When I talk to other trainers I tend to get specific and speak in scientific terms. Things like Differential Reinforcement, Discriminative Stimulus, the Four Quadrants, etc. When I speak to customers, students, or the public at large, I tend to speak in much broader, general terms. It's kind of like me not needing to know about programming when I get my computer worked on. I did want to get a little more specific though about something I mentioned in my last blog: clicker or marker training.
In my last blog, I grossly generalized clicker training. Now that we have the basics, I wanted to get a little more specific. Basically how it works is you use a unique stimulus to mark a behavior. This is a way to tell the dog he has done something correctly at the moment it occurs. Because the marker (click or a specific word) has been paired with a reinforcer (food, praise, etc.), it tells the dog that when he hears the marker a reinforcer is coming. It really isn't so much a training technique as it is a system of reinforcement.
I like to use a clicker because it is a unique sound that carries very well over most noise. In class we tend to use the word "yesssss," drawing out the "s" to make the sound unique so as not to be confused with just saying "yes" in regular conversation. This saves you from having to carry around something extra in your pocket. For the purposes of the rest of the blog I'll refer to "click" as whatever marker you like to use. I will also refer to any reinforcer you care to use as "treat" even though it can be praise, a favorite toy, or anything the dog enjoys.
First you have to start by "charging" the clicker. You simply have the dog close to you, click and treat. The treat should follow the click within about a second to be the most effective. The more the dog wants the treat, the faster this will work. It should only take very few repetitions before he starts to associate the sound with the reinforcer. Once this pairing has taken place you will be able to delay the delivery of the treat because the click will elicit the same emotional response as the treat. I was able to charge the clicker with my dogs in about 2 minutes, but would start any training for about 2 more weeks by charging, just to make sure it was stable.
Now, if I am working on a sit, I can ask Duncan to sit, and as soon as his bottom hits the floor, click and treat. If I'm working on target training, as soon as his nose hits the target, click and treat. Whatever the training, as soon as it happens, I can mark it and he knows a treat is coming.
The two most common questions I hear at this point are "How do you get the behavior for a longer period of time?" and "What if you don't have a clicker or treat handy?"
To get the behavior to last longer, we'll take the sit for an example. Have the dog sit, and as soon as he's sitting, click and treat. Once he's sitting without any prompting five to ten times in a row, you can delay the click and treat for about a second. Once he's doing that amount of time consistently, you can delay the click and treat for two seconds, then three, and so on.
Something to remember at this point is to keep the training sessions short at first. You may only want to do about five minutes at first, and in that time you may only get a three second sit inconsistently. In the next training session you'll want to back up one or two steps before you move forward. Once he gets used to the training routine you can go for about 15 minutes, but usually it's not recommended to go longer than that. You can do two or three training sessions a day, but be aware of your dog's boredom and/or stress levels.
The point of training a behavior is so the dog will respond in real-life situations. In real-life, you won't always have a clicker or treats on hand. If you don't have a treat, don't click. The click means a treat is coming and if you that pairing doesn't happen, it will begin to lose its affect.
When you ask for a behavior - we'll use the sit again - if the behavior has a history of being reinforced, he is more likely to offer the behavior even if it does not get the click and treat. You don't want to do this too often, but enough that you will get a response in those real-life instances. Think of a slot machine. A person may not be reinforced - hit a jackpot - every time they put a quarter in, but it happens often enough that some people will still offer the behavior - keep playing - for hours on end. As long as the behavior has a history of reinforcement, there's no reason you can't get the behavior at any time.
Something to keep in mind with clicker training: it was popularized with training sea mammals. This is how killer whales and dolphins are trainer. It has also been shown to work by coaches training athletes such as swimmers and gymnasts. It can work on any vertebrate from fish to lizards to dogs.
Maybe later I'll do an entry about why clicker training works so well, but this is more about method, so I'll stop there for now.
No comments:
Post a Comment