In Defence of Dogs

by John Bradshaw

  In the means–end test, most dogs pull on the string that is nearest to the food, not the one actually connected to it

  Dogs do better in another aspect of ‘folk physics’, the ability to count. Because this ability is regarded as an indicator of intelligence, scientists have examined it in a wide variety of animals, including dogs. It is clear that dogs can tell the difference between a half-full bowl of biscuits and one that is a quarter full, but do they actually count the biscuits or simply judge the size of the pile? Researchers have attempted to answer this question using a technique first developed for the study of human infants.9 When babies as young as five months are shown first one doll, and then a second, and then, after a brief gap, three dolls in the place where logically there ought to be two, they look at the three dolls for longer than expected – they seem surprised that the third doll has apparently come from nowhere. Their reaction suggests that they had added together one doll plus one doll, and were therefore expecting only two. It seems logical that dogs should also be able to do this; a bitch that sees two of her puppies go momentarily out of sight should presumably be surprised (and get up to investigate) if, say, three, or only one, reappears.

  To test dogs’ ability to count, the researchers in this study used food treats rather than dolls. The dogs were shown first one treat being placed in front of them, and then a second. Next, a screen was placed between them and the treats. When the screen was removed and either one treat or three treats were revealed, they stared at the treats for a long time, as if in disbelief. If there were two, as there should have been, they glanced at them only briefly.

  In summary, although rather few studies have been conducted, it seems that dogs have little intuitive grasp of how the world around them works. This was Thorndike’s conclusion from his puzzle-box experiments, and one that has been confirmed by all subsequent experiments. Of course, dogs can easily learn how to manipulate some specific aspect of their surroundings in order to get what they want, but they do not appear to understand why the manipulation works, just that it does.

  Dogs, as the descendants of social animals, are likely to be much more attuned to one another than they are to the physics of their surroundings. Accumulated experience evidently plays a big part in enabling canids such as dogs to exploit their environment and find enough to eat; indeed, it would seem inefficient if each animal had to learn everything from scratch. Evolution should favour the transfer of skills from parent to offspring, and the longer the young are dependent on the parents, as in the canid ‘family pack’, the more frequent are the opportunities for this to occur. Thus there is good reason to conclude that dogs should have inherited some potential for learning from one another.

  This is not to say that one dog deliberately teaches another in the way that we teach our children. Biologists who study how one animal learns from another usually try to use simple explanations rather than those that require a capacity for complex mental processes. Start digging your garden, and your dog may try to ‘help’ by digging alongside you. Is the dog really imitating what you are doing? If so, why does not it try to pick the spade up in its paws? More probably, your digging has simply drawn its attention to the soft earth, and digging is a normal thing for a dog to do in soft earth. Biologists refer to this as stimulus enhancement: it is sensible for one animal to keep an eye on what other animals are up to, in case it sees something useful. But the observant animal’s subsequent behaviour is more likely to be dictated by what it would normally do than by an urge to imitate the other’s actions precisely.

  Some studies have found no evidence for copying. Thorndike himself researched the question of whether his dogs could learn to escape from his puzzle-boxes by watching other dogs that had already learned the trick, but he concluded that there was no evidence that they learned anything from observing the behaviour of the other dogs. In another study, pet dogs watched through a transparent gate as a trained German shepherd cross called Mora performed one of two tricks: lying down on her belly, or ‘playing dead’.10 Both are tricks that many dogs can do, and indeed might have already been trained to do. The twist was that the trainers issuing the commands to Mora used arbitrary words as the cues (their own names), which the observer dogs had presumably never heard before: ‘tennie’ for lying on the belly, and ‘josep’ for playing dead. Each observer dog was allowed to watch (and hear) the demonstration going on five times. The researchers then tested to see if the observer dogs had learned the commands from watching Mora. Apparently not; when they were tested with the commands, none seemed to know what to do. A few lay down on their bellies, but not in response to ‘tennie’, and none ‘played dead’, with or without the ‘josep’ command. Children gain the ability to learn by imitation at about eighteen months, so the dogs did not do very well at this task, if judged by human standards.

  However, other research has shown not only that dogs can copy other dogs, but also that they are selective and logical about what they copy.11 In one study, dogs between twelve months and twelve years old were trained to obtain food from a box that opened when a wooden handle was pulled down. Most dogs would naturally do this by grabbing the handle in their teeth and pulling it, but these particular dogs were trained to pull it down with their paws, as shown in the picture below. Next, other dogs were allowed to watch the dogs performing their new trick. If all the trick did was draw attention to the handle and the food, then most of the dogs should simply have pulled the handle using their mouths. But some of the dogs started using their paws, as the demonstrator dogs had, suggesting that they were copying the action itself.

  A dog demonstrating pulling a handle with its paw, while holding a ball in its mouth

  The experimenters then added a twist to this test that seems to show why the dogs sometimes copied the action of the demonstrator. The dogs that were demonstrating the handle-pawing action had been trained to hold a ball in their mouth while doing so. When actually demonstrating, they were sometimes given the ball to hold, at other times not. When the demonstrator was holding a ball (as in the picture), the other dogs often used their mouths; it was as though they were thinking, ‘That dog is using its paw only because its mouth is full – I’ll use my mouth, it’s easier.’ In contrast, when the demonstrator was not holding a ball in its mouth, most of the observers used their paws, as if thinking: ‘That dog is using its paw, not its mouth, so that must be the only way to get the food.’ This experiment, if understood correctly, suggests that dogs are capable of quite sophisticated reasoning (although human children can make deductions of this type as young as fourteen months). Experiments like this one may in the future lead to more insight into what dogs are capable of thinking. Perhaps we shall find out that dogs are better at imitating when they are trying to get hold of something they can see than when doing something simply to please their owners, such as lying on their bellies or playing dead.

  In any case, dogs’ abilities to solve problems and to learn from observing other dogs should not be explained glibly in terms of the emergence of similar abilities during the development of human infants. Dogs develop skills and insights that are appropriate for their own species; we do likewise. For example, human infants begin to acquire language skills at an early age, something dogs never achieve, but children are older than dogs when they acquire the ability to imitate selectively. Dogs, in turn, have inherited a set of learning skills from their canid ancestors that evolved over millions of years to allow co-operative hunting and rearing of young. Hunting as a pack can be only as efficient as the least experienced member of the pack, since it will be the weakest link that allows the prey to escape. Young canids must therefore pick up skills from their parents as quickly as possible if they are not to jeopardize the hunting success of the pack. Biologists are only just beginning to understand dogs’ ability to learn from each other: the necessary experiments are difficult to design and the results are often open to more than one interpretation.

  Unlike the other canids, domestic dogs al
so have the opportunity to learn from people. Indeed, the close co-operation that can occur between man and dog suggests that domestication is likely to have enhanced this aspect of the dog’s intelligence. Most of the abilities discussed so far in this chapter are probably held in common with wolves and other canids; it is simply that we know more about domestic dogs because they are easier to study. Yet the fact remains that dogs are domesticated and other canids are not, so the question arises: is there anything about the dog’s intelligence that is a unique product of the domestication process, one that has enabled dogs to interact with us at a level of sophistication no other animal can match? In proportion to body size, the dog’s brain is somewhat smaller than that of its ancestor, the wolf, so it is unlikely that dogs are simply smarter than wolves. Nevertheless, dogs can outperform even chimpanzees, probably the most intelligent mammals apart from man, in certain selected tasks. Nowadays biologists tend to think of animals not in terms of whether they are more or less intelligent than one another but rather in terms of how their mental processes match the demands of their lifestyles. The dog’s lifestyle is so intimately connected with our own that it is reasonable to look for special intellectual capacities that they may have gained during the long process of their domestication.

  One area in which dogs especially outdo chimpanzees is in their ability to extract information from what humans are doing, in particular their ability to read human faces and gestures. Even hand-reared chimpanzees take a long time to learn this. And they can become confused when, for example, the person who first trained them is replaced, even if that person does his or her best to behave in exactly the same way as the original person. In contrast, dogs trained by one person can quickly learn another person’s version of the same commands. Dogs are particularly good at following pointing gestures, outperforming even chimpanzees. There is currently some argument among scientists about whether or not this ability is a unique product of domestication, but it seems most likely that it was present to begin with, pre-domestication, and has been refined and developed since (though what its function might have been for wild wolves is unclear).

  Dogs’ ability to follow pointing gestures (and, by inference, other human gestures as well) is not quite what we would expect intuitively. Like humans, dogs can follow pointing not only with the arm nearest to the target (in the tests, usually a pot covering a food treat) but also cross-pointing, with the opposite arm (as shown). They prefer to follow a pointing arm even when the person pointing is nearer to, or moving towards, the ‘wrong’ target. However, dogs do have their limitations. They take much less notice of a pointing finger than a pointing arm (children are happy with either) – but they do follow a pointing leg! The rule seems to be: ‘Take the direction indicated by whatever whole limb is obviously pointing somewhere.’

  Following a cross-point

  There is also some argument about whether dogs are born with the ability to follow pointing gestures, or, conversely, have to learn it. Six-week-old puppies seem to know what pointing means, but are less good at following it than adult dogs are – puppies may just be too easily distracted. There is little doubt that this ability can be refined by training; for example, gundogs that are highly attentive to their handlers test better than other types of dog. It is also possible that domestication has prepared dogs to learn the significance of pointing very quickly, without giving them an instinctive ability that they can use the very first time they need it. But the response to pointing is clearly not universal: in some studies of pet dogs, over half do not respond spontaneously to pointing. Clearly, a strong component of learning is involved. Some dogs seem to find it difficult to learn to follow pointing even when rewarded for doing so, but many of these are dogs from shelters, and may have become fearful of outstretched hands due to physical punishment that they have received in the past. The unanswered questions are (a) whether the underlying ability is universal, but some dogs learn not to respond to hands because they deliver punishment, or (b) whether all dogs have to learn the significance of pointing, and some simply find the task easier than others.

  Although pointing has become the scientists’ favourite experimental tool, it is by no means the only activity to which dogs are particularly attentive. They also follow gestures such as nodding and hand movements much more attentively than most animals do. In addition, dogs seem fascinated by people’s eyes and faces: they will follow the direction of their owner’s gaze almost as reliably as they will follow pointing.

  Domestication’s main effect on dogs seems to be that it has rendered humans their most relevant source of information. For example, most dogs faced with the impossible problem of getting a tasty food treat out of a locked box turn to the nearest person for help within a few seconds, whereas even hand-raised wolves just keep scrabbling at the box. However, wolves outscore dogs when quick decisions have to be made, whereas dogs will tend to repeat what they have learned from people, even when it is obviously the wrong thing to do.

  In one experiment that supports the idea that dogs are hyper-dependent on people, a comparison was made between the abilities of dogs, ten-month-old human infants, and wolves, in following the progress of a ball between two screens.12 The experimenter first hid the ball behind one of the screens four times, each time talking to and maintaining eye contact with the dogs. Each time, the dogs were then allowed to retrieve the ball from behind the screen and play with it. Next, the experimenter walked behind the first screen, but quite obviously left the ball behind the other screen, finally showing the dogs her empty hands. The dogs nonetheless continued to look for the ball behind the first screen, even though they had seen it disappear behind the second one. Clearly, the attention they had received from the experimenter on the first four occasions had marked the first screen as the best place to try first. The ten-month-old infants made exactly the same mistake, again apparently prioritizing social cues. The wolves, however, believed their eyes and went straight to the second screen. Giving a low priority to cues given by humans, they instead relied on their interpretation of the physical world, in this case presumably a skill that would enable them to guess where a prey animal was most likely to be hiding.

  Interestingly, however, the dogs were not incapable of understanding where the ball had gone. They did get the right answer when the ball was moved from one screen to another by an invisible string rather than by a person. Evidently their first priority was always to do what a human had encouraged them to do in the past. Even hand-raised wolves maintain a sufficient degree of independence from people to keep their minds on the problem as presented to them; dogs are too easily distracted, too eager to please a human. Just because they can follow a human’s gaze or hand gesture does not necessarily mean that dogs understand what that person is thinking. They might simply be using the person’s eyes as a convenient but otherwise arbitrary pointer as to what they should attend to next. Two remarkable studies done in France have exposed some of the dog’s limitations.13 Detailed comparisons were made between guide dogs, owned by blind people for several years, and ordinary pet dogs living with sighted owners. First, the researcher studied how the dogs tried to get food from their owners. They all used the standard doggie routine of looking forlornly at their owner, then at their bowl, and then back again: the guide dogs gave no indication that they knew their owners were blind. The only difference was that the guide dogs made louder slurping noises, which their blind owners could and did attend to. This tactic can be accounted for by simple associative learning – the dogs had learned that food followed when they made the noises – and does not prove that the dogs understood anything about blindness. Next, the study examined how dogs draw the attention of their owners when they are trying to get at a toy made inaccessible behind a heavy wooden box. Here, too, the same pantomime of looking back and forth between the owner and the toy went on, irrespective of whether the owner could see it or not. Other scenarios produced similar results, such as the owner offering the dog a different toy instead o
f the one it wanted: again, no difference. There was no indication that the guide dogs knew that their owners were blind, and that they thus had to rely on other cues, such as sounds, to tell them which way to point their heads.

  These experiments tell us that dogs have a passion for following the human gaze, but they do not tell us whether the dogs are born with this obsession, or whether they learn it. The guide dogs that were studied had not always lived with blind people; they had been raised by sighted families for the first twelve to eighteen months of their lives, before starting guide-dog training. Perhaps these are the critical months in which the habit of following a person’s gaze is learned, after which it is very difficult to break. However, it is remarkable that the guide dogs seemed hardly to have altered their behaviour at all in the four years, on average, that they had been living with a blind person.

  This apparent over-reliance on human eyes and arms raises the question of whether dogs actually have the ability to understand what we are thinking. They are clearly responsive to what they see us doing, but that is not at all the same as understanding that each person has a mind, one that, furthermore, is different from their own. Such an idea must seem like heresy to the majority of dog owners, but science has thus far failed to demonstrate such abilities in any species apart from our own. Even the many experiments that have been done on chimpanzees have failed to provide conclusive evidence that the great apes know that other minds exist.

  Part of the problem is that it is very difficult to design experiments that test whether dogs are able to tell what we are thinking, in other words, experiments that exclude simpler explanations for their behaviour. For example, in a study examining whether dogs understand what people can and cannot see, eleven dogs that had been trained not to steal food were taken into a room where there was a piece of food on the floor.14 When the experimenter told the dogs not to eat the food when she was facing them, they usually left it, but when she was facing away from them while issuing the command, they often did take the food. The dogs might have ‘known’ that they could not be seen, but the simpler explanation is that they associated seeing a human face with obeying a command – ‘If there is no face, then the verbal command doesn’t mean anything’. Given that dogs are so sensitive to human faces and the expressions on them, it is very difficult to think of experiments that would rule out such an explanation.