Clicker Agility Training
With the ever growing popularity of positive reinforcement in dog training, we, as agility enthusiasts, should be sure as we are not left behind. Karen Pryor, a pioneer in the concept of using conditioned reinforcers to train animals, wrote a book "Don't Shoot The Dog" which is a must-read for any motivational dog trainer today.
By way of a brief introduction, a primary reinforcer is anything the dog wants or desires. This may be different for each dog. Examples are food (for those chow hounds out there), or toys. To repeat, a primary reinforcer is a reward the animal will work for. A secondary or conditioned reinforcer is something the animal learns is a symbol for their primary reinforcer. This may be a "click" from a clicker, a whistle , a word (example "yes" or "ready") or a physical touch like a pat on the head . It is probably better to use the term conditioned reinforcer (c/r) rather than secondary reinforcer as we don't want to imply the click comes after the treat. The click must come before the treat. Used correctly, the dog will quickly learn that when they hear the conditioned reinforcer they have performed the task correctly and a reward is coming.
It was Karen Pryor who, in the 1950's, pioneered this technique to train dolphins to perform in Sea World performances, but the concept was stumbled upon long before. In 1924, the researcher Pavlov was working on an experiment and discovered the fact that dogs would salivated when they heard a bell (a conditioned reinforcer) if it was followed by the reward of food (a primary reinforcer). He found most dogs became conditioned to the bell and started to salivate after 20-50 repetitions of this pattern (bell--food--salivation). When he tried a new group of dogs with the reverse pattern of giving food--then ringing the bell--then more food -- after 500 repetitions the dogs still weren't conditioned to salivate on the bell. So he then took this same seemingly "untrainable" group of dogs, tried the original pattern of a conditioned reinforcer (the bell) before the food and voila the dogs caught on after only 25 repetitions! Try to remember this when you lament "my dog will only work for me if I have food or am carrying his favorite squeaky toy". Ask yourself "...have I given my dog any reason to work for me without his primary motivator?" For food etc. to be most effective when used in training we must first pair them with a conditioned reinforcer. Personally, I find the use of a clicker and/or a one syllable word example "yes" to be most effective.
Once you have reliably established a conditioned reinforcer, you will find your dog will work harder and longer for a "click" than he ever would for a piece of food dangled in front of his nose. The results are that the clicker, or whatever your c/r is, becomes a stronger motivator to your dog than the food ever could be. It is like the anticipation of something wonderful. Now that the dog understands this you can use it to shape any behavior you want the animal to perform or improve. For example if you have a dog that cuts out of the last weave pole you can click while he is still weaving--i.e. before he cuts out. As long as the dog understand the use of the clicker and you are able to click prior to the dog cutting out---you will be able to shape the dog's behavior into a perfectly performed weave.
Behaviorist B.F. Skinner's proved that the more frequently a behavior is reinforced, the more frequently an animal will repeat it in hopes of further reinforcement. The conditioned reinforcer acts as a pointer, allowing you to indicate to the dog which part of the behavior you want repeated. In the case of the weave poles, you can inform the dog, staying in the poles is good and that is where you will receive your click. Eventually the dog will continue weaving a little longer (as you raise the criteria and hold off clicking until they do one more pole) and then finally until you have perfectly performed weave poles with a click happening at the end. The advantage of this over using force to teach the same thing is that the dog is making the decision himself and the repeatability is a lot higher especially when you get greater distance between you and the dog, plus it is a lot more fun for both of you.
To learn more about the use of c/r read Karen's Don't Shoot the Dog, try to attend a seminar on shaping behavior (Karen gives seminars with Gary Wilkes--they are excellent and well worth the money) or contact Sunshine Books (1-800-47-CLICK) to obtain one of Gary's "Clicker Starter Kits" (includes a video, a booklet and 2 clickers a deal at $49.95).
The clicker can be an effective method of backchaining contacts for fast and reliable performance. The dog is first taught to touch the target on the ground. You simply drop it in front of him and c/t when the dog looks at it/lowers his head/touches it --in a normal shaping progression. Once the dog reliably touches the target, go for a twofer--that is do not click until he hits it twice (be patient). From there you can add your "target" cue or command as long as you are sure the dog is going to go and touch the target. Only say it once and wait. Once he will touch it on cue, stop reinforcing him if he touches it without a cue or command--this is how you gain stimulus control. You can now move the target to different rooms of the house; changing locations helps make the behaviour stronger.
Then put the target at the end of a flight of stairs to mimic the angles of a contact obstacle. You start the dog from the last stair so he only has to go a very short distance to touch the target--have the target far enough out in front of the dog so he has to put his front feet on the floor to reach the target but not so far that he lets his back feet come off the stairs. Backchain on the stairs until you have the dog taking up his contact position from the top of the flight of stairs, on cue. If you really want a laugh, once this behaviour is learned, ring the door bell before you say "target" (providing you have stairs near the door). If you keep preceding your cue target with the ringing of the door bell you will teach the dog that both signals mean target- kind of like having a signal and voice command for the same behaviour. If you don't want to do the door bell thing (it is optional but very amusing when instead of jumping all over visitors you have three dogs in their contact position at the bottom of a flight of stairs) move the target now to the bottom of your contact obstacle. Now backchain again the same way you did on the stairs, but this time start one wrung up from the ground on a very low A Frame (3 foot apex).
Set a table beside the A Frame for the dog to climb up on so you don't have to pick the dog up to get him at the bottom the the A Frame. Hold the dog's collar, get him all hyped up and say "target" to which he will spring forward to touch his nose to the target. Nose is better than feet, as when you have the dog use his feet he sometimes leaps at the target and sends it flying--which also conditions a little jump when he hears the word "target". You don't want him to jump at the target as he may end up jumping over the contact zone.
All of the concepts I teach to the dog are done through operant conditioning. The dogs are clicker trained and are reinforced when they touch the target with their noses, with their front feet on the ground and their back feet on the contact obstacle. Once the dog has mastered that part of the contact training and has the "cue" target under complete stimulus control, I add part two to the contact training. The dog, on a flexi, is put in a sit not near agility equipment. He is taught a *motivational* wait. In doing this you ask the dog to wait and, standing 1 foot in front of the dog, you pull gently on the flexi, hoping that the dog will resist and pull back without moving. If the dog breaks you simply say a quiet "wrong" and ask the dog to sit again, back in the same location. Make this a really fun game for the dog so you can say exciting words for him to enjoy it. No correction should be given to the dog. If the dog can wait for a couple of seconds I say --very motivationally...one...two..three go!!!! On the word go the dog leaps forward to tug on a waiting favorite toy. Once the dog masters this game and is truly enjoying it I do the same thing from and stand and then move it to the contact. This accomplishes two things. It gets the dog all hyped up to explode from the end of the contact and move on with alot of speed to the next obstacle, and the second thing is it makes it clear in the dog's mind----this is your contact position and this is not. You must never anticipate the *BIG RELEASE GAME* until you are given the cue to *go*---that may be the word "go" just as the game or as the dog gains agility experience it is simply the name of the next obstacle.
Obviously the written description of the game is not as exciting as the real thing. Getting the dog all hyped up about leaving his contact position is like mimicking the excitment the dog will feel at a trial when he wants to blow off the contact whenever he likes. They do like the game and they will try to anticipate it--so you can allow the dog to learn, regardless of how much fun you think there lies ahead for you, you are never allowed to leave the end of the contact obstacle until I give the cue to do so.
You really don't need a target or a lure or any other prop to teach the *concept* of distance work with a clicker. Most clicker addicts find that distance work is where the silly contraption really outshines the competition. Like most elements of skill, it's best to teach the distance concept away from the agility course at first.
You could start by teaching some simple distance tricks . Favorites that are all easy to do indoors or out, include "back up," "look back," arm signals for "this way, that way," and "go." The idea is to expand the dog's distance comfort zone first, then bring that newly-expanded skill to the agility course.
Decide which trick you're going to build on. It is best to change other variables around the one task rather than throw multiple tasks at the dog while they are dealing with the new context. If you choose "go," for example, and the dog is going out 30 feet or more in other contexts but only 15 feet in agility, then set them initially LESS than 15 feet from their favorite obstacle to begin your agility distance session. It might seem that you should click the dog when they reach and commit to the obstacle, or even when the dog completes it correctly, but if you are working on *distance*, click "movement away" rather than the obstacle performance per se right now. There's plenty of time to delay your clicks until the dog commits/completes, and you'll need to do that, but not yet! That's a later step.
In the first session on course, concentrate on increasing distance and clicking the dog's willingness to travel further away. Even in small increments, the distance will increase very rapidly this way. You should vary things like your relative position and which obstacle. This is one of many cases where the foundation work sounds like it will take too long, but it really goes quickly and lasts forever.
After a session of clicking only distance, you'll be ready to complicate the job by moving closer to the obstacles again and clicking for commitment and/or completion (mix it up, and sometimes click only for distance, to avoid patterning). At first, the dog might look back at you waiting for a click, but remain neutral until they commit; click that and jackpot on the spot. It doesn't matter whether they complete the obstacle. You'll get to that. Now raise your criteria again and require more distance, or further commitment, etc. You're increasing the challenge in increments that the dog can figure out and control in order to train you to give them the cookies. The important part of OC is that *you both want the same thing*. If you encounter resistance, you've gone too fast and need to go back several steps.
Only two cautions: 1) Because it seems to be so easy, people tend to keep going after they should end the session. Let latent learning help you. Let that lesson sink in for awhile before you begin another with an abbreviated review and more progress. 2) Don't forget to vary lots of peripheral variables while you keep the dog's main focus the same. You do not want to make them dependent on some unexpected element (like the famous "handler position" problem in weave poles or the "luring" requirement for contacts). If a pattern is too often repeated, the dog will naturally grow to count on it, so don't get stuck on any of your steps too long.
If you'd like your dog to do reliable go outs without using food or toys as targets, here is an idea. It is a clicker training method.
- Put a dog blanket on the floor. Step onto the blanket with the dog. When his feet hit the blanket click and treat.
- As soon as the dog is stepping onto the blanket before you do, start to stop at the edge. When he goes on click and treat.
- Begin to stop farther away from the blanket and let him go on, click when he hits the blanket and treat.
- At first you are taking the treat to him. Sometime during the process he will start coming to you for the treat.
- When you are practically certain that the dog will go out to the blanket when you and the dog start moving toward it, you can begin to add a command.
- Once he understands the command, then start using a variable schedule of clicking and treating for go outs. For example, faster, straighter, etc. and begin to fade the blanket.
- Even a Velcro non-retrieving breed like a Scottie will be doing fast go outs to a blanket in about 10 minutes. Then you place a dumbbell on the blanket and click/treat for picking it up, then picking it up and bringing toward you, then closer to you, then to you and releasing it into your hands. This is all possible in one 30 minute training session.
If you teach your dog to pick up the dumbbell in a separate session before you add the go-out, teaching the retrieve will be much faster. When you teach the go out, getting a complete retrieve is easy because picking up a dumbbell is a skill your dog already knows.
Clicker training also works great for sending your dog over jumps, through tunnels, and weaves also. Remember, clicker methods don't have to be slow. Just remember to shape one behavior at a time, then chain them together.
|« Working Ahead|||||FAQ Home||||