Alternative Ways To Train Dogs

In my dog training classes, I use a combination of operant conditioning and Pavlovian conditioning to train the participants. Operant conditioning involves reinforcement for a correct behavior or response, and Pavlovian conditioning involves pairing two stimuli together so that the dog learns to link the two in her mind. For example, when I say “Sit” and the dog sits, she gets a reward, either as a food treat or as praise. The sitting part and the reward are operant conditioning. And when I pair the word “Sit” with a hand signal for “Sit,” that is Pavlovian conditioning. This has been a tried and true method for training dogs.

However, there might be other ways of training dogs. In an article published in 2003, S. McKinley and R. J. Young tried training dogs with the model-rival method (McKinley, S. and R. J. Young. 2003. The efficacy of the model-rival method when compared with operant conditioning for training domestic dogs to perform a retrieval-selection task. Applied Animal Behaviour Science 81: 357-365). The model-rival method has been used extensively by Irene Pepperberg in training her African Grey Parrot Alex to learn to recognize different shapes and different colors, and to say the names of those shapes and colors when he was asked about them.

The basic method of the model-rival technique involves two people talking about an object in the presence of the animal. One person asks the other some question about the object, and the other person answers and is given the object if the answer is correct. Meanwhile, the animal observes the interaction. Then the animal is asked to join in by providing the correct answer.

In the case of dogs, the two experimenters compared the results of using operant conditioning and using the model-rival technique in training 6 male and 3 female dogs to retrieve an object that had a distinct name. In the model-rival part of the experiment, rubber toys that had a similar appearance were used. There were three red rubber toys (a boot, a fire extinguisher, and a strawberry) and three yellow rubber toys (a saxophone, a toothbrush, and a hammer). The experimenters chose a toy at random, and if it was a red one, they called it “Socks,” and if it was a yellow one, they called it “Cross.” Then, while a dog was tethered on a leash 0.5 meters (1.5 feet) away from them, the two experimenters had a conversation along these lines: “Can you see the Socks?” (handing the toy to the other person). “Yes, these are great Socks,” (handing the toy back), and so forth for 2 minutes. Then the dog was told, “Go get the Socks,” and the dog had to go and retrieve the object that the experimenters had called Socks, from among the other two similar toys that did not have any labels.

In the operant conditioning part of the experiment, the dogs were trained to retrieve a single toy whose name was given, when no other toys were available. Here the experimenters used the standard tool of shaping, or rewarding the dog at first if she came close to the object, and then later rewarding the dog for touching the object, and eventually only rewarding the dog for retrieving the object.

The experimenters found that the dogs on average learned to bring back the correct toy in the model-rival method as quickly as they learned to retrieve the toy with the operant conditioning method.

Another article published in 2008 repeated these studies, and extended them somewhat (Cracknell, N. R., D. S. Mills, and P. Kaulfuss. 2008. Can stimulus enhancement explain the apparent success of the model-rival technique in the domestic dog (Canis familiaris)? Applied Animal Behaviour Science 114: 461-472). In addition to using the model-rival technique, these experimenters used a technique called direct enhancement, in which one person held the correct object in her hands and kept looking at it, while the other person spent the whole time looking at the dog. This study confirmed that the model-rival method worked well in training dogs to retrieve objects, but also found that the direct enhancement method, where the experimenter spent the time looking at the object, worked as well as the model-rival technique.

What these results show is that there may be other ways to train dogs. Some of these ways might be better than what we are using today. In the model-rival method, the dogs actually learned the labels that were given to the objects that they were expected to retrieve. With operant conditioning, it is not clear that they learned the labels for the objects, or simply learned that a particular word (e.g., Socks) stood for a food reward that they were given at the conclusion of their task.

These kinds of experiments suggest that dogs might be learning much more from us than we have given them credit for. Our theories about animal learning mostly come from laboratory rats that live their entire lives in cages. But our dogs are social, interacting creatures living in a rich world of experience, and it may well be that we have just begun to scratch the surface of their abilities to learn from us.

--Con Slobodchikoff

Comments

[Vibram Five Fingers]

Hi,What a great post! Thanks for article. Everytime like to read you.Have a nice day!

[Susan Kuchinskas]

This is fascinating. Maybe dogs could learn a lot more from observing us if we did things they also could do. Playing soccer is pretty easy to get your dog to do, for example.

My mother-in-law says when she first got her dog, he would carefully watch her make dinner. "Finally," she says, "he realized he wasn't going to have to do it himself, and he stopped paying attention."