Inside of a Dog: What Dogs See, Smell, and Know (25 page)

Two experiments have tested just this imitative understanding. The first asked what exactly dogs see in others' behavior: the means or the end. A good imitator would see both, but would also see if the particular means isn't the most expedient way to the end. From a young age, human infants can do just that. They will religiously imitate—sometimes to a fault*

—but they can also be astute. For instance, in one classic experiment, after watching an adult turn on a light in an unusual way—with his head—the infant subjects could imitate this novel action, if asked to do so. But they did not spontaneously imitate if the adult was grasping something in his hands, making him unable to use them to turn on the light: the infants used their hands, reasonably enough. If the adult held nothing in his hands, infants were more likely to turn on the light with their heads, too—inferring, perhaps, that there must be good reason, besides one's hands being full, for this new maneuver. They seemed to realize that the adult's actions
could
be imitated, and they selectively imitated them only insofar as it seemed necessary to do so.

In the dog variation on this paradigm, a wooden rod taking the place of a light, one "demonstrator" dog was taught to press the rod with his paw to release a treat from a spring-loaded dispenser. The researchers then had the demonstrator dog perform his newfound trick in front of other dogs who were being restrained to watch. In one trial the demonstrator pressed the rod while holding a ball in his mouth; in the other, he had no ball. Finally, the observer dogs were let at the apparatus.

It should be noted that dogs are not naturally drawn to mechanical dispensers, even ones with wooden rods. And
pressing
is not the first approach of most dogs when facing a problem: dogs can use their paws handily, but they typically go at the world mouth first and paws second. Though they can be trained to push or press an object, dogs' first approach at an object such as this one is not one of intuitive understanding. They will bump it, mouth it, knock into it. If they can, they will push it over, dig at it, jump on it. But they do not consider the scene for a moment and then calmly press the rod. Thus the first approach of the observer dogs was particularly interesting: Would the demonstration change their behavior?

These dog subjects behaved just like the human infants with the light switches: The group that saw the demonstration with no ball imitated faithfully, pressing the rod to release the treat. The group that saw the demonstrator acting while holding a ball in his mouth also learned how to get the treat, but used their (ball-less) mouths instead of paws.

That the dogs so imitated is remarkable. This is no mere mimicry, copying for copying's sake. Nor is it just an attraction to the source of activity. It looks more like the behavior of an animal who is considering what another animal is doing: what his intention is, and how—or how much—to reproduce that behavior themselves, if they have the same intent.

If these experiments represent the performance of all dogs, it looks as though we could say that dogs are, at the very least, able to learn by watching others in particular social contexts—when food is at stake, for instance. One final experiment suggests something even more impressive: that dogs may actually understand the
concept
of imitation. The single subject, an assistant dog trained to work with the blind, had already learned by operant conditioning to do a number of non-obvious actions on command: to lie down, turn around in a circle, put a bottle in a box. What the experimenters wondered was whether he could do these actions not just to a command, but after seeing someone else do the action themselves. Sure enough, the dog ably learned to turn around in a circle not after the
Turn around in a circle
command, but on simply seeing a human do such a thing, followed by the imitation request
Do it!
They then examined what he would do when seeing a human do a new, completely odd action, such as running off to push a swing, tossing a bottle, or suddenly walking around someone else and returning to their starting spot.

He did it. It was as though this dog had learned the concept
imitate,
and, given that notion, could apply it more or less in any direction. To do this, he had to map his body onto a human's: where a person tossed a bottle by hand, the dog used his mouth; he used his nose to push the swing. This is not the final word about imitation (just ask your dog to copy your swing-pushing, and you can see how results do not always generalize), but these dogs' abilities are suggestive of something besides mindless mimicry. Dogs may be enabled to imitate by the same ability—almost compulsion—to look at us that allows them to use us to learn how to act. That is what I see in Pump's morning stretch alongside me.

I open the door stealthily and Pump's there, not two feet away, walking toward the rug with something in her mouth. She stops in her tracks and looks over her shoulder at me, her ears down, her eyes wide. In her mouth is an unidentifiable curved form. As I slowly approach, she wags low, ducks her head, and in the moment that she opens her mouth to get a better grip on her find I see it: the cheese left out on the counter to warm. The brie. The entire enormous round of brie. She gulps two gulps and it's gone, down the gullet.

Think of the dog caught in the act of stealing food from the table … or looking at you squarely in the eyes with a plea to go out, be fed, be tickled. When I see Pump, mouth full of brie, seeing me, I know she's going to make a move; when she sees me seeing her, does she know I'm going to try to thwart it? My strong impression is that she does: the moment I open the door and she looks at me, we both know what the other is going to do.

The study of animal cognition reaches its pinnacle in addressing just this kind of scene: one raising the question of whether an animal conceives of others as independent creatures with their own, separate minds. This ability seems more than any other skill, habit, or behavior to capture what it is like to be a human: we think about what others are thinking. This is called having a
theory of mind.

Even if you've never heard of theory of mind, chances are you nonetheless have a very advanced one. It allows you to realize that others have perspectives different from your own, and therefore have their own beliefs; different things they know and don't know; a distinct understanding of the world. Without one, others' behavior, even the simplest acts, would be utterly mysterious, arising from unknown motivations and leading to unpredictable consequences. Trying to guess what a man approaching you, mouth agape, arm raised high, hand waving frantically, is going to do is greatly aided by having a theory of mind. It's called a
theory
because minds are not directly observable, so we extrapolate backward from actions or utterances to the mind that prompted that act or remark.

We aren't born thinking about others' minds, of course. It is quite likely that we aren't born thinking about much at all, even our own minds. But each normal child develops a theory of mind eventually, and it appears that it is developed through the very processes discussed so far: through attending to others, and then noticing their attention. Children with autism often don't develop some or any of these precursory skills: they may not make eye contact, point, or engage in joint attention—and many don't seem to have a theory of mind. For most people, it is but one large theoretical step from an awareness of the role of gaze and attention to realizing that there is a mind there.

The gold standard experiment for theory of mind is called the
false belief
test. In this design, the subject, typically a child, is presented with a minidrama played out by puppets. One puppet places a marble in a basket in front of her, in full view of the subject and a second puppet. Then the first puppet leaves the room. Promptly, the second puppet wickedly moves the marble over to her basket. As the first puppet returns, the subject is quizzed: Where will the first puppet go looking for her marble?

By age four, children answer correctly, realizing that they and the puppet know different things. Before that age, though, children surprisingly and unambiguously fail. They say the puppet will look for the marble where the marble actually is—in the second basket—showing that they aren't thinking about what the first puppet really knows.

To design a verbal false belief task for animals, who cannot be expected to communicate their answers (nor be engaged by a puppet marble-switching drama) is nigh impossible, so nonverbal tests have been developed. Many take their cue from anecdotal reports of compellingly mindful animal behavior seen in the wild: of deception or clever competitive strategies. Chimpanzees are the most frequent subjects since, as close relatives to humans, one might expect that they would have the most similar cognitive abilities.

While the results with chimpanzees have been equivocal, lending credibility to the notion that only humans have a fully developed theory of mind, a wrench has been thrown in the experimental works. That wrench is the dog: whose attention to attention, whose seeming mind reading, looks anecdotally just like what we call acting with a theory of mind. To go from my living-room theorizing about a dog's understanding of mind to solid scientific standing, researchers have begun to run dogs on the same tests used with chimps.

Here's what one dog, an unsuspecting experimental subject, found awaiting him at home one day. Instead of the usual ready availability of his favorite tennis balls, every ball in the house had been collected, and an extra lot of people were standing about gazing at him. Fine so far: Philip, the three-year-old Belgian Tervuren in question, didn't freak out. But he might have been bemused when, one by one, the balls were shown to him and then placed in one of three boxes and locked up. This was new stuff. Whether game or threat, what was clear was that the balls were being methodically placed somewhere other than his favorite place: right in his mouth.

When released by his owner, Philip went, naturally, straight toward the box where he saw a ball hidden, and he nuzzled the box. This turned out to be the right thing to do, for it prompted the humans to exclaim merrily, open the box, and give him the ball. Despite just having his mouth on the ball, the dog found that the people around him kept taking it away and securing it in one or the other box—so he kept playing along. Then they started locking the boxes and putting the key elsewhere, so the whole thing took even longer after he selected the right box: someone must find the key, bring it to the box, and open it. The final twist involved one person who locked the box, hid the key, then left the room. Another person came in—surely one who, like all other people around, would be able to use these key-things to open these lock-things.

This was the moment the experimenters were waiting for: they wanted to know if the dog saw the new person as unknowledgeable about the location of the key. If so, then not only should Philip indicate which box has the beloved ball, he should also help the person find the key that would enable access to that ball.

On repeated trials, that's more or less what the dog did: ever patient, Philip looked toward the spot where the key had been hidden, or headed that way. Note that he didn't actually take it in his mouth and open the box: that'd be some trick, but even the most ardent dog enthusiast will admit it's unlikely. Instead, Philip used his eyes and his body as communications.

Philip's behavior could be interpreted in three ways: one functional, one intentional, one conservative. The functional interpretation is this: the dog's gaze served as information for the person, whether the dog meant it to or not. The intentional: the dog did in fact mean it to: he looked because he knew the person was ignorant of the key's location. The conservative: the dog looked reflexively, since someone was recently over there where the key was.

The data do the interpreting. They show that the functional is definitely true: gaze did serve as information to the person nearby. But the intentional take is also true: the dog looked at the location of the key more often when the person in the room with them was ignorant where it was—as if meaning to inform the person with his gaze. That nixes the conservative interpretation. Philip seemed to be thinking about these crazy experimenters' minds.

This is but one dog—maybe a particularly astute one. Remember the begging experiment run with chimps and dogs? Unlike chimps, all the tested dogs immediately followed the knower's (non-blindfolded or bucketed person's) advice as to which box was baited with food. Hoorah for these dogs, who thus all found food inside. This looks good for the theory of dog mind: they acted as though thinking about the knowledge states of the strange people pointing in front of them. But after this seeming cognitive accomplishment, a strange thing happened. When run again and again on the same test, these dogs changed their strategies. They began to pick the guesser just about as often as the knower. Does this mean they were prescient and then grew dimwitted? Although dogs will do impressive convolutions for food, this doesn't make sense as an explanation. Perhaps it indicates that the first round was a fluke.