The widespread use of punishment-based dog training is usually traced back to Colonel Konrad Most, whose highly influential book Training Dogs: A Manual first appeared in 1910 (in German) and was translated into English, due to popular demand, in 1944. Most was emphatic that the relationship between man and dog was not only hierarchical—with only one “winner”—but could be established only through physical force, by an actual struggle in which the man was instantaneously victorious. The dog had to be convinced of the absolute physical superiority of the man.4 This approach demands that the owner constantly maintain and reinforce his or her position at the head of the family pack. In this conception the dog perceives the people it lives with as fellow-members of its pack, and misbehavior is construed as being due to a failure of the owners to maintain their dominance over the dog. Training methods are accordingly designed to lower the dog’s position in the hierarchy or “pecking order,” as it is sometimes referred to.5
The Monks of New Skete, best-selling authors of dog training manuals for over thirty years, are highly influential promoters of this philosophy. They maintain that understanding wolf behavior will help owners to understand their dogs, and that books about wolves are often of more use to owners who wish to understand and appreciate their dog’s behavior than dog-training manuals.6
The Monks are very specific in turning this principle into practice. For instance, for aggressive dogs they recommend the “Alpha-wolf roll-over.” This is a disciplinary technique nicknamed for the way the lead wolf is supposed to punish misbehaving members of the pack,7 whereby the dog is grasped firmly by the scruff of its neck and vigorously rolled over onto its back. For puppies, they recommend the “shakedown method,” which they claim resembles what the mother does to her pups to keep order in the litter:8 The puppy is grasped by the loose skin on either side of its neck, lifted off its front feet, and shaken.
Those dog trainers, such as Dr. Ian Dunbar, who promote reward-based methods regard this “dominance reduction” as both unnecessarily cruel and based on a complete misconception; they fundamentally reject the assumption that because an animal is misbehaving it must mean that the misbehavior is motivated by a desire to have high rank. Instead, they rely on much simpler explanations based on the science of animal learning, emphasizing that many of the behaviors that animals perform are performed simply because those behaviors have been rewarded many times in the past.9 They rarely make special reference to the dog’s origins as a wolf, since the effects of reward on behavior are universal among vertebrate animals.
Some experts go even further, asserting that training techniques derived from the dominance concept can actually harm the dogs they’re applied to. Their uppermost concern is that punishment-based methods, often used in an attempt to cure a supposed “dominance problem,” may initially suppress the behavior but can then cause the dog to become depressed and withdrawn.10 Even worse is what can occur if the “dominance reduction schedule” does not work: If the misbehavior continues, the owners may come to think that they are not asserting their position strongly enough and become more and more aggressive in their attitude. Eventually the dog may become so fearful of them that it bites them in self-defense.11
Personally, I’m delighted that the most recent scientific evidence backs up an approach to managing dogs that I’m comfortable with. As a scientist as well as a dog lover, I am dedicated to assessing the best evidence available and then deciding on the most logical approach to adopt. If wild wolf packs had turned out to be as fraught with tension as their counterparts in zoos, I’d have to agree that the dominance approach had merit. I’d still be reluctant to adopt punishment rather than reward as my philosophy for training my dog, because for me the whole point of having a dog is the companionship it brings, and domination and companionship don’t jibe for me. As a dog owner, I was relieved by the discrediting of the wolf-pack idea, since I could then explain to myself and, more importantly, to others why routinely punishing a dog is not only unnecessary but also counterproductive.
Both sides of the debate over proper techniques for dog-training claim that their approaches are based on serious science. Dog owners, not surprisingly, have trouble assessing which claims are true and, therefore, how best to train their dogs. The issue of how closely dogs’ behavior aligns with the behavior of their wolf ancestors turns out to be something of a distraction in this regard, because when it comes to training, the most important question is really How do dogs learn?
First of all, it’s important to stress that dogs are learning all the time—not just during formal training. Or, to put it another way, dog owners often do things that train their dogs without being aware that they’re doing them. Dogs learn especially fast while they’re growing up; they can modify the “instinctive” ways in which they communicate with one another and with us; they learn how to get on with other dogs and with the people they meet. From a young dog’s perspective, there’s not much difference between the training class and everyday life; the dog will learn all the time. However good the training session may have been, owners need to bear in mind all the opportunities their dogs have for learning, not just those formally labeled as “training.”
Dogs learn in much the same way that other mammals—including humans—learn. However, species vary slightly in terms of what they find easiest to learn and what motivates their learning. One reason that domestic dogs fit into human communities so well is that they find human contact very rewarding and, conversely, become anxious when separated from their human companions. Thus they are strongly motivated to do things that please their owners or, if they can’t work out what those are, to at least get their owners’ attention.
However, if you want to train any animal to do something, it’s easiest to start with a behavior that the animal would do anyway. Obviously, not all animals are equally easy to train, either in general or with regard to particular behaviors. Biological heritage does matter. On the one hand, many of the things that we train dogs to do, such as rounding up sheep and retrieving game, make use of pieces of behavior that evolved millions of years ago as part of their canid ancestors’ natural hunting behavior. On the other hand, as predators, dogs do not naturally run away from things unless they’re scared of them, so it’s much more difficult to train them to, say, pull a cart than it is to train prey animals such as horses to do so.12 Nevertheless, there are fundamental disagreements among trainers about how dogs are motivated to learn. Old-school advocates, supported only by tradition, think dogs need to learn their place in the pack; modernists, supported by scientific evidence, think dogs learn to please their owners.
What, then, does it mean for a dog to learn? We can usually say that a dog has learned something when its typical reaction to a particular situation changes. Short-term changes that can be ascribed to internal processes such as hunger don’t count. The more time that passes since a dog’s last meal, the hungrier it will get, but its interest in food wanes after it has eaten. That’s not learning. However, when a dog suddenly gets excited when it hears its food bowl being taken out of the cupboard and then repeats this behavior every day, we can be sure that it has learned something.
The simplest kind of learning is habituation, defined as the waning of a response to an event that turns out to have no consequences. Most animals have sense organs that pick up far more information about the world than they can possibly attend to, and dogs are no exception to this rule. To avoid wasting time, animals need a mechanism that allows them to avoid responding over and over again to something that their senses are telling them they might want to attend to but does not actually need to be bothered with. It’s a very primitive and universal ability: Even animals without nervous systems can do this, or something like it.
Habituation explains, for example, why dogs can rapidly lose interest in a particular toy. If dogs are repeatedly offered a favorite toy—a squishy teddy bear, for example—they will stop playing with it, usually after only five or six presentations. But if the to
y is then swapped for one that is only very slightly different—identical in appearance except for a different color or odor, say—they will start playing with the new toy just as excitedly as they had with the first one. Of course, they quickly get bored with the second toy as well, since there is nothing intrinsically more exciting about it compared to the first toy.
Why do they show this rapid loss of interest? We don’t know, but it’s tempting to speculate that it’s connected to the dog’s origins as hunters. Something taken into the mouth and maybe tossed in the air is worth persisting with only if it produces food, or at least starts to break apart and so might eventually yield food13—which may be why many dogs love to rip their toys apart. But something that remains unchanged even after repeated chewing is probably not worth bothering with.
Habituation can be useful, for both dogs and their owners, because it reduces a dog’s anxiety in reaction to unexpected events. Of course, the technique works only if the trigger for the fear (say, the sound of fireworks) has no actual consequences for the dog (is not actually painful). As a training technique, the stressor has to be presented at a level that is just high enough to be detected by the dog but not high enough to frighten it. Commercially available recordings of gunfire and firework noises are a very practical way of reducing the anxiety associated with such noises. Once this level has been established, the intensity of the sound can be increased very, very gradually, with quite long gaps in between, such that the dog becomes more and more habituated to it, eventually reaching the point where even the “normal” everyday intensity is no longer important enough to make the dog feel frightened.
Bear in mind that great care has to be taken to avoid raising the intensity of the stimulus to the point where the dog becomes even slightly frightened. If this line is crossed, the process goes back several stages before you can try again.
The opposite process, sensitization, occurs when the dog panics because it cannot escape from whatever is scaring it—this is a common cause of so-called firework phobia. Many dogs are frightened of loud noises; some will gradually get used to them, if their initial exposure is not too intense, either by good fortune or because their owner has had the foresight to deliberately habituate them. Others, those that may be intrinsically nervous or whose first exposure is especially intense, on finding that nothing they do makes the noise go away, react more and more intensely on each successive exposure. Once this has happened, even very low levels of the stimulus will trigger the feeling of fear, making habituation almost impossible. The technique known as flooding (exposure to extreme intensities of unavoidable fear-inducing stimuli) can be successful in the treatment of irrational phobias in humans, but when used on dogs and other animals less rational than ourselves it is much more likely to make the fear even more deep-seated.
Both habituation and sensitization are forms of learning—both change the way the dog responds emotionally to situations. Each combination of external events, as recognized by the dog, triggers one emotion—in this case, fear. The specific combination may be important; for example, a dog may lose its sensitivity to loud noises when it’s at home but still become frightened when it’s out in the car. The dog may still not like the sound of fireworks but has learned that nothing bad follows when it’s at home. The car introduces a new context, one in which the dog has never previously heard fireworks, and so the fear resurfaces.
The role of context is essential also to understanding much more complicated forms of learning, including associative learning. Associative learning occurs when two hitherto unlinked events get connected together in the dog’s mind. A dog may learn that “I get fed soon after my owner gets my bowl from the cupboard,” that “soon after the doorbell rings the door opens and people come through it,” that “there are rabbits in that wood,” that “when my owner says ‘sit’ it’s fun to sit down,” that “if I fetch my leash my owner may take me for a walk,” and so on. These are all simple associations, either between two sets of information (bowl = food, bell = people, wood = rabbits to chase) or between doing one thing and achieving something (sit = praise from my owner, fetch leash = walk).
In psychologists’ jargon, these two types of associations occur in what are respectively referred to as classical and operant conditioning. Classical conditioning is also sometimes referred to as Pavlovian con-ditioning, after Ivan Pavlov’s famous experiments conducted in the 1900s. Having noted that dogs anticipated the arrival of food by drooling, Pavlov was able to show in these experiments that if the arrival of food was always preceded by a bell, the dogs would start to salivate whenever they heard the bell—whether or not the aroma of food was also present in the air. Thus he helped to establish the fact that animals such as dogs were able to quickly learn the significance of artificial cues that evolution could not have prepared them for. Subsequent studies have demonstrated that something that reliably predicts a mealtime, such as the bowl coming out of the cupboard, doesn’t just make the dog drool (and the quantity of drool is precisely what Pavlov measured) but actually conjures up some kind of mental picture of food in the dog’s mind. (Knowing dogs, it’s probably an olfactory representation rather than an imaginary snapshot of some brown stuff in a bowl.) Thus something that the dog instinctively likes (in this case, the smell of food) gets linked to something arbitrary, something that would otherwise not mean much—the owner getting something out of a cupboard.
Classical conditioning is automatic; it doesn’t involve the dog’s reflection on what has just happened. For this reason it works well only when the arbitrary stimulus comes immediately—within one or two seconds—before the stimulus that the dog is already programmed to respond to. If the bowl appears and then the owner is distracted, the dog will probably drool until the food eventually arrives. If for some reason the owner changes her routine and starts getting the bowl out long before she fills it, the dog will, after a period of frustration (and drooling), start to unlearn the association—a process referred to as extinction. On the other hand, given that food is very important to dogs, they may be conditioned to another predictor that is present at the same time—perhaps the owner getting the pack of food from the cupboard.
Classical conditioning also works in the opposite direction, if the association is with something the dog doesn’t like. If a dog treads on a thorn and hurts its foot, it will immediately associate the pain with the place where it got hurt and avoid it for a while. Such aversions are generally quite long-lasting, partly because the dog isn’t disadvantaged much if it stays away from that place for a while—in biological parlance, it has an alternative strategy available. Dogs also don’t like electric shocks and will rapidly learn to predict when they’re going to occur, if they possibly can. This is the concept behind a product for dogs called the “pet fence,” which involves a collar that delivers a mild shock to the dog’s neck. The shock is triggered when the dog gets close to a buried wire emitting a radio signal, visually marked out by a line of flags or something similar. The collar also makes a beeping sound just before it’s about to deliver the shock; the dog rapidly learns that the beep means the shock is about to happen and also associates the shock with the location in which it previously happened. The dog should then be able to learn to turn away whenever it hears the beep, thus avoiding the shock itself—in short, it is given an alternative strategy.
In the context of applying the principles of classical conditioning to dog training, it’s crucial to appreciate that dogs live in the here-and-now to a much greater extent than humans do and, therefore, that they may associate any punishment (or reward) with something that we humans may not expect. For example, many owners punish their dogs, verbally or physically, when they come home to find that the dog has done something wrong. They assume that the dog will be able to think back to whatever that deed was and thereby associate the punishment with it. However, as noted earlier, dogs don’t do mental time-travel at all well. What the dog actually does in such instances is to associate the immediate situatio
n—the owner’s return—with the owner’s angry words and any physical punishment that follows. In short, the dog associates events that happen immediately one after the other. The mess in the room is highly relevant to the owner but much less relevant as far as the dog is concerned; the dog is incapable of reflecting on what the owner is angry about. What is different from the “pet fence” situation is that here the dog has no alternative strategy; it has no means of avoiding punishment, because it does not understand what has precipitated the punishment in the first place, nor has it had any warning that the punishment is imminent. Because it does not understand the causation, the dog is unable to predict when its owner is going to come home angry and when not. It’s like a rat in a cage, being shocked at random. Researchers have established that rats can become quite tolerant of mild electric shocks, even in situations where they can’t avoid them, if they are given a reliable warning of when they’re going to occur. However, a rat receiving exactly the same shocks, but this time without the warnings, becomes progressively more anxious and stressed. The same is true of dogs.
Although almost all learning in dogs takes place between events a second or two apart, there is one major exception—namely, when the “punishment” takes the form of an upset stomach. An inexperienced young dog may pick up the rotting corpse of some animal on a walk, mistakenly think that it is good to eat, and then feel nauseous and vomit it up an hour or so later. Although these events seem to be much too far apart in time for conventional classical conditioning to work, it would be very useful to that puppy if it could learn not to eat such things again, and indeed this is what happens. There is a special rule for food—connect taste/odor of last meal with painful stomach—but it is confined to food, and doesn’t seem to apply to any other learned associations. This necessary relaxation of the usual rules can have unintended consequences: Animals (this applies to humans too) can “go off” a particular food if they are affected by a gastric virus within a few hours of eating it, even if the food did not actually cause the stomach problems. But this confusion is presumably a price worth paying for learning to avoid genuinely toxic foods.
Dog Sense Page 13