If teleology is the answer, what was the question?
Josef Perner is one of the leading developmental psychologists of mindreading (or mental state attribution). His contribution to the subject, including his influential (1991) book, has been huge. It started back in 1983, when he and Heinz Wimmer reported the very first study of false-belief attribution in human childhood: Maxi and Maxi’s mother (a pair of puppets) store some desirable object (e.g. a chocolate bar) in a blue cupboard for later use and Maxi leaves the scene. While Maxi is away, Maxi’s mother removes the chocolate bar, uses some of it to bake a cake, places it into a green cupboard and leaves. Maxi comes back and wants some chocolate. Participants are asked the prediction question: “Where will Maxi look for his chocolate?”
Wimmer & Perner (1983) found that 4-5-year-olds were below chance in answering the prediction question. Subsequent studies showed that on average most 4,5-year-olds correctly point to the empty location where a mistaken agent believes her desired object to be, but most preschoolers (3-year-olds on average) incorrectly point to the object’s actual location (cf. Wellman et al., 2001). A few years later, Perner was also the first to report that most 3-year-olds, who incorrectly point to the object’s actual location when directly asked to predict the likely action of a mistaken agent, nonetheless correctly look at the empty location in anticipation of the agent’s action (cf. Clements & Perner, 1994).
Clements and Perner’s (1994) insight was further corroborated by novel evidence based on non-verbal tests, such as participants’ looking behavior, helping behavior and brain responses. This evidence strongly suggests that preverbal infants, who cannot understand, let alone respond to, a direct linguistically encoded question, have expectations about an agent’s likely action based on their representation of the content of the agent’s belief (whether true or false). For example, Southgate et al. (2007) used an eye-tracker and found that 2-year-olds correctly gaze at the empty location in anticipation of a mistaken agent’s action (thereby confirming Clements & Perner’s earlier observation). In their landmark study based on the violation-of-expectation, Onishi & Baillargeon (2005) found that while an agent was facing a pair of boxes in front of them, 15-month-olds looked longer whenever the agent did not reach for the box where she either truly or falsely believed her desired toy to be (cf. Baillargeon et al., 2016 for review of much other evidence based on non-verbal tests).
When Onishi & Baillargeon’s (2005) study appeared, Josef Perner offered a deflationary rebuttal to the mentalistic interpretation of the findings, published in the very same issue of Science (cf. Perner & Ruffman, 2005). He has explored several non-mentalistic interpretations of the infant findings, but on the whole he has remained faithful to his skepticism about the reliability of the evidence for false-belief attribution in human infancy based on non-verbal tests (cf. Perner, 2010).
There are presently over thirty studies that report evidence for false-belief attribution in children from 6 to 36 months of age. New studies of false-belief attribution in both infancy and early childhood keep cropping up (cf. Tauzin & Gergely, 2017; Király et al., 2018). A recent study by Burnside et al. (2019) reports that 16-month-old infants attribute false beliefs to a toy crane — which could be taken either to show that whatever they do infants do not really attribute false beliefs to others (because crane toys don’t have beliefs) or on the contrary that, just like adults, they tend to over-attribute false beliefs to other agents, including a toy crane. However, in the past couple of years, several failed replications of some of the studies of early false-belief attribution based on non-verbal tests have also been reported. Thus, one of the main questions currently faced by the study of mindreading in human childhood is: should findings based on non-verbal false-belief tests be regarded as reliable evidence for false-belief attribution in human infancy? Or else should only findings based on verbal false-belief tasks be regarded as reliable evidence?
Recently, Perner and several of his collaborators (in particular the philosopher Johannes Roessler) have proposed to provide a principled strategy to answer the question why it is at best unlikely that findings based on non-verbal tests could be reliable evidence for false-belief attribution — let alone by preverbal human infants. The strategy is to give primacy to something they call teleology over mindreading (i.e. mental state attribution). Teleology has a rich pedigree in the philosophy of mind and action and it is supposed to capture human adults’ commonsense or folk psychological understanding of intentional actions, i.e. what it takes to understand what an agent does when he or she does it for a reason.
Teleology is a modified version of the philosopher Donald Davidson’s (1963, p. 685) picture of what he took to be “the ancient — and commonsense — position that rationalization is a species of causal explanation.” In short, a rationalization in Davidson’s (1963) sense is both a causal explanation and a reason explanation of an agent’s intentional action: it shows how the agent’s reason for acting is also the cause of the agent’s action. What made Davidson’s (1963) picture of rationalizing so appealing to philosophers is that it seems to provide a framework for reconciling the puzzling ontology of reasons and the naturalistic ontology of causes. What Perner and collaborators call teleology-in-perspective (or sophisticated teleology) matches Davidson’s picture of rationalization.
Perner and collaborators’ fundamental strategy is to give developmental primacy to the teleological understanding of intentional action over the mindreading capacity to attribute true and especially false beliefs to others. To justify their strategy, Perner and collaborators make two related fundamental psychological assumptions. First, following Davidson, they assume that teleology-in-perspective is what captures human adults’ mature folk psychological understanding of intentional action, i.e. what an agent does for a reason. Secondly, they assume that teleology-in-perspective is a later mentalistic developmental stage of understanding intentional actions, which is preceded by an earlier non-mentalistic developmental stage, which they call pure teleology (cf. Perner & Roessler, 2010; Perner & Esken, 2015; Perner et al., 2018; Priewasser et al., 2012; Roessler & Perner, 2013). What is the difference between teleology-in-perspective and pure teleology?
Teleology is supposed to enable humans to understand that what is characteristic of an intentional action is that it is something done for a reason: teleological explanations are reason explanations of intentional actions. Teleology-in-perspective enables older children and human adults to grasp the distinction between an agent’s objective reason and her subjective reason for acting, and to thereby rationalize the agent’s intentional action. The agent’s objective reason is a non-mental fact that supports the agent’s decision to perform the action. The agent’s subjective reason is the agent’s accurate or inaccurate mental representation of what the agent takes to be her objective reason for acting. If the agent’s subjective reason represents what turns out not to be an objective reason for acting, then objective and subjective reasons diverge. Pure teleology is non-mentalistic: it is supposed to enable young children to understand the agent’s objective reason for acting, at the expense of her subjective reason. Pure teleologists are supposed to explain an agent’s intentional action exclusively by reference to the non-mental fact that constitutes her objective reason.
The developmental primacy of non-mentalistic teleology over mindreading has the following fundamental consequence. Children could not ascend from pure non-mentalistic teleology to mindreading, as it is embedded within teleology-in-perspective, unless they could reason by supposition and entertain counterfactual propositions, both which are necessary to understand what Roessler & Perner (2013, pp. 38-40) call introduction and elimination rules for the concept BELIEF — which in turn they take to be necessary conditions for genuine mindreading. For example, if Maxi’s mother had not removed the chocolate bar from the blue cupboard, i.e. if it still were in the blue cupboard, then Maxi’s belief that the chocolate is in the blue cupboard would be justified. Conversely, if Maxi’s belief that the chocolate is in the blue cupboard were true, then Maxi would have good reasons to look for the chocolate in the blue cupboard (cf. Perner et al., 2018, p. 107). In short, one could not attribute beliefs to others unless one could understand how both beliefs are justified and how beliefs justify an agent’s action.
On the one hand, Perner and Roessler’s teleology-first strategy makes it highly unlikely at best that preverbal infants could attribute beliefs to others if belief attribution requires mastery of the concept BELIEF. On the other hand, Perner and Roessler’s teleology-first strategy provides them with a plausible answer to the question: why do most preschoolers incorrectly point to the object’s actual location when asked to predict the mistaken agent’s likely action (cf. Perner & Roessler 2010; 2012; Perner et al., 2018; and Roessler & Perner, 2013)? As pure teleologists, preschoolers explain what an agent does by appeal to her objective reason to act, i.e. the fact about the relevant object’s actual location, not by the agent’s subjective reason, i.e. her misrepresentation of the object’s location. Older children and adults point to the empty location because as teleologists-in-perspective, they can draw the distinction between the mistaken agent’s objective reason and her subjective reason for acting, and they understand that what is relevant to answering the question is the agent’s subjective reason (because it is also the cause of the agent’s action).
The teleology-first strategy faces, in my view, two basic problems, the first of which is the least psychological, and the most philosophical, of the two. First, it is very unclear whether a pure teleological explanation, i.e. an entirely fact-based non-mentalistic explanation, of an agent’s intentional action, could be construed as a viable causal explanation. Only an agent’s occurrent mental state (or event) can be a cause of the agent’s action. A non-mentally represented fact can justify an agent’s action (or decision to act), but the former cannot cause the latter. This is why Davidson required that in a rationalization of an agent’s action the agent’s “primary” reason for acting be identified as the cause of the agent’s action, namely the relevant pair of the agent’s belief and desire. A pure non-mentalistic teleological explanation of an agent’s action is committed to the assumption that the agent’s action is based on her objective reason. Thus, it is at least an open question whether a pure fact-based teleological explanation can be a genuine causal explanation of an agent’s action.
Perner & Roessler (2010, pp. 208-209) gesture towards an answer to this objection by endorsing a “minimalist” account according to which a counterfactual dependence between a pair of facts is a sufficient condition for making an explanation a causal explanation of one fact by the other. The reason why I am not convinced is that there are clearly counterfactual dependencies between pairs of logical and mathematical facts. But I think that such counterfactual dependencies should not be construed as instances of causal explanations of one logical or mathematical fact by another.
Secondly, what children are directly asked to do in standard versions of false-belief tasks of the kind first designed by Wimmer & Perner (1983) is to answer the prediction question: “Where will Maxi look for the chocolate?” They are asked to predict a forthcoming action that has not yet been executed. They are not asked a why-question: they are not asked to explain why something happened, e.g. why Maxi’s action failed to fulfill his desire. Only if Maxi had in fact unsuccessfully looked for the chocolate in the blue cupboard would it make sense to ask children to explain his failure to find the chocolate there. Arguably the duality between Maxi’s objective reason to look for his chocolate at its actual location and his subjective reason to look for it where he falsely believes the chocolate to be is relevant to explaining why Maxi failed to find the chocolate when he looked for it where he falsely believed it was. Arguably this duality might also be relevant if children were further challenged to justify their prediction. But in standard verbal false-belief tasks, children are neither asked to explain a mistaken agent’s failed action, nor challenged to justify their prediction; they are merely asked to predict a mistaken agent’s likely action.
To endorse the teleology-first strategy is to assume that understanding practical and epistemic reasons and justifications precedes and makes possible mindreading and mental state attribution. If so, then belief attribution derives from and depends on grasping what justifies beliefs and which actions an agent’s beliefs justify. The alternative is that the capacity for mental state attribution is independent from, and precedes the emergence of, the capacity for understanding justifications and for attributing reasons to self and others (cf. Mercier & Sperber, 2017 for arguments). One major problem for the teleology-first strategy is that the question it sensibly purports to answer is not the question raised by Wimmer & Perner’s (1983) landmark study, namely: “why do preschoolers fail to predict a mistaken agent’s likely action?” The question it purports to answer is the question: “why do preschoolers fail to explain the failure of a mistaken agent’s action”? On the one hand, it is logically possible to deny the existence of any strong asymmetry between explaining and predicting an agent’s action or any event — as some prominent philosophers of science have. If so, then from a pure teleological point of view, one could both explain an agent’s executed action and predict an agent’s forthcoming action by appealing to the agent’s objective reason. But my point is that Perner and Roessler have no choice: if young children are pure teleologists, then they are committed to the strong claim that for young children, prediction and explanation are symmetrical. On the other hand, Perner and Roessler’s endorsement of the teleology-first strategy also commits them to the strong empirical claim that young children are insensitive to the difference between an agent’s successful goal-directed action and her failed goal-directed action. Further empirical work is needed to determine both whether indeed preschoolers are insensitive to the difference between an agent’s successful and failed goal-directed action and whether for young children, prediction and explanation are symmetrical tasks. 
Baillargeon, R., Scott, R.M., & Bian, L. (2016). Psychological reasoning in infancy. Annual Review in Psychology, 67, 159-186.
Burnside, K., Severdjila, V. & Poulin-Dubois, D. (2019). Infants attribute false beliefs to a toy crane. Developmental Science, https://doi.org/10.1111/desc.12887
Clements, W.A. & Perner, J. (1994). Implicit understanding of belief. Cognitive Development, 9, 4, 377-395.
Davidson, D. (1963). Actions, reasons and causes. The Journal of Philosophy, 60, 685-700.
Mercier, H. & Sperber, D. (2017). The Enigma of Reason, Cambridge, Mass.: Harvard University Press.
Perner, J. (1991). Understanding the Representational Mind. Cambridge, Mass: MIT Press.
Perner, J. & Esken, F. (2015). Evolution of Human Cooperation in Homo heidelbergensis: Teleology Versus Mentalism. Developmental Review, 38, 69–88.
Perner, J., Priewasse, B. & Roessler, J. (2018). The practical other: teleology and its development. Interdisciplinary Science Reviews, 43, 2, 99-114.
Perner, J. & Roessler, J. (2010). Teleology and Causal Reasoning in Children’s Theory of Mind. In Causing Human Action: New Perspectives on the Causal Theory of Action, edited by J. Aguilar and A. Buckareff, 199–228. Cambridge, MA: MIT Press.
Perner, J. & Roessler, J. (2012). From Infants’ to Children’s Appreciation of Belief. Trends in Cognitive Sciences, 16, 10, 519–525.
Roessler, J. & Perner, J. (2013). “Teleology: Belief as Perspective.” In UOM-3: Understanding Other Minds, edited by S. Baron-Cohen, M. Lombardo, and H. Tager-Flusberg, 3rd ed., 35–50. Oxford: Oxford University Press.
Wimmer, H. & Perner, J. (1983). Beliefs About Beliefs: Representation and Constraining Function of Wrong Beliefs in Young Children’s Understanding of Deception. Cognition, 13, 103–128.
Wellman, H. M., D. Cross, & Watson, J. (2001). Meta-analysis of Theory of Mind Development: The Truth About False Belief. Child Development, 72, 655–684.
 Thanks to Dan Sperber for his comments.