to Add the Cue
Since Helix invoked my name (smile), let me comment; it depends somewhat on the animal, the environment, and the behavior. By and large, when you get the behavior to the level of 80 percent of the time it, the behavior, is what you want (the final behavior), you should begin the cue training. This is one of the infamous "rules of thumb," where there is great latitude in when and how to apply it.
Here are some of the factors to consider: If the behavior is simple and easy for the animal, especially if it is something the animal naturally does on its own, such as a dog's SIT, I would cue it earlier rather than later. If the behavior is a precision behavior, that might be somewhat difficult to do, I would do it a bit later, when the final behavior is a bit stronger.
When we begin to cue is a function of strength of behavior and the rate of extinction. We spend time building strength of behavior (usually the number of correct and reinforced repetitions) using OC's principle of reinforcement. When we begin cue training, we will employ the principle of extinction to reduce both the uncued responses and those responses not meeting criteria (not what we want). The stronger we build the responses, especially responses not making criteria, the more extinction trials, and the longer it takes.
The reason I emphasize the trainer error of strengthening poor responses is that the animal learns best when there is the greatest clarity during training. If the animal fails for more than one reason, then the more difficult it will be for the animal to discern why it failed; the more reasons for failure, the muddier the picture, and the longer it will take for the animal to learn what you want.
We might say that there different kinds of errors. For now, consider CONTROL errors and PERFORMANCE errors. Control errors are those times when the animal gives you the behavior you want, but there is no cue, thus the response will not be reinforced. Performance errors are those trials where the animal's behavior does not meet criteria, and there will be no reinforcement, even if the behavior occurs after the cue is given.
Can we accept that the trainer's task, especially in early training, is to maximize success and minimize error? If this is so, then the trainer's objective should be to make it abundantly clear, and easy for the animal, to give a proper response and only after the cue is given. So, the behavior should be accurate enough (be what you want) most of the time, with fewer poor responses, so few trials are performance errors or a combination of performance and control errors. It also means that we must give frequent enough "cue-trials" to reduce control errors to an acceptable level.
Most certainly, we want to start with, and maintain, enough strength of the wanted behavior that we don't risk losing what we started with. Thus, we should maintain a high enough rate of reinforcement to keep the animal "playing our game."
Please bear in mind that much of cue training is made of judgment calls, and the world will not end if the trainer begins a little earlier or later. There are trade-offs, and what we should strive for is speed of learning accurate behavior.
I have repeated myself, and I meant to. The judgment factor in cueing opens words and statements to wide interpretation. As some of you know, much of my intermediate level operant conditioning classes is concerned with cueing. Most trained behaviors are elicited by cues, either specifically given by the trainer, or context cues, from the environment. So, cueing is important, as well as being subject to trainer judgment.
Copyright of all posts is the property of the original author. Please obtain permission from the original author before copying, quoting, or forwarding.
List and Site Owner: Melissa Alexander, mca @ clickersolutions.com