Is your dog 100% reliable? How do we train "reliability"? How might we train a dog to reliably avoid a snake or to come when called using positive techniques?
Warning, this article is a little more technical than most of the articles I write for Positive Petzine.
No matter how you train, all behaviors you want to see more of must be reinforced. That means if you teach your dog to come when called, you must reinforce it on some sort of schedule. It doesn't matter if you use a clicker, an e-collar, a check chain, dog whispering, dominance, voodoo, flower essences, animal communication, or secret alien technologies the KGB sold you in 1984 - the very definition of reinforcement is that an operant has been increased.
This could be positive reinforcement (we added something the dog wanted) or negative reinforcement (we took something away that the dog wanted to avoid). Here is an example of a recall trained with +R and an example of a recall trained with -R:
1. Positive Reinforcement - call dog, dog comes, ball is thrown; dog comes when called more often. If the dog doesn't come when called more often then the ball was not a positive reinforcer and the intended conditioning did not happen.
2. Negative Reinforcement - call dog, press button on remote shock collar at low level, dog comes, release button on remote collar; dog comes when called more often. If the dog doesn't come when called more often then the timing was out, or the shock was not felt and the intended conditioning did not happen.
In both cases there is a reinforcer. The reinforcer in the Positive Reinforcement (+R) example is the thrown ball. The reinforcer in the Negative Reinforcement (-R) example is the electric shock. Please note that I specifically stated that the reinforcer was the shock itself, NOT the removal of the shock.
(And yes, I do call it a shock because that's what it is. Please don't write to me and try to convince me that it's not a shock because I'm Australian and I call a spade a spade; unless it's a shovel, which is used for something else and if you're stuck using a spade when you need a shovel, you've got your work cut out for you!)
Any time you stop reinforcing a previously reinforced behavior something called "Extinction" occurs. Extinction usually results in the behavior getting stronger temporarily, then it diminishes, then it stops altogether.
If you train a recall and stop reinforcing it, eventually you will not have a recall any more. It doesn't matter if you used dog whispering or secret alien technologies sold to you by the KGB - if you stop reinforcement the behavior will eventually extinguish. (Please note that sometimes a behavior isn't reinforced by you and unless you prevent it, the behavior will be maintained.)
One of the touted advantages of -R training is that we can, usually quite quickly, move from "escape conditioning" to "avoidance conditioning". What this means is that we move from teaching the dog to escape the aversive to having the dog avoid the aversive. This seems great because the aversive is no longer required and we are training without any reinforcer.
However, what doesn't get written on the packet your fancy remote training collar came in is that avoidance conditioning is actually an extinction procedure - after all, you are training without a reinforcer.
That's right, if you do it enough times eventually the behavior will start to diminish and then you have to use your reinforcer again.
There is no such thing as a free lunch.
The same thing happens when you use schedules in +R training. You can fade the food rewards, but you still have to use them as often as is required to maintain the behavior by +R. This is known as a "schedule of reinforcement".
What we do know from thousands of repeated experiments is that some schedules of reinforcement are more resistant to extinction than others. Most notably, a Variable Schedule of Reinforcement is highly resistant to extinction. If your dog doesn't know when he is going to earn his next pay-cheque but knows that he will definitely get one, he keeps trying to earn it.
Another schedule that works really well if you're constantly aiming for improvement is, as I was taught it, known as "Continuous Reinforcement (Differential)". What this means is that we keep increasing our criteria and reinforcing on a continuous schedule. Every time the dog meets our criteria we reinforce. The catch is we keep raising the criteria a little. We stop reinforcing sub-par responses.
This is the best schedule for competition or service behaviors. I don't compete, but I would forget about "fading food from the ring" and just worry about training the required criteria on a continuous schedule of reinforcement. The behaviors are not going to fall apart because of one unreinforced trial that takes from 2-10 minutes.
The big important lesson from this article, no matter what method of training you choose, is this:
Any operant behavior you want to maintain or increase must be reinforced on a schedule or it will eventually extinguish.
Remember that and soon I'll post Part 2 of this article...