From American Scientist:

Does your dog know what you are thinking? Can a chimpanzee understand what another sees? In the three and a half decades since David Premack and Guy Woodruff first asked whether chimpanzees have a “theory of mind,” a considerable empirical and philosophical literature has sprung up around what has come to be called “mind reading” in animals. Theory of mind, as Premack and Woodruff defined it, is the ability to attribute perceptual and cognitive states to others. This is not about telepathy, but about whether any animals besides humans have the capacity to attribute such states to others. Numerous experimental tests and other observations have been offered in favor of animal mind reading, and although many scientists are skeptical, others assert that humans are not the only species capable of representing what others do and don’t perceive and know.

Robert Lurz, a philosopher at Brooklyn College, CUNY, surveys the experiments at the heart of the debate and finds that not one of them solves what he calls “the logical problem” in animal mind-reading research. The logical problem is that for any mind-reading hypothesis, it seems possible to construct a complementary “behavior-reading” hypothesis that makes exactly the same predictions but is assumed to be less cognitively demanding. The basic point is that whatever mind reading is, it is not magic, and thus depends on ordinary, perceivable cues—but these same cues are then available as a basis for expectations about actions that an animal might take without making any mental attribution. If you see me gazing at a piece of cake with a certain look on my face, you may infer that I’m thinking about eating it, or you might instead directly form some expectations about my cake-directed behavior without imagining what I might be thinking about it.

Lurz argues that all previous experiments conducted or proposed to test mind reading in animals suffer from some version of this problem. For example, he notes, Brian Hare and colleagues ran an experiment in which a subordinate and a dominant chimpanzee could compete for two pieces of food. One of the pieces was hidden so that the dominant animal could not see it because of an opaque barrier blocking the line of gaze to the food item. The subordinate chimpanzee, who had an unobstructed view from the other side of the barrier, preferentially went for the piece that the dominant could not see, leading the researchers to argue that the subordinate chimpanzee attributed to the dominant a state of ignorance about the presence of this food item. But Lurz points out that the subordinate need only make the behavioral prediction that the dominant will not attempt to retrieve a piece of food for which it lacks a direct line of gaze. The subordinate animal does not need to attribute any states of seeing or knowing to the dominant to make this behavioral prediction.

The fact that Lurz finds all prior experiments lacking does not, however, make him an outright skeptic about animal mind reading.

“Do I See What You See?”, Colin Allen, American Scientist