Anyone who has ever lived with a dog or a cat (or any other intelligent social animal) will attest to the occasional uncanny feeling that one's pet knows what one is thinking. But is that really possible? Can animals really know or even grasp the idea of what is going on in another's mind? And even if they can, how would we ever know? After all, we can't just ask them. This is the puzzle of animal mindreading.
The first scientific attempt to address this puzzle occurred back in the late '70s by David Premack and Guy Woodruff at the University of Pennsylvania. In their study, a chimpanzee named Sarah was shown videos of one of her trainers encountering a particular problem, such as trying to reach a banana hanging by a string from the ceiling high overhead or shivering next to an unlit gas heater in a cold room. The videos were paused before the trainer solved the problem, and Sarah was then presented with a pair of photos to examine and select. One of the photos always showed the trainer doing something, or in some cases just an object, that would aid the trainer in solving the problem in the video, while the other photo showed the trainer doing something, or just an object, that would not aid in solving the problem. Sarah's photo selection was quite striking. In nearly all the test trials, she picked the photo that would lead to solving the trainer's problem. Premack and Woodruff believed that Sarah's performance demonstrated her understanding of the trainer's intention in the video (e.g., the trainer's intention to get the banana) and her desire to select the photo that depicted an event or object that would lead to the intention's fulfillment.
Not everyone in the field was so convinced, however. Subsequent reviews of the videos and photos persuaded some researchers that Sarah may have succeeded in selecting the correct photos without understanding anything about the trainer's underlying intention. It was argued that Sarah may have selected the correct photo on some trials merely by matching the object in the photo (e.g., a horizontal stick on the floor) with the same object prominently displayed in the video. And on other trials, it was argued that Sarah may have simply picked the photo that showed a predictable stage in a sequence of human behavior with which she was familiar (e.g., Sarah may have learned after fourteen years of living in close contact with trainers and experimenters that humans when faced with an unobtainable piece of food before them, will sometimes use an available tool or remove a constraining obstacle to gain access to it).
Despite its contentious results, Premack and Woodruff's study proved a landmark in the field of comparative psychology and gave birth to an industry of scientific research into mental state attribution in animals and children. It also incited a longstanding, and at times quite heated, dispute among empirical researchers and philosophers about the very existence and possibility of animal mindreading. There is now a large volume of research on mental state attribution in animals. The results of many of these new studies have been interpreted by a number of prominent researchers as demonstrating that some animals (e.g., apes, monkeys, dolphins, dogs, goats, ravens, and scrub jays) are quite capable of attributing perceptual states (seeing and hearing), state of knowledge and ignorance, intentions and goals, and in some cases even false beliefs to other agents.
The best known and most influential of these new studies is Hare, Call, Agnetta, and Tomasello's competitive paradigm study published in 2000. In their study, a subordinate and dominant chimpanzee competed for two pieces of food placed in a room between their cages. The subordinate chimpanzee, however, was always allowed to observe that one of these pieces of food was hidden from the competitor's line of sight while the other piece of food was out in the open or behind a transparent barrier. It was discovered that when the chimpanzees were allowed to compete, the subordinate chimpanzee nearly always attempted to steal only the food that the dominant could not see. The researchers took the data as compelling evidence that chimpanzees understand the mental state of seeing and use this understanding strategically to predict what their fellow chimpanzees are likely to do in competitive situations.
Some researchers, however, have been considerably less sanguine about how to interpret the data from these new studies. Most notable among these skeptics is Daniel Povinelli at the University of Louisiana. Povinelli has long argued that all such experiments, by virtue of their very design, simply fail to control for the very real possibility that animals are just clever behavior-readers, that they simply represent and adaptively respond to various behavior-predicting cues in social encounters (e.g., the cue that a dominant conspecific has an unobstructed line of sight to the food) without understanding the cues themselves as signs of underlying mental states (e.g., that the dominant conspecific sees the food). In addition, Povinelli contends, there is presently positive and independent grounds to doubt the very idea of mindreading in animals. Mindreading, he argues, involves representing hidden, internal causes of other agents' behavior, and animals have shown absolutely no sign of understanding such a notion of causation when tested in non-social contexts (e.g., tool use).
Not surprisingly, given its importance to issues in philosophy of mind, philosophers too have weighed in on the debate. Some, such as Donald Davidson and José Bermúdez, have argued quite forcefully that animals simply cannot be understood to represent other agents' beliefs, desires, knowledge, or other propositional attitudes. To represent such states, they argue, would require having some idea of what propositions are (i.e., abstract bearers of truth and falsity) as well as a way to represent various relevant logical relations among them (especially, conditional or 'if-then' relations), and neither of these abilities is possible without an understanding and use of a structurally complex natural language - which animals do not possess.
So from this brief sketch, it would appear that the animal mindreading debate is at an intellectual standstill. Some have even suggested that it is irresolvable. This would be most unfortunate if true, since important theoretical and practical questions hang on it. Knowing whether some animals are mindreaders, for instance, is quite relevant to ethical questions regarding whether certain animals deserve special moral status and rights, to biological questions concerning the evolutionary origins and selection pressures leading to mindreading and related psychological abilities in humans, and to age-old questions in psychology and philosophy regarding the role of public language in the acquisition and use of abstract (non-observational) concepts.
Is there, then, really no chance of advancing the debate beyond its current state of stalemate and in a direction that might resolve it? Despite the way things may appear at the moment, I believe that the animal minreading debate can be resolved. In Mindreading Animals, I provide a comprehensive overview of the debate and, in an attempt to move it forward, defend a radically new approach to understanding mindreading in animals, called the appearance-reality mindreading (ARM) theory. This new way of thinking about mental state attribution in animals, I argue, escapes many of the standard objections leveled against the idea of animals as mindreaders (for the theory does not require animals to attribute inner causes or represent propositions) and opens up an entirely new way of testing for mindreading in animals that overcomes the methodological problem of controlling for Povinelli's clever behavior-reader hypothesis.
According to the ARM theory, mindreading evolved in the animal kingdom to provide animals with an advantage in predicting other agents' (e.g., predators, prey, competing conspecifics) behavior in illusory environmental settings. In illusory environmental settings, distal objects appear differently from what they really are, and being able to predict a naive agent's behavior in terms of how these objects appear to it (contrary to the way they really are) has the potential to contribute to an animal's overall level of fitness. Knowing, for example, that a competing conspecific mistakenly sees the distant piece of fruit on the forest floor as a small dark shadow (rather than as a piece of edible fruit), would enable an animal to predict that its competitor is not going to try to eat the fruit in this situation. Such a prediction would, in turn, provide the animal with a tactical reason to postpone trying to eat the fruit until its competitor has left - for trying to eat it now would only alert the competitor to the existence of the fruit (as such) and likely lead to an agonistic encounter that is best avoided.
Thus, the ARM theory hypothesizes that mental state attribution in animals evolved as a result of animals coming to introspect their own ability to distinguish appearance from reality, and then using this introspective knowledge to their benefit in predicting naive agents' behavior in illusory settings. At the moment, there is no direct empirical evidence supporting the theory other than the fact that chimpanzees, of all the species of animal that have been studied, have shown the strongest (though equivocal) evidence for mindreading and have also shown themselves capable of discriminating perceptual appearances from reality (see Krachun et al. 2009 "Can chimpanzees discriminate appearance from reality," Cognition, 112, pp. 435-450). To its credit, the ARM theory predicts the co-occurrence of these two cognitive abilities in chimpanzees, whereas other theories of primate social cognition do not. More direct methods of testing the theory are needed, of course. And in the book, I provide a handful of innovative protocols for directly testing the theory on a variety of animals (e.g., chimpanzees, monkeys, ravens, and dogs) and show that the protocols, unlike any currently used in the field, can successfully control for Povinelli's clever behavior-reader hypothesis.
The deeply contentious debate in science and philosophy over animal mindreading can be resolved, I argue. Actually resolving it, though, will depend upon researchers adopting the theoretical framework and empirical test procedures argued for in the book - and, of course, on the performance of the animals in such tests. And so regarding one's pet, the verdict is still out (at least, scientifically speaking) on whether he or she can truly know what you're thinking. As they say in television, 'stay tuned.'