The click is clear
The click means one thing only: "Bingo. You win."
It's like when you're waiting for a special call and you hear the phone ring. It's not the reward itself; it's just the clear-cut, simple message: "You got it."
The click is more than just a conditioned reinforcer. It's what engineers call feedback. It marks or identifies precisely which behavior is paying off and should be done again. That's why we call it a marker. (FYI, this use of the word "marker" is not a clicker training invention; the phrase "event marker" was first coined by Skinner's protégé, the late Ogden Lindsley, when he used a sound to teach his pet donkey to open a mailbox.)
The voice is a wonderful tool for conveying the trainer's pleasure and pride, and for giving warnings, instructions, and cues—and we use it for all those purposes. But because it has all these functions, the voice does not make a very good marker. The brain must parse out all that other information.
The result seems to be that "clicker is quicker." We know this from practical experience. For example, the complaint I heard many times in the early days of clicker training's growth in the dog world was this:
"I converted all my classes to clicker training, and now my students are going through my six-week curriculum in four weeks."
That gave the teacher a new problem—how to keep the students busy during the extra two weeks they'd paid for!
But is this difference in speed of learning real? And, if so, is it due just to the click? Psychologist and clicker trainer Lindsay Wood decided to take a look at the question experimentally. She trained a number of naive shelter dogs to cross the room and bump a target, using the click as a marker, and taught another group of dogs to perform the same behavior using the word "yes" as a marker. Same trainer, same skills, similar clueless dogs, but the clicker dogs learned the behavior roughly 40% faster.
And Lindsay teased out another very interesting piece of information—the clicker only made that difference during the learning phase. The word "yes" was as good as a click for maintaining the behavior once the dog already knew what to do. It was only when the dog was trying to figure out what worked that the click displayed its power.
With the author's permission, we've posted Lindsay Wood's master's thesis on our website so you can see just how the experiment was set up and go over the data for yourself.
The next big question, of course, is WHY the difference occurs. That's due to how reinforcers are processed in the brain. Neuroscientists know something about that, and I'm reporting on it in my upcoming book, Reaching the Animal Mind. The manuscript has at last gone to press and will be in bookstores in June of 2009.
Click is better "mathematically" than voice as a marker as well
I have been giving a lot of thought to the Click vs. Voice for several weeks. My thoughts run like this:
- I need to minimize the number of inputs to the animal so that he knows what the marker actually is. I need to minimize all body movement so that there is no variability. The marker needs to be the click and that is all. Animals don't know instinctually to pay attention to the noise, but ignore all body movement.
- A voice has a wide variability based on my emotion and environment. When the animal performs especially well, or when it perfoms marginally, I cannot help but vary my voice in ways that I may not even be aware of.
If my goal is to eliminate unnecessary inputs in order to focus communication with the animal, a clicker is better than a voice because it has less variability.
I have heard people claim that they have experimented (in a loose sense) and found that a marker or bridging stimulus isn't even necessary - that the animal was trained just as well simply by delivering the primary reinforcer quickly. I suspect that their experiment was flawed - I highly suspect that they had established a marker with that animal in the form of voluntary or involuntary body movements, breathing, or something else. What they found was that their technique and mechanics were no more effective if they used a clicker or not. Bad mechanics don't prove anything.
When I can use a clicker, it is better than a voice for several reasons - one is that it minimizes extraneous inputs and focus communication between me and the animal. The way the brain processes different types of inputs is another fascinating aspect.
Post new comment