While it is still officially the year of the pig, for us 2007 was the year of the dog. Juno was already seven months old when we met her in December of 2006, and we took over full time ownership of her in February of 2007. Young retrievers are a handful, and despite her age, Juno still wasn’t fully house-trained when we got her. So the first task at hand was to learn something about dog training.
Dog training is a big business. Just about everyone seems to think they are a qualified dog-trainer these days. As a result there are an abundance of dog training books out there. But after going through about a dozen dog training books (here’s a list of my favorites) it quickly became clear that the basic principles of positive dog training are fairly well agreed upon. Below I try to explain my understanding of dog training based on my reading and my limited experience with one dog. While Juno is much better behaved than before, she wouldn’t win any dog shows, and she certainly isn’t ready for any free-style competition, so don’t mistake my random musings for the advice of an expert!
There are numerous benefits to having a well trained dog. Most importantly, it teaches the dog self-control, which leads to overall better behavior. It also helps build new neural pathways by providing the dog with a stimulating environment. And, of course, it makes it easier to live together with your dog. But dogs are not robots and a lot of dog training actually involves training humans to be better owners. Dogs need exercise. A large dog will need at least 90 minutes a day running around outside. If you can’t provide the dog with regular exercise and play you have no business owning a dog. Similarly, when you are away from home either put your dog in a crate, or hide anything that she might damage. It isn’t her fault she’s bored when you are away. Better yet, buy a Kong toy and stuff it with treats to keep her busy till you get back.
The current thinking about dog training owes a tremendous amount to two students of B. F. Skinner, Marian and Keller Breland. While their ideas were widely applied at zoos and ocean parks, etc. since the 1940s, it took a while for a more scientific approach to be adopted by dog trainers. All training involves some kind of operant conditioning, but it is only recently that dog trainers have begun to rely upon research about how dogs respond to such conditioning, as opposed to personal experience and industry lore.
Operant conditioning defines four possible consequences which can be used to shape a dog’s behavior:
- Positive reinforcement adds something to the situation to increase the chance of the behaviour being exhibited again (for example, giving a dog a treat when he sits.)
- Negative reinforcement removes something from the situation to increase the chance of the behaviour being exhibited again (for example, releasing the tension on an uncomfortable training collar when the dog stops pulling on the leash).
- Positive punishment adds something to the situation to decrease the chance of the behaviour being exhibited again (for example, verbally growling at a dog to make it stop jumping up).
- Negative punishment removes something from the situation to decrease the chance of the behaviour being exhibited again (for example, walking away from a dog who jumps up).
The scientific consensus, broadly supported by the literature, is that positive reinforcement is the most effective means of training your dog. That does not mean that there are not situations wither other types of consequences could or should be used, but simply that 99% of the time the same desired effect could be better achieved by using positive reinforcement. For instance, rather than hitting your dog every time she chews on your slippers, give her a treat every time she chews on her rubber dog toy. Pretty soon that dog toy will seem a lot more interesting than the slipper. And, amazingly, you can slowly fade out the treats and the dog will often still continue to prefer her dog toy over the slippers.
Thoughts on Punishment
While positive reinforcement has its limitations (see below), positive punishment can have very serious negative consequences for both dog and owner. For instance, an owner who hits a dog on the nose every time she jumps on guests might find himself with a dog who is afraid of guests. She may even start snarling and biting guests, which is worse than the problem he set out to solve. Juno had this problem, and we found the best result was a combination of negative punishment (having guests ignore her while she was jumping and pulling) combined with positive reward (petting her and giving her attention when she sat still). The other problem with positive punishment is that the dog may not understand what she is being punished for. Dogs have poor memories and if you don’t catch them in the act, punishment won’t accomplish very much.
I personally differ somewhat from the Skinner folks on the issue of negative punishment. While I think it is generally a bad idea, and I’m opposed to anything that might hurt a dog, I do think that the objections to negative punishment depend upon the same faulty dog-as-automaton theories which led a previous generation of trainers to overly rely upon such methods.. Unlike other researchers who observe dog behavior in the home and around other dogs, the behaviorists tend to restrict their analysis to the laboratory. For this reason they tend to see dogs as incapable of having a theory of mind with respect to other dogs or their owners. But I’ve always liked David Hume’s observation that a dog can distinguish between being kicked and being tripped over. Moreover, some recent research shows that dogs are capable of more complex reasoning than previously thought.
Juno’s previous owner trained her to understand a simple “No!” command (actually the Chinese “Bu!”) which we still use to mean “stop what you are doing right now” and only use when we catch her in the act of being bad. We always try an alternative positive command first, like “sit” or “drop it,” but if she still doesn’t behave we use “No!” If that doesn’t work I hold her on her back with my hand around her snout and look into her eyes. I think it is important for the negative punishment to be a consequence of not listening to the “No!” command, and not to say “No!” after punishing, which would confuse the dog.
The idea about holding a dog down and looking into its eyes comes from the dominance theory of dog training — and is based on the idea that dogs respect authority. Many authors are now very critical of the “dominance” theory of dog training for good reasons. For one thing, humans are not dogs and it isn’t clear that they will ever think of us as dominant dogs. Also, any loving pet owner will allow the dog to dominate some situations, thus making it very confusing for the dog just who is dominant in what situation. Despite that, I do think Juno understands this action as a form of non-violent “negative punishment.” If I’ve yelled “No!” and she hasn’t responded and I move towards her, she will sometimes immediately roll on her back into the position I put her when she gets such punishment.
In short, if (a) the negative punishment is not harmful to the dog, and (b) the dog has already been well trained to execute a positive substitute behavior, and (c) the negative punishment is a direct consequence of not obeying the “No!” command uttered while the bad behavior is ongoing, I think it is OK to use negative punishment. Considering that this is a lot of ifs, ands, and buts, I understand why many dog training books advise against using any negative punishment. But because it is hard to avoid using negative punishment altogether, I think it is useful to have a theory about when and how to use it appropriately. For very similar reasons we use a head-halter when walking the dog in places with lots of distractions.
Notes on Positive Reward Training
The biggest problem with positive reward training, one that is not well discussed in many books, is that it is a skill. It isn’t enough to just read a dog book. One needs practice and one needs to watch experienced dog trainers in action. I was getting very frustrated with some of these methods until I was able to find a good dog training film online. There are a lot of great clips on YouTube, but I ended up buying this nice little video (warning, video uses annoying DRM technology). It really helps to watch a good dog trainer in action. I read about one study which compared expert trainers and novices and found remarkable differences. Novices were much less able to suppress extraneous sounds and movements which confused the dog. They were also were far more stingy in doling out treats. You don’t spoil the dog by treating her, you condition her.
Another difficulty with positive reward training is getting your dog to associate the treat with the specific behavior you are working on. This is accomplished with a “clicker.” I bought a bunch of noisy light switches at the dollar store and they do just fine. The important thing is that you make a unique sound at the exact moment the dog does the behavior you wish to reward, and the treat comes after the sound. The dog quickly learns to associate the sound with both the behavior and the reward. When I first started using a clicker Juno loved it so much she would bring the clicker in her mouth to ask me to play. You only need the clicker to train new behaviors. Once the dog has learned the command you can slowly phase out the clicker.
Finally, another major obstacle to positive reward training is that dogs are not very good at generalizing. Something they do very well in the kitchen they might not do at all when you are outside. It takes constant practice under a variety of conditions. Unfortunately, some conditions cannot easily be controlled — like behavior towards strangers and guests. No matter what, you inevitably end up with situations where your dog is getting petted by strangers as a positive reward for jumping up on them, exactly the behavior you are trying to avoid. Even worse, when you try to remedy the situation the strangers will tell you that they don’t mind, as if it were all about them and not about the dog. It helps if you have a regular visitor to your house whom you can work with, or if you can find friends the dog hasn’t yet met to be “strangers” and train them on how to behave towards your dog before she meets them. A good side to the dog’s inability to generalize is that you can allow her up onto one couch while still keeping her off all the other furniture.
Goals for 2008
Currently, Juno knows the following commands: sit, stay, down, fetch, leave-it, touch (she’ll touch her nose to your hand), shake, wait (i.e. wait at a door before going out), easy (don’t pull on the leash), and heel. She doesn’t know them particularly well, for instance she’ll heel for about 15 seconds before forging ahead, but its a start. Our goal for 2008 is to work on duration, getting her to do sit-stays and heels for minutes rather than seconds. Also, she can either run and fetch a frisbee or jump and catch one from sitting, but she can’t yet run and jump to catch one in the air …