Dog Behavior Blog

A Rudimentary Theory Of Mind In Dogs And Other Animals

Note from Con Slobodchikoff: This is a guest post by Dr. Alexandra Horowitz, whose work on dog attention and theory of mind was featured in a post on this blog on October 8, 2008. Dr. Horowitz has written a book, "Inside Of A Dog," that will be published in September 2009 by Scribner. It is about what a dog's point of view might be like, using scientific results to help draw the picture.

It is notoriously difficult to test for theory of mind in animals. "Theory of mind" (ToM) refers to the attribution of mental states--particularly belief, desire, and knowledge--to others. Research shows that human children are not born with this understanding, but they develop it by age four. Many claim that ToM is wrapped up in what enables us humans to be (at times) so cooperative, so empathetic--in a word, so human. As a result, there is great interest in determining whether any non-human animal has a ToM.

The question of non-human ToM is vexed in a couple of ways. For one thing, tests to determine when children develop a ToM--for instance, the false-belief test, which asks if a child realizes that someone who's been out of the room doesn't know about any machinations that happened inside the room--require that the children use language to tell researchers what they know. So much for testing animals. For another, with the non-verbal tests that have been developed, the results with animals (mostly chimpanzees) have been mixed. Sometimes chimps act with seeming understanding of what others know; sometimes they act as though utterly oblivious.

It occurred to me that there is a problem with our concept of ToM. The implicit premise is that ToM is an "all or nothing", on-or-off ability, which pops "on" early in childhood, rather than developing in stages. This kind of conceptual definition reminds me of the Cartesian notion that an animate creature is either self-aware and conscious, or is merely an unthinking material form. Descartes led us to dualism, which we spent hundreds of years struggling out from under; but we retain his binary template of an animal being either insightful or mindless.

I propose, by contrast, that ToM is not best considered a unitary ability that one either has in spades, or one does not have at all. Instead, an individual may be able to realize that something connects all the behavior you see someone else doing--without specifically postulating a "mind" behind it. This step could be described as having a "rudimentary" theory of mind: more than acting instinctively, or as a behaviorist, but less than acting with the theoretical understanding characteristic of humans.

Formulated in terms of what we need to attribute to animals in order to explain their social behavior, we get the following hierarchy:

Theory of behavior: the understanding that certain behaviors lead to certain outcomes
Rudimentary theory of mind: the appreciation that there some overarching element which leads to seen behaviors
Theory of mind: extrapolating that what others' behaviors have in common is their root in a mind

Much of animal behavior requires only a "theory of behavior" to explain: an animal's action is based on simple associative conditioning, having learned that one event follows another. For instance, a dog who learns to sit at attention when a familiar person approaches him with a biscuit need not have posited anything about the mind of the biscuit-carrier; he need only have associated the fact of the biscuit being carried with his ultimate reward (biscuit-eating). The five-year-old child sitting waiting for some similarly valued reward from a parent has drawn the same association, surely--but also has a head full of conjectures about what the parent intends, thinks--in other words, what is on her mind.

The intermediate level allows for a middle ground between these poles. The dog who has learned not just to sit before the biscuit-carrier, but also to continue to adjust his sitting to always see her face; the dog who has generalized from the one biscuit-handling person to investigate whether other humans carry biscuits, but learns that not everyone can have biscuits extracted from them by sitting dutifully--these dogs have a more advanced theory of behavior. They are sensitive to salient elements of the behavior--the attention of the human, for instance--and use this knowledge to generalize to future contexts.

As Con has described in an earlier post, when I studied dogs' behaviors in dyadic rough-and-tumble play, I saw dogs behave in a way that I think is best explained by imputing to them a rudimentary theory of mind. They play-signaled--a request to play--only to dogs facing them, thus able to receive that request. When a dog paused in play, his play-partner used attention-getters to redirect him before signaling again. And, even more impressive, dogs used attention-getters matched to their partner's state of inattention: if a dog is looking at you, you needn't bite him on the rump to get his attention; a minor leap up will do. But if he is playing with another dog, the leap-up is ineffective; something on the order of a rump-bite is required.

What this demonstrates, I have suggested, is that the dogs were acting as though they understood that not only must you react to the presence of other dogs, not only must you communicate to the faces of other dogs--but you must also attend to an invisible feature of them--a feature that between humans we call "attention"--in order to communicate, and thus play together. Acting with sensitivity to attention is much more sophisticated than simply acting in response to seeing a stimulus: play-signaling whenever you see another dog, or using any attention-getter if you can't see a face. We simply cannot tell from the data whether the dogs actually have a theory of mind--a real understanding of the minds of others--but their behavior plainly shows us that they have more than just a theory of behavior.

I realize that many dog owners have no reluctance in asserting that their dogs have a theory of mind--and much more sophisticated cognition besides. "Dogs have the intelligence of three-year-olds", it's claimed. As discerning and sensitive companions as dogs are, I haven't seen the evidence yet. And I'd suggest that we owners, in turn, don't truly treat dogs as theorists of mind: we wouldn't dream of leaving a three-year-old alone with little to do for hours on end, the way we do our dogs; we ignore them when we're busy with other people or with work; we do nearly nothing to advance their intellect. Dogs are best served not by over-attributing to them but by acknowledging what they do know, and acting accordingly.

On dogs and wolves: The origin of the differences

Note from Con Slobodchikoff: This is a guest post from Dr. Adam Miklosi, who has been researching the comparative behavior of dogs and wolves. Dr. Miklosi's recent extensive and informative book, "Dog Behaviour, Evolution, and Cognition (Oxford University Press), is available through Amazon.

Researchers know two sorts of facts. First, there was a time when dogs did not exist, but only wolf-like creatures lived across the Northern Hemisphere. Second, relatively suddenly (in terms of an evolutionary time scale) a very similar type of animal emerged that was usually found in association with humans: the dog.

In recent times geneticists have established that the dog’s closest living relative is the wolf, and zoo-archaeologists date the dog-like animals from about 17,000 BP. All the rest is still a mater of debate, including the exact date and the location of domestication, and whether dogs domesticated themselves by joining human groups, and to what extent humans played an active role in “selecting” for certain types with regard to form and behavior.

In recent years there has been a huge interest in the behavior of dogs, and that has included comparative studies that were aimed at looking at dog-wolf differences. Frank (1980) and his colleagues were the first who suggested that dogs and wolves differ in the way that  they process environmental information. According to them, wolves are more bound to certain type of stimuli for any behavioral response, but at the same time they are able to show flexible problem solving behavior. In contrast, dogs lack complex problem solving skills but they can learn in very flexible ways about the connection between arbitrary stimuli (e.g. Sit!) and some action (Bottom touches ground). Although Frank’s ideas were very interesting, he worked with very few animals, did not account for differences in socialization, and utilized a very restricted experimental methodology.

The main problem is that it is easy to find behavioral difference between dogs and wolves, but an ethologist looks for the origin of these differences. Living in a different physical and social environment makes it more than likely that the two species will display divergent skills. Since dogs look very different from wolves, most researchers agree that domestication of dogs was accompanied by some genetic changes, but what is so obvious in morphology is much more difficult to be revealed in the case of behavior.

Most comparative behavioral research assumes that dogs have been selected to live in close association with humans. In other words, dogs had a greater chance to survive in the human (anthropogenic) environment if they acquired human-like social skills. If one wants to understand more about the genetic underpinning of these skills, one way to reveal them is to perform experiments on dogs and wolves that are raised in similar social environments. Our research group at the Eötvös Loránd University was the first one that raised many dog and wolf pups individually in the same way by a human caretaker (one caretaker lived with only one pup at the same time) with the intent to start a comparative research program.

At present there are two main theories that try to account for behavioral differences between dogs and wolves: selection for cognitive skills, and changes in behavior. If dogs lived for long enough in human society, then selection could have changed their cognitive skills markedly. For example, researchers found that dogs showed high level performance in communicative tasks when they had to find a piece of hidden food on the basis of human gestures. Since wolves were not very successful at this task, in contrast to dog puppies without much human experience, researchers assumed that dogs might have gained some genetic advantage during domestication that affected directly their social cognitive skills.

However, there is also an alternative way to explain this. In order to be successful in a communicative task, not only cognitive skills are important, but also changes in other behaviors. To be a successful companion to a human, a dog has to stand or sit quietly for a few seconds, it should look at the person in order to see the communicative action (gesture), it should allow itself to be controlled by a signal from the other instead of making its own decision, etc. This means that changes in other aspects of behavior like attention span, interest in humans, or temperament could also improve performance, and one does not need to assume that the core-system of social cognition is affected. Thus dogs might be very similar to wolves with regard to cognitive skills in general but in certain tasks they have a better chance to reveal those cognitive skills because of changes in other behavior systems.

In our research program we emphasize dog-human parallels in behavior, assuming not only that the human environment selected for some particular social skills in dogs, but actually, that in functional terms (what is the outcome of the behavior) dog and human behavior shares similar features. Over the years we have collected some evidence for this view by studying various aspects of social behavior in dogs and humans in parallel. In comparative experiments we have found that 4-month-old dogs show attachment behavior toward their human caretaker, which is not the case in socialized wolves. Similar direction of difference was revealed in the case of communication abilities. Dogs proved to be more skilled at an earlier age than wolves in comprehending human gestures, and they also seem to be relatively flexible in utilizing unfamiliar gestural signals. In both cases dogs share many features of their social behavior with human infants. This means that from a functional perspective dogs can be compared to 1.5-2-year-old infants, depending on the task. However, we do not claim that necessarily similar mental mechanisms are at work in these two species. Some of the mechanisms might overlap since both species (dog and human) share a common mammalian heritage and the ability to live in social groups but it is also very likely that human evolution endowed us with some special skills which are not present in dogs.

The increased flexibility in emotional behavior in dogs might also have contributed the success of dogs in the human community. Although it has been paid little attention, dogs and wolves have similarly complex emotional behavior; however, in the case of dogs the possible behavioral combinations are larger. Barking is an obvious example for this. Both wolves and dogs bark, but in wolves barking is used mostly as a “warning” signal. In contrast, dogs utilize barking in a wide range of contexts, including fear, aggression and play. Moreover, humans are quite skilled in discriminating these barking sounds. Considering that we can also use our vocal behavior to modify the behavior of the dogs, it seems that dogs and humans have evolved a novel communicative “channel” for interaction.
    
Personality is another interesting aspect of dog behavior in which human parallels might be possible. Interacting individuals living in the same social group may be at an advantage if the range of their personalities overlaps. Dogs and humans might share some personality traits which facilitate their interaction within a group. Interaction with wolves depends in most cases on the skill of the human to adapt to the “needs” of the individual animal, while with dogs, relationships can be based on mutual acclimatization. At the moment we do not know how dogs and humans affect each other’s personality, but the possibility of such an influence is much greater than in the case of wolves.

Dogs Pay Attention To Where Other Dogs Are Looking During Play

Recent research has shown that dogs can understand where people are looking, but can dogs understand where other dogs are gazing? A recent article in the journal Animal Cognition suggests that dogs can do just that. The article is by Alexandra Horowitz (Horowitz, A. 2008. Attention to attention in domestic dog (Canis familiaris) dyadic play. Animal Cogntion, in press, DOI 10.1007/s10071-008-0175-y).  Horowitz videotaped the interactions between 39 pairs of dogs engaged in play behavior in a park in southern California where dogs were allowed to run off leash. She then analyzed her videotapes to find out what behaviors the dogs were using during play.

Playing dogs used several different signals to indicate to each other that they were playing. These included a play-bow, an open-mouth play face, a head bow, a play slap with the forelegs slapping the ground, and a leap-on or jumping up and putting the front paws around the other dog’s head.

But sometimes one of the dogs looked away and seemingly lost interest. Then the other dog used attention-getting signals that most of the time caused the other dog to restart the play behavior. These signals included placing the body near the other dog’s face, making a backwards leap, turning around so that the dog’s rear end was next to the other dog’s face, bumping into the dog, touching the other dog with nose and closed mouth, biting the other dog with a soft bite, making biting movements in the air, pawing at the other dog, and barking.

When these attention-getting signals were used, play restarted 88 percent of the time. When the other dog who wanted to continue to play used a play signal or some behavior other than an attention-getting one, play restarted only 20 percent of the time.

Sometimes the dog who stopped playing was not looking at the dog who wanted to continue to play. When that happened, the dog who wanted the play to continue did not use visual attention-getting signals (e.g., placing the body near the dog’s face or making a backwards leap) but instead used tactile signals (e.g., bumping, touching the nose to the other dog’s body, biting with a soft bite).

The results suggest that dogs can understand that when another dog is not looking at them, it is better to use touch as a communication tool rather than using visual signals. This relates to what philosophers and psychologists call Theory of Mind. The question is, does the dog understand what is going on in the mind of the other dog?  Horowitz takes a cautious approach and says that more research is needed in this area. Those of us who have been around dogs a lot are probably willing to be less cautious about that determination.

--Con Slobodchikoff