One of the lab and lecture series I attended at clicker expo was given by Cecilie Koste on chaining behaviours. Cecilie is Norwegian and is a top level obedience/working trials competitor in Europe and also is very active training Search and Rescue dogs. She owns clicker based training schools in Norway and her dog of choice is a flat coated retriever. The videos we saw of her working her dogs were amazing!
I suppose I knew about back-chaining and have used it as guided by my instructor but I don't think I had ever really thought out the principles and applications. We are pretty much set as humans to go forward from A to B so this was a valuable session for me. Cues that are clicker trained have strong reinforcement history so can actually be used to "click" another behaviour. In a back chained exercise the next behaviour in the chain will always have a higher probability than the preceding one. This helps to reinforce and maintain every behaviour in the chain.
Backchaining is based on Premack's principle which states that high probability behaviours can be used to reinforce low probability behaviours. It is important to understand our dog's preferences and use them to our advantage. When creating a chain we need to make sure that the first behaviour we use to train the chain has a high probability. This means the behaviour needs to be either (a) fluent (b) have a strong history of reinforcement (c) have usually produced a high quality reinforcer or (d) have been reinforced recently. A fluent behaviour that is recently reinforced will have a very high probability.
Cecilie showed lots of video to demonstrate various chaining (retrieve, roll up in a blanket etc) and then in the lab we all did the chain she wanted first and then chose one of our own to create. The chain she chose was for the dogs to do a foot target and then sit. This would be how she would train a go-out for obedience. To begin we needed to train the dogs to go to a foot target and then return to us for reinforcement. Stella defaulted to grabbing the target and retrieving it to me so we tried a smaller target (flat coaster) and she still picked up that so we then switched to me holding the target in my hand and asking her for a paw and clicking when she hit the target. I gradually moved the target closer to the floor and before long she had the behaviour. Next step was to reinforce the sit which we did by resetting with a treat toss and asking for a sit. Sit is a very strong default behaviour for Stella so this was pretty easy for her. The final step is to do a few more foot targets with click/treat and then at the exact moment you would normally click then instead ask for a sit. Once she sits then she gets a click/treat and voila soon a behaviour chain is created!
My chosen chain was a spin, down, nose touch (jump up). Stella enjoys the nose touch behaviour so this was my starting behaviour to train the chain. I worked each of the behaviours separately and then combined the down, touch and finally the spin, down, touch. Stella did awesome!
Backchained behaviours have a high reliability because the animal always knows what behaviour is coming next so is prepared. It is less stressful and easier learning for the dog. During training there is usually a "testing" phase where the animal will see if there is an easier way to the reinforcer. It is important to terminate the behaviour if a mistake is made to prevent the mistake being reinforced. Try again but if the dog continues to make the same mistake after another 2-3 tries then go back and retrain only that problem part before chaining it again. Cecilie uses backchaining for obedience so considers anticipation by the dog to be a failure as we only want the dog to do the behaviour when it is cued. It should be noted that some cues are environmental and not handler cues. An example would be in a retrieve where the act of picking up the dumbbell is an environmental cue to turn around and return to the owner.
I really enjoyed this topic and Cecilie is a great presenter. Hopefully I have presented all her wonderful information correctly. Lots to think about with this and I think it will be very useful for me. One of my big issues in training is I tend to not break things down enough when training a new behaviour. I need to start physically writing down all the steps needed and then figure out how to chain that. Since returning home I have started to backchain the dumbbell retrieve and I am pleased with the results so far.