Did Dogs Gain Their Social Intelligence By Accident?

I will be reposting some dog-related posts from the archives in the coming few weeks as I prepare for the course I'm teaching this semester on dog cognition. Please let me know if you find something inaccurate or unclear.

ResearchBlogging.orgDomesticated dogs seem to have an uncanny ability to understand human communicative gestures (see here). If you point to something the dog zeroes in on the object or location you're pointing to (whether it's a toy, or food, or to get his in-need-of-a-bath butt off your damn bed and back onto his damn bed). Put another way, if your attention is on something, or if your attention is directed to somewhere, dogs seem to be able to turn their attention onto that thing or location as well.

Amazingly, dogs may be better at this than primates (including our nearest cousins, the chimpanzees) and better than their nearest cousins, wild wolves.

And so it was that biological anthropologist Brian Hare, director of the of Duke University Canine Cognition Center wondered: did dogs get so smart because of direct selection for this ability during the domestication of dogs, or did this apparent social intelligence evolve, in a sense, by accident, because of selection against fear and aggression?

Remember the Russian foxes? In that experiment, by selecting for tameness and against fear and aggression, a number of other seemingly unrelated traits began to express themselves in the domesticated foxes. They appeared to enjoy hanging out with humans. They sniffed and licked their caretakers. They wagged their tails when excited (remember this video?). Their tails got curlier and shorter. Their ears got floppier. Their skull-size (relative to body-size) increased. Their fur coloration patterns changed. They also showed evidence of physiological changes, such as in the pituitary-adrenal pathway.


Figure 1: Time for a picture of a dog. Some dogs are clearly smarter than others.

Hare and his colleagues pitted these two hypotheses against eachother. The "Selection for Communication" Hypothesis suggests that predicting and manipulating human behavior by reading human communicative signals (such as eye gaze or arm pointing) was explicitly selected in the domestication of dogs. Those dogs who were most skilled at comprehending such gestures were more likely to survive and reproduce. The "Correlated By-Product" Hypothesis predicts that the ability to read human communicative gestures was not directly selected for during the domestication of dogs. Instead, fear and aggression were selected against (which is the same as saying that tameness was selected), and as a result, those individuals with less stress, who had positive interactions with humans had the highest evolutionary fitness. Moreover, those same dogs, because of the "changes responsible for [their] high levels of tameness... were no longer constrained [by fear or disinterest, for example] in applying previously existing social problem-solving skills to humans" in interactions between the two species.

Hare and his colleagues went to study the artificially domesticated fox infants in Siberia to find out if they were as skillful as dog puppies in using human gestures, and if they were more skilled than the control-group foxes, who were not domesticated. Neither group of foxes had ever been tested for use of human gestures, and both groups of foxes had limited interactions with their human caretakers, so it is more likely that - if they did have such skills - they would be innate and not learned.

Experiment One: Do Domesticated Foxes Use Human Communicative Cues to Find Food?
Eleven experimental foxes (from Belyaev's fox farm) and eleven dog puppies were compared. All animals were between 2-4 months of age when tested. This way, the dog puppies' responses could also be assumed to be innate, and not learned, given such little opportunity to learn.

First, each animal was brought into a testing room for 3-4 play sessions in which they were free to interact with the experimenter for 45 minutes. This ensured that the animals were accustomed to the experimenter and the testing room (in every case, the experimenter waited for the animal to initiate the interactions). The experiment began after the subject took food from the experiment's hand, and out of the bowls that were to be used in the task.

Two round bowls were placed apart in the middle of the room. A line between the two bowls (equidistant from either bowl) was drawn on the floor. The animal was shown that food could potentially be found in either bowl: Experimenter 1 placed a piece of food in one of the bowls while the animal watched, while a second experimenter held the animal at a spot equally distant from either bowl. Once the food was hidden, the animal was allowed to retrieve the food. If the animal touched the correct bowl first, it received the food - if the animal touched the incorrect bowl first, it was shown the food but was not allowed to enjoy the snack. This was repeated for both bowls.

Then, on a different day, the animal was brought back to the same testing room. After showing the food to the animal, experimenter 1 pretended to bait both bowls, but only actually left the food in one of them. Experimenter 2 was holding the animal during this phase, as in the familiarization phase. The animal therefore knew there was food to be found, but not in which bowl. Then, experimenter 1 attracted the attention of the animal (such as by snapping), and then pointed and gazed towards the correct bowl. Then, experimenter 2 released the animal, while experimenter 1 continued gazing and pointing towards the correct bowl. If the animal chose the proper bowl, it got its snack. If not, it was shown the proper location, but did not get the food.

A control condition was identical except that experimenter 1 looked at the animal, and not at either bowl. Each animal received 18 trials of each condition in separate sessions.

Both the experimental foxes and the dog puppies found the hidden food well above chance levels in the point-and-gaze situation, and were equally successful in this task. The ten experimental foxes tested in the control situation (no point or gaze) were unable to reliably find the hidden food. Furthermore, there was no effect of accumulated experience across the 18 trials. It should be noted that adult dogs are unable to find hidden food in a similar task, even when scent-related cues are available.


Figure 2: Baby foxes and dog puppies are equally good at finding the food when given a point-and-gaze cue.

Experiment 2: Do Domesticated or Control Foxes Use Human Gestures More?
Seventeen experimental foxes, and seventeen control foxes participated in this experiment, and all were 3-4 months old.

In the familiarization phase, the experimenter approached the subject's pen and then placed the testing apparatus (two toys made of flexible metal which, after manipulating, would return to its original position and would make a small noise) against the front of the animal's pen. Half of the subjects were presented with the experimenter first, and half with the toy first.

Later, on a different day, the toy was presented and the experimenter sat behind it. Once the animal was in the center of its pen, the experimenter looked at, pointed at, and touched one of the toys so that it moved and made a small noise. Then, the toys were moved closer to the animal. Which toy did the animal touch first, with its mouth or paw? How long did it take before the animal approached the objects?

All animals, in both groups, manipulated one of the two toys. However, the domesticated foxes touched the same toy as the experimenter significantly above chance, while the control foxes did not.


Figure 3: Experimental foxes touched the manipulated toy more than the control foxes.

Experiment 3: Are Domesticated Foxes More Skilled than Controls in All Human-Led Tasks?

This experiment was identical to experiment 2, except it appeared as if an object was manipulating a toy, instead of a human. In this experiment, there was no difference between the two groups for choosing between the two toys (despite the fact that it looks like there was, it's not statistically significant). The control foxes, however, actually preferred to play with the toy that had been manipulated by the object, compared with playing with the toy that had been manipulated by the experimenter.


Figure 4: Experimental foxes manipulated the same toy as the experimenter, control foxes manipulated the same toy as an object more than when manipulated by the experimenter.

Experiment 4: Are Domesticated Foxes More Skillful with Human Communicative Cues than Controls?

This was identical to experiment 1, except instead of comparing domesticated foxes with dog puppies, they were compared with control foxes.

Both the control and the experimental foxes used the point-and-gaze cue above chance levels. However, the experimental foxes used the human communicative cue significantly more than did the control foxes. As in experiment 1, there was no effect of experience.


Figure 5: Both groups used the human communicative cue, but experimental foxes did it more often than the control foxes.

What do these experiments tell us?
(1) Domesticated foxes are as good as dog puppies at finding hidden food when offered communicative signals from humans.
(2) Domesticated foxes are more interested in playing with a toy that a human has played with, than are control foxes.
(3) Although control foxes can use human communicative gestures, the domesticated foxes do it more readily.
(4) Although domesticated foxes are no more likely to approach a novel human or a novel object than the control foxes, when they do approach, they do it much faster.

Taken together, these results appear to support the correlated by-product hypothesis, and not the selection for communication hypothesis. It suggests that the evolution of social cognitive abilities in domesticated dogs mirrors that process observed in the experimentally domesticated silver foxes, and that it was a by-product of selection against fear and aggression. To really really get at this question, a study of wolves should be conducted as well.

More broadly, the social intelligence hypothesis (which is another way of framing the selection for communication hypothesis) asserts that primate (and human) intelligence was driven by the need to predict and manipulate the behavior of others, by reading subtle cues in their behavior. These findings suggest that human intelligence may have evolved, instead, as a by-product of selection against fear of and aggression towards others.

Hare, B., Plyusnina, I., Ignacio, N., Schepina, O., Stepika, A., Wrangham, R., & Trut, L. (2005). Social Cognitive Evolution in Captive Foxes Is a Correlated By-Product of Experimental Domestication Current Biology, 15 (3), 226-230 DOI: 10.1016/j.cub.2005.01.040

More like this

Thanks for this. It've been curious about this for a while and it continues to intrique me.
I have seen comparisons between dogs and wolves, in which the communication instincts between dogs and wolves are displayed, but never between dogs and pre-adult wolves.
Have wolves ever been selected in a way similar to the fox experiment. I also wonder what would happen if a similar experiment were to be conducted using non-human primates.

This is awesome. I want to take your canine cognition class.

I recently visited a wolf sanctuary that was home to not only pure wolves, but also a number of wolf/dog hybrids. Our guide (a wolf-lover, but not a biologist) mentioned that they don't really have a test for verifying hybrid status, just a couple of key physiological markers (eye color, ear shape, size) and make up the difference with a couple weeks of behavioral observation. He also told an anecdote about a caretaker pointing out a treat while in a wolf enclosure and being promptly attacked because the wolf took the pointing gesture as a sign of dominance, the caretaker making himself look bigger. Made me wonder how innate that response is in wolves. Is dominance something that manifests later, or will a wolf and dog puppy interpret pointing in noticeably different ways?


I'm a writer and dog trainer living in New York City. I also write about canine behavior for PsychologyToday.com.

I just came across your blog through your article on the directed-gaze behaviors found in reptiles, which I found to be very interesting and helpful in understanding where a dog's social behaviors may or may not come from.

I wanted to point out a few things that I've observed, and been thinking about, in relation to this idea that dogs are able to follow wherever a human points. In my observations this is only true (meaning it only happens often enough to be relevant) when the dog is engaged in a hunting game with the researchers: "Find the treat!" or "Find the toy!" When you point randomly behind most dogs, for instance, and say, "Look! Look, behind you!" they very rarely look where you're pointing. So following where a human looks or points is not necessarily an innate behavior.

Secondly, I disagree with both the social intelligence and the correlated by-product hypotheses. I don't think enough thought has been given to the role that oxytocin must have played. In fact, I believe that the dog's social instincts are inherited specifically from the wolf's need to hunt large, dangerous prey animals, which probably evolved as a result of oxytocin: The Key to Canine Evolution: The Love Hormone [see below].*

With that in mind, I think early humans probably selected for the dogs who produced more oxytocin rather than for those who produced less adrenaline (though, obviously, the two things could be correlated). For various reasons, this was probably done more by the tribe's women rather than by the men. (Of course, this would still fall under the general rubric of the correlated by-product hypothesis.)

Also, there are still several unanswered questions about how relevant the Belyaev experiments are in terms of how dogs became domesticated. In another recent blog article for PsychologyToday.com I wrote the following:

A little over 20 years ago I noticed an interesting anomaly: more and more Labs and golden retrievers -- breeds usually known for their friendly, easygoing nature -- were becoming aggressive, while at the same time more and more pit bulls were becoming overfriendly! Could breeding for "friendliness" -- as Belyaev did with foxes, and as Lab and golden breeders had been doing with their litters -- actually create more aggression? And could breeding dogs for fighting somehow create overfriendliness? [2]

Temperament testing in puppies -- though no longer considered a valid means of predicting an adult dog's personality -- still shows one thing: in every litter of pups there is a spectrum of temperament types, from what we used to call "dominant" to what we used to call "submissive." It's literally impossible to find a litter of puppies where every single pup has the exact same behavioral tendencies; there is always a spectrum, as if Nature were acting as a quality control technician for Crayola, ensuring that each box of crayons contained an array of colors so that your child wouldn't open a new box and find all red, yellow, or blue.

But why would Nature want this spectrum to exist?

I can only think of one reason, and it goes back to the ideas that 1) dogs and wolves are genetically related, and 2) the wolf pack is not a dominance hierarchy but a mechanism for hunting large prey by working in harmony. If all wolves had a direct approach to prey, the hunt would surely fail; the same is true if they all had an indirect approach. And this box-of-crayons operating system is still seen in every litter of puppies, in every breed imaginable, across the board. It's as if each litter of pups is still seen as a potential hunting unit, at least as far as Nature is concerned.

I'm glad I found your blog! Keep up the good work!


*(It's also interesting to note that of all land mammals only dogs and humans have an evolutionary history of routinely hunting animals that are larger and more dangerous than themselves; and that the only other species that do this are members of the dolphin family.)