Reinforcement and Punishment

No, don't click away! It's not going to be yet another article about the good positive trainers and the bad guys with prong collars. It's about what the terms mean, how we use these principles in real life and how to make them more effective. And about differences between reinforcement and reward, as well as between punishment and correction.

Let's start with Reinforcement. It's very common nowadays to teach dogs by using reinforcement principle, meaning that when the dog does something desirable, he gets a treat, or praise, or a toy. It feels good, the dog is usually happy, and the training goes on. I can't deny that it is my favourite method of working with dogs. But in everyday life we also use the word reward, very often in exactly the same situation as reinforcer. And the problem is, they are not the same.

Reinforcement is a procedure that increases the frequency (or magnitude, intensity, duration) of a behaviour. It has a specific goal, and is defined by the outcome. If I give my dog a piece of sausage and his behaviour, let's say Sit, now happens more often, I reinforced this behaviour. If I give my dog a piece of sausage for sitting, but I don't see him sitting any more than before, than my piece of sausage was a reward. The dog liked it, wanted it, but it didn't necessarily influence his future behaviour. You can be dishing out sausage and cheese all day long and still notice that your dog doesn't learn anything, or very little. Sounds familiar? That's because it's so common. I do it as well. It makes me feel good, I just rewarded my dog, so cute. But not every treat my dogs get is actually a reinforcer. To qualify as such it must bring results.

There are a few conditions that must be met for that piece of sausage to bring you the best result they can.

Timing: the reinforcer should follow the behaviour pretty much immediately. The longer you take to deliver that treat, the more likely your dog will do something else in the meantime and then what are you really reinforcing? Is it something that you are trying to teach or a random behaviour? If your dog doesn't seem to get it then your timing may need tweaking. Maybe start using some marker signal?

Contingency: this is a difficult word. At least for me. If you are training Sit and sometimes you feed your dog for sitting, and sometimes just for standing and looking at you, there is no contingency here. From your dog's point of view sitting doesn't make much sense because he gets the treats anyway. It's really the difference between reinforcement and reward again. The dog needs to know that only sitting will bring him the treat, only then the learning will happen quickly.

Rate of Reinforcement: This is a big one. As a general rule an average dog owner treats the dog too little in a training session. One of the biggest hurdles for trainers is to persuade their clients to give much more food in training. Or play more often, and more vigorous, or praise for longer. No, the dog will not get fat. The brain uses a lot of glucose when it's working, all the calories will be burnt. The best is to do several repetitions in quick succession, feeding each one of them well. Especially when you are in a distracting environment, where a lot is happening around you and your dog is struggling to keep attention on you – feed more! Toss food on the ground, when the dog looks up – ask for a simple behaviour and feed, then repeat.

Quality of Reinforcement: Ah, one of my favourites! You want quick and long lasting results? Pay better. Your dog knows what you have in your pocket or treat pouch. Even before you give him the first treat. Remember they have very good noses. If you use toys, you may have a surprise effect here, but only the first time. After that they know. So if it's a dry kibble, and they are not particularly fond of it, they may or may not put effort to get it. One of my own dogs is like that. He'll do easy, well known exercises for low value treats but will get lazy if I ask for something requiring more effort. For those behaviours, and for teaching completely new ones, I need sausage, cheese or liver. Outside as well. And let's not forget that it is your dog who decides what is low and what is high value for him.

Now let's look at punishment. What I mean by that is a procedure that decreases the frequency (or magnitude, intensity, duration) of a behaviour. So we could say it's the opposite to reinforcement. Punishment is kind of a bad word now, nobody wants to be the punishment trainer. But we shouldn't be scared of this word, everybody uses punishment in their everyday life, and are subjected to it too. As a scientific procedure, punishment, like reinforcement, is defined by its outcome. It was a punishment if it actually led to diminishing of a behaviour. That's it. However a correction is an action intended to put a stop to an annoying behaviour, right now, for a period of time, but its effect doesn't necessarily carry on into the future. Correction is the bad word. It's like retribution, you lunge at this dog, I yank your leash hard. You bark, I whack you with a newspaper. You get the gist. I am greatly opposed to endless corrections in dog training. And they often are endless. If you don't see any long lasting results from your corrections, it's time to rethink what you're doing. It's time to look at the conditions for effective punishment, or better yet - switch to a proper positive training, where punishment will become almost unnecessary.

The conditions for effective punishment are

Timing: no surprise here. As with reinforcer, a punisher must be delivered immediately after the behaviour. Punishing the dog long after he's done something is an abuse, not even a correction. Dogs do not understand that you are kicking and screaming because three hours before they chewed your shoe. They only connect the most recent event to it, which may be you coming home. Ouch! You've now taught your dog to fear your coming home.

Contingency: Yes, this too. You will need to deliver the punisher to each and every instance of the behaviour in question, and not at any other time. It is actually very difficult to do. If you are working on your dog's barking and want to spray him with water for it, and you want to be effective and fix it quickly, you will have to follow your dog's every step with a bottle in your hand. If you miss a few barks here and there, you are actually shooting yourself in the foot, as the dog will figure out that he should only be quiet when you are next to him. If you can't reach the bottle even if he is standing next to you, you're missing the opportunity too. The barking behaviour will be sometimes corrected, sometimes not, so the result may be that it will not go away at all. If you continue to spray your dog from time to time, you are wasting your energy, and guess what? You are correcting, getting even, but you are not applying effective punishment.

Magnitude: this is one of the biggest problems with punishment. Generally, to be really quick and effective, it should be sufficiently strong the first time you do it. The dog would need to think, 'holly crap, I'm not doing that again!' This would be really the most humane use of it. Instead, we start gently, a tap on the head, a little pull on lead, slightly raised voice... With time we escalate little by little, since the dog stops paying attention to it. If we increase the corrections at this rate, the dog gets desensitized to them and we need to escalate even more. This may lead to unthinkable levels of aversives used, and still they will not be effective.

Relevance: an important question to ask whenever your dog is doing something undesirable is: what is he getting out of it? What is the function of this behaviour? Remember that behaviours don't just happen, the dog does things to have an effect on his environment. Animals can't, and don't, waste energy for nothing, so if your dog is chewing on your table, it must satisfy a need that he feels. If you scream at him but let him stay in place, he will start chewing again. Shouting was not a relevant correction for chewing. Putting him in a crate could be, as he would be removed and denied the access to the table. Which is what he wanted. If he barks at you for attention, looking at him and saying 'quiet!' is not likely to stop him, but ignoring will. He wants your interaction, and by talking to him, you're interacting with him. Even if you are shouting, he's still getting what he wanted. Ask yourself what would be the most appropriate punishment for this particular behaviour at this particular time. And you'll start be more effective. Or, better yet – think how you may satisfy your dog's basic needs and what you can teach him so he can be happier and more enjoyable to be around.

I could write much more on this topic but hopefully this is enough for a start. Any omissions and oversimplifications are there to make the article digestible. Hope it tasted good!