Tuesday, July 16, 2013

THE KILLER WHALE AND THE KETTLE



Just today, someone sent me the link to Blackfish, the documentary film that re-examines the deaths of three whale trainers attributed to Tillikum, the killer whale.  This documentary led me to do some reading again about the work that is done with killer whales, and what the industry thinks about punishment and aggression and how these things are linked.  I have also been getting a lot of posts on my Facebook page about how we don’t use punishment with zoo animals and thus we ought not use punishment with our dogs.  Most of the time, I agree that punishment is not the right tool for the job.  Some of the time though, it may be the best alternative.  To be clear, punishment is anything that decreases behaviour.  I don’t touch the hot kettle on the stove because it will hurt.  My touching of hot kettles is a very low frequency behaviour because I understand intimately and deeply that it will hurt if I do so and I really don’t want that outcome to occur.

I would not reach out and touch this hot kettle; I know that I would hurt my hand if I did.  I also don't worry about it if I am sitting next to it.  I understand that I can control if I will get hurt or not.  When I touch such an item by accident, I am not traumatized, in part because I am flexible and can cope with a certain amount of unpleasant experience.  Image credit: comzeal / 123RF Stock Photo


If you read through my blogs you will see that I DO use positive punishment and sometimes intentionally and sometimes fairly heavy positive punishment. If you are on my farm, regardless of who you are; a dog, a squirrel, a bird, a raccoon or a coyote, or a person, and you touch the hot wire on my fence, you will get hurt.  The hot wire is an electric fence wire that runs on the top of my fence to keep dogs in and critters out.  It is a simple rule that no one on the farm gets particularly stressed about. It is very much like the rule about hot kettles.  If you touch the kettle when it is hot, it will hurt. It is a simple rule that everyone in the house understands and to my knowledge no one in the house is stressed about.

When I am training, I always ask the animal about how he experiences the process, and there are tons of questions to ask. Did you understand me? I have seen many, many trainers who don't understand that confusion is incredibly aversive, and they are busy feeling great about the fact that they are only using R+ but the dog is more and more stressed because he is confused. At a seminar John went to once he came home and told me about a clicker trainer on stage trying to get a dog to do something; the dog was confused and stressed and the trainer kept talking about breaking the behaviour down into smaller increments; the problem wasn't the method of the training; the problem was that no one had asked the dog if he was comfortable or happy.  Heck no one had asked the dog if he wanted to be on stage!

When I was training my dog Crow, as I write about in my blog (http://tinyurl.com/blc89ce) we had one big stumbling block.  Leash manners. He was terrible on leash. After two years of trying to use only R+ and P-, I “resorted” to a prong collar. We were both so much happier after that, that I learned that I will never ever “resort” to pain as a training tool again. Now if I choose to use pain, I CHOOSE. I choose to use pain, and yes, sometimes making the lesson clear is much more important than avoiding all pain.  Pain can be less aversive than confusion to many learners.

What most folks, (including the author of Coerecion and its Fallout) miss is that pain in isolation is stressful.  Pain that happens repeatedly without warning and that you cannot control is very stressful, but pain that you understand and can avoid is not. If you are not walking around in fear of your electrical sockets, your tea kettle or your woodstove then you understand this on a much deeper level than you realize. I am well educated about contingent control over behaviour. Non contingent punishment is extremely stressful. We have a special name for this; we call it torture. I do not torture my learners. But I do sometimes use P+. When I give D'fer the hairy eyeball when we are in an airport and he is in a goofy mood and is thinking about doing something funny in security, I am using P+. He understands this, and it is not a conditioned P+; he understands that my dirty look is disapproval of his behaviour, and he tones down. He is not stressed because he can control that disapproval. When I am in a bad temper though and I am stomping around a hotel room in a funk, packing and worrying about being late, he IS stressed because there is nothing he can do. He is helpless.

Coming back to the killer whales, and the other marine mammals in captivity that are being so successfully trained using only positive reinforcement, we need to understand something about the difference between a captive zoo animal and a household pet.  This is that most people cannot provide that much structure to their dogs.  If you are a complete control nerd, and you really like ensuring that your dog never ever encounters any stimulus that you don’t have control over, then yes, you can likely train absolutely 100% without any unpleasant consequences.  The problem is, who lives like this?  My life with dogs is fairly structured, but my clients certainly don’t keep their dogs in the sort of controlled environment that zoos and aquaria keep their animals within.  This means that some of the time, we are faced with situations where the dog is going to have opportunities to behave in ways that aren’t what we want. 



What is interesting to me is that I started out as a jerk and treat trainer and I changed as I learned more. I started out using pain to gain control over behaviours. I then switched over to being a completley R+ trainer. I was as R+ as I could be. The more I learned and listened and thought and researched and studied, the more that I learned that P+ can sometimes have a place.  It is as important as any of the other quadrants, as important as classical conditioning and as important as extinction (if you want to see a stressed animal have a watch at an animal going through extinction-they are often very distressed and it is highly unpleasant for the learner and for the trainer), and yet it is trampled down and labeled as inhumane. There are a lot worse things in my life than remembering not to touch the hot wire or using a tool that will cause pain to resolve a behaviour quickly and efficiently. When I had a dog who was predatory to my chickens, yes, I used a shock collar. I guarantee that my dog was less stressed than the two chickens he disemboweled and killed. I used a shock collar on a dog who had been through several surgeries to remove rocks from his gut. That was far less stressful than the surgeries were. 

I think that in the world of training we can easily become academically lazy. If you can train a killer whale without shock, why not the dog who is eating rocks? Let's look at that. We are comparing apples and oranges.  If I had a killer whale who was eating non food items, it would be easy enough to put the whale in a tank without anything non edible in it. If I wanted to teach a wild killer whale not to eat non food items using positive reinforcement only, I would be entirely unsuccessful.  The world itself would put an end to this behaviour the day that the learner ate the first deadly thing.  Shooting the dog is indeed, as Karen Pryor so strongly pointed out, the ultimate in end games for behaviours.  Dead whales don’t eat non food things ever.


It is an exciting time to be involved with dog behaviour consulting, practice and research.  So much is available to us in terms of information, research and data.  The last thing we can afford to do is to be academically lazy.  We must question what we know, and examine what we do each and every day.  Image credit: lightwise / 123RF Stock Photo

I cannot put the dog in a position where he is never exposed to rocks; he lives in a house with kids who bring things inside of the house. He goes on walks outside of the house with rocks in the environment.  There are rocks everywhere and I have few choices in how to stop him from eating them.  A muzzle is a good first step and in the short term, this will work, but in high heat, this will kill my learner and that is not the outcome I am looking for.  In that case, I will choose to use punishment, and likely a pretty significant one.  What I want to do though is to set up a contingency that the dog can learn as clearly as I understand that touching my electric hot wire is going to hurt.  I make mistakes from time to time, but I don’t repeatedly touch the hot wire.  The frequency of my touching the fence is about once a month or less.  I never intentionally touch the wire when it is on, and I am not afraid of the fence because I understand the outcome of doing so.

And let's look at the zoo animals. In case anyone is interested, I am as strongly opposed to keeping wild animals in captivity as most of the R+ trainers are against the use of P+. I think it is morally reprehensible and I will not be a party to it, but it does provide an interesting contrast to what we do with dogs. Let's consider the life of your average killer whale in a tank. Do you know that they cannot properly ecolocate within the echoing concrete of the tanks they live within? Do you realize that these animals typically travel several hundred miles a day, and we keep them in tanks of less than ten acres in size? Do you realize that they are environmentally enriched less than 50% of their days? Do you know that they are likely on par with us in terms of our intellectual capacity? Can you imagine what it would be like to live in your kitchen for the rest of your life with exciting trips to the living room once a day to "interact" with a trainer? I think that what we do to zoo animals is far more inhumane and horrendous than what I do with positive punishment.  Yes, we can get spectacular responses from all kinds of animals when the only game in town is what we offer them for the few hours a day that they are able to come out and train with us.

This sterile environment is where captive whales spend most of their time, with little to do and a tank that is so small that they cannot echolocate and communicate normally.  We never use pain to train these incredible creatures, but we keep them in such appalling conditions, that using no pain to train seems like a small nod indeed to humane treatment of the learner.  Image credit: ozbandit / 123RF Stock Photo

I want people to think about never ever again being able to say "thank you, no, I don't want that" to your child, your spouse, your parent or your sibling. In training, punishment is the tool we use in place of "no. don't do that". If you want an interesting week, frame every conversation you have for the week in terms of thank you for trying, I would like something else. Never ever say no, just reframe your request in terms of "thank you, please do something different." It doesn’t work any better if you frame everything you do in terms of only attending to the things you like without any other information.  Sometimes that will work, and more often than not, it will not. What often happens in the classroom when teachers try this is that the willing students will be successful and willing, the students who don't care will not advance at all and the students who are not compliant will become more and more and more creative in ways to try and get the information about what is not allowed.  Often this tactic results in a very frustrated trainer and an even more frustrated learner.

There are two interesting and competing theories in Applied Behaviour Analysis that do not get discussed nearly often enough.  They are called two factor and one factory theory.  Two factor theory says that when negative reinforcement is used, both operant learning and classical learning happen at the same time.  This means that not only will the learner increase his behaviour, but he will develop a classical association with the signals that the aversive stimulus may occur, and will suffer distress when the signal is present.  This is the most common argument that I have heard for why people eschew the use of both negative reinforcement and positive punishment.  By extension, when this concept is applied to positive reinforcement, not only will the behaviour increase, but the signal that a appetitive or pleasant outcome is available will produce good feelings in the learner.  The problem is that there is some good research out there indicating that this may not in fact be happening.  Pavlov may NOT be part of every single interaction you have with your learner.  One factor theory says that only one thing is occurring at a time; that if you increase a behaviour through negative reinforcement or decrease a behaviour through positive punishment, the only thing that is happening is an increase or decrease in behaviour; the procedure is entirely operant.  People who believe strictly in one factor theory would argue that you are simply increasing or decreasing the frequency of a behaviour when you train.  Two factor theorists would argue that the learner's well being is attached to the choice of method of behavioural change.

I would like to propose a third option; that is that some of the time, two factor theory is relevant and some of the time, one factor theory prevails.  From my observations of over thirty years of training, when the animal has work that is meaningful to him, and an environment that he can predict and control through his behaviour, then one factor theory is going to control.  When the animal is uncertain, his welfare is dicey and his work is not meaningful, then two factor theory will be the more important scenario.  In my opinion the difference between the whale in the tank and the dog in my home is that with the domestic dog who evolved to fill the niche that was created by the detritus of the human environment is that the dog is usually a stable individual in his native environment.  Of course two factor theory will prevail when you are talking about training animals who live in horrendously suboptimal environments.

Punishment isn't something I use often, but I do use it and I support the use of it in some situations. Understanding about two factor and one factor theory has helped me to see that learners experience both or either operant and classical conditioning some of the time depending a lot on their state of mind and welfare.    Most of the time, punishment is misused because it is very poorly understood.  Most of the time it is not the right tool for the job. Some of the time though, I believe it is.    When the conditions are right, and the one factor theory is in effect, it can be very helpful.  Punishment doesn't have to have baggage with it; sometimes it is just a kettle we should not touch.  Done properly, that is exactly the effect.

1 comment: