s helpful to be aware of them, especially when you are using training to solve a problem.
In the first of this series I explained a formula for solving behavior problems.
Antecedent -> Behavior -> Consequence
So far I have covered managing antecedents and introduced the idea of training one behavior to replace another. Now it’s time to talk about the C: consequences.
If there is a behavior that we want to eliminate, chances are that means it is being reinforced well enough to last on its own. I.E. it’s what we call durable.
What makes a behavior durable?
- Adequate reinforcer(s)
- Effective reinforcement schedule
An adequate reinforcer is a reinforcer that is "worth" whatever the cost of the behavior is. This calculation is different for every being. The reward for jumping up — attention and maybe licking a face — is worth the expense: sometimes the attention is scary or maybe even a little painful. The reward for counter surfing — the occasional windfall like a slice of pizza or a peanut butter and jelly sandwich — is worth the expense: nothing most of the time and getting yelled at if a human catches you.
The reinforcement schedule is how often reinforcers are delivered. The rules for how schedules work are a little counterintuitive and can be quite surprising.
The most durable behaviors are those rewarded continuously or on a regular schedule. In the human world, think of a paycheck. You get paid a some sort of regular and predictable interval. But there’s a big downside to this schedule. What would happen the the behavior “going to work” if a paycheck was completely missed or suddenly very disappointing?
This is a pitfall many people fall into when training with food: they start out rewarding their dog every time he does something and then stop completely after a while since “he knows it now.” The behavior very quickly extinguishes.
So what’s the best schedule for a durable behavior in the "real world" where rewards are rarely available every time? Intermittent, or in other words random. If a dog is rewarded for jumping up on counters with the occasional human treat, he’ll keep checking them. (Which is why controlling antecedents is such a huge part of stopping counter surfing!) If a dog is periodically rewarding for jumping up with an "It’s OK! I love dogs!" he’ll keep jumping up.
These intermittent schedules are very compelling because if you think about it, it’s how nature works. Hunting and scavenging aren’t rewarded for every chase of a rabbit or every check of a trash can.
So, when we are trying to replace a behavior with another we need to make sure of two things:
- Is the reinforcer we are using for the new behavior better than the reinforcer for the behavior we want to end?
- Are we presenting it often enough to strengthen our new behavior, and how we will continue to reward it going forward?
Of course being truly random is impossible too….but I need to save something for next time.
Photo credit: einalem
ABCs Part 4: Matters of Consequence is a post written by Eric Goebelbecker . You can see the actual post at Dog Training in Bergen County New Jersey
Dog Spelled Forward Website and Blog