There is an argument among dog trainers which doesn't matter to the average dog owner. But for those of us with sight hounds, it's important. I've written before about why punishment doesn't work in general and specifically with Borzoi. I advocate for positive reinforcement (R+) exclusively when training any new behavior. Even better if you can use R+ to capture a behavior the dog offers without prompting. If you acknowledge (called "marking" in behavior lingo) and reinforce (reward) that behavior several times in short order, your dog will begin doing it for you to get the reinforcement. Then you should name that behavior and ask for it. If your dog offers the behavior on cue, keep rewarding him until he's very good at it. Then take it on the road. Try asking in new places and from farther away. That is dog training.
But once a dog responds to a cue, how do you respond when you give the cue and the dog doesn't offer the behavior? Ah, this is the argument that will divide the room. Trainers who claim they are R+ trainers will say you simply refuse to give the reward. Trainers who claim to be balanced trainers will offer the dog a correction, or punishment of some sort. For example, a collar check or even a verbal sound that indicates a mistake or that no reward will be forthcoming.
Some R+ trainers fear that aversive actions on the part of the trainer are too costly to the relationship between dog and trainer. Some trainers who use punishment fear that after a dog has learned a cue and the trainer is sure the dog heard the cue and could respond, if he does not and there is no consequence the dog might choose between responding to a cue and just ignoring the trainer when distracted enough. R+ trainers believe that if you set up distractions during the training process when the dog is ready, punishment would never be necessary because the dog will respond to the cue after a long history of reward for a behavior.
Here is the problem. When you are setting up those steps of distance, duration and distraction to ensure your dog will respond to your cue anytime, anywhere you will have incidents in which the dog fails, makes a mistake, doesn't perform. If you with hold the reward, which is what R+ trainers recommend, you are, in fact, punishing the dog. (insert gnashing of teeth here).
R+ Trainers claim the withholding of the treat is a "negative reinforcer"* or is neutral and under the control of the dog which somehow makes it not punishment, even though all punishment is technically under the control of the dog if he could have chosen another behavior to avoid it. But this is semantics. The scientific difference between *negative reinforcement and punishment can only be assessed from the subject's behavior. If the dog is more likely to offer the behavior in the future, the action taken by the trainer is reinforcement (negative means something was taken away). If the dog stops a behavior, the trainer's action was punishment.
So what is the behavior? If we cue a dog to sit, and the dog sits, he gets a reward. Reinforcement since the next time the dog is asked to sit, he is more likely to do so. Positive because something was given to the dog because of his sitting. If we cue a dog to sit and he does not sit, there is no option to take something away (negative) to reinforce his NOT sitting. That is the behavior he offered and the only one that can be reinforced. Reinforcement does not act on any behavior other than that offered.
So we are actually STOPPING the not sitting which is punishment. An example of negatively reinforcing sitting would be putting pressure on a dog's collar (assuming he dislikes this) and relieving/removing that pressure as soon as he sits. An R+ trainer would not do this because using an action the dog works to AVOID is not acceptable to the model of training without force or even aversives (things dogs don't like).
However, dogs certainly don't like not getting a treat. So most trainers agree that it's impossible to avoid all things dogs don't like. They like to run after cars. They like to roll in rotten smelling garbage and not allowing these activities could be considered restrictive or even aversive to a dog. There are always extreme examples.
But the argument has become crazed among some training groups. But lets' go over this again. A dog who is asked to sit and receives something he likes is being positively reinforced for being RIGHT. A dog who is asked to sit and has something taken away that he doesn't like is being negatively reinforced for being RIGHT. A dog who has receives something he does NOT like for being WRONG is being punished and a dog who does NOT receive something he likes for being WRONG is being punished.
Some trainers will argue that if the dog is asked to sit and he does not sit, and the reward is withheld, will sometimes sit late. And then the trainer rewards the dog. So he is not being punished. This is not true. The dog is punished by the removal of the treat for NOT sitting which REDUCES the behavior of not sitting. That is the definition of punishers--they stop the offered behavior. If the dog is rewarded for sitting late, he is rewarded for sitting but he was still punished for standing.
Why does this matter? Well, it really doesn't matter to dogs. They figure out how to get the rewards they want when the distractions are not too much better than the reward they are offered. But it does matter to dog trainers. This argument has not only divided dog trainers but it's set in motion accusations that dog trainers who understand this "believe in punishing dogs" and there fore cannot call themselves humane trainers. So, that makes them in-humane which no dog trainer will ever accept.
For the record, I am a member of the R+ crowd. I do believe in teaching a new behavior by capture and reward, putting it on cue, and proofing through stages of duration, distance and distraction. I know it works. But I am also a student of psychology who understands that the definition of punishment (behavior stopper) is not the same as painful, cruel, harsh or even physical. Any action a trainer uses to stop a behavior without body blocking, is a punisher specifically when that action makes it less likely that the dog would offer the same response in the future. Therefore, with holding a reward is a punisher for the wrong response. Since the dog probably doesn't like that, it's also aversive.
The dividing line between trainers who claim to use no aversives and those who claim it's not possible is now imaginary so lets all get back to working together to solve problems for dogs and their owners. The idea that learning can be stress free is taking a toll on training and education. Kelly McGonigal has a great TED talk that might stop this argument for good. Watch it. It will change your life.