When should you shape a behavior, use differential reinforcement or back chaining as the preferred protocol?
Niki Tudge Copyright 2011
Differential reinforcement is an appropriate procedure to choose when the target behavior is already present but requires a dimension of the behavior to be reduced, such as the intensity of barking or the frequency of jumping. The problematic behavior is targeted for extinction while another behavior is simultaneously positively reinforced. Differential reinforcement is also used when training a simple behavior that is already in the dogs repertoire but needs to be put under stimulus control, a cue. For example you would reinforce a sit over another behavior such as looking at you or sniffing the ground. The behavior that is reinforced is then more frequently offered. Don’t confuse differential reinforcement for reinforcement schedules. Reinforcement schedules are the frequency, of when you actually offer up a primary or secondary reinforcement for a correct response/s. For example you may decide to reinforce every third correct sit, fixed ratio 1:3 or every 10 seconds of a sit behavior, fixed interval.
Shaping is the most appropriate procedure to choose when the target behaviors are not in the learner’s current repertoire and they cannot be captured or are too complicated to lure (Chance 2008). Shaping is also an appropriate procedure to choose if the animal lacks social confidence or offers rigid behaviors as a result of previously harsh or aversive experiences. For example if you want to teach your dog to open the fridge door and take out a soda. It is highly unlikely that you can capture this behavior as the dog will not do the behavior until it has been trained, nor can you lure the behavior. So small approximations of the behavior are reinforced until gradually the dog is exhibiting the entire behavior. You may start by reinforcing glances to the fridge then targeting the fridge door with the nose, then mouthing the handle, then pulling the door, then opening the door etc etc.
Chaining is an appropriate procedure to employ when training a sequence of behaviors, two or more component behaviors that will be performed in a sequence. Because in the sequence each response functions as a conditioned reinforcer for the previous behavior and the discriminative stimulus for the next response. Chaining is used when training dogs to retrieve over a jump or down on a recall. The decision then is whether you forward train the sequence or backward train the sequence. If you look at the behavior chain retrieve over a jump you have the following behaviors that are all performed in a sequence with one cue.
Sit at heel – Run to retrieve – Grab dumbbell – Jump -Return to handler – Sit – Hand over dumbbell – Heel sit