For many species of animal, the ability to predict the behavior of others is vital to their well-being and reproductive success.1 In the field of animal social cognition, there are two generally recognized types of strategies that animals are understood to use to make such predictions. Behavior-reading is one type of strategy. This strategy involves predicting the behavior of others on the basis of observable cues that are perceived, believed, or otherwise represented to obtain without interpreting those cues as signs of underlying mental states (Lurz 2009, 2011; Povinelli and Vonk 2003). The observable cues can include bodily appearances (e.g., threatening posture), behaviors (e.g., reaching toward a particular object or place), and environmental relations (e.g., looking in the direction of a particular object or place); and the predictive process itself can be the result of individual learning or innate mechanisms. The other behavior-predicting strategy is mindreading (aka theory of mind). This strategy involves inferring others’ mental states, such as sensory experiences, desires, and beliefs, from represented observable cues, and using this information about others’ mental states to predict their behavior (Premack and Woodruff 1978). Here, too, the inferential and predictive processes involved may be the result of individual learning or innate mechanisms.2
Although mindreading and behavior-reading are different predictive strategies, they are not mutually exclusive. Humans are capable of both (Apperly and Butterfill 2009; Doherty 2011), and some animals are capable of behavior-reading (Lurz, Kanet and Krachun 2014). But are any animals capable of mindreading? And if some are, what sort of empirical tests would validly demonstrate this? This last question, in particular, has been a central question in animal social cognition research for over four decades. During that time, researchers have tested a range of animals on a number of different types of mindreading tests. Although the tests have varied along a number of dimensions (e.g., the apparatuses used or whether the tests involved a cooperative or competitive task), they all followed a standard methodology.
On the standard methodology, an animal A1 is given an experimental and a control test.3 In the experimental test, A1 is presented with an observable cue C that is a sign that another animal A2 is in some type of mental state M. For example, A1 might observe that A2 is looking at a piece of food on the ground, which is a sign that A2 sees the food.4 In the experimental test, there is also an expected behavior B of A2 that is contingent upon A2 being in the mental state M. The expected behavior, for example, might be that A2 will go for the food, since A2 sees it. The control test is just like the experimental test except that the observable cue C is absent, as well as A2’s mental state M and expected behavior B. For example, in the control test, A1 might observe that a piece of food on the ground is not in A2’s line of gaze (since it is hidden behind an opaque barrier), indicating that A2 cannot see the food and, as a result, will not go for it. On the standard methodology, if A1 successfully predicts that A2 will do B in the experimental test but not in the control test, and there is no evidence that A1 has learned to make such a prediction during the course of testing (e.g., by being rewarded for making such predictions), then A1 is taken to pass the mindreading test.
On some tests that employ the standard methodology, animals have failed (e.g., Call and Tomasello 1999; Povinelli and Eddy 1996), while on others they have passed (e.g., Hare, Call, Agnetta, and Tomasello 2000; Hare, Call, and Tomasello 2001). Does the fact that some animals have passed such tests provide sufficient grounds to believe that they are capable of mindreading? The answer, of course, depends upon the validity of the tests. If the tests are such that they could just as plausibly be passed by animals using a behavior-reading strategy, then they are not valid and passing them should not be taken as sufficient grounds for believing that the animal in question is capable of mindreading.
A handful of researchers (Heyes 1998; Perner 2008; Povinelli and Vonk 2003; Lurz 2009; Lurz and Krachun 2011) have argued that tests that employ the standard methodology are invalid. The reasons for this are that (a) the observable cues used in these tests are confounded with the mental states being investigated, and that (b) the design of the tests leaves open the reasonable possibility that the animals know in advance of the experimental test that these cues correlate with the type of behavior they are asked to predict. To illustrate, consider the example of the standard methodology above. In that example, A2’s line of gaze to the food (observable cue) is confounded with A2’s seeing the food (mental state). Furthermore, if A1 is like most social animals that compete for food, it is quite possible that A1 knows (either from past experience or innately) in advance of the experimental test that A2 (or animals like A2) typically go for food in their line of gaze. As a result of the confound and reasonable possibility mentioned, we cannot say whether A1’s successful prediction of A2’s behavior in the experimental test is due to A1 understanding that A2 sees the food (mindreading) or to A1 understanding that A2 (or animals like A2) typically go for food in their line of gaze (behavior-reading). This problem of experimentally ruling out such plausible behavior-reading explanations is what Povinelli and Vonk (2003) call the logical problem (aka Povinelli’s problem). Povinelli and Vonk, along with other researchers (Heyes 1998; Lurz 2009; Lurz and Krachun 2011), argue that empirically answering the question of whether animals mindread requires solving the logical problem, and that solving the logical problem requires designing tests that use a methodology that is fundamentally different from the standard methodology. It is important to note that the logical problem is presented as an instance of the rather common methodological problem in science of confounding variables; it is not presented as the insoluble problem of designing an experiment that can rule out every conceivable behavior-reading hypothesis (see Halina, Chapter 22 in this volume).5 Many fields of empirical research face similar problems of confounding variables, and there are different strategies that researchers employ to control for confounding variables. One such strategy is designing alternative test procedures in which the confounding variables are dissociated. This is the strategy that Povinelli, Vonk, and others recommend using to solve the logical problem.
In very general terms, it is perhaps not too difficult to see what this alternative test procedure might be. It would involve testing an animal A1 to see if the animal would predict that another animal A2 will perform some type of behavior B when a novel observable cue C is represented as obtaining, where the following conditions are satisfied:
If the no confounding cue condition is satisfied, then A1 will not be capable of using a behavior-reading strategy to successfully predict that A2 will do B in the experimental test. However, if the novel cue, learning, and prior knowledge conditions are satisfied, A1 could use a mindreading strategy to predict that A2 will do B. With A1’s prior knowledge that A2 (or others like A2) tend to perform B-type behaviors when in an M-type mental state, and with A1’s newly acquired knowledge that the novel observable cue C is a sign for the mental state M, A1 could infer that A2 is in the mental state M when C is represented as obtaining and is likely to perform a B-type behavior in the experimental test. Thus, with a test where these four conditions are satisfied, and where A1 succeeds in predicting that A2 will do B in the experimental test as a result of representing that observable cue C obtains, we can be confident that A1 is making this prediction using a mindreading strategy and not a behavior-reading one. Such a test, then, would be a valid mindreading test and would solve the logical problem. But what sort of test would this actually be?
A number of researchers have argued that an experience-projection test with transparent and opaque barriers would satisfy the four conditions above and solve the logical problem (Bugnyar, Reber, Buckner 2016; Heyes 1998; Karg, Schmelz, Call and Tomasello 2015; Povinelli and Vonk 2003). So far, just three such tests have been conducted – two with chimpanzees (Karg et al. 2015; Vonk and Povinelli 2011) and one with ravens (Bugnyar et al. 2016). The test with ravens is the most recent, and because of its rather elegant design and clear argumentation by the researchers, it provides an excellent illustration of the methodology of the experience-projection test and the reasons some researchers believe that such a test solves the logical problem.
In their study, Bugnyar and colleagues allowed a group of ravens (N = 10) to cache food under three different test conditions. In the window-open test, a (focal) raven was given food to hide in a caching room while another raven (competitor) watched from an adjacent room through an open window. After the focal raven hid its food, the competitor was let into the caching room and allowed to search for the hidden food in the presence of the focal raven. The window-open test was used as a baseline in which the focal ravens’ caching strategies would be compared to those used in the experimental and control tests (described below). If the ravens’ caching strategies in the window-open (baseline) test were found to be significantly similar to those used in the experimental test and significantly different from those used in the control test, the ravens would be taken to pass the mindreading test.
The window-closed test (control test) was just like the window-open test except that the window between the rooms was covered by an opaque panel which prevented the competitor from seeing the caching room. Results from the two tests showed that the focal ravens were significantly more likely to hide their food quicker and to return to cache sites less often in the window-open test than in the window-closed test. Such caching strategies, the researchers point out, make ecological sense, since food that is cached quickly and cache sites that are not returned to while competitors are watching are less likely to be noticed by competitors and, therefore, less likely to be pilfered later.
On the assumption that ravens are capable of mindreading and possess the mental state concept see, it is plausible to suppose, as the researchers do, that the focal ravens employ different cache strategies in these two tests because they understand that the competitor can see them caching in the window-open condition but not in the window-closed condition, and they know from prior experience that caches that are seen by competitors are less likely to be pilfered if the caching is done quickly and the cache site is not returned to while competitors are watching. And so it would appear that at this point, we can say that Bugnyar and colleagues’ study satisfies the prior knowledge condition.
After completing the window-open and window-closed tests, the focal ravens were given a familiarization trial. In the trial, the ravens were moved into the competitor’s room while the competitor was absent and allowed to look through a peephole cut into the panel that covered the window separating the competitor’s room and the caching room. While the raven peered through the peephole, an experimenter hid a piece of food in the caching room, after which the raven was allowed back into the caching room to find the hidden food. The objective of the trial was to introduce the focal ravens to a novel observable cue – the peephole – that, on the assumption that the ravens are capable of mindreading and possess the mental state concept see, they would reasonably interpret as a sign for seeing the caching room after their experience of looking through the peephole and seeing the caching room. And so it would appear that Bugnyar and colleagues’ study satisfies the learning condition, too.
Once the ravens passed the familiarization trial, they were given the peephole test (experimental test). In the test, the panel with the peephole covered the window separating the two rooms, and a focal raven was given food to hide in the caching room while the researchers played prerecorded sounds of a familiar competitor raven through a loudspeaker in the competitor’s room. After the focal raven hid its food, the researcher stopped playing the recording and allowed the actual competitor raven, from whom the sounds were recorded, to enter the caching room and search for hidden food in the presence of the focal raven.
According to the researchers, the peephole in the experimental test is a novel observable cue, since the ravens “lack a specific associative history of caching in the presence of peepholes” (Bugnyar et al. 2016: 4). That is, the focal ravens have had no experience, prior to the peephole test, with competitor ravens pilfering food from caches that were made behind opaque barriers with peepholes, and thus it is unlikely that the ravens could have any knowledge prior to the experimental test about how competitors would behave toward caches made behind an opaque barrier with a peephole.8 Thus, according to the researchers, the peephole test satisfies the novel cue condition.
The researchers also claim that there is no confounding cue in the peephole test, since there is “no actual competitor whose gaze could be read” (Bugnyar et al. 2016: 3). Without an actual competitor raven present in the peephole test, the researchers maintain, there is no confounding observable cue, such as the line of gaze of the competitor raven, that the focal ravens could plausibly be taken to represent and use to predict pilfering behavior by the competitor similar to that observed in the window-open test. And so, according to the researchers, the no confounding cue condition is also satisfied.
On the assumption that the researchers’ arguments above are sound, it would appear that their experience-projection test satisfies the four conditions of a valid mindreading test and, therefore, solves the logical problem. Since the peephole test supposedly satisfies the no confounding cue condition, it is unlikely that the focal ravens, were they behavior-readers, would be able to predict similar kinds of behavior from the (real/imagined) competitor in both the peephole and open-window tests, since there is (apparently) no common observable cue in these two test conditions that the focal ravens could represent and use to predict similar kinds of behavior from the competitor. But this is not unlikely if the ravens are mindreaders and possess the mental state concept see. For in both tests, the ravens could infer that the (real/imagined) competitor can see the caching room – an inference which, in the peephole test, could be based on the knowledge learned in the familiarization trial that peepholes afford seeing the caching room and which, in the window-open test, could be based on prior knowledge that competitors that have a line of gaze to the caching room can see the caching room. The ravens could then apply their prior knowledge that caches that are seen by competitors are less likely to be pilfered if the caching is done quickly and the cache site is not revisited to undertake similar caching strategies in both the window-open and peephole tests, which is precisely what the ravens did. In both the window-open and peephole tests, the average duration to caching a piece of food and the average number of returns to cache sites were nearly the same and significantly lower than in the window-closed test. From these results, the researchers concluded that
ravens treat the [peephole] test condition like the [window-open] test condition, indicating that they can generalize from their own experience using the peephole as a pilferer [in the familiarization trial] and predict that audible [imagined] competitor could potentially see their caches.
(Bugnyar et al. 2016: 3, emphasis added)9
A number of researchers have argued that the visual experience-projection test with transparent and opaque barriers does not solve the logical problem precisely because it fails to satisfy the no confounding cue condition (Andrews 2005; Hurley and Nudds 2006; Lurz 2009, 2011; Perner 2012). I am afraid that Bugnyar and colleagues’ study is no different on this score from earlier visual experience-projection tests using transparent and opaque barriers. Contrary to what the researchers claim, the fact that there is no actual competitor in the peephole test does not mean that there is no confounding observable cue, such as line of gaze, that the ravens could reasonably be taken to represent and use to predict similar behavior from the (real/imagined) competitor in both the peephole and window-open tests. Although there was no actual competitor in the peephole test, the focal ravens were made to think that there was and, therefore, they might well have thought that this (imagined) competitor could potentially have a line of gaze to the caching room through the peephole. If asked why the focal ravens might think this, the answer is that the ravens could have learned in the familiarization trial that peering through the peephole affords a line of gaze to the caching room and use this knowledge to infer that the (imagined) competitor could potentially have a line of gaze to the caching room if it peered through the peephole.10 This explanation incidentally is analogous to, and thus no less plausible than, the one that Bugnyar and colleagues give to the similar question: why might the focal ravens think that the (imagined) competitor could potentially see the caching room, given that there was no actual competitor for the ravens to observe? The researchers’ answer is that the focal ravens, from their experience with peering through the peephole, learn that the peephole affords seeing the caching room and use this knowledge to infer that the (imagined) competitor could potentially see the caching room if it peered through the peephole.
Therefore, it appears that the focal ravens could have just as easily passed Bugnyar and colleagues’ test if they had used a behavior-reading strategy. Due to their past experience with caching food in competitive contexts, it is plausible that the focal ravens had knowledge prior to the peephole test that caches that competitor ravens have a line of gaze to are less likely to be pilfered if the caching is done quickly and the cache site is not returned to; and they could use this prior knowledge, together with the knowledge they learned in the familiarization trial, that the peephole affords a line of gaze to the caching room, to predict the same kind of behavior from the (imagined/real) competitor in the peephole and window-open tests.
Designing a mindreading test for animals in which all of the four conditions above are satisfied is difficult but not impossible (see Lurz 2009, 2011; Lurz and Krachun 2011). Although I do not believe that Bugnyar and colleagues’ study solves the logical problem, I do believe that a modified version of it can. What is needed is an experimental test in which the competitor can see but does not have a line of gaze to the caching room. One way to achieve this is through the use of mirrors, since mirrors allow one to see things, such as one’s face or the room behind one’s head, that one does not or cannot have a line of gaze to.11
Let us imagine, then, that the ravens are given the following three tests. The window-open test (baseline) is just like the one given to the ravens in Bugnyar and colleagues’ study in which the competitor can easily peer through the window and into the caching room while the focal raven hides its food. The window-up test (control), however, involves placing the window high up on the wall separating the competitor’s room and the caching room so that the competitor cannot peer through the window and into the caching room. In both tests, the focal raven is given food to hide in the caching room while a competitor is in the adjacent room, after which the competitor is released into the caching room to find the hidden food in the presence of the focal raven.
After taking these two tests, the ravens would be given a mirror familiarization trial. In this trial, the focal ravens are transferred to the competitor’s room while the competitor is absent. In the room, the window is high up on the wall, preventing the focal ravens from peering through the window and into the caching room. However, a mirror is placed high up on the wall opposite to the window and angled downward, allowing the focal ravens to see the caching room when they look at the mirror.12
After the mirror familiarization trial, the ravens are given the mirror test (experimental test). In this test, the focal ravens are given food to hide in the caching room while the window is high up on the wall and the competitor is in the adjacent room. In addition, the mirror is placed in the competitor’s room, as it was in the familiarization trial. The placement of the mirror performs two functions. It allows the competitor to see the caching room by looking at the mirror, and it allows the focal raven in the caching room to see the mirror through the open window. Since the window is placed high up on the wall, just like in the window-up test, the focal ravens can see that the competitor is prevented from having a line of gaze to the caching room. Furthermore, during the mirror familiarization trial, the focal ravens were given no reason to think that looking at the mirror affords a line of gaze to the caching room. When the ravens looked at the mirror, it was the mirror, not the caching room, to which they had a line of gaze. What the mirror allowed the focal raven to do was to see the caching room, not to have a line of gaze to the caching room. Thus, unlike Bugnyar and colleagues’ peephole test, it is not plausible to suppose that the focal ravens in the mirror test might think that the competitor in the adjacent room could potentially have a line of gaze to the caching room. On the reasonable assumption that line of gaze is the only confounding cue that the focal ravens might plausibly be understood to represent and use to predict the same kind of behavior from the competitor in the window-open and mirror tests, it would appear that the mirror test satisfies the no confounding cue condition.
Without a confounding observable cue for the focal ravens to represent in the mirror test, they cannot employ a behavior-reading strategy to predict the same kind of behavior from the competitor in the window-open and mirror tests. But the ravens could make such a prediction if they are capable of mindreading and possessed the mental state concept see. For in both tests, the ravens could infer that the competitor in the adjacent room can see the caching room – an inference which, in the mirror test, could be based on the knowledge learned in the mirror familiarization trial that looking at the mirror affords seeing the caching room and which, in the window-open test, could be based on prior knowledge that competitors that have a line of gaze to the caching room can see the caching room. The ravens could then apply their prior knowledge, that caches that are seen by competitors are less likely to be pilfered if the caching is done quickly and the cache site is not revisited, to undertake similar caching strategies in both the window-open and mirror tests.
Thus, there are ways of designing a valid mindreading test for animals – the logical problem has a solution. Yet these types of tests have not been used to assess animals’ mindreading capacities, and therefore we do not know whether animals are capable of mindreading. Until such tests are used and animals pass them, we should remain agnostic – though, optimistic – about the possibility of animal mindreading (Lurz, Kanet and Krachun 2014).
1 Throughout, ‘animal’ is used to stand for nonhuman animals.
2 Vincent and Gallagher (Chapter 26 of this volume) put forward a third type of strategy, the interaction theory, which holds that chimpanzees predict the behavior of others by directly perceiving their mental states. Interaction theory and mindreading agree that chimpanzees represent the mental state of others; they disagree over whether such representations take the form of perception or inferred belief. In this essay, I follow tradition and present the ‘logical problem’ as existing between behavior-reading and mindreading accounts of animal social behavior. The problem could just as well be presented as existing between behavior-reading and the interaction accounts (Cf. Gallagher and Povinelli 2012).
3 Typically, groups of animals are tested. However, for easy of explaining the standard methodology, I use an individual animal.
4 ‘Line of gaze to’ and ‘looking at’ are used throughout as synonyms for the observable spatial relation that holds between a subject’s eyes and non-occluded objects in front of the subject’s eyes. It is important to note that line of gaze/looking at is not seeing. Seeing is a state of awareness and, thus, a mental state; line of gaze/looking at is a spatial relation, not a mental state. Although line of gaze/looking at is not seeing, it is an important observable cue used to infer what someone is seeing.
5 After all, most possible behavior-reading hypotheses are not even antecedently plausible and, thus, do not need to be ruled out by a test procedure.
6 It is important to note that A1 does not learn that C is a sign for M by learning that C is correlated with a type of behavior B that A1 knows to be caused by M. Rather, A1 must learn that C is a sign for M by learning, via introspection, that C correlates with A1’s own mental state M.
7 That is, A1 has no reason other than what A1 can infer from what it has learned in the learning condition about the relation between C and M, and what it supposedly knows from the prior knowledge condition about the relation between M and behavior B in others.
8 It is also unlikely that the ravens might instinctively know how competitors would behave in such a condition, given the novelty of the peephole situation.
9 In contrast to this mindreading proposal, Bugnyar and colleagues at one point argue for the more “ecumenical proposal” that the focal ravens attribute an “intervening variable” to the competitor (Cf. Whiten 1996). On this proposal, the focal ravens are hypothesized to expect similar types of behavior from competitors in the window-open and peephole tests because they attribute a common intervening variable that they understand to cause such behaviors in “perceptually dissimilar situations” (p. 4). The researchers argue that since the window-open and peephole tests are perceptually dissimilar, the intervening variable proposal offers a better account of the focal ravens’ behavior than any behavior-reading proposal. The researchers are mistaken, however, that the window-open and peephole tests are perceptually dissimilar, or so I argue. If my argument is correct, their study not only fails to provide convincing evidence that ravens are mindreaders rather than behavior-readers, but also that ravens are attributors of intervening variables rather than behavior-readers.
10 Previous studies have shown that ravens are capable of representing others’ line of gaze (Bugnyar, Stöwe and Heinrich 2004; Schloegel, Kotrschal and Bugnyar 2007). It is quite plausible, therefore, that the focal ravens possess the concept line-of-gaze and use it to represent this spatial relation holding between their own eyes and the caching room when they peer through the peephole.
11 Recall that line of gaze is the observable spatial relation that one bears to non-occluded objects in front of one’s eyes. Hence, by looking into a mirror, one does not have a line of gaze to one’s face or the room behind one’s head, since these non-occluded objects are not in front of one’s eyes.
12 Ravens are corvids and some corvids (e.g., magpies and crows) have been shown to understand the reflective properties of mirrors (Medina, Taylor, Hunt and Gray 2011; Prior, Schwarz and Güntürkün 2008). It is plausible, then, that the ravens, upon looking at the mirror, take themselves to be seeing the real caching room behind them and not some virtual caching room behind the mirror.
K. Andrews, Do Apes Read Minds? (Cambridge, MA: MIT Press, 2012) is an excellent book on the question of whether animals have a theory of mind and the different uses of theory of mind in humans and animals. T. Suddendorf, The Gap: The Science of What Separates Us From Other Animals (New York: Basic Books, 2013) is an equally excellent book on theory of mind in animals as well as related questions of self-recognition and mental time travel in animals.
Andrews, K. (2005) “Chimpanzee Theory of Mind: Looking in All the Wrong Places?” Mind & Language 20: 521–536.
Apperly, I., and Butterfill, S. (2009) “Do Humans Have Two Systems to Track Beliefs and Belief-Like States?” Psychological Review 4: 953–970.
Bugnyar, T., Reber, S., and Buckner, C. (2016) “Ravens Attribute Visual Access to Unseen Competitors.” Nature Communications 7: 10506. doi:10.1038/ncomms10506.
Bugnyar, T., Stöwe M., and Heinrich, B. (2004) “Ravens, Corvus corax, Follow Gaze Direction of Humans Around Obstacles.” Proceedings of the Royal Society B 271: 1331–1336.
Call, J., and Tomasello, M. (1999) “A Nonverbal False Belief Task: The Performance of Children and Great Apes.” Child Development 70: 381–395.
Doherty, M. (2011) “A Two-System Theory of Social Cognition.” In J. Roessler, H. Lerman and N. Eilan (eds.), Perception, Causation, and Objectivity. Oxford: Oxford University Press.
Gallagher, S., and Povinelli, D. (2012) “Enactive and Behavioral Abstraction Accounts of Social Understanding in Chimpanzees, Infants, and Adults.” Review of Philosophy and Psychology 3: 145–169.
Hare, B., Call, J., Agnetta, B., and Tomasello, M. (2000) “Chimpanzees Know What Conspecifics Do and Do Not See.” Animal Behavior 59: 771–785.
Hare, B., Call, J., and Tomasello, M. (2001) “Do Chimpanzees Know What Conspecifics Know?” Animal Behavior 61: 139–151.
Heyes, C. (1998) “Theory of Mind in Nonhuman Primates.” Behavioral and Brain Sciences 21: 101–148.
Hurley, S., and Nudds, M. (2006) Rational Animals? Oxford: Oxford University Press.
Karg, K., Schmelz, M., Call, J., and Tomasello, M. (2015) “The Goggles Experiment: Can Chimpanzees Use Self-Experience to Infer What a Competitor Can See?” Animal Behavior 105: 211–221.
Lurz, R. (2009) “If Chimpanzees Are Mindreaders, Could Behavioral Science Tell? Toward a Solution to the Logical Problem.” Philosophical Psychology 22: 305–328.
Lurz, R. (2011) Mindreading Animals: The Debate Over What Animals Know About Other Minds. Cambridge, MA: MIT Press.
Lurz, R., Kanet, S., and Krachun, C. (2014) “Animal Mindreading: A Defense of Optimistic Agnosticism.” Mind & Language 29: 428–454.
Lurz, R., and Krachun, C. (2011) “How Could We Know Whether Nonhuman Primates Understand Others’ Internal Goals and Intentions? Solving Povinelli’s Problem.” Review of Philosophy and Psychology 2: 449–481.
Medina, F., Taylor, A., Hunt, G., and Gray, R. (2011) “New Caledonian Crows’ Responses to Mirrors.” Animal Behaviour 82: 981–993. doi:10.1016/j.anbehav.2011.07.033.
Perner, J. (2008) “Who Took the Cog Out of Cognitive Science? Mentalism in an Era of Anti-Cognitivism.” In P. A. Frensch and R. Schwarzer (eds.), International Congress of Psychology: 2008 Proceedings. Hove: Psychology Press.
Perner, J. (2012) “Mini-Meta: In Search of Minimal Criteria for Metacognition.” In M. Beran, J. Brandl, J. Perner and J. Proust (eds.), The Foundations of Metacognition. Oxford: Oxford University Press.
Povinelli, D., and Eddy, T. (1996) “What Young Chimpanzees Know About Seeing.” Monographs of the Society for Research in Child Development 61: 1–152.
Povinelli, D., and Vonk, J. (2003) “Chimpanzee Minds: Suspiciously Human?” Trends in Cognitive Sciences 7: 157–160.
Premack, D., and Woodruff, G. (1978) “Does the Chimpanzee Have a Theory of Mind?” Behavioral and Brain Sciences 1: 515–526.
Prior, H., Schwarz, A., and Güntürkün, O. (2008) “Mirror-Induced Behavior in the Magpie (Pica pica): Evidence of Self-Recognition.” PLoS Biology 6: e202. doi:10.1371/journal. pbio.0060202.
Schloegl, C., Kotrschal, K., and Bugnyar, T. (2007) “Gaze Following in Common Ravens (Corvus corax): Ontogeny and Habituation.” Animal Cognition 74: 769–778.
Vonk, J., and Povinelli, D. (2011) “Social and Physical Reasoning in Human-Reared Chimpanzees: Preliminary Studies.” In J. Roessler, H. Lerman and N. Eilan (eds.), Perception, Causation, and Objectivity. Oxford: Oxford University Press.
Whiten, A. (1996) “When Does Smart Behavior-Reading Become Mind-Reading?” In P. Carruthers and P. Smith (eds.), Theories of Theories of Mind. Cambridge: Cambridge University Press.