Commands vs. Cues and Signals
Almost every hobby, discipline or science has a way of adopting words to suit it's own purpose. Often, these special terms have older and more basic meanings in general conversation. For instance, in the context of obedience regulations, a command is a verbal cue and a signal is a visual cue. In order to fully discuss how dogs learn to respond to specific cues, we need to use these words in their more general meanings. So, for the purpose of this column we will stick to Webster's definitions.
Command -- "to give an order"
Signal -- "anything that occasions a certain action or response."
Cue - "Anything serving as a signal to do something"
First of all, the term "command" implies a "commander" and that the "commanded" knows how to follow the order. For instance, if you don't know how to speak Russian, it won't do much do good for someone to command you to translate a Russian newspaper. From the perspective of clicker training, a command is something that is developed late in the process of training, (after the behavior is in existence and almost to a performance level) if at all. We will cover commands in a later column.
The two words that are really important to this discussion are cues and signals. A "Cue" is anything that triggers a behavior. This term might apply to a learned behavior, like roll-over, an instinctive behavior like chasing a frisbee or a basic response like salivation or trembling. It has roughly the same meaning as the scientific term, stimulus, but has the advantages of less letters and a simpler form of pluralization - just add S.
As the definition implies, behind the occurrence of every behavior is a cue. For instance, if you use a food lure to get a dog to sit, the movement of the food is the cue that initiates the behavior. This same cue might also trigger sniffing the air or looking on the ground for a treat. In essence, cues are those environmental things which cause the dog to respond.
In contrast to the more general term cue, a signal describes a more specific association between a particular cue and a particular behavior. While a target stick or a food lure can trigger many variations of "targeting", the signal "sit" should trigger only one behavior. For clicker trainers, this is the heart of the matter. Teaching a dog that a specific behavior is connected to a specific signal is the anchor of dependable performance.
TV's, Sienfeld, The Cable Guy and Remote Control.
One of the cornerstones of clicker training is the knowledge that the signal doesn't cause the behavior to happen. A signal is a trigger - like a TV remote control -- nothing more. We know that the TV remote doesn't "cause" Seinfeld. The remote merely selects one of the options from a large system of television set, network production, and local affiliate. If the television is unplugged, the program is canceled, you don't get Seinfeld, no matter how much you monkey with the remote.
The same logic applies to obedience signals. A signal can only trigger behaviors that already exist. Chanting a word while the dog learns a behavior is like punching the buttons on a remote while the cable guy is still trying to hook up the cable. Indeed, if you keep pushing the buttons long enough, something will happen - but the remote didn't "cause" it to happen.
The Zenith Motto -- "The quality goes in before the name goes on."
The second thing we need to know about signals is that teaching a behavior and attaching a signal are two different training tasks. Since a signal without a behavior is a pretty useless item, it is logical that the behavior must be shaped before the signal is attached. Not only is this simple sequence important, but later precision and reliability is also affected by when the signal is attached. I.e. If you attach a perfectly good signal to a crappy version of the behavior, you will get a crappy behavior when you give the signal. If you then improve the behavior, but retain the signal that was connected to the crappy behavior, you run the risk that the dog will revert to the older, crappy version of the behavior. In effect, you have created multiple definitions of the same word or hand signal. This is a great way to confuse the heck out of your dog and a great way to screw up a performance.
For more experienced "traditional trainers," refraining from using a "command" word may feel really uncomfortable at first. If you find that you are having trouble keeping your lip zipped, here is a compromise that won't screw up your dog while you learn this new method. - teach your dog a "working" cue while shaping the behavior, that you will later replace with a "performance" cue. If you use the term "drop" for drop on recall, make your working cue "flat" or "place" Once the behavior gets polished enough to satisfy your taste, you can easily attach the real cue -- "drop."
Now that we have some of the talk-talk out of the way, it's time to put this stuff into a real training situation. Let's start with the basic routine that occurs during shaping.
1) You get the dog to do the behavior.
4) Repeat 1-3
Depending on the complexity of the behavior, this sequence might take one or more training sessions. Once you have a pretty good version of the behavior, try to get the cycle short enough so that there is almost no hesitation between repetitions. Now it's time to switch to the pattern for making a cue become a signal. Once the behavior is on a short cycle, inject the cue between steps 3 and 1.
1) You get the dog to do the behavior.
2) Click and give the treat
3) The dog eats the treat
4) The dog offers the behavior again
5) Present the signal that you want to associate with the behavior.
6) The dog offers the behavior
9) Dog eats treat
10) Repeat 5-9 as needed.
Correct repetitions receive a click and treat, while incorrect attempts cause a "wrong." Make sure you say "wrong" in a totally neutral tone of voice. It is meant to communicate that a particular behavior didn't work, rather than a scolding for doing something bad.
If you follow this simple pattern, after between 20-50 repetitions, the link between the cue and the behavior will start to form. To give you an idea of how long this takes, I once counted the number of repetitions in a 15 minute period with a client's 12 week old husky. The behavior was "down at a distance" a la utility signals. The puppy offered 62 repetitions in 15 minutes. So, 200 repetitions might take three 15 minute training sessions. Your mileage may vary.
To see how easy this process is, here is a simple project for you -
1) Use your target stick to get the dog to bump an object - like a key ring.
2) Reinforce key touching until it is on about a 5-10 second cycle.
3) Start saying "keys" between the last gulp of the treat (end of repetition) and the next bumping of the keys. (start of next repetition)
4) Do at least 50 repetitions of this sequence - "Keys" -- dog bumps keys - click - treat.
5) Teach at least 3 more items.
6) Lay the items on the ground and start asking for them by name. Use clicks and treats for correct repetitions, and "wrong" for mistakes. Try to keep the success rate rather high at first, then gradually start asking for several correct identifications in a row, without a click and treat. If you lose the behavior, drop your standards and make it easier for the dog to earn a reinforcement.
Additional Rules for Attaching signals - clicker style.
1) A good way of determining the proper time for attaching the cue is to imagine that you want to show off for a friend. You just called the friend in to the back yard to see Fido do his newly learned "trick." Here's the test -- if you don't think the dog will "make you proud" and do the behavior within the next 10 seconds, it is too soon to start adding the cue.
2) If you find yourself saying "wrong" a lot, stop talking, drop your standards and get the behavior on a shorter time cycle.
3) Don't give the signal twice twice. If the dog fails to offer the behavior after you use the signal, say "wrong" and end the repetition. You must make sure that you don't get in the habit of multiple commands, or the dog will wait for the second signal before responding.
4) Give the cue in a normal tone of voice or slightly softer than normal. The idea is to make the particular word important rather than a particular volume or inflection.
5) If you are going to teach a hand or arm signal, make sure it is in silhouette - away from your body. To see what I mean, stand next to a wall in the sunlight and run through your hand signals. If your shadow doesn't change, that means your silhouette isn't changing. What ever signal you are using may be very difficult or impossible for your dog to see if you are back lighted.
6) If you wish to add a new signal to a behavior, try this sequence. New signal -- Old signal - Behavior - Click - Treat. After about 20 repetitions, stop giving the old signal and watch what happens. The odds are very high that the dog will continue to perform the behavior with the new signal. If not, bring back the old signal for awhile until the dog has a chance to complete more successful repetitions.
Go to the next article | Go to Prior Article