Dogs Take Our Visual Perspective Into Account When Obeying Our Commands

Dogs have been shown in numerous experiments to be very good at following a human’s pointing gestures to find hidden food. They have even been shown to follow the movements of our eyes. This is generally thought to prove that in the process of domestication they have learned to interpret certain signals given by humans. Most other species, including apes and wolves, do not follow our gestures as well, though there are some studies of wolves that indicate they may be able to understand pointing gestures just as well, or even better in certain settings, than dogs. A recent study by scientists at Cambridge University and the Max Planck Institute in Leipzig asked a different question. Do dogs consider a human’s visual perspective in attempting to carry out that human’s command? Put another way, do dogs interpret the command by taking into account what they know the human can see and what they know he cannot see?

The experiment was designed as follows. A human stood at one end of a room and a dog at the other end. There were two barriers closer to the human. On the side of the barriers where the dog stood were two toys. One of the barriers was opaque and prevented the human from seeing the toy on the dog’s side of the barrier. The other barrier was transparent and though it was between the human and the toy, the human could see through it. The human, without looking at either toy and without signaling in any way, commands the dog to FETCH. The dog will be rewarded if he brings either toy to the human. That is the first situation depicted in the figure (click on the figure to see an enlargement).

In the second situation, the human stands behind the dog and both of them are able to see both of the toys. Each toy is beside a barrier as before, and one of the barriers is opaque and one transparent, but neither block the human’s view of the toy since the human has the same perspective as the dog. Here again, the command is to FETCH.

In the third situation, the human is back on the side of the barrier where he could only see the toy beside the transparent barrier, but his back is turned so that he sees neither toy, nor the barriers, nor the dog. Again he tells the dog to FETCH.

In each situation the dog could bring either toy to the human, yet the choices the dogs made in the first situation were not random. In 70% or more of the cases the dogs brought the toy that the human could see to him on the command to FETCH. In the third situation, where the human was looking away from everything in the room, the dogs brought the toy beside the transparent barrier about 60% of the time, as if expecting that the human’s command would apply to the toy he could see if he did turn around. Only in the second situation, where the dog was aware that he and the human could see both toys did the dogs choose almost randomly, with slightly more than half of the dogs in the experiment choosing the toy beside the transparent barrier. Different dogs were used for each of the three situations.

The researchers concluded that dogs take into account the perspectives of humans in deciding how to interpret their commands. They do not just choose from what they themselves can see. They take into account what we can see. When I began taking lessons with Chloe, Rick Manley of the Phoenix Field and Obedience Club told me not to let my frustration get to me. "She does want you to be happy. You have to learn to understand each other." Perhaps this research provides a scientific basis for his statement. Juliane Kaminski, Juliane Brauer, Josep Call, and Michael Tomasello, “Domestic Dogs Are Sensitive to a Human’s Perspective,” 146 Behaviour 978-998 (2009)

Empathy Responses of Dogs to Other Dogs?

Wolves and feral dogs live and hunt or scavenge in groups. Conflicts can arise between members of a group over food, efforts to mate, and for other reasons. Resolving these conflicts becomes important, since continuing hostility can reduce the effectiveness of the group to act as a unit. Curiously, conflict resolution has not often been studied in dogs, so recently published research by three Belgian scientists at the University of Antwerp is particularly interesting. The researchers used dogs that were housed at a pet food production company, but the dogs were not used to test food during the period of the research. The dogs were housed in pens that held two or three dogs and all were between three and ten years old. During the day the dogs in the group that was being studied were released for many hours to a meadow surrounded by a fence where their interactions were recorded by observers. The first and second groups studied consisted of seven dogs, four males and three females. The third group was formed from two Labrador Retrievers from the first group and four from the second, four of them male and two female. The groups are depicted in the figure.

When a group was released to the meadow, observers recorded conflict interactions of the dogs, dividing their behavior into light aggression (growling, directed barking, bending over, inhibited biting, head-prodding), grave aggression (biting, fighting, knocking down), submission (active submission, passive submission, averting eyes), affiliation (greeting, sitting/lying down together, anogenital sniffing, playing, licking), and neutral behaviors (sitting/lying down alone, drinking, throwing ground, following). If grave aggression did not end within several seconds, an observer intervened to prevent serious injury. Most conflicts were brief, lasting only a few seconds, in which one of the opponents often made a submissive gesture. The most common form of physical conflict arose when a male paid more attention to a female than she cared for.

The observers recorded 1711 conflicts, with former opponents making up by some affiliation behavior in 606 cases (a little over a third), and third parties providing some affiliation in 621 cases (also over a third). The remaining cases ended without reported affiliation behavior. In these cases, the opponents usually avoided each other for several minutes or more. Where third parties—dogs not involved in the conflict—engaged in affiliation behavior, loser directed affiliation was the most common behavior, and in these situations the intervention was more often initiated by the third party rather than by the loser. This meant that true consolation seemed to be involved, as opposed to solicited consolation by the loser whimpering or seeking attention. Affiliative behavior was generally observed more between dogs that shared pens, supporting the “valuable relationship hypothesis,” which predicts that reconciliation will be more frequent between opponents whose relationship is of high biological value. This arguably is present between dogs that share the same pen and must learn to get along.

The authors admit that their study does not prove cognitive empathy in dogs, but they believe that “dogs do possess significant social cognition.” Given that wolf packs are often larger than packs of feral dogs, it would be important (though very difficult) to determine the extent of conflict resolution in the group behavior of wolves. Annemieke K.A. Cools, Alain J.-M. Van Hout, and Mark H.J. Nelissen, “Canine Reconciliation and Third-Party Initiated Postconflict Affiliation: Do Peacemaking Social Mechanisms in Dogs Rival Those of Higher Primates?” 114 Ethology 53-63 (2008).

Additional Note. Ronald Keats, a Shiba Inu owner, has been following the plight of dogs in Japan following the 2011 earthquake and sent me a YouTube link, showing a dog that would not leave his friend's side. Hopefully some of the people filming this took some action.

Why Do Dogs Respond to Our Social Cues?

A focus of research on dog behavior has concerned the ability of dogs to respond to our social cues. In object choice tasks, a dog will have to decide which of two or more containers contains food. With certain controls, the dog unaided by a human will only perform at chance levels. If, however, a human points to the correct container, the dogs will much more likely go to it than the others. Even a human moving his eyes from the dog to the correct container and back and forth in this manner will help the dogs make the right choice, though not as often as pointing. Dogs will also choose the container a human is pointing at even if this is the wrong container, they’ve had a chance to smell the container with the food, and their noses should tell them to ignore the human’s pointing gesture. They trust us even when they shouldn’t. These findings, long known among canine behaviorists, lead to another question. Why do they follow our gestures when other animals, such as apes, don’t? Do they know we are trying to help them? Have they learned to trust us in the process of domestication? Pamela J. Reid of the ASPCA’s Animal Behavior Center in Urbana, Illinois, recently reviewed the literature on this topic, looking at the various hypotheses that have been propounded to explain why dogs read us so well. She distilled the conclusions of prior researchers down to four basic theories:

1. Dogs learn to respond to human social cues through basic conditioning processes.
2. Dogs reduced their fear of humans in the process of domestication and began to apply all-purpose problem-solving skills to their interactions with people.
3. Dogs’ co-evolution with humans equipped them with the cognitive machinery to respond to human social cues and understand our mental states (sometimes called a “theory of mind”).
4. Dogs are adaptively predisposed to learn about human communication gestures.

Reid herself favors the fourth explanation.

As to the first of the four theories, Reid notes that although dogs respond better to social cues after training, they perform significantly above chance even without prior experience with a pointing cue. Wolves raised the same way as dogs will not succeed as well as dogs in responding to pointing gestures.

Some researchers have argued that selection pressures placed on dogs for tameness and other desirable traits may have also been a stimulus for dogs to develop a specialized set of social skills. Dogs have learned to eat in the presence of humans and to accept restraint and in the process of domestication became expert readers of human social cues. Most (but not all) studies have found that dogs respond to social cues from humans far better than do wolves, their undomesticated ancestors. (Contrary research has recently been published by Monique Udell of the University of Florida and her colleagues.) Reid notes that Belyaev’s foxes, domesticated artificially, can follow pointing gestures as well as dog puppies of the same age, and are better at it than wild foxes. New Guinea Singing Dogs, similar to Australian dingoes, also perform of above chance, indicating that the ability to respond to human cues probably began in the earliest phases of domestication, since these unique dogs had only a rudimentary level of domestication before losing human contact. Perhaps dogs have learned to apply general rules of thumb such as “always approach the closest extension of the person,” or “always approach the person’s movement.” Such a mechanism would generally lead to the correct container in object choice tasks where a piece of food is placed in one of two or three containers and a human points to the correct container.

Reid attributes the third theory on the list above to the work of Adam Miklosi and Jozsef Topal and their colleagues, who argue that dogs and humans have evolved together to such an extent that human-like social skills have materialized in the dog in a process sometimes termed convergent evolution. This comes close to arguing that a dog understands that the human is trying to convey the location of food with a gesture. Miklosi has shown that dogs faced with an insoluble problem will look at a nearby human as though soliciting assistance. Reid argues that those scientists that take this “theory of mind” approach have found more sophisticated cognition in dogs than is justified by their research. It is, of course, likely that most pet owners would prefer to believe that there is a high level of human understanding in their pets.

Reid thinks the answer lies in the biological status of the dog as a scavenger, a niche that requires that an animal be acutely aware of other individuals in the social group that are also looking for opportunities to scrounge. They respond to our gestures towards food sources just as they respond to the other members of a pack that may be on the trail of a food source. This skill, combined with a tendency to learn, provided dogs with the skill to respond appropriately to our gestures, she argues. Reid admits that further research will be needed to narrow down the possible explanations as to why dogs are able to respond to our social cues. Pamela J. Reid, “Adapting to the Human World: Dogs’ Responsiveness to Our Social Cues,” 80 Behavioural Processes 325-333 (2009).

Dogs Identify Murder Suspect from Manure on Shoe

Something of a precursor to the modern scent lineup is described by Adee Schoon and Ruud Haak in their book, K9 Suspect Discrimination. In 1918, a farmer was murdered near the town of Breezand in Holland. Fingerprints taken from the crime scene were found to belong to one suspect, who confessed and implicated two other men, one of whom was also apprehended. This second suspect had slipped when climbing through a barn window and his right foot went into a manure gutter. When asked about the manure still on his shoes, the suspect said that he had recently gone to a cattle market in the town of Purmerend, where he had stepped in manure. The investigator in the case, Van Ledden Hulsebosch, wondered if a dog could distinguish between manure obtained from different locations. He asked the gendarme from Breezand to bring him cow manure from 12 different locations, including the place where the murder had occurred. The manure from the different locations was put into 12 clean jam jars, which were numbered and labeled indicating the cowshed from which the manure was obtained. Hulsebosch took 72 pieces of paper and divided them into six groups of 12, labeling each of the 12 sheets according to the numbers on the jam jars holding the manure. Manure was then spread on the papers, which were then laid out in the courtyard of the police headquarters. The inspector let dogs sniff the manure from one barn, then released them in the courtyard to find the sheet with manure from the same source. The dogs alerted to the correct piece of paper. The experiment was repeated a number of times, with the correct paper identified every time. The next phase involved the shoes of the suspect. The suspect’s left shoe had no manure on it and the dogs did not react to it. When they were presented with the right shoe, the one with manure on it, they began searching until they found the piece of paper in the courtyard with the same scent on it. That piece of paper held manure taken from the barn where the farmer had been killed. The dogs had identified a second participant in the crime. Adee Schoon and Ruud Haak, K9 Suspect Discrimination, 27-28 (Detselig Enterprizes 2002).

The Cancer Sniffers

Training dogs to detect lung and breast cancers requires a rigorous training approach. A team that used 3 Labrador Retrievers and 2 Portuguese Water Dogs taught the dogs to distinguish a tube with a cancer breath sample and food from four blank tubes. In this, the first stage, the dogs were taught to SIT before the correct sample, then received food and praise. In the second stage, the dogs still distinguished a tube with a cancer breath sample and food from blank tubes, but the handler did not know which tube was correct. The dogs got food and praise on sitting before the correct tube. In the third stage they received the same rewards, but there was no food in the tube with the cancer breath sample. They were also reprimanded with a No! if they chose the wrong sample in this stage. In the fourth phase, each tube had a breath sample, but only one had a cancer breath sample. In the final phase, it was possible that all five tubes had cancer-free samples, so that alerting to any of them would be a fail. One tube might have a cancer breath sample, requiring an alert. Neither the handler or the experimenter knew if there was a cancer sample in this final stage, and dogs received no reward at this stage for a correct alert until they left the experimental room. (Click on the table to the right, showing the stages of training, for a larger image.) A dog was not fully trained until it made no mistake for 30 consecutive trials. The dogs could not detect patients in remission, but for one individual thought to be in remission, the dogs sat before her breath and the cancer was later found to be active again. Dogs detected lung cancer with 99% consistency with biopsy-confirmed diagnoses, and breast cancer with 88% consistency. Michael McCulloch, Tadeusz Jezierski, Michael Broffman, Alan Hubbard, Kirk Turner, and Teresa Janecki, “Diagnostic Accuracy of Canine Scent Detection in Early- and Late-Stage Lung and Breast Cancers,” 5(1) Integrative Cancer Therapies 30-39 (March 2006). There are a number of legal or at least ethical issues that will arise if this dog function becomes more prevalent. As Mary Elizabeth Thurston noted in her wonderful book, Lost History of the Canine Race (p. 58 in the Avon edition), sniffer dogs might be useful in poor countries. I have previously suggested that they might be taken to remote locations in parts of the U.S. Will this mean they are owned by facilities and are essentially laboratory dogs? What protections can they be given? Will facilities that use them have to cover them with malpractice insurance?

Service Dogs for Autistic Children Also Help Their Families

Service dogs for autistic children are proving beneficial for both the children and their families. National Service Dogs of Cambridge, Ontario, uses a double-leash system where the autistic child is attached to the dog by a leash that is attached to the child’s belt. A parent or other person holds a separate leash. The dog obeys the parent’s commands, but is also trained to prevent the child from bolting away from the parent, or into traffic. Golden and Labrador Retrievers are preferred by National Service Dogs for this work. Three researchers studying ten families where such dogs had been placed found that in addition to keeping the child safe, autistic children began to walk at the dog’s speed, making going places more pleasant for the whole family. Families reported that they generally let the dog sleep with the child, and both children and families often slept better. The autistic children were reported to learn new skills with the dog. “Parents reported a wide variety of physical tasks that were facilitated by the dog’s participation. One child learned how to take the lid off a dog food container, pour the food into the dog’s bowl, place the bowl on the floor, and look at the parent to give the dog a command to eat the food. A number of children learned how to pick up and throw a ball for their dog. Motor function improvements were gained by the parents sitting with the child and helping him/her learn how to pet the dog. Many of the children struggled to control their movements to pet the dog, and by using the dog as a therapeutic tool the parents were able to teach the children to pet or caress the dogs gently. Children learned to throw a ball and manipulated grooming tools.” The Canadian researchers reported that families began to consider taking vacations after getting a service dog because traveling with the autistic child became easier. People encountering families with service dogs were also friendlier, and the families were less inclined to avoid interactions for fear that the autistic child would have a tantrum. As with many service dogs, there is more demand than there is supply, and it is to be hoped that more service dogs can be trained to live with families with autistic children. Kristen E. Burrows, Cindy L. Adams, and Jude Spiers, “Sentinels of Safety: Service Dogs Ensure Safety and Enhance Freedom and Well-Being for Families with Autistic Children,” 18(12) Qualitative Health Research 1642-1649 (2008).