In Defence of Dogs

by John Bradshaw

  Domestication has not adapted dogs to human environments; it has merely given them the means to adapt. Exposure to both people and man-made environments must occur, in a gentle and gradual way, to enable them to learn how to cope. This process starts in about the fourth week of their life and goes on for several months. If the exposure is either deficient or defective, the dog will develop deep-seated fears or anxieties that can be very difficult to eradicate later. Although some details of precisely how this happens have not yet been scientifically explored, the overall process of this ‘socialization’ is well charted, and it is a tragedy that so many puppies do not receive enough experience of everyday life to allow them to cope adequately with their life among humans.

  In 1961, a short paper appeared in Science that completely revolutionized our thinking about the bond between man and dog.1 In order to study when puppies are most sensitive to exposure to people, the researchers raised five litters of cocker spaniels and three litters of beagles in fields surrounded by a high fence, so that they never saw people – food and water were provided through holes in the fence. Then every other week from the time the puppies were two weeks old until they were two months old, a few of them were taken out to live indoors for a week’s ‘holiday’, receiving an hour and a half of intensive contact with people each day. At the end of their week’s socialization, they were put back in their field with their mother and littermates.

  The timing of the ‘holiday’ was absolutely crucial to how the puppies reacted to being handled. At two weeks old, they were too immature and sleepy to interact much, but the puppies taken out at three weeks were instantly attracted to the person looking after them. They would paw and mouth the researcher and play with the hem of his lab coat. Five-week-old puppies were wary for a few minutes, but soon started boisterously playing with the person. Seven-week-old puppies needed two days of coaxing before they could be persuaded to play, and nine-week-old puppies took even longer, only becoming friendly in the second half of their week’s holiday.

  The timing of their first introduction to human contact was absolutely crucial to how the puppies reacted to people later on. All the puppies in the experiment were taken out of the field when they were fourteen weeks old, and at that point began to live with people like normal dogs. The five puppies that had spent all of their lives in the field never learned to trust people, even after months of intensive handling. The six that had been taken on ‘holiday’ when they were only two weeks old, and were then returned to the field for eleven weeks, fared better: though they were initially quite wary of people, they became somewhat friendly after a couple more weeks of gentle attention. All the other puppies were instantly friendly – remarkable, given that some had last seen a human over half their lifetime ago. The six that had not seen a person for ten weeks were initially difficult to leash-train, but training the others was straightforward.

  Overall, the results indicated that puppies need some (but not very much) contact with people if they are to react in a friendly way towards them. There also seems to be an optimum age for this contact to be effective. Two weeks old appears to be too early. Twelve weeks old is definitely too late; by this age the puppies observed in the study had become fearful of anything they had never been exposed to when they were younger. This implies a window of opportunity between about three weeks and ten or eleven weeks of age – what the scientists referred to as their ‘critical period’.

  The idea of a ‘critical period’ derives from a 1930s study by Nobel Prize-winning biologist Konrad Lorenz. Suspecting that some animals have to learn their mother’s identity, rather than knowing it instinctively, Lorenz hand-raised a clutch of goslings. His prediction proved to be correct: having never seen their mother, they adopted him as their ‘parent’, following him around like a pack of faithful hounds, and paying no attention to their biological mother.

  What Lorenz had discovered was the process now known as filial imprinting, whereby young animals learn the characteristics of their parents.2 Geese will imprint on to the first moving object of about the right size that they encounter between twelve and sixteen hours after hatching. In the wild, this is so likely to be the mother goose that the chances of anything going wrong are remote. It is essential to their survival that goslings know what their mothers look like, otherwise they could easily stray away from the nest and perish. But why do they need to learn this? Would it not be more sensible if they hatched with the mother’s image already burned into their brains? Biologists have no definitive answer to such questions, but perhaps learning is simply easier; three-dimensional images are probably difficult to encode in DNA. In fact, studies show that young birds are born with some inbuilt guidelines for what to look out for: something that moves, makes bird-type noises, and has a head and neck. (But not much more than this: for example, if they are prevented from seeing their mother, domestic chicks will readily imprint on to a stuffed ferret.)

  Lorenz originally conceptualized his ‘critical period’ as a rigid timetable of events. In the gosling example this enables a young bird that is mobile within a few short hours of hatching, and unlikely to survive for long on its own, to quickly latch on to its mother. It is now known that there is more flexibility in this type of learning than was first thought: subsequent research has shown, for instance, that a gosling hatched in an incubator and kept away from its mother until it is thirty-six hours old can nonetheless bond to her immediately. Thus the timetable is not quite as rigid as Lorenz originally thought, and for this reason these windows of opportunity for learning are nowadays usually referred to as ‘sensitive periods’. They seem to be modifiable according to circumstances, rather than coming to an abrupt end when a clock in the brain says they should. Nevertheless, it is true that eventually, after a few days, imprinting cannot be reactivated. The young bird’s brain does not wait indefinitely for the mother bird to show up; that door does eventually close as its brain matures.

  Once the young bird has finished learning, persuading it to change its attachment to its mother is almost impossible. It will usually flee from other animals – a very sensible thing to do, considering that some might want to catch and eat it. The blocking that prevents the young bird from accidentally forgetting its mother and latching on to something else is called ‘competitive exclusion’: once a complete picture of the mother has built up, the imprinting process terminates automatically, preventing the bird from accidentally attaching itself to another goose if its mother is temporarily absent.

  This ‘sensitive period’ concept explains the behaviour of many young animals. For example, it can explain why hand-raised rhesus monkeys, taken from their own mothers soon after birth, prefer to be with their surrogate ‘mother’ rather than with real monkeys, even when the surrogate is only a cloth-covered, unresponsive dummy.

  Dogs do this too: they imprint on to their mothers, and vice versa, and they do this using their number one sense, olfaction. In one set of experiments, researchers collected scents from two-year-old dogs by placing cloths in their beds for three consecutive nights. The dogs had all been separated from their mothers since they were twelve weeks old or even younger. Nonetheless, when their mothers were presented with a selection of these cloths, they were much more interested in their offspring’s scent than in the scent of unrelated but otherwise similar dogs. Likewise, the young dogs’ behaviour showed that they recognized their mothers’ scents. A second experiment done at the same time showed that the two-year-old dogs were able to recognize their littermates by odour alone, but only if they were currently living with another member of the same litter.3 This suggests the existence of a ‘family odour’ that reminded each dog of the brother or sister it was currently living with, even though the odour was coming from a dog living in a completely different household. Similar tests given to four–five-week-old puppies showed that, even at that young age, they had already learned their litter odour. Unexpected abilities such as these serve to remind us that we still have a great deal to learn
about how much information dogs get from odours, especially those that are completely imperceptible to us.

  However, the ‘competitive exclusion’ principle does not seem to apply to imprinting in the domestic dog. Puppies ‘imprint’ not only on to their own mothers and littermates, but also on to people. (Strictly speaking, this phenomenon is slightly different from true imprinting in that it does not seem to be restricted to one individual person, even at the start.) In fact, puppies can also ‘imprint’ on to other animals that they have friendly encounters with during their sensitive period, such as cats. One of the beauties of the capacity for multiple socialization among dogs and cats is that they only become fearful of one another if they first meet in adulthood. I have usually kept both dogs and cats at home, and if they are introduced to one another, carefully, when they are young, they can become great friends. The accompanying picture shows one of my cats, Splodge, performing a tail-up rub, a sign of social bonding, on my Labrador retriever Bruno, while he is wagging his tail in greeting (although I suspect neither understood much of what the other was saying).

  Inter-species socialization expressed as species-typical greeting behaviour

  Some traditional uses of dogs exploit this flexibility. The sheep-guarding breeds, such as the Great Pyrenees and the Anatolian karabash, if raised with sheep, grow up to behave as if the flock is their family, although of course they behave like dogs rather than like sheep. I say ‘of course’, because the dog’s capacity to bond to two or even more species at the same time is so obvious that we take it for granted. Such a capability is, however, highly unusual in the animal kingdom as a whole. Most animals are programmed by evolution to learn about just their own species, and no other. Indeed, hand-raised animals often have great difficulty adjusting to living with their own kind, as zookeepers once found, to their dismay, when first trying to breed many endangered species of carnivores, especially some of the wild cats.

  Sheep-guarding dog with its flock

  Domestic dogs do not appear to lose their species identity, even as they form attachments to humans. Not only do they learn about how to interact with other species, but there is no evidence to suggest that this in any way disadvantages them in terms of how they interact with other dogs. The capacity to adopt multiple identities is unusual, but its origins must lie in regular biological processes. Likewise, since the capacity for interaction with humans cannot have sprung from nowhere, its antecedents must lie in the social behaviour of the wolf. While I am highly critical of the old ‘lupomorph’ model when it is applied to social structures, as a biologist my instinct is to look for something pre-existing for evolution to work on. Since dogs are neotenized wolves, it is logical to look for the answer in the behaviour of wolf cubs and juveniles, rather than in that of adult wolves.

  When wolf cubs are born, they are looked after by other wolves. They learn the characteristics of those individuals, based on the eminently reasonable assumption that they must be their parents, or, in a large pack with existing helpers, their close relatives.4 When puppies are born, they are usually looked after by both their mother and their owner. Their mother’s characteristics are slotted into the ‘parent’ category, simply because she is there and looking after them. This learning will be retained throughout life, forming the basis for one set of social preferences – namely, for members of their own species (as happens in both wolves and feral dogs). Their owners’ characteristics, since they do not match this first category, will be slotted into a second category, generated spontaneously, because they are there and are also caring for them. Apart from this parallel recognition arrangement, there is no reason why everything else about these early social preferences should not be based on the same model: that of parent and offspring. Indeed, it is hard to imagine what an alternative model might be.

  Put another way, we humans hijack dogs’ normal kin recognition mechanisms. The domestic dog puppy’s unusual capacity for multiple socialization is the mechanism whereby we can insert ourselves into their social milieu and substitute ourselves into a role that in the wild would be served by their parents. Until weaning is complete, the bond with the human owner is probably weaker than with the mother, but after that, attachment to humans is reinforced every day, when we feed our dogs, play games with them, and reward them during training. There are fewer opportunities for pet dogs to reinforce attachments to one another. Even in a multi-dog household, it is the humans that act as the parental figures, providing food, controlling where the dogs are at any given time of day, and so on. (There are of course a few intentional exceptions to this scenario, such as hunting dogs housed in packs, and sled dogs, where leadership from one dog is essential to the co-ordination of the team.)

  Is this really imprinting? Dogs certainly imprint on to their mothers but science has yet to address whether they also imprint on to their first owners. Imprinting, in the narrow sense of a bond formed to a primary carer, cannot account for the generally outgoing nature of dogs and, more specifically, how easy it is for many dogs to change their allegiance from one owner to another. During the early stages of their lives, most mammals learn both the general identity of their species, and the specific identities of the individuals around them, especially those that look after them. Normally, the latter characteristic leads to the more powerful attachment – but not so in the dog, which shows a much greater flexibility, presumably as a consequence of domestication. However, imprinting, or something very like it, does play a strong role in directing the dog’s preferences for whom to approach, and whom to avoid. These preferences are set up early on in the dog’s life – specifically, during the socialization period.

  Even though dogs can possess several ‘friendly’ categories simultaneously (a capacity unusual among mammals), each category has its boundaries. This is well exemplified by some dogs’ distrust of children. How do dogs know that children are little humans and not another species entirely? The answer appears to be that they do not. Children are distinctly different from adult humans in a number of respects, in the way they move, the sounds they make, and – probably of particular significance to dogs – in the way they smell. Dogs that were never exposed to children during puppyhood can be very wary of them when they first meet them as adults, although being dogs they can easily be trained to overcome this initial reluctance. On the other hand, if their first encounter with a child involves the pulling of their tail and ears, such dogs can easily become irritable and snappy with other children. The dog generalizes between children, treating them as a category rather than as individuals. In the same way, during socialization puppies must generalize between one adult human and another. Although puppies undoubtedly come to recognize some people as individuals, unfamiliar people are presumably categorized as ‘friendly’, based on their similarity to the first handful of people that the puppy has met.

  This is why it is so important to (gently) introduce puppies to as wide a selection of people as possible; men as well as women, people wearing different kinds of clothes, men with beards as well as clean shaven, and so on.5 This process serves to expand the boundaries of what the dog categorizes as ‘adult human’. If the template is left too narrow, perhaps because the puppy meets only one or two kennel-maids during the whole of its sensitive period, then later on the appearance of men, in that instance, will trigger fear and anxiety. This is one of the several reasons why owners have difficulties with dogs from puppy farms and pet shops; such puppies’ concept of what the human race looks like is often too narrow, and they default to fearful avoidance of every other two-legged animal they meet.

  As we have seen, the organization of the dog’s social brain is different from that of most other mammals. It can form multiple, on-demand, representational spaces for each of the species that the puppy encounters during the socialization period.6 The change in the dog’s social brain may have been a product of domestication; alternatively, it may even have arisen as a pre-adaptation to domestication, in certain wolves that no longer exist in t
he wild today. Regardless, any dog that only meets other dogs until it is fourteen weeks old develops only one such space, defined initially by its littermates and mother, because they are all that is available, but potentially extendable to all types of dog a little later on in its life. The evidence suggests that a dog born in a human household develops two such spaces, one for dogs and one for humans (again, each with the capacity to accommodate other types of dog, and other types of human, that do not respectively look/sound/smell like the owner’s family). A dog born into a human household also containing a dog-friendly cat may develop three such spaces. And it is possible that dogs born into households with small children develop yet another space to cover the children – or perhaps they learn to generalize between adult and infant humans, and thus essentially conceive of them as part of the same continuum of two-legged animals.

  In short, dogs therefore have very unusual brains, which allow them to construct several social milieux simultaneously. It is this capacity that enables them to be so useful to us; to cite just two examples, hunting dogs can run in a pack, and sled-dogs can run in a race, while remaining under the control of their human handlers. Dogs are born with the potential to develop multiple identities, but all the detail and context has to be provided by experience. One might argue that this is the only way that domestication could have been made to work. Evolution does not possess foresight, so it could not have provided built-in knowledge of what humans are and how they function in advance; the best that natural selection is likely to be able to provide is the machinery to acquire this knowledge. Socialization to other dogs, on the other hand, is most likely based upon mechanisms set up in the early evolution of the carnivores, millions of years before domestication, and may therefore involve some additional processes.