This article is from the Training Your Dog FAQ, posted to rec.pets.dogs newsgroup. Maintained by Cindy Tittle Moore with numerous contributions by others.
Keep these firmly in mind:
* _A REWARD results in an increase in the selected behavior_.
* _A CORRECTION results in a decrease in the selected behavior_.
Well that seems obvious enough, why did I bother putting those down? Because all too often, obvious as they may be, an astonishing number of people ignore them. How many times have you seen someone call their dog over and over and over again while the dog blithly ignores them? How many people wind up automatically rewarding their dog all the time until they find that the dog is either bored and wanders off, or won't do a thing unless the food is held in front of them? How many people smack their puppies when he soils in the house but never wind up with a housetrained dog?
Let's examine each of these scenarios in detail. The person who calls their dog repeatedly without doing anything is in fact teaching their dog that the "Come" command is meaningless. The dog is neither being rewarded for the correct behavior nor being corrected for the unwanted behavior. Therefore "Come" has no particular meaning for this dog.
If you consistently reward the dog no matter how he performs the selected behavior, you will have two things happen. First, the behavior will never _improve_ as the dog has no feedback on which is "better". Second, the dog learns that he always get rewarded, so the incentive to keep working (unless the dog is _very_ food motivated) will decrease. Or, if the dog is strongly food motivated, he may flat out refuse to do anything the moment he realizes that he will not get food. In this latter case food has stopped being a reward and is now an entitlement and no longer will increased selected behavior.
A puppy that is smacked for soiling in the house has no way of associating the correction with the action, particularly if it happens well after the act. Furthermore, hitting a dog is interpreted by the dog as aggressive rather than corrective and so will not reduce the selected behavior.
Back to rewards. Rewards should be given in such a way as to increase the behavior in question. This means, to begin with, that it should be something your dog enjoys and is motivated by. For some (many) dogs, food will do. Toys, squeakies, tug toys, tennis balls, are often good bets. A few dogs seem to be motivated by verbal praise, although to be honest, not so many as people would like to think. In most cases dogs learn to accept verbal praise as a secondary reward, through association with a primary reward. You can also use multiple reward methods, especially if that interests your dog.
(A _primary_ reward is something that is _inherently_ rewarding to your dog -- food, petting, toys, etc. A _secondary_ reward is something that the dog _learns_ is a reward. For example "Good Dog!", a click, clapping. The technical term for a reward is _positive reinforcer_.)
When you reward a dog, it should be directly associated with the selected behavior. A reward is ineffective if you apply it at the wrong time. However, the most common problem with rewards is that people will inadvertantly reward a dog for unwanted behaviors. Here is an example: Your dog growls or barks when he sees other dogs. Since you think he is afraid, you pet him to calm him down. "It's OK," you say. "Nothing bad is going to happen." OK, so what happened? The dog growled, you rewarded him. He's no dummy; he'll growl again in the hope of a reward next time.
Corrections are equally full of pitfalls. First of all, what constitutes a correction? That's even more difficult to answer than for rewards. For some dogs, the tone of voice will do it, for others they'll never notice it. Many typical corrections are really secondary (eg, learned) corrections. And, many typical corrections really don't do anything other than make the dog afraid of you, or, when applied inconsistently, cause the dog to lose trust in you. Here is another classic example. Your dog is on the far edge of a field, and you call him. He doesn't come. You call him again. He doesn't come. No matter how often you call him, he doesn't come, so you march over and start to correct him. Or, he finally comes over and by this time you're so mad you correct him. So what happens? In the first instance, the dog may well have no idea what you're mad about. If he's never learned the "come" command (even if you think he knows it) then going over and popping him a couple of good ones will teach him that it's really bad when you go near him! If he _did_ come over to you and you popped him a good one, what do you think he'll remember next time you call him to come? That's right, you just applied a correction to a behavior (coming to you) in order to _decrease_ it!
People very frequently misuse rewards and corrections in this way because many people seem to think that dogs really do know which are good and bad behaviors and will correctly associate one behavior (out of several) with the punishment. This simply is not the case. Dogs will association what they _most recently did_ with the correction or reward.