Punishment and Dog Training
Lots and lots of words have been written on this subject by lots of people, across many disciplines, including psychology, law, dog training, you get the picture. I am going to have a go at it now and in the context of dog training. This article is really the third out of three that complete the operant conditioning picture, so I recommend you read the two previous articles, though of the three, this one is more easily understood as a standalone than the others.
As always, we need defined terms in order to formulate a coherent argument, so lets define what punishment is, and as in my two previous articles I shall source different definitions from different places and then present the definiton that is correct.
The AKC defines punishment as:
“Punishment (anything that makes the behavior less likely)”
in this article; Positive Reinforcement Dog Training: The Science Behind Operant Conditioning by Stephanie Gibeault MSc CPDT.
The American Psycological Association (APA) defines punishment as;
A physically or psychologically painful, unwanted, or undesirable event or circumstance imposed as a penalty on an actual or perceived wrongdoer.
In operant conditioning, the process in which the relationship, or contingency, between a response and some stimulus or circumstance results in the response becoming less probable. For example, a pigeon’s pecks on a key may at first occasionally be followed by presentation of food; this will establish some probability of pecking. Next, each peck produces a brief electric shock (while the other conditions remain as before). If pecking declines as a result, then punishment is said to have occurred, and the shock is called a punisher. —punish vb.
Link to the APA Dictionary of Psychology for their above definition.
B.F. Skinner did not really ever come up with a definition of punishment as a single sentence or paragraph like he did for reinforcement or operant conditioning, so I’ll have to do a mix of quoting and paraphrasing, but the following is from Skinner;
Responses from the environment that decrease the likelihood of a behavior being repeated. Punishment is designed to remove a behaviour from a repertoire.
Punishing contingencies are the reverse of reinforcing, but, the effect on behaviour is not the reverse of reinforcement; punished behaviour is displaced by incompatible behaviour (escape or avoidance) but the punished person (or animal) remains inclined to behave in a punishable way, the punishment is avoided by doing something else, sometimes nothing more than stubbornly doing nothing.
What a punished person or animal feels when in a situation in which they have been punished or when they have engaged in previously punished behaviour depends on the type of punishment issued, and this depends in turn on the punishing agent or institution. If a person or animal acts to avoid further punishment, they do not act because the internal feelings toward the behaviour have changed, they act because of the punishing contingencies to which they have been exposed.
That’s the end of Skinner, so let’s have a quick review of the differences between the various definitions because they are important. All three sources agree that punishment is designed to and does reduce the likelihood of a behaviour recurring or removes it from happening at all. What is missing from the AKC and APA but is included from Skinner is the failure of punishment to change / reduce / eliminate the desire to do the behaviour that is being punished. This is key because that failure to change the desire is the reason that as soon as the punishing contingencies (the application and / or threat of punishment) are removed, the original behaviour returns, and often stronger than before it was punished. In short, punishment only works when the constant application or constant threat of application exists. I’ll get into real examples later so this can be understood more clearly.
One additional key point I need to make here that links back to the definition of positive reinforcement and the peculiarities we have with the English language is using the word “punishment” this context. The removal of positive reinforcement is a negative reinforcement, i.e. it generates behaviour. If all animals are in a constant state of seeking the ultimate positive reinforcement (they are) and something is preventing them from obtaining that state of positive reinforcement, by stopping a behaviour, that prevention is by definition, a punishment. But that is a very narrow window, because the moment a behaviour occurs to sidestep or work around or destroy the thing doing the punishment, we are now back into the reinforcement space.
So behaviourally it is best to think of punishment as the thing that makes you (or your dog) stop what you are trying to do or get to because either it hurts (physically or psychologically), or, you can tell it isn’t working, or both. It is the necessary step making you consider an alternative course of action, which once you act would be a reinforcement.
Punishment in the behavioural context is therefore something that prevents you from obtaining your desire(s), that can be physically or psychologically, or both. This fits with the definition of all of the sources above in which punishment is designed to reduce and / or remove a behaviour from a repertoire.
I’ve said it a couple of times already, but it bears repeating, again; what punishment does not do is change your desire to get at or do whatever it was you were trying to get at or do. i.e. your goals and desires do not change despite the punishment, even if that means you stubbornly do nothing in the face of your desire and some form of punishment. This is what Skinner was able to prove with many of his experiments across his career and summarised in the paragraph I quoted above.
I’d just like to touch on the APA’s point 1 of their definition of punishment; this is what I called in an earlier article “a bug with the English language”. That punishment has these two commonly used definitions is a problem for us in this context. The application of a physically or psychologically painful or unwanted event or circumstance as a punishment is really above and beyond what a punishment is - depending on what or who is doing the application and how it is being applied. Unlike reinforcement, punishment is not only intrinsically interpreted, it can also be applied, but there are limits as to what constitutes punishment before it steps over a line into something more like retribution. More on this later.
It’s probably about time I linked all this theory into something tangible and related to dogs, dog behaviour and dog training. So lets get into examples of the theory in action, beginning with a demonstration or two of how punishment can stop a behaviour but it cannot change a dog’s intent to do a behaviour;
Prong collars work. They stop pulling and they stop lunging. The evidence is everywhere. Same for slip leashes. But what happens when you take those tools off and switch back to a flat collar? The pulling returns immediately, as does the lunging and barking and other unwanted behaviour. Why? Because punishment cannot change the desire to do the behaviour. The punishment and threat of it have been removed with the removal of the prong collar and so out comes the animals desire - visible as behaviour, and often unwanted behaviour, which is why the tools were employed in the first place.
The same can be said of e-collars, but there are some well documented studies that show effectiveness of e-collars beyond wearing of the device. This happens when the dog did not understand what or how the shock was delivered and the shock was of sufficient power to remain as a constant threat of punishment for a very significant period of time, sometimes for life i.e. the dog’s brain paired the situation with the shock not the collar and shock and the situation.
Muzzles are another good one. The wearing of a muzzle is a punishment to a dog. How good at preventing the attempts at biting are muzzles? Useless. Why? Punishment cannot change the desire to do the behaviour, even when the dog knows after trying that it can’t bite, it will keep on trying if the situation is assessed as requiring that course of action, hence the expression in the industry of “muzzle punching”. I’ve been muzzle punched many many times, it is not fun, and each one is a bite attempt. Joel Beckman’s YouTube channel is a great example of this in action. Muzzled dogs everywhere, constantly trying to bite. It’s the negative reinforcement of his dog Prince returning fire that stops the dogs attempting to bite, not the muzzle.
Broken record time: punishment cannot change the desire to do a behaviour. The only way to change the desire to do a behaviour is reinforcement, either positive or negative.
Even the psychologist who first coined what are (incorrectly) called the quadrants of operant conditioning, Dr R. Miltenberger, a gentleman who has published several textbooks for psychology degree studies with lenghty chapters on punishment, albeit focussed on people with autism spectrum disorder, had to acknowledge: “whilst punishment can supress behaviour in the short term, it is generally less effective and can have negative side effects compared to reinforcement”. If you’ve read to this point, you already know why. Skinner was ahead of his time in that regard. What are the negative side effects Dr Miltenberger refers to then?
These side effects are the bursting out of the supressed behaviour, generally it finds an outlet in another area of life, but it does find an outlet. Examples I have seen recently include resource guarding against owners, new instances of aggression with other dogs, general increases in disobedience or challenging of boundaries, reduction in willingness to make a decision. There are myriad options is this space, these are just ones I’ve seen recently. These are the end result of long-term use of aversive tools, or said another way, the end result of long-term suppression of behaviour.
I mentioned retribution earlier, lets expand upon that for a minute. When someone punishes their dog by preventing it from doing a behaviour i.e. using a leash to restrain the dog from moving forward, say reactive lunging at a bike going past, if the dog does not acquiesce and immediately stop and the person does something like pop the leash or shout at the dog or grab the dog and pin them or shock with an e-collar, or a combination of those, the additional action above the restraining of the dog is not extra punishment, it is retribution. It is frustration / anger bubbling over in the human and they are taking it out on the dog who has exactly zero idea what they did wrong - remember behaviour is not under conscious control, it is instinctual (Trigger causes chemical reaction in limbic brain (“feelings” or “emotions”) and nervous system reacts, causes behaviour. ) and punishment does not have the ability to change the emotions that cause the behaviour, so the layering on of retribution just harms your relationship with your dog. Imagine being slapped about by someone you trust for doing something in an attempt to keep yourself safe like trying to escape from someone threating violence on you.
The above paragraph should explain why in some circustances despite punishments, unwanted behaviour continues to appear - the thing the dog is reacting to causes a bigger emotional reaction in the limbic system than the threat of your punishment, and so the dog does what you are trying to stop with punishment and you are confused and angry. A behaviour occurs due to prior reinforcement, therefore; Reinforcement > Punishment. Always.
Don’t go thinking that if you don’t use aversive tools but carry around a bag full of treats all the time you are not operating in the punishment space. You are. There is no “training system” that is punishment free. Any that claim to be such are either ignorant or lying. Treat “trainers” are operating in the psychological punishment space, not physical. I’ll leave it to the reader to consider which is worse.
When using treats to set up a change in behaviour, lets use the treats to distract from a trigger scenario, you are setting up competing motivators. The dogs desires remain driven by its seeking of positive reinforcement, in each and every situation. Remember the removal of a negative reinforcement is a postive reinforcement. So if a dog is likely to react at the presence of another dog entering the environment, the dog’s positive reinforcement is to make the other dog go away, removing the negative reinforcement, this presents as reactivity;
If the human in said situation is now putting food into the mix, the human is seeking to alter the desire of the dog onto something else it likes, the competing motivator of food. By definition this is punishment because the positive reinforcement of making the other dog go away is being denied. It raises some very interesting questions about what exactly the dog’s brain equates the food with, we don’t know exactly what dog’s are thinking when we shove the food in their face in fear situations. They might be considering murder. They might be thinking why isn’t my human listening to my fear cues? They might be thinking do I look hungry? Or more likely, why is my human a moron?
What is occurring is they are just wanting to get back to being calm, and that happens when the other dog goes away, or, our dog rationalises the situation into non-threatening. Food presentation in fear situations makes zero sense. It is not a reinforcement. At best it is distraction and prevention of learning. In the middle, its bribery and manipulation, I’ve written about this at length in previous articles. At worst, it is a reward for the emotional reaction of fear toward another dog. Fear is a punishment - it stops behaviour. So treats in a fear situation are rewarding punishment. That is a confusing psychological punishment, and not something we ought to be doing. For the results of this, see Zak George’s dogs, anxious nervous messes, psychologically manipulated and confused.
Many dogs choose to ignore the food and continue to focus on the trigger in the environment. This is why many dogs wrongly carry the label “not food motivated”. Yes they are, they are just more motivated by the desire to control the environment around them and establish that they are safe from unknown apex predators that happen to be looking right at them. Part of the problem is that the human is stopping them from doing that by interrupting the learning process, the other part of the problem is the human is attempting to bribe the dog with the competing motivator, which harms the relationship between dog and human. It is also a commentary on how much the dog trusts the human in the situation.
With all of that said, punishment remains a necessary part of having a great relationship with your dog. The enforcement of rules, boundaries and expectations means the dogs desires will sometimes be denied, and that is punishment. Even something as simple as not allowing your dog to choose which direction to go on a walk at a crossroads is a punishment.
You’ll see me frequently using punishment when working with client’s dogs, I will use the leash to hold my ground whilst the dog reacts at something. I'll wait them out until they calm down. That is using punishment to kickstart the rationalisation and reinforcement process. You’ll see me tell a dog on a walk with my body language that we are not going the direction they want, I face a different direction to the dog and hold my ground as they pull on the leash, again I wait them out, this is a punishment, I am denying them their desire in that moment, but crucially, I am allowing them to show me their desires and I will grant them their wishes when I can, give and take is part of any good relationship. I cannot overstress the importantance in a healthy two-way communicative relationship that all parties are given the opportunity to present their wants, this inclusion fosters trust which is vital.
What you won’t see me do is seek retribution when the dog does something wrong, nor will you see me using food to psychologically punish dogs by rewarding punishment.