Given this book’s title, which includes both ‘mindreader’ and ‘Theory of Mind’, you might think that it is about children and their developing ability to understand others as bearers of mental states (beliefs, desires, intentions, perceptions etc.). In light of the fact that most research on “mindreading” among humans has been carried out with children as subjects, this would be a reasonable bet. While this book discusses mindreading research on children in some detail, it also presents innovative experimental research on adult subjects (often conducted by Apperly and his collaborators). In the Introduction, Apperly tells us that he has often been asked by fellow researchers why he is studying adults. After all, “[d]on’t they already have a theory of mind?”. This question reveals both a narrow focus on the issue of when humans become mindreaders, and a way of thinking about this developmental milestone in terms of a child’s acquisition of a new concept or “folk theory”.

The aim of mindreading research in developmental psychology has typically been to find out when children pass some version of a “false-belief task”, the litmus test of whether a child has acquired the concept of ‘belief’, or has a “representational theory of mind”. In one well-known version of this test, subjects watch two characters (puppets or cartoon figures) who are standing in front of two boxes. Let us call the characters Andy and Sally. In Sally’s plain sight, Andy first puts an object, a ball say, in the right box. Sally then leaves the scene and while Sally is away, Andy transfers the ball to the left box. Finally, Sally returns and the subject of the experiment is asked where Sally will look for the ball. The “correct” answer, of course, is that Sally will look in the right box since this is where she (falsely) believes it is. Before 3 to 5 years of age, children fail to make this prediction, however, thinking that Sally will look in the left box where they know the ball is located, which indicates that they lack the concept of belief. The results of this strand of research is certainly interesting (even if it remains controversial how the results should be interpreted), but as Apperly rightly points out, they do not tell us much about how mindreading is achieved. In Mindreaders, Apperly shifts the focus to the question of how it is possible for humans to mindread, and to what the cognitive basis for this ability is — the architecture, representations and processes underpinning it.

While we ascribe many types of mental states to others in interpreting, explaining or predicting their behavior (e.g. intentions, desires, emotions and character traits), Apperly restricts his scope to the ascription of epistemic states (that is, states such as beliefs and perceptions which have a mind-to-world direction of fit). The bulk of the book, five of its seven chapters, consists of a systematic survey of mindreading research focused on the ascription of such states. Four chapters summarizes experimental research on mindreading in children (ch. 2), infants and non-human animals (ch. 3), and adults (ch. 5), and one chapter goes through neuroimaging research on, and neuropsychological case studies of, mindreading in adults (ch. 4). The experimental studies conducted by Apperly and his collaborators are primarily presented in chapters 5 and 7. The survey sets the stage for the last two chapters where Apperly presents his own “two-systems” account of mindreading (developed together with the philosopher Stephen Butterfill, see Apperly & Butterfill 2009). He characterizes this as “a very simple and somewhat speculative sketch that […] provides structure for further investigation.” (p. 133)

On the two-systems account, adult human beings’ mindreading abilities rely on two systems, a low-level system that operates automatically and efficiently, and a high-level system is more flexible but also demanding on our limited executive resources. The low-level system is shared with infants and some non-human animals, whereas the high-level system is probably unique to human beings. The main motivation for this two-systems account comes the fact that empirical findings as well as theoretical considerations put conflicting demands on the cognitive basis of mindreading. Research on both children and adult subjects shows that mindreading depends on cognitive control, executive function and working memory. But at the same time, an equally robust finding is that some mindreading competence seems to be wholly independent of executive function. For example, while children’s performance on tests of executive function predicts their later performance on ordinary verbal false belief tasks, infants with only very rudimentary executive function reliably form correct expectations about another agent’s behavior that are sensitive to the agent’s false beliefs. Apperly emphasizes that his two-systems account isn’t merely a description of the dual demand profile of efficient-but-inflexible and inefficient-but-flexible mindreading. By appealing to his own research, he suggests that there are specific “signature limits” on the type of contents that are processed by the low-level system. In chapter 7, Apperly presents results that show that while both adults and children appear to automatically compute which objects an observed perceiver sees (a Level-1-perspective), they do not automatically compute in what way these objects are seen by the perceiver (a Level-2-perspective). (However, Apperly argues that what is automatic is not merely the computation of the observed perceiver’s line of sight, but also attribution of mental content to the perceiver.) The automatic perspective-taking system thus appears to be sensitive to a restricted type of inputs and only performs a limited range of operations on that input.

Besides accommodating an unusually wide range of empirical research, the two-systems account is motivated by theoretical reflection on the so-called “frame problem” of “problem of relevance”. In a mindreading context, this is the problem of constraining which aspects and opportunities of another agent’s situation that may be relevant for interpreting or predicting his or her behavior. Take the example of Andy and Sally in the false-belief task. Apperly points out that the fact that it seems so obvious to us that Sally (if she is a rational agent) will look in the right box where she falsely believes it is, obscures that this prediction is only obvious given a host of taken-for-granted assumptions. Just because Sally saw Andy put the ball in the right box and wants to retain it doesn’t actually mean that it is rational for Sally to look in the right box. For example, perhaps Andy has been mischievous in the past, and has a habit of removing objects from where he thinks Sally thinks they are. Perhaps Sally is herself in a mischievous mood and wants to fool Andy into thinking that she has just gone mad, so she decides to look in the left box first even though he she actually believes it is in the right box. Or perhaps Sally falsely believes that the two boxes are the two ends of a” teleporter” that automatically transfers objects between the boxes. Perhaps Sally is very curious about what might be in the left box, so despite thinking it very likely that the ball is still in the right box, she decides to look in the left box first (that way, she gets to satisfy her curiosity while also finding the ball in case Andy has moved the ball to the left box, a possibility which she doesn’t rule out). What it makes sense for Sally to do does not simply depend on one belief-desire pair but on a whole web of beliefs and desires. The take home message is that without any constraints on how to interpret Sally’s situation and her likely behaviour, the answer to the question of where Sally will look for the ball is computationally intractable.

The signature limits of the low-level system allow this computational complexity to be avoided, and Apperly suggests that is handled by the high-level system by reliance of on general background knowledge about behavioral norms, typical situations, social roles, scripts etc. According to Apperly, such knowledge serves two functions, it makes mindreading unnecessary in many situations (to successfully order a meal in a restaurant, you do not need to ascribe beliefs and desires to the waiter) and it makes it computationally tractable in other situations, by constraining the range of possibilities that are deemed to be relevant. What allows high-level mindreading to actually work then, on Apperly’s account, is that mindreaders typically have much in common with those they interact with on a daily basis: language, culture, habits and norms.

While the range of empirical research that Apperly draws on is broad, it should be noted that his approach is that of a mainstream cognitive psychologist. Almost all of the empirical research that is discussed involve subjects who, from a disengaged “third-person standpoint” observe representations of other agents in the captivity of the laboratory (in the form of short narratives that are written, presented as cartoons, enacted by puppets, are perhaps presented in pictures or with videos of actual people). While Apperly emphasizes the role of social interaction in creating a convergence of social expectations which enables mindreading to get off the ground in the face of computational complexity, mindreading “in the wild” often occurs in the midst of social interaction, and perhaps processes of mindreading may even be constituted partly by social interaction (see, e.g., De Jaegher et al. 2010). The targets of mindreading are not like the targets of ordinary reading, they respond to our interpretations of their behavior, and we, in turn, respond back. In Apperly’s defense, he does not equate mindreading with social cognition or social understanding in general — these are broader concepts — so he probably wouldn’t disagree that social interaction has this synchronic role as well.

Much of this book’s value lies in the details. The fascinating surveys of empirical work that make up the first five chapters are shot through with interesting theoretical and methodological reflections. As someone with only a cursory familiarity with one or two corners of the vast mindreading literature, I found this part of the book very rewarding. As far as I can see, the two-systems “sketch” given at the end of the book is well supported in light of Apperly’s review of the landscape of empirical findings, and accomplishes Apperly’s intention of providing a “structure for further investigation.” Mindreaders is of a terrific blend of rich empirical detail and sophisticated theory-building. I would highly recommend it to anyone who has an interest in the science of human social cognition.

References

Apperly, Ian & Stephen Butterfill (2009). “Do Humans Have Two Systems to Track Beliefs and Belief-Like States?”, Psychological Review, 116(4):953–970.

De Jaegher, Hanne, Ezequiel Di Paolo & Shaun Gallagher (2010). “Does social interaction constitute social cognition?”, Trends in Cognitive Sciences, 14(10):441-447.

© 2011 Olle Blomberg

Olle Blomberg is a Ph.D. student in Philosophy at University of Edinburgh (UK) and a freelance journalist. He is interested in the philosophy of social and cognitive science, the philosophy of technology, as well as science and technology journalism. For information about his freelance writing, see http://www.olleblomberg.com/english.html. Information about his Ph.D. research can found on his University of Edinburgh web page [http://www.philosophy.ed.ac.uk/postgraduate/students/phd/OlleBlomberg.html]