If I would ask you, every time you come home from work you get something very good? How fast would you be coming home? Giving you the confidence to know where you are going is most of the time the way to go. Most likely because you have done that road many times before. But are you just going to try and see if you can make it without knowing the direction to go home? Or would you slowly increase the length of the road back home? What about practising increasing distance to your home instead? Would that give you more success?
You might wonder Peter why all those questions? What I’m going to talk about in this blog is all about forwards or backwards training. Let me explain you the difference.
When we train a behaviour we start at the very beginning most of the time. From there we add steps and we try to reach our goal. How those steps are taken are sometimes more important then you might think. For example when training a speed swim with a sea lion I send the animal to a B person and he runs with the target the first 2 meters away from the reinforcing trainer. After we did this approximation we do this again but now we let him run from 2 meters away another 4 meters and so on. Eventually you reach the end and you will have your speed swim. Reading this might make me wonder when does the animal swim quickest, when the animal goes away from the reinforcing trainer or when the animal goes back to the reinforcing trainer? Most likely the last one because of the placement of the consequence. This is the big difference with backwards training. Backwards means you start at the end. Instead of sending the animal away we are going to let the reinforcement person stay on the end position and make person B send the animal to the reinforcement trainer. This works in such a high rate that the animal will swim quicker and will have more confidence to do the behaviour because it works its way to something the animal already knows. This gives you a big change in the learning curve of the animal.
In this video you see a speed swim from a sea lion in Ouwehands Zoo in the Netherlands. This behaviour is trained backwards by starting at the end. The course the animal swims is build step by step to the start of the whole behaviour.
Another example is agility with dogs, you know the sport where the dog and trainer run a course and the dog has to go through a tunnel and a slalom etc as quick as he can. The way you can make the dog move like a bullet is the following, the whole course is most likely chained together by backwards training. The trainer started at the very end of the chain. Jump on the box and you get a great reward, now slalom and then jump on the box, now tunnel, slalom and the box… and so on. We add steps before the last goal. This way the animal works its way nonstop to something he already knows.
I’ve been thought to use as less as possible in your training sessions. This has a drastic effect on your approximations you need to take later. When you add trainers and targets etc who are actually unnecessary you also add approximations later in your training just because you need to train all the prompts away again. When I was working in Canada we didn’t have a big team in winter time but we would still train big behaviours such as a perimeter bow with Cleveland (Who you met in the previous blog). We would start at the end and that is making her jump on a signal. We thought her a very nice high bow. After this step we took it where she had to do the perimeter where we started at the last position first. The challenge was that we only had 3 trainers present, 1 on stage, 2nd on the last position and the 3th on the bow position before the very last position. Before we would introduce trainer 3 we would send the animal with the bow signal to trainer 2, she then would slap the water with a target what meant jump. We had trained this in previous session when we had to shape the actually jump itself. When she got the last position we would train this target slap away first and then the target. The trainer was the prompt at this moment. After this step we added the 3th trainer who would use a target as well. Because we added a criteria, trainer 3 would slap the water again to loosen up another for more success. This meant that we would have 2 jumps. The next step was a lot trickier because we had to train trainer 2 completely away and add her on the next position next to trainer 3. Now she had to jump on the last position who didn’t have a prompt anymore while on the other positions there would be a trainer. Challenging but fun to them working this out. Then we would train trainer 3 away and so on till we would be completely without trainers except the stage trainer. At the end we would have 8 bows trained backwards with 3 trainers.
This is the result:
If we would train this forwards we would have less success and stay on some appoximations a lot longer as we did. Not to imagine the positions the animal had to jump on. When the reinforcement comes behind the animal the behaviour won’t have the same outcome as when the animal sees the reinforcement position in front of them.
Try it next time when you train an A – B with your animals or when you fly your birds from station to station.
Are you a forwards trainer or a backwards trainer?
“Thinking Outside the Zoo”