Animals learn through associating environments which as we know can also be called classical conditioning. When we teach our animals a signal we associate the signal to the behaviour we want the animal to respond to. Because animals are great observers there is a very high chance that the animal did not only pick up to the presented signal but to the whole environment the signal was presented in.
In cases where we as trainers try to free shape a behaviour this is something we need to look out for all the time. When we free shape a behaviour we reinforce the animal of doing a behaviour we like to see. When this behaviour occurs more often we add a signal to the environment for the animal to discriminate. This way the animal understands when to preform the behaviour. But, what we often forget is the last little part when we free shape which is crucial for the signal. We have to generalise the signal in the environment. Read more about free shaping here.
Free shaping or capturing – strengthen spontaneously occurring behaviour by adding a reinforcer when the behaviour occurs.
Most likely the behaviour you have been free shaping seems to happen very often even when you haven’t presented the signal. Well at least in your eyes. But why is the animal preforming the behaviour regardless? The reason is that the animal observes predictable signals from the environment to preform the behaviour. Which means that there are plenty of other signals occurring before the actual signal is presented by the trainer. The animal observes predictable events. The trainer might have no idea but gives a signal to the animal to make the behaviour happen. If it is not the behaviour it has been something else.
Predictable Events Which Trigger Behaviour
Animals amaze me since I started working with them. Over the past years I became more interested in predictable events and what this has for an effect to the learning curve of our animals. It seems like animals always try to find a red line through all we do. What I mean with this is situations they see as the same. Routines are a perfect example of such predictable events. But imagine consent behaviours when we have an environment where the behaviour and the feeling that come with it are as similar as possible in the situation we put them in like the dog in the following video from the animal training center in Austria.
We can deliberate plan predictable events for the animal to make it more likely a behaviour is preformed well. This doesn’t always means that the animal understand the signal we ask for. The signal might be the whole environment and not just one component of the environment. Just like the video above. There are a lot of signals that tells the dog which behaviour he is asked to do. It is not just jump on the table, it is the person taking the blood sitting at the end of the table, the plate the dog has to step into and the trainer standing behind the person taking the blood even the placement of the table is part of the signal.
The Front Flip With a Dolphin
I remember in my dolphin days that generalisation was an important part of training new behaviours. Once we were conditioning a front flip behaviour. We trained this behaviour at one position in the exhibit. The dolphin jumped high and rotated well just like in the video below. We decided that the behaviour was finished. But it wasn’t after all. It didn’t mean that when the dolphin could do this behaviour in this specific positioning that the dolphin could do it everywhere in the exhibit.
A huge part of conditioning such a behaviour is generalisation. In psychology generalisation means “transfer of a response learned to one stimulus to a similar stimulus”. In easy words we have to teach the animal that one specific signal connects to one specific behaviour. In the case of the dolphins front flip we still had quite a long way to go. We first had to slowly change the position to another position in the same exhibit. Then we had to transfer it to another exhibit area. Then we had to add the lights and sound used in the show, we had to add water sprayers and other animals to join. At last we had to introduce this behaviour to the team. All of this was part of generalising the signal which represents the front flip.
To be able to properly generalise a signal we have to understand that everything in the environment where we train a behaviour will be connected to the signal we present. Therefore going to another area might be very difficult for the animal to understand what we mean with that specific signal. We have to slowly change the environment for the animal to understand which signal the animal has to respond too.
The 35 Fallow Deers in Sweden
I remember that when we trained a group of about 35 fallow deers in Sweden for a recall. Before we even gave the signal (which was a whistle) the fallow deers already came when they saw the car. What happened was that predictable events made this behaviour happen. This is how it looked like:
- The keeper drove into the 6 acre exhibit.
- Drove to the nearest group of fallow deers.
- Stepped out of the car.
- Grabbed the buckets with reinforcement.
- Blow the whistle.
This predictable chain of events became the signal to the fallow deers coming. Eventually the fallow deers already came when the car just entered the exhibit. We can therefore say that the whistle signal had changed to the car signal. Not even to talk about the time of day which was very similar each day. The keepers didn’t want this anymore and wanted the signal they panned to become the signal again.
The question was how would we solve such a problem?
There are many different options. We first have to observe what triggers the behaviour. Especially because not only the whistle is the signal there are many more signals present the animals had observed. What we can do is go into the exhibit without the car. We break the predictable event by doing so which allows the fallow deers not to respond or at least most likely not to respond. Another option is to go into the exhibit without the car and then followed by the car. This way one keeper can already reinforce the fallow deers for not responding to the car at all.
Changing the time you go to the exhibit, the place of reinforcement, and ignoring the animals when you come into the exhibit with the car are all valid options to solve this challenge.
Another great example is when the animals are already waiting for us to come. We see this often in zoological facilities or horse stables. We can conclude that this behaviour is occurring due to a predictable chain of events. We give a lot of signals away to the animals. This is what we can call anticipatory behaviour but it might as well be that the animals have understood that a specific signal given which predicts the reinforcer to come makes that behaviour stronger.
Generalising a Target
Recently I’ve started to help out one of the my friends daughters who wanted to train her horse with a clicker. I think it’s very courages that a person chooses for this strategy to train her horse. Just because there is such a set way of working in the horse world. We are at session 7 and started to introduce a target. After a couple minutes the horse understood what the target meant and touched it when the target was present to receive a reinforcer.
Our next step was generalise that you can only touch the target when we say target. Which means we reinforce for standing still and not respond to the target even though the target is in plain sight and close enough to touch it. When we want the horse to touch the target we point at the target and say “target” which helps the horse to understand what to do which is touching the target. Our next step will be only to use the word “target”. I was standing in the middle of the training area and asked Lisa to pass me while the target was pointed at the horse. The horse had to ignore the target, which she did. Walking with the trainer had a high history of reinforcement and was therefore more likely to happen in comparison to touch the target.
The next time she passed she asked the horse to touch the target which she did. This allowed us to conclude that the horse started to understood that when we ask target you the target was touched and when we don’t she doesn’t. This is generalisation.
Our Biggest Challenge Is To Teach Signals Which Discriminate Behaviours!
In todays training world the biggest challenge is not teaching a new behaviour. It is understanding what is part of the signal that triggers the behaviour, deleting all the variables that might trigger the behaviour and focus on the only signal you want to have. This is a huge challenge for many. At the end we want behavioural control. To reach this goal we need clear signals to behaviour and an understanding what triggers the behaviour.
Have fun training!