Why Do Some Dogs Act "Sneaky" When Stealing Food?

If Dogs "Steal" Food Does That Automatically Mean They Have a Theory of Mind?

I've Got a Case of the Munchies—I Hope No One's Looking...

Perspective Taking?

In a recent post here I discussed a 2013 study which purports to show that dogs only steal food when the lights are off. It was based on several previous research projects designed to determine if dogs have the ability to entertain points-of-view other than their own—in other words, the capacity to see things from another dog's or person's perspective which is also referred to as having a Theory of Mind, a theoretical construct used by cognitive researchers.

In 2010 article on his blog at PsychologyToday.com, Dr. Marc Bekoff says that “dogs show evidence of having a theory of mind,” . (“Domesticated dogs (Canis familiaris) react to what others can and cannot hear,” Shannon M.A. Kundey, et al, Applied Animal Behavior Science, August 2010.)

There are three levels to Theories of Mind: 1) The capacity to know what kind of perceptual experiences another person or animal is capable of having: sight, hearing, etc, 2) The capacity to know what kind of feelings or emotions another person or animal is capable of having, and 3) The capacity to know what kind of thoughts and mental processes another person is capable of having.

In the study Dr. Bekoff cites, what’s ostensibly been proven is that dogs have the first of three levels of a ToM, the conscious awareness of the perceptual states of other beings. Dr. Bekoff writes, “We’ve ... learned that dogs know what others can and cannot hear.”

Is this true?

No. It's not.

Is it possible that dogs have this ability? Perhaps, but I’ve read the paper in its entirety, and I don’t think it comes anywhere near to proving that they do as Bekoff seems to believe. The researchers have assumed that this is so, but I don’t think that, within the confines of their study, they’ve showed any real evidence of it. (Another study—which shows that guide dogs don’t know that their owners are blind—completely contradicts the idea that dogs are capable of understanding the perceptual states of others.)

How The Study Was Done

Here’s the gist: after a brief (5 min.) inhibition task, where the dogs were individually shown a treat and told “Wait!” or “No!” (or whatever commands they knew),1 they were then given a choice of taking a treat from a silent or a noisy container. Both containers had 18 small bells strung across their openings. The bells on the silent container had their ringers removed. When treats were put in each container, the researchers made it a point to make sure the dogs could hear the bells on the noisy container, and to demonstrate that the other container was silent.

Each dog was tested separately, and given only one trial. They were all given permission to go to the containers and take a treat. An experimenter—someone the dogs had never met before the inhibition test—sat in front of the containers, with one container on either side of her. In some cases the experimenter had her eyes open (Looking Position), in others she hugged her knees to her chest, and hid her face (Not Looking Position).

All of the dogs who were allowed to grab a treat while the experimenter was in the Not Looking Position chose the “silent” container. When she was looking, the dog’s choice seemed to have been made entirely at random. (I’m not sure there is such thing as a random behavior for dogs, but that’s another story.)

“This suggests,” the researchers conclude, “that [the] dogs’ pattern of approach in the Not Looking condition was not due to either a general preference for the silent container or an aversion to the noisy container. Instead, dogs appeared to prefer the silent container only when the experimenter was not looking and therefore did not have knowledge of their approach. This suggests dogs took into account the noise caused by their approach only when that noise could change what the experimenter knew about their actions.”

The Path of Least Resistance

I don’t see any clear evidence for this conclusion. It’s entirely possible that one could also conclude that the “dogs significantly preferred the silent container only when a strange person sat in front of them in a strange position, with her eyes hidden from view.” And one could paraphrase the conclusion as follows: “Dogs appeared to prefer the silent container only when the experimenter was acting in a strange manner, with her eyes hidden, which from the dog’s point of view, created feelings of uncertainty. This suggests dogs took into account the noise caused by taking the treat only when that noise would further impact their own feelings of insecurity.”

Also, the researchers don’t give us any insight into why the dogs felt they needed to be quiet. Since the dogs were encouraged to get a treat, there’s no logical reason for them to be “sneaky” about it. So what was their motive in “making less noise?” It’s unclear.

Dogs make great guinea pigs for this kind of study because they’re the most social animal on the planet. And as such, they have a unique way of seeing things, a unique way of doing things, and a unique way of “thinking,” based largely on their evolutionary history as group predators, a history that has given them a predilection for being able to read the emotional states of their fellow hunters, a gift they expanded on exponentially with their human packmates during the domestication process.

It’s well known that when wolves hunt they’ll target the “weakest member” of the herd. However, if we look at this from the wolf’s pov, and keep in mind that all animals have to obey the laws of nature, including the first law of thermodynamics (the conservation of energy), perhaps we can see what the wolves are really doing, when they focus their energy on the “weakest member,” is that they’re finding the path of least resistance.

Knowledge States or Feeling States?

In the ToM study, the researchers themselves stated that the noisy container could have been aversive, and since the level of aversion a dog might feel is dependent, in part, on their baseline levels of anxiousness or uncertainty, the path of least resistance when the dogs couldn’t see the researcher’s eyes, might have been to choose the container that was the least jarring to them individually. (Bells may have a pleasant sound for humans, but not so for dogs.) If this is true, then the dogs’ behavior had nothing to do with what they “believed” about the human’s knowledge states. It was specifically about the dog’s own feeling states.

Another angle to consider is that from the first day a puppy comes home she learns to read the patterns in her owner’s body language, eye contact, emotions, etc. She does this to find cues as to which of her behaviors will cause the human to prevent her from gaining access to things she desires and which won’t. This ability in dogs, to tell when our attention is focused on them, and when it’s not, is a form of pattern recognition, and I think it has a clear adaptive purpose.

How so?

It seems to me that there’s a significant difference in 1) having the ability to tell when another animal is focusing its attention on you, which, again, has a clear adaptive purpose and is based, in part, on pattern recognition, and 2) having an ability to form an awareness of another animal’s “knowledge states,” which a) lies higher on the evolutionary scale, b) requires more brain power (i.e., a much bigger prefrontal cortex with more neurons, more bells and whistles), and c) isn’t as economical (it depletes more energy).

One reason we don’t jump to the conclusion that roaches “know we can see them” when we turn on the lights (and that’s why they scatter), or believe that a spider who pulls a leaf over himself to hide from his prey does so because he “believes he’s invisible,” is that insects and arachnids have tiny brains, while the dog’s brain is much closer to ours in size and shape. But sometimes even the smallest difference in size or shape, can have a huge impact on cognition. There are certain features of the human brain that simply do not exist in dogs.

Reading Minds or Feeling Emotions?

I believe it’s 100% true that dogs are mind readers, in that they read our emotions, not our “knowledge states.” That’s why some dogs are able to find their way home when they get lost. That’s why all dogs are able to read hidden desires that even we didn’t know we had.

Years ago I advised some clients that they should never pick their puppy up when she was about to do her business inside the house. “You have to interrupt her before she’s already in the act. Otherwise, she’ll start doing it only in places where you can’t reach her.” They ignored my advice, and sure enough, the dog began peeing under the coffee table. If this dog was concerned about her owners’ ability to see her while she was doing her business, that wasn’t the best choice. But if she was only concerned about whether they could grab her, it worked like a charm.

So to a dog, it’s not necessarily about what your owner can see or hear you doing, it’s about what they can do to you, do for you, or best of all, do with you. Remember, dogs are not only the most social species on the planet, they’re also the most loving.

Bad Science Hurts Good Dogs

Some readers may feel I’m being too meticulous about these issues, that I’m “assigning” lesser cognitive functions to dogs than they really have or “deserve.”

Yes, I am being meticulous. Why? Because all this make believe about canine cognition hurts dogs. For some people it feels grand and wonderful to think of dogs as having brains and minds quite like ours, and it puts them more on “our level,” but there’s also a down side. The more some people believe that dogs do things for reasons, or that they have an awareness of our mental states, or that they’re capable of understanding human language, the more they’ll tend to punish their dogs because “he’s doing it to piss me off!” or “I know she did it on purpose!” or “she knows perfectly well that what she did was wrong!” etc. (Alexendra Horowitz published a study showing that when people believe their dog has done something wrong, they tend to interpret the dog’s body language and facial expressions as a consciousness of guilt, even when the animal hasn't misbehaved!)

Then, when dog owners scold their dogs, or try to reason with them, and don’t get the results they’re looking for (which some scientists are insinuating that they should be getting), such owners get frustrated and angry. Some will mistreat their dogs terribly. Others will simply abandon them. And not a single one of those poor misunderstood doggies will have any idea as to why they’ve been beaten and battered, or why they’ve been left alone by the side of the road, or why the purity of their love has been scorned and denied by the people they loved so deeply, so completely, and so much.

Our shelters are overloaded. So I think it’s not only more scientific, but better for our dogs, if we’re a little more parsimonious in how we look at canine cognition.


“Life Is an Adventure—Where Will Your Dog Take You?”


1) I elided any mention of the inhibition task in the initial version of this article because it didn’t seem to have any relevance to what happened later in the study. It was a brief, one time set-up, which I’m guessing was designed to make the dogs see the experimenter as “the gatekeeper” to whatever happened later. I contacted the lead researcher via e-mail to give her a chance to clear up any misunderstandings I might’ve had, and she suggested that the inhibition task was a necessary part of the proceedings.

If the purpose of setting herself up as the gatekeeper was to establish a certain level of inhibition in the dogs when it came time to allow them to take a treat from one of the containers, then this part of the study only bolsters my position that the dogs were not, in fact, acting as they did because the dogs had a knowledge of the gatekeeper’s perceptual states, but rather they were even more inhibited by her “strange behavior,” and so chose the silent container because it was the path of least resistance for them individually.

2) “Associative memory is vast, effortless and quick; logical processing is limited, painful and slow. In all likelihood, the human brain evolved with a strong bias towards pattern recognition rather than deductive reasoning. This neural bias would emerge early in the biological record, well before the development of primates, or mammals, or even vertebrates. Homo sapiens is endowed with sophisticated pattern recognition capabilities honed through eons of evolution, and it is unsurprising that this capacity is put to use in social behavior. Deductive reasoning, in contrast, is a comparatively recent development and is much more difficult. While we are very proud of deductive reasoning, it is not necessarily more useful, particularly when dealing with social behaviors which may also have some evolutionary roots.” (“A New Kind of Social Science,” Valerie M. Hudson, Brigham Young University, Phillip A. Schrodt, University of Kansas, 2004.)

Featured Posts
Recent Posts


  • Facebook Social Icon
  • Twitter Social Icon
  • LinkedIn Social Icon
  • Black Facebook Icon
  • Instagram - Black Circle
  • Black Vimeo Icon

© 2016 by Lee Charles Kelley.