ResearchBlogging.orgDogs are particularly good at tasks that involve communicating or cooperating with humans, which has led some researchers to speculate that they are really good at solving social tasks, more generally. For example, dogs can figure out where a human's attention is, are really good at picking up on eye-gaze and finger pointing cues, distinguish among different individual humans (by contrast, humans are really bad at distinguishing among different individual monkeys, for example), and at least in one outstanding case, are capable of "fast mapping."

Relative to non-human primates, domestic dogs indeed seem to have exceptional social skills. For example, previous research has demonstrated that dogs are able to use human social cues to find hidden food while non-human primates do not. Furthermore, cross-sectional studies of dogs and puppies of different ages, as well as longitudinal studies which track the development of individual puppies, have indicated that dogs do not require extensive exposure to humans to skillfully use those cues (though training enhances their skill). By contrast, wolf pups do require extensive exposure to humans to be able to extract meaning from human social cues such as eye-gaze and finger-pointing. So, while both dogs and wolves are able to understand human social cues, the domestication of dogs seems to have selected for this trait and allowed it to emerge early in development without much experience.

But these studies of social cognition in dogs have had one common theme, which is that they all tested social cognition in the context of a communicative-cooperative task. But do dogs' social skills extend beyond this narrow context? In non-communicative or non-cooperative social tasks, are dogs' social skills otherwise unremarkable? The distinction is not trivial; social information comes in various forms beyond explicit communication. For example, various non-human primate species are known to alter their behavior when trying to steal food from a human, according to whether or not that human is watching them. This is surely a social problem, but one devoid of explicit communication or cooperation.

Two researchers with whom the regular reader of this blog should now be familiar, Victoria Wobber (who ran the bonobo testosterone study I mentioned in the review of Bonobo Handshake) and Brian Hare, wonder to what extent dogs can reason about the social world more broadly. Specifically, would their impressive social skills persist in a task that did not involve cooperative communication? They compared dogs and chimpanzees in two versions of a reversal learning task: non-social and social.

The reversal-learning task goes something like this: you present the individual with two containers. In one container there is a reward (like a peanut or a dog biscuit), and in the other container, no reward. If the subject chooses correctly, the reward is given; if not, the reward is revealed in the alternate location but is not given. The reward is always going to be found in the same location, so after a few trials, the participant should learn which container to choose to receive the reward. Then, once it becomes clear that the participant has learned the task (in this study, 84% accuracy), then the rewarded container switches. How quickly does the participant learn to inhibit the previously-correct choice, and begin choosing the alternative container which now contains the reward?

In the non-social condition, the reward was hidden underneath cups or bowls. The experimenter takes a treat, hides it under one of the bowls, and sham-hides it under the other bowl. Then the experimenter backs away and stares directly at the individual instead of at either of the bowls and the individual is allowed to choose.

In the social condition, instead of cups or bowls, human experimenters hid the rewards in their hands. Both experimenters went to the box of treats: the first took one and hid it in her hand, and the second only pretended to take one and hide it in her hand. Then both experimenters approached, reached out and offered their closed fists to the participant. As in the non-social condition, the participant chooses by pointing or approaching one of the individuals. This task requires that the participant form an association between one individual human and the potential reward, but does not require the participant to interpret any cooperative-communicative cues. Compared to the non-social task, this task is highly social: the participant must use social information about an individual's previous behavior to succeed.

It is well-known that in humans, placing an otherwise non-social problem in a social context increases performance. It is therefore possible that chimpanzees and dogs may be better able to learn when identical information is presented in a social context. Wobber and Hare describe their predictions in this way:

If neither species showed a different between the two contexts, this would suggest that the ability to inhibit prior learning is independent of the stimuli being learned. However, if chimpanzees showed increased performance in the social task, this would create two alternative outcomes for the dogs. If dogs' exceptional social skills extend beyond the use of cooperative-communicative contexts, they should show relatively more skill in the social reversal learning task than the non-social task. If, however, their skills are limited to interpreting human signals, they should show no distinction between the two tasks.

Critically, the goal here isn't two compare chimpanzees to dogs. The predictions concern within-species differences between the two tasks. In that sense, the chimpanzees serve as a control for the dogs. If the chimpanzees do not show a task-dependent difference, the data from the dogs can't really be interpreted. However, if the chimpanzees do show a task-dependent difference, then the data from the dogs is interpretable in terms of the importance of the social context.

Twenty-two chimpanzees participated as well as twenty-four dogs. Each individual participated in either the social or non-social condition, but not both. The chimp research took place at the Tchimpounga Chimpanzee Sanctuary in the Congo Republic (chimpanzee congo, not bonobo congo), and the dog research took place at the Max Planck Institute for Evolutionary Anthropology, in Leipzig, Germany (all dogs were pets).

First, how quickly did the participants learn the initial association, in either condition? For both species, there were no differences between tasks in how quickly they were able to form the initial association.

Mean trials to reach 84% criterion in the first part of the experiment. Bars are standard error. No difference between conditions in either species.

Next, how did the participants do in the second part of the experiment, after the reward location had been reversed? There was a significant difference between conditions for the chimpanzees, with chimpanzees in the social condition performing better. There was no significant difference for the dogs.

Mean number of trials where the accurate choice was made. Bars are standard error.

Furthermore, when comparing the first ten and last ten trials of the reversed part of the task, it is clear that while the dogs in general improved throughout the course of the experiment, there was no distinction between the two conditions. In contrast, for the chimpanzees, there is a clear difference between conditions: they improved over the course of the experiment in the non-social task, but were near ceiling the entire time in the social task.

Performance across the twenty trials of the reversal. Dogs, above; chimps, below.

Taken together, the data suggest that while chimpanzees were able to inhibit their learned response in the social condition, the dogs were unable to inhibit their learned response in either condition. The social context allowed the chimps to quickly alter their behavioral response. In contrast, the social stimuli devoid of communicative-cooperative information, was not helpful for the dogs. This supports the hypothesis that dogs' exceptional social abilities are specific to contexts involving cooperation and communication with humans, not that dogs are more skilled with social problems more broadly. By contrast, chimpanzees were more successful in the social condition than the non-social condition. While dogs may be more successful when given communicative-cooperative information by humans, chimps are clearly more successful than dogs in the absence of such information.

What allowed the chimpanzees to excel in this task? It is known that chimpanzees (as well as rhesus macaques and human children) keep track of who their friends and enemies are, so it may be that they formed "reputations" for the two human experimenters in the social task. It is not yet known whether dogs do this as well.

Another possibility is that the presence of the two human experimenters in the social task was too distracting for the dogs to really perform to their potential in this task. However, neither the chimps nor the dogs differed between conditions in terms of their ability to learn the initial association, suggesting that the task was a valid measure of learning, and that both species were sufficiently motivated.

So overall while the data don't completely contradict the hypothesis that dogs have superior social problem-solving skills overall, it does suggest that their domestication has particularly enhanced their problem-solving skills in communicative-cooperative contexts. More specifically, even if they are unusually competent with social tasks in general, they seem to be especially good at social tasks involving communication or cooperation with humans.

Wobber, V., & Hare, B. (2009). Testing the social dog hypothesis: Are dogs also more skilled than chimpanzees in non-communicative social tasks? Behavioural Processes, 81 (3), 423-428 DOI: 10.1016/j.beproc.2009.04.003