Dogs are pretty smart. They can have huge vocabularies, they can infer meaning in the growls of other dogs, and they can effortlessly figure out if other dogs want to play or fight with them. But their intelligence might be limited to the social domain; indeed, while they outperform chimpanzees in social tasks, chimpanzees outperform them in many other tasks. And they might have developed their impressive social skills as merely an accident of natural and artificial selection.

Previous research has shown that dogs can use lots of different forms of human communicative signals to find food, and they can also inform humans of the location of hidden food, by looking back and forth between that human and a second location. But what is it about dogs that allows them to comprehend and invoke human social communication?

ResearchBlogging.orgHungarian canine cognition researcher Adam Miklosi has written that "the genetic divergence of the dog from its ancestor was [presumably] accompanied by important behavioral changes that could have a genetic basis because of a selection pressure for dogs that were able to adapt better to the human social setting." If that was the case (and it probably is), then it follows that while some natural variation in performance on a given task requiring the use of human social communicative cues may be seen in wolves, it should be a lot stronger in dogs. Comparing wolves and dogs is the obvious way to address this question, but wolves and dogs differ both in genetics as well as in environment. While most dogs are raised in houses as pets, most wolves live in the wild, or in zoos, or occasionally in conservation parks. Adam Miklosi and colleagues, of Eotvos University in Budapest, took advantage of a very unique situation. Thirteen wolves were hand-raised and socialized in human homes, just as dogs would be. By raising both dogs and wolves similar contexts, the effects of rearing environment can be minimized, allowing the researchers to infer that differences in behavior are more likely due to genetic differences.

In the first experiment, four of the socialized wolves were tested in the standard two-way hidden food task. The experimenters would hide food in one of two containers, and then using one of three gestures, they would indicate to the wolf the location of the food: distal pointing (the human's finger is about 50cm from the container), proximal pointing (the human's finger is 5-10cm from the container), and making physical contact with the container with the index finger.


Figure 1: Mean performance (with standard error) of the four wolves on the first experiment. The "distal" cue is separated into first and last twenty trials, indicating a slight increase in performance over time.

In addition to the overall analysis, which indicates that the wolves performed significantly above chance one condition, they looked at the performance of each wolf individually. For each wolf, performance on the distal pointing condition was at chance, but one wolf increased his performance such that by the end of the experiment, he chose correctly on 80% of trials. All individuals performed significantly above chance on the touching condition, as is evident in the aggregate data above. In the proximal pointing condition, two of the four wolves consistently performed above chance. Taken together, it appears that given dog-like rearing, wolves can learn something about human social communication. Despite this, compared to similar studies with dogs, the performance of the wolves was worse and more variable.

To succeed in the two conditions which resulted in the highest number of successful trials, the "touching" and "proximal pointing" conditions, the wolves only needed to attend to the immediate space around the container. This may explain the higher success in those conditions. In the distal pointing condition, they would need to attend to two locations in space: the container as well as the human experimenter. If wolves do not pay attention to the human, they would be unable to determine the direction that the hand is pointing, and the task could become, in a way, unsolvable.

In order to investigate this a little further, the experimenters designed a second study, involving two behavioral tests: bin-opening and rope-pulling. Upon successful completion of either task, the individual received a piece of meat as a reward. Both pet dogs as well as the socialized wolves were allowed to learn how to solve either problem during a training phase. Both groups of animals were equally able to solve the task, and did so following an equivalent number of trials, suggesting similar motivation. Once the individuals had learned the task, they were presented with what appeared to be the same task, but was unsolvable. The key variable was where, how quickly, and how long the individual would look after attempting and failing to complete the task.


Figure 2: How quickly did the animals look at the human after finding the task unsolvable? Line represent the median, boxes show the middle 50%, and bars indicate the total variance.

In both tasks, dogs looked back at humans earlier than the wolves did, and for greater duration. In the bin-opening task, specifically, dogs spend more time overall gazing towards the human experimenter, and did so significantly earlier than the wolves did. In fact, only two of the seven wolves tested looked towards the human at all during the insolvable trial, while five of seven dogs did so. On average, the dogs began to look towards the human experimenter after one minute of attempting to solve the task, while the wolves all but ignored the presence of the experimenter. This is not to say that wolves are not intelligent. In fact, wolves can be quite intelligent, they just don't care too much about humans, and therefore perform poorly in tasks that require them to engage socially with us.

Experiment one demonstrated that under dog-like rearing conditions, wolves could understand some human social communicative gestures. However, experiment two suggested that only dogs regularly attempt to initiate communication with humans, by attempting to make face or eye contact with them. Based on the result of the second experiment, the researchers inferred that the relative failure of the socialized wolves to succeed in the first experiment was due to their unwillingness, or put more neutrally, their disinterest, in looking at the humans. They also inferred that, since the wolves and dogs had similar upbringing, the dogs' preference for looking at humans was due to a genetic predisposition. Further, they speculated that, of the first steps in the domestication of the dog was the selection for "human-like" communicative behaviors. As we found some behavioral variability in our wolves, this species might have been predisposed for successful selection to take place. Since in humans taking up eye/face contact is understood as initialization and maintenance of a communicative interaction, we suppose that the corresponding behavior in dogs provides the foundation on which developmentally canalized complex communicative interactions can emerge between man and dog.

What does this mean exactly? Selection (whether natural or artificial) occurs because there is natural variation in a given trait in a population. Certain environmental constraints make it such that certain variations of a given trait are more adaptive than others. In the wolves, for example, at least some portion of the population must have displayed an ability to understand and initiate communication with humans. These individuals would be most adapted to life with humans and would therefore be more likely to breed. What would eventually emerge was the domesticated dog.

Dogs' social skills could therefore be encoded in their genes. Canalization is a process wherein genetics limits the variation in possible developmental outcomes, regardless of environmental specifics. In the passage quoted above, Miklosi and colleagues suggest that the near universal ability of dogs to engage socially with humans is the result of such a strong genetic predisposition that even differential rearing environments would not significantly alter the outcome. In a sense, while wolves may have limited abilities to socially engage with humans, domesticated dogs are specialized for the task. And this would suggest that dogs are a uniquely suited species to help us understand our own human social cognition, whether we are interested in investigating attachment between individuals, cooperation, social learning, or even pedagogy.