Shaping: Guidelines For Success

Shaping is defined as the teaching by differential reinforcement of new behaviours by systematically reinforcing successive approximations toward a final target behaviour.

In dog training if a behaviour never occurs, or does not occur on a frequent basis, we say that it is not in the dog’s repertoire of behaviours.  Shaping is a way of adding behaviours to a dog’s repertoire.  Shaping is used when the target behaviour does not yet exist; for example weaving poles,  or when the target behaviour occurs infrequently; for example lying down with front paws crossed. In shaping, what is reinforced is some approximation of the target behaviour.

Approximation means any behaviour that resembles the desired behaviour or takes the dog closer to the desired behaviour. Successive approximations are steps toward the target behaviour, the behaviour you want to shape.

In playing the game of “Hot & Cold”, you reinforce any movement that takes the player closer to the prize.  Each of those successive movements is a closer approximation of the desired behaviour.  If the prize is under the sofa, and the player is moving toward the sofa, every time the player takes a step toward the sofa, you are yelling “hotter”, and you are reinforcing the behaviour.  If the player moves away from the sofa, you would yell, “colder” (non-reinforcing).

The general guideline is that you are reinforcing any behaviour that is a closer approximation of the target behaviour than the behaviour you reinforced last.  If a new approximation does not occur, you reinforce the last approximation again.  If an approximation is repeated and reinforced three to four times, you can withhold reinforcement the next time that behaviour appears. By witholding reinforcement you are increasing the criteria, you are asking the dog to figure out what more it needs to offer for reinforcement.  This is why I love shaping over other ways of getting behaviours, shaping truly does make a dog think.

Guidelines for shaping

• Define the target behaviour:  The behaviour you want hasn’t occurred yet; it’s the goal at the end of the process, so you must decide what behaviour is to be “shaped up”.  To get to the target behaviour, you must have a clear idea of what it is.

• Reinforce successive approximations of the target behaviour:  The target behaviour is ‘shaped up’ by reinforcing the nearest approximations of that behaviour.  You decide what these nearest approximations are, you set the criteria for each approximation. If the dog gets stuck at a particular step, you have two choices; you can go back a step and reinforce the previous approximation, or you can withhold reinforcement in order to induce variability in the dog’s behaviour.

• Reinforce an approximation several times or until a closer approximation appears, whichever comes first:  If no new approximation has appeared after several reinforcements, withhold reinforcement until a new approximation occurs.  In general, shaping progresses more rapidly when the increases in the criteria for reinforcement are small.  When you hold out for something better, the something better should be only a very slight improvement.  If an approximation appears that is a big advance, reinforce it, but don’t hold out for big advances.

• Monitor results:  The only way you can gauge how successful you are being at shaping behaviour is by noting what changes in behaviour are occurring.  Are you seeing progress toward the target behaviour?  Is the behaviour that occurs now closer to the target behaviour than the behaviour you got earlier?  Is it time to increase criteria for a closer approximation of the target behaviour? Has the behaviour begun to break down? Should you move back to a previous level?  Do you need to increase the value of the reinforcer? Are you able to keep the rate of reinforcement high?  These are questions you must constantly ask yourself while shaping behaviour and you can only answer them by paying close attention to changes in behaviour offered or not offered by the dog.  Taking mental data on each step, keeping track of approximations, criteria, and rate of reinforcement, help you determine answers.

Note: You will have to make judgments about when to raise the criteria and by how much; sometimes, you will be wrong.  It is best to err on the side of caution, reinforcing behaviour at a given step for just a few more trials and make very small increases in the criteria for reinforcement.  If you make the mistake of moving too quickly, then progress will break down and you will need to move back to a previous level.

Here is a really good example of shaping and clicker training. Jessie an assistant dog learning the task of turning on a light switch in under 4 minutes:

http://www.youtube.com/watch?v=6DWbV5VKZxc