More on Commands and Signals
One of the apparent differences between clicker training and other methods is the way signals are attached to behaviors. In traditional forms of training, the command is usually chanted during the learning process before the behavior is learned. In clicker training, the behavior is shaped first and the command is added later. The underlying principle that allows both methods to work is basic and natural. When a signal consistently precedes a significant event an animal will make an association between the two. It's that simple.
About 100 years ago, Ivan Pavlov studied the way the brain connects information and forms associations. His method of study was to make associations between sights, sounds, smells and touches with things that triggered natural reflex actions like salivation and muscular contraction. Though it was not his primary interest, he also came up with some solid information for trainers. For instance, his most famous experiment uncovered the "signal first" rule. He created a sequence where a signal (the sound of a bell) was followed by a significant event. (meat paste) Within 20-50 repetitions, the dog would make the association between bell and meat and salivate at the mere sound of the bell. To fully test his new knowledge, Pavlov did what good scientists do - he reversed the process. He grabbed a likely dog and tried to have the significant event (meat paste) come first, then the signal. (ring the bell) Six hundred repetitions later, the dog still didn't show any interest in the bell. The same dog was then taught the association using the A) ring the bell B) feed the dog sequence. Within about 20 repetitions, the bell was starting to cause salivation. The conclusion is pretty easy to see. If you want to connect a signal to a consequence, the signal comes first.
Even though Pavlov wasn't connecting signals to behaviors, the information is still valid. To connect signals to behaviors, all you have to do is "Oreo" the behavior between the signal and the consequence. This requires that you know when the behavior is going to happen. That's why it is important to maintain a high rate of response when you are attaching a cue to a behavior. High rates of response tend to create cycles that lead in a predictable pattern from behavior to reinforcement, over and over.
Here are some pointers for connecting signals to behaviors...
1) If you want to connect a signal/cue/command to a behavior, try to create a response cycle that starts with the dog initiating the behavior and ends with a click and treat. Once the sequence is predictable, start giving the cue just before the behavior starts -- which you can also think of as just after you presented the actual reinforcement for the last repetition. Try not to have more than a 10 second delay between the time you give the cue and the time the dog starts the behavior. If the dog's cycle of response is slower than that, just keep your lip zipped for a few more repetitions. If the dog is performing the cycle so quickly that you can't squeeze a word in edge-wise, start tossing the treat a few feet away to create a momentary break in between each repetition.
2) If you have a hand signal and a spoken command that you wish to attach to a behavior, put them in sequence. To create a hand signal for any behavior, you need to have the dog looking at you (a high probability when you have your response cycle up to speed) and put the new hand signal in front of the spoken command. After about 20 repetitions, drop out the spoken command and continue with the hand signal. Over a series of training sessions, start alternating between hand signal and spoken command. Remember, hand signals work best when they include movement that changes your silhouette. If you want to learn a lot about hand signals, set up a slide-projector and screen. Stand between the light and the screen and practice your motions. The high contrast silhouette that you see will be a pretty good likeness of what your dog sees every time you are back-lit. (which happens a lot when your dog is looking up at you)
3) Say commands softly when you attach them to a behavior: As sound travels through the air, it becomes softer. No matter how you can bellow up close, your dog is going to hear a softer version at a distance. To make sure there isn't any confusion later, start training with normal tone of voice to soft tones so your dog specifically learns the sound of the words rather than the tone of your voice.
4) Don't "Cry Wolf": Only say the command once. There is a natural human tendency to repeat a command if there is any delay on the part of the dog. This is a guaranteed way to screw up the proceedings. Just like the boy who cried "wolf", if you make noises that don't accurately predict events or consequences, your dog will learn to ignore your noises. To avoid this problem, anytime a command does not trigger an immediate response, give a "you blew that one" signal and connect it to an appropriate consequences. If you think the error is the result of trying to attach the cue too early, then way "wrong" and do some more "cue-less" repetitions. If you think the dog understands perfectly well, but is merely blowing you off, then you must say "NO!" and attach your consequence of choice for failure.
Understanding the basic principles of how animals respond to cues and signals is a cornerstone of good training. Developing good habits for connecting signals to behaviors allows you to help your dog learn to respond correctly the first time - so you'll never have to say it twice twice.
Go to the first article | Go to Prior Article | Go to Next Article | Back to Clicker Project Index