I attended two interesting sessions by Ken Ramirez from Chicago’s Shedd Aquarium. At Ken’s Smart Reinforcement lab session one of the topics was how to condition a new primary reinforcer. As we know in clicker training food most often is the primary reinforcer and the click is the secondary reinforcer. How about if you would like to have a brand new primary reinforcer, for example a touch/hand clapping/play, that you can use when you don’t have food handy or in obedience competitions where food is not allowed? Well, you can train this to your dog! Many of us actually have inadvertently done this. How many times have you started talking to your dog in a happy, high-pitched voice between the click and the treat (even though the rules say that you should be quiet between the click and the treat :-))? I have been guilty of this! Basically there is your recipe in teaching your dog a novel primary reinforcer.
It is a good idea to pick something that the the dog already seems to enjoy and you should train it with an easy, well-established behavior. The recipe goes:
1.New Reinforcer-Primary Reinforcer. For example, Clap-Treat.
2.Easy Behavior-Click-New Reinforcer-Primary Reinforcer. These sequences are introduced into training sessions gradually.
3.Easy Behavior-Click-New Reinforcer. This should be done very carefully, max 3 per session, and the number increased gradually.
Since the whole process takes time, only the first step was practiced at the lab session, most people paired hand clapping with treats and the dogs picked it up really fast. It would be interesting to see the following steps as well, I definitely intend to experiment with this for getting a new reinforcer for the obedience ring.
Another topic at the lab was using play as a reinforcer. Some dogs don’t naturally enjoy playing tug or fetching tennis balls but you can definitely teach this as a behavior. Then two things can happen: either the dog will play happily for food or the play will become reinforcing in itself. You can definitely use clicker with both food or toys, but the variety should be built gradually. Someone asked a very good question: should you have a different secondary reinforcer for each primary reinforcer, for example, click always means food but a whistle means a tennis ball? Ken’s reply was sure, but you should be aware that very often one reinforcer exceeds the others, therefore one marker signal will exceed others, and the others become disappointing. But if you combine all the reinforcers to the same marker signal, this brings in unpredictability and variability and the dog never knows what is coming next.
It was a joy to watch dogs playing, isn’t play just such a wonderful way of bonding with your dog!
The inner scientist in me cannot help writing about Ken’s Concept Training lecture, where he presented some very interesting unpublished data from a study on ‘Can dogs can learn new behaviors by copying other dogs?’. Mostly an animal doing what other animals do is called social habituation, for example a husky puppy learns to run along with the pack. The concept of whether an animal can be taught to copy what another animal does, even if it is a completely new behavior, has puzzled scientists quite some time. But now we are told that it is possible! It may not be a particularly useful way of teaching behaviors because if the behavior falls apart and the other animal is not present anymore, then you have lost the behavior. But it is interesting, nevertheless.
To test the hypothesis, Ken’s animal trainer group had two dogs that got along well and who had the same shared repertoire of behaviors. They began by training the two dogs in close proximity side by side. The ‘decoy’ dog was cued to do a behavior, then the test dog was given the ‘copy’ cue followed by the actual cue. This was then repeated throughly with different behaviors so that the ‘copy’ cue was not associated with only one behavior. The setup was repeated until the test dog started anticipating and offering behaviors on cue ‘copy’. Next the behavior cue was faded out. All of this worked really well. The experimental set-up was well optimized, the trainers did not see each other, the trainer of the test dog did not know what the the other trainer was cueing etc. The setup worked with different decoy dogs as well. Then the ultimate test: can the test dog copy a behavior that he doesn’t already know? Well, it turned out that yes…but I’m thinking kind of. The decoy dog barked, and the test dog kind of barked as well. The decoy dog rolled over, and the test dog kind of rolled over as well. Fair enough, the test dog probably tried!!! Maybe more behaviors to be tested and with more dogs?! Would anyone fund this kind of a research project for me :-)?