The H-ELP training team is often asked about the structure of our training sessions when we are in the field. Our sessions are always driven by our mission and that is twofold: to improve the welfare of captive elephants and the safety of their handlers (mahouts). This means that our ultimate goal is to produce an elephant that is able to complete her job calmly and contentedly because this not only ensures her welfare it also ensures the safety of her mahout.
To achieve this we employ a range of evidence-based training techniques. We use subtraction, addition and combined reinforcement, classical conditioning, and shaping, as well as a number of gradual desensitisation techniques when required. So what do these terms mean and how do we use them in training?
Subtraction reinforcement is the removal of a mildly annoying pressure at the onset of a desired behaviour. It increases the likelihood that the behaviour will be repeated. So, for example a trainer might apply pressure to the elephant’s chest with their palm until she takes a step backwards at which point the trainer’s hand is removed. Elephants (and horses, dogs, camels etc.) soon learn that stepping backwards removes the pressure. When subtraction reinforcement is used the pressure always starts very lightly and increases over a predictable timeframe. This gives the elephant a sense of control over the process because she is able to avoid any annoying pressure by performing the behaviour.
Addition reinforcement is giving the elephant something that she wants (like a banana or gentle stroking) when she performs a desired behaviour. For example, if the elephant stands quietly to have her feet checked we will often reward her with either stroking or something delicious. This increases the likelihood that she will do the same thing again.
Combined reinforcement is when subtraction and addition reinforcement are used together. It also increases the likelihood that the reinforced behaviour will occur again. For example, if the elephant is being trained to step forwards the trainer will often apply gentle pressure at the base of her ear and as soon as she takes a step they will both release the pressure and either stroke her or give her a treat (like a sugar cane).
Classical conditioning occurs when two seemingly unrelated things become associated. So, for example, once the elephant has been taught to step backwards from light pressure, the mahout will then start using a voice cue. He will say the word for back and then apply the pressure cue that has already been trained by subtraction reinforcement. After several repetitions, the elephant learns that the word predicts the onset of pressure and she starts to step backwards from the word alone.
Shaping is the gradual change of behaviour. Desirable behaviours often take weeks and even months to achieve. Captive elephants need to accept the handling of their feet and legs because they can suffer from small wounds caused by sharp sticks in the jungle and these need to be treated to avoid infection. Elephants are surprisingly sensitive and often don’t enjoy their feet being touched. Desensitization must always occur very gradually because rushing can cause the elephant to become stressed. A shaping program for foot care will often include walking the elephant slowly over a small log and rewarding her when the chosen foot touches the log. Pretty soon she will place her foot on the log when she is near it because she has learned that this behaviour leads to delicious treats. She will then be rewarded for allowing a mahout to touch her foot while it is on the log. This process will continue until she is happy to allow her foot to be touched and stroked while it is resting on the log.
When we give animals (and humans) a sense of control and an ability to predict certain things in their environment they become calmer and more content. Using evidence-based training techniques provides this. Subtraction reinforcement gives the elephant a sense of control because she can remove the mildly annoying pressure of the trainer’s palm at any time by producing the trained behaviour. Classical conditioning gives the elephant the ability to predict events because the mahout’s voice command precedes and therefore predicts the onset of pressure. A very gradual shaping process ensures that there are few incorrect answers in any training session which means lots of rewards, happy elephants and safe mahouts.