Home » working dogs

Back Chaining - Morten and Cecilie Style

Back chaining is something they do to connect the basic behaviors for competition.  They only start back chaining AFTER the basic cues needed are put on cue (so they are all fully fluent) and they said if a cue has been poisoned, it can't effectively be used in a chain.  The reason is, that when it's done correctly, each behavior in the chain reinforces the previously performed bahavior.  If a cue has been poisoned, then it fails to reinforce the behavior that comes before it in the chain and the chain will start to fall apart there.
For multi-step behaviors, like a directed retrieve for example, they break the chain into smaller ones for training.
To make it easier to put those small chains together into larger chains, they include some over lap.
So, what the heck does THAT all mean?  I think it is best explained by using an example.  So let's look at the directed retrieve (sending the dog away from to you get them to pick-up and bring back an obedience dumbell. Which is an object, not a clueless person LOL)
The full behavior, as seen in U.S. competition is this:
Dog starts by sitting in heel position.
Dog is cued to STAY (in some venues)
Handler throws the dumbell
Handler cues the dog to go get it ("TAKE IT")
Dog goes out to the dumbell, picks it up and returns to the handler in "front" position (in the U.S.)
Dog holds the dumbell till asked by the handler to RELEASE it
After handler has the dumbell, the dog is cued to return to heel position ("FINISH")
I put the cues in capitals, but you can see there are many behaviors that comprise the cues, especially "Take it."  So those will be broken down and taught and perfected as a chain (or chains) before the final cue for all those behaviors is added.  There will be many chains taught to the dog before cues are added and before this exercise is ever done all the way through.
By breaking down each behavior and building them up to fluency before adding other behaviors, you are sure that the dog understands each component.  Plus, each component is fun for the dog and motivating so it can act later as a reward during the chain.
The reason for teaching the behaviors in a backward order has to do with reward histories.  You want the final behavior of the chain to have the strongest reward history so the dog is most motivated to DO that final behavior.  By starting with short chains, the dog learns that the only way to GET to do that final behavior of the chain is to first do the others correctly.  Because the final behavior is most well known, the dog is always moving from his least reinforced behavior to the most reinforced one (relative to this practice session and later generalized to the practice related to the whole chain.) It is the Premack principle, so it's a very strong motivator for completing the chain.
Here is how I would break up the above complex behavior.  I would be working on these mini chains simultaneously, but in a rough order, so they can all build fluency and can later be combined with each other.
The forward order of the behaviors with numbers I'll refer to in the back chain:
1) Sit/Stay in heel while dumbell is thrown
2) Send out to dumbell
3) Pick-up dumbell
4) Return to handler
5) Hold dumbell while moving
6) Front position
7) Hold dumbell while sitting
8) Release the dumbell
9) Move from "front" to heel position "finish"
That is a LOT of behaviors!  By breaking them down and teaching them backwards, the dog always knows what is coming next.  A big revelation in the session for me was that, OF COURSE the dog is going to try to anticipate! It is not bad, it is simply information you can give the dog: "no, that's not going to earn a click, only waiting for my cue will earn a click even though you know what is next."  They said you should work through this anticipation stage with each mini chain BEFORE you combine the chain with others.  The dog needs to be very clear that even though it knows what you are about to ask for, it can't respond till you cue it.  This is why the behaviors need to be fully fluent BEFORE you start chaining them!  The great thing is that because the dog DOES know what is next in the chain, when he has to wait for a cue, he is "on the edge of his seat" in anticipation and when the cue IS given, his response is so FAST it looks nearly like anticipation. There is NO latency.
So here are some chains I'd work on and build to fluency simultaneously:
Behavior 9: moving from "front" to heel position.  Dog has to learn to wait for the "finish" cue. Once that is established, Behavior 9 "finish" can be combined with other behaviors.
Behavior 7 (hold dumbell while sitting) then Behavior 8- (release it on cue).  Dog has to learn to wait for the release word and not cue off the motion of your hands toward the dumbell.
Behavior 4 (return to handler) should mean return to behavior 6 (front position) so I'd start with the dog in "front" and take a step back- dog should then move to front for the reward.  Gradually build the steps taken before the dog is released by a helper to "front."  Dog is held by a helper as you move away so you can build up distance, but the dog is allowed to pull toward you (if he chooses) so he shoots forward as soon as he's released.  You WANT the dog driving on the return.
Behavior 7 (hold dumbell while sitting) and Behavior 6 (front)  Basically the same exercise as above, but now the dumbell is involved.  The front position is what is clicked, so the dog has to learn duration and not to spit it out till it hears the click.
Behavior 5 (hold dumbell while moving) and Behavior chain 6/7 (return to front and sit) A helper holds the dog while you move away.  Helper hands the dumbell to the dog so the dog can carry the dumbell back to you.  Again, it's the sitting in front holding the dumbell that gets clicked.  Dog has to release it to get the reward.  You aren't asking for or cueing the release yet.
But once Behaviors 5, 6 and 7 are a fluent chain, you could add in behavior 8 (release on cue) by STARTING the practice with release on cue (which is already a fluent behavior) several times so it now becomes the most reinforced behavior, then ask for 7: hold while sitting before you ask for the release and click/reward the release on cue. Behavior 7 becomes a way for the dog to GET to do #8 (release on cue.)  Once that is well established (6 is part of 7 because the dog is holding while in front) you can add #5 with a helpers help.  By now, the dog is driving toward you with the dumbell because he knows he has to get to you, sit in front, and hold till the cue is given to get the reward.  But the beautiful thing is that each of those behaviors becomes rewarding because they get him closer to the final goal in this chain.
To add #3 (pick-up the dumbell)  instead of the helper handing the dumbell to the dog, have the helper drop the dumbell right in front of the dog so the dog has to pick it up, then return to the handler and do #6/7 before the click/reward.  Later, have the helper that is holding the dog, toss the dumbell behind the dog, so the dog starts to move away from the handler to pick-up the dumbell before they can return with it.
I think Behavior 1 (stay while dumbell is thrown) should be worked on by itself and proofed before you start sending the dog for the dumbell.  Have a helper that can pick-up the tossed dumbell if the dog breaks his stay.
Now you can add behavior 2: the send-out to get the dumbell, after behavior #1 (wait as it's thrown.)  Start with short distances and gradually increase the distance the dog needs to go.  When it's fluent in all environments and with all distractions, you can start adding this to other chains.
Just a note, so far in all these chains, the only cues used so far are "release" the dumbell and "finish" after those behavior chains are fluent.  The cue for the "take it" hasn't been used because that chain hasn't been combined yet.  So now you can teach "take it" by combining those behaviors in this order:
4- 5/6 several times then
3 - 4 - 5/6 several times then
2 - 3 - 4 - 5/6 several times then
1 - 2 - 3 - 4 - 5/6  (once this is strong and fluent, the cue for "take it" can be added)
So now the dog knows how to stay in heel while the dumbell is tossed, can move away from the handler to pick-up the dumbell, bring it back, sit in front, hold till cued to release and can finish. But these chains haven't been put together fully.  When you have "proofed" all these mini chains and the dog can do them reliably in any environment and has gone through the anticipation stage with each chain, then you can start making longer chains.  This might take several months or maybe even years- depending on the dog and practice schedule! DON'T RUSH IT!  If any behavior in the chain is weak, it will compromise the whole chain. 
If at any time the dog makes a mistake- stop the chain immediately without fan fare.  Since each behavior in the chain rewards the previous behavior, you don't want the mistake to be rewarded by letting the dog do the next behavior.
Always start with the last behavior first!  Putting it all together:
9 several times then
6/7 -8 - 9 several times then
4/5 - 6/7 -8 -9 several times then
3 - 4/5 - 6/7 - 8 -9 several times then
2- 3 -4/5 - 6/7 - 8 - 9 several times then finally
1 - 2 - 3 - 4/5 - 6/7 - 8 - 9
By working it this way, the dog is highly motivated to get to the finish position because that has the strongest reward history.  But he has learned that the only way to get there is by doing the other behaviors as fast and as correctly as possible.
It's a beautiful thing.

Chris Puls "Scoutdogs"
Troop Leader for Dog Scout Troop 107 (OH-KY-IN tri-state)
Captain of the Dearborn County Animal Response Team
My K9 boys: www.Dog-GoneHappy.itgo.com
My book: www.SafeHomeVisits.com
My art & action photos of my dogs: www.PictureTrail.com/scoutdogs