I got a question about how I add a cue to a shaped behavior, and why I do it that way. When you add a cue, you can choose to add the cue before the behavior, as the dog performs the behavior, or after the behavior (just before the click and/or reward). As I pointed out in this post last summer, I choose to add the cue before the behavior. It just makes much more sense to me than any other way of doing it, and it seems like my dogs learn faster than they did before (when I did differently).
A cue, or a discriminative stimulus, is information for the dog. It tells him what behavior will be reinforced (or punished) right there and then. Even before we’ve added a cue, there are discriminative stimuli at work. When there’s a behavior, there is a stimulus preceding it. When we add a verbal cue, there are already stimuli that cue the behavior (our position, what we’ve rewarded recently, the way we reward etc.). Our goal is to get the verbal cue to become the most important one and override the other stimuli. In order to do that, we need to make the new cue valuable to the dog, it needs to give information.
In theory, I think this is the most important thing. I try to make the cue mean something to the dog as soon as possible. When I only reward the behavior if the cue has been given, the cue becomes valuable to the dog. I also want to mix in other cues pretty soon, so that the cue is not only valuable, but also contains information on which behavior I want. I think this is the important part, and it can only be done with the cue happening before the behavior.
Pairing the cue with the behavior for a long time is probably not as important, but I do that as well. I can’t really explain it in theory, but it seems that you can add a cue to an operant behavior, using a procedure more like classical conditioning (pairing the cue with the behavior, not really focusing on the consequence). If we are using a classical conditioning procedure, it is absolutely most efficient if the cue is presented before the behavior (just like we would click before we give the treat, when we want to condition the clicker). In both operant and classical conditioning, the stimulus comes before the behavior.
If we add the cue at the same time, or after the dog performs the behavior, we’re not effectively pairing the cue with the behavior, and other stimuli will still be what gives the dog information about what behavior to perform. Of course, people teach their dogs cues in many different ways, so they all work. I just don’t think that adding the cue when the behavior already is happening is teaching the dog much at all. Eventually, most trainers change the timing and give the cue earlier and earlier, why not do it with perfect timing right from the start?).
Please write a comment and tell me what you think!