A philosophical perspective on the Infant Mindreading puzzle
‘I know a small child who replied when asked “Are you learning the violin”? with a denial: ‘No, I already know how to play the violin; I’m learning how to play it better’
(Millikan 2000: 55).
Over the past 35 years or so, there has been a continuous and intense focus within the cognitive sciences upon the nature of our everyday, commonsense psychology. How is it that we can effortlessly predict and understand other people’s behaviour in most everyday situations?
If I see my colleague hiding behind a statue of Gustav Mahler while our mutual boss walks by on the way to a meeting that we should all be going to, I understand that she does not want to go to the meeting, that she does not want our boss to see her because our boss might then order her to go the meeting, and that she believes that hiding behind a statue will prevent our boss from becoming aware of her presence because our boss cannot see through statues. The tactic makes sense to me.
What this case illustrates is that we often make sense of others’ behaviour by exercising an ability to ascribe mental states to them, such as beliefs, desires and intentions. This capacity has variously been dubbed ‘theory of mind’, ‘mentalising’, or ‘mindreading’ (I will use the latter term here, as it is currently the most prevalent).
One general strategy that has been pursued in investigating the nature of mindreading has been to compare different kinds of agents with different mindreading abilities – ie different species and also human children at different stages of development – and to attempt to correlate these differences with other cognitive and/or social differences in those agents. The task that has most commonly been used in assessing whether an agent is capable of mindreading has been the so-called false belief task.
In a standard verbal false belief task, a child (or any other test subject) observes as an agent places an object in location A and then temporarily departs, whereupon a second agent arrives on the scene and transfers the object to location B. When the first agent returns, the child is asked where they are likely to search for the object. The correct answer, of course, is that the agent is likely to search in location A, because that is where they falsely believe the object to be located (Wimmer and Perner, 1983; Baron-Cohen et al., 1985).
This task – first proposed by the philosopher Daniel Dennett (Dennett 1978) – has been regarded as a litmus test for the capacity to represent the beliefs of other agents because the child can’t use her own knowledge concerning the location of the object to predict where the agent will search. Rather, the child must distinguish the agent’s belief about the location of the object from her own knowledge, and generate a prediction about where the agent is going to search on the basis of this belief.
For many years, it has been a robust finding that children under the age of four years tend to fail this task (Wellman et al. 2001), which has motivated the widespread view that children younger than about four don’t represent beliefs. And yet, more recently, non-verbal studies employing various implicit measures have now produced a wealth of evidence that infants are sensitive to others’ false beliefs soon after their first birthdays (Onishi and Baillargeon 2005; Buttelammn et al. 2009; Southgate et al. 2010) or even by the middle of their first year (Kovacs et al 2011; Southgate et al. 2014; for an overview, see Christensen and Michael 2015).
The debate about how to account for this puzzling pattern of findings (i.e. the ‘infant mindreading puzzle’) has been structured by a contrast between rich and lean accounts. Rich accounts (e.g. Baillargeon et al., 2010; Kovacs et al. 2011) maintain that infants represent others’ beliefs by around one year or earlier, and then offer some explanation to account for the lag in performance on explicit verbal false belief tasks, while lean accounts (Perner and Ruffman 2005; Zawidzki 2013) deny that children represent beliefs before about four, and offer deflationary explanations of the infant data that do not impute the ability to represent beliefs to infants.
Midway between these two extremes, Apperly and Butterfill (2009; see also Butterfill and Apperly 2013) have proposed a two-systems account of mindreading, which explains the infant data by positing an early-emerging system (ie system 1) that enables infants and young children to track and reason about some belief-like cognitive states of other agents, but not beliefs per se (until around the age of four, when the more sophisticated system 2 is in place).
Arguably, all three of these theoretical options are internally consistent, independently motivated by sensible theoretical considerations, and consistent with at least most of the existing data. How, then, should we adjudicate among them?
In situations such as this, philosophy can be of use in identifying theoretical commitments which implicitly shape a debate or a methodology, in making the terms of the debate clear, and in opening up overlooked alternatives. In this vein, I would propose taking a fresh philosophical look at the infant mindreading debate against the background of philosophical conceptions of mental content.
Specifically, I would like to point out that this debate has been and continues to be shaped by an implicit commitment to a descriptivist conception of mental content. The basic idea underlying descriptivism is that the contents of a mental representation are individuated by the set of descriptions with which that representation is associated, and which underpin the role that the representation plays in categorization and in inferential reasoning (Harman 1987; Block 1987; Frege 1892; Russell 1905).
So, for example, a representation may count as a representation of foxes if it is associated with a set of descriptions that are satisfied by foxes and which thereby enable the bearer of the representation to categorizes foxes correctly and to draw appropriate inferences about them (eg they are mammals, they have four legs, etc.).
To see how the current debate is shaped by a commitment to descriptivism, let us consider Butterfill and Apperly’s strategy (2013; cf. Apperly and Butterfill 2009). They propose that it should be possible to corroborate a theory about the cognitive mechanisms underlying an ability (such as infants’ ability to succeed at implicit false belief tasks) by finding evidence of particular limits on that ability that one would expect on the basis of the theory in question (but which one should not otherwise expect).
Thus, their theory about the cognitive mechanisms underlying infants’ mindreading abilities entails that infants’ mindreading abilities should be subject to the following signature limits: for example, infants should be unable to attribute beliefs to an agent when it would require the agent to draw inferences involving identity relations, an inability to track beliefs that involve quantifiers (e.g. the belief that there is no object at a location), an inability to perform level‑2 perspective-taking (i.e. to take modes of presentation into account), and a lack of understanding of the interactions between beliefs and other psychological states in producing actions.
In contrast, if one endorses a theory according to which infants represent beliefs per se, according to Butterfill and Apperly, one should not expect their mindreading abilities to exhibit these signature limits.
For example, imagine that an infant knows that some agent has observed that an object O is currently at a location L, that the infant also knows that the agent has witnessed evidence that O is identical with O¢, but that the infant fails to infer that the agent believes that O¢ is currently at L. In such a case, we would have evidence that the infant lacked some core knowledge about how beliefs combine with each other to form other beliefs and/or to guide action.
Butterfill and Apperly go further, though, and submit that in such a case we would have evidence that the infant were not able to represent beliefs at all. The underlying thought here is that the content of a representation is fixed by the knowledge that is associated with that representation (i.e. the knowledge about how beliefs combine with other mental states and thereby contribute to form other beliefs and/or to guide action).
In other words, rather than conclude that infants make faulty inferences about beliefs in some cases, Butterfill and Apperly conclude that infants do not make inferences about beliefs at all.
Thus, it appears that the research strategy proposed by Butterfill and Apperly depends upon a prior decision to link the judgment as to whether children are able to represent beliefs at a particular age to the question of how well children understand beliefs at that age (I believe that the other theoretical alternatives are similarly shaped by descriptivism, although I do not have the space to argue for this claim here). Although there is nothing wrong with this decision, it is important to realize that it is a non-mandatory decision, to which there are legitimate alternatives.
For example, alternative conceptions of mental content within the broad class of causal-historical views do not appeal to the knowledge associated with representations in order to determine their contents. Instead, what is decisive on such accounts is that the representation in question stand, or have in the past stood, in a particular kind of causal relation with the object, property, relation or kind that is its content. It is important to note that causal-historical accounts differ substantially in how they specify the particular kind of causal relation that determines content, and also with respect to the further conditions they introduce.
According to one particular causal-historical theory, namely Ruth Millikan’s (1984; 2000; 2013) teleosemantic approach, what is decisive is that the current existence of a representation be explained by the fact that it, at some point in phylogeny or ontogeny, had the function of co-varying with, and thereby providing information about, its content. As a result, Millikan’s approach does not demand that the representation currently stand in any kind of causal relation with its content whatsoever, and need not ever have stood in a reliable causal relation – as long as it co-varied with it often enough for some consumer system in the organism to be able to benefit from using it as a source of information about that content.
When the infant mindreading debate is approached from a causal-historical perspective, and more specifically from a teleosemantic perspective, the question of whether infants and young children represent others’ beliefs does not hinge upon their possessing or deploying some particular knowledge about beliefs. Rather, it hinges upon their having representational vehicles that have the function of co-varying with other agents’ beliefs, and thereby serve as a source of information about others’ beliefs.
Thus, infants and young children may be said to represent beliefs even though they fail to draw inferences about others’ behaviour that one would draw as an adult who fully masters the concept of belief. As a result, causal-historical theories – in particular of the teleosemantic sort – make it possible to take the infant mindreading data at face value, and accordingly to conclude that infants have the ability to represent and reason about beliefs by as early as six months, and yet at the same time to acknowledge that this ability undergoes substantial further construction throughout childhood.
One way of understanding this point would be to note that abilities can be improved without thereby becoming different abilities, as is illustrated by Ruth Millikan’s charming anecdote (quoted above) about the child learning to play the violin better.
Adams, F. and Aizawa, K., (2010). “Causal Theories of Mental Content”, The Stanford Encyclopedia of Philosophy (Spring 2010 Edition), Edward N. Zalta (ed.), http://plato.stanford.edu/archives/spr2010/entries/content-causal/
Apperly, I. A., & Butterfill, S. A. (2009). Do humans have two systems to track beliefs and belief-like states? Psychological Review, 116(4), 953–970. doi:10.1037/a0016923
Baillargeon, R. (1998). Infants’ understanding of the physical world. In M. Sabourin, F. Craik, & M. Robert (Eds.), Advances in psychological science, Vol. 2: Biological and cognitive aspects (pp. 503–529). Hove, England: Psychology Press/Erlbaum (UK) Taylor & Francis.
Baillargeon, R., Scott, R. M., & He, Z. (2010). False-belief understanding in infants. Trends in Cognitive Sciences, 14, 110–118.
Block, N. 1987, Functional Role and Truth Conditions, Proceedings of the Aristotelian Society LXI, 1987, 157–181
Buttelmann, D., Carpenter, M., & Tomasello, M. (2009). Eighteen-month-old infants show false belief understanding in an active helping paradigm. Cognition, 112(2), 337–342. doi:10.1016/j.cognition.2009.05.006
Butterfill, S. A., & Apperly, I. A. (2013). How to Construct a Minimal Theory of Mind. Mind & Language, 28(5), 606–637. doi:10.1111/mila.12036.
Carlson, S. M. & Moses L. J. (2001). Individuals differences in inhibitory control and children’s theory of mind. Child Development, 72, 1032–1053.
Call, J., & Tomasello, M. (2008). Does the chimpanzee have a theory of mind? 30 years later. Trends in Cognitive Sciences, 12(5), 187–192. doi:10.1016/j.tics.2008.02.010
Carey, S. (2009). The Origin of Concepts. Oxford University Press.
Christensen, W. & Michael, J. (2015) From two systems to a multi-systems architecture for mindreading. New Ideas in Psychology. Accepted on 1 January 2015. DOI: http://dx.doi.org/10.1016/j.newideapsych.2015.01.003
Clements, W. A., & Perner, J. (1994). Implicit understanding of belief. Cognitive Development, 9(4), 377–395. doi:10.1016/0885–2014(94)90012–4
Dennett, Daniel C. ‘Beliefs about Beliefs’ Behavioral and Brain Sciences 1 (1978): 568–70.
Dretske, F. (1981). Knowledge and the Flow of Information. Cambridge, MA: MIT Press.
— (1988). Explaining Behavior: Reasons in a World of Causes. Cambridge: The MIT Press.
Frege, G. 1892. Über Sinn und Bedeutung. Zeitschrift für Philosophie und philosophische Kritik, C: 25–50.
Harman, G. (1987). (Non-solipsistic) Conceptual Role Semantics. In New Directions in Semantics, ed. Lepore, E. London: Academic Press.
Kovács, Á. M., Téglás, E., & Endress, A. D. (2010). The Social Sense: Susceptibility to Others’ Beliefs in Human Infants and Adults. Science, 330 (6012), 1830–1834. doi:10.1126/science.1190792
Millikan, R. (1984). Language, Thought, and Other Biological Categories. Cambridge: MIT Press.
— (2000). On Clear and Confused Ideas: An Essay about Substance Concepts. Cambridge: Cambridge University Press.
— (2013). An Epistemology for Phenomenology? In: Richard Brown (ed.), Consciousness Inside and Out: Phenomenology, Neuroscience, and the Nature of Experience, Springer’s series Studies in Brain and Mind.
Onishi, K. H., & Baillargeon, R. (2005). Do 15-Month-Old Infants Understand False Beliefs? Science, 308(5719), 255–258. doi:10.1126/science.1107621
Papineau, D. (1984). Representation and Explanation, Philosophy of Science 51: 550–72.
Perner, J., and T. Ruffman. (2005). Infants’ Insight into the Mind: How Deep? Science 308:212–14.
Russell, B. 1905. On Denoting. Mind 14:479–93.
Scott, R. M., Baillargeon, R., Song, H., & Leslie, A. M. (2010). Attributing false beliefs about non-obvious properties at 18 months. Cognitive Psychology, 61(4), 366–395. doi:10.1016/j.cogpsych.2010.09.001
Southgate, V., Chavalier, C., & Csibra, G. (2010). Seventeen-month-olds appeal to false beliefs to interpret others’ referential communication. Developmental Science, 13, 907–912.
Southgate, V., Johnson, M. H., Osborne, T., & Csibra, G. (2009). Predictive motor activation during action observation in human infants. Biology Letters, 5(6), 769–772.
Southgate, V., Johnson, M., El Karoui, I., & Csibra, G. (2010). Motor system activation reveals infants’ online prediction of others’ goals. Psychological Science, 21, 355–359.
Wellman, H. M., Cross, D., & Watson, J. (2001). Meta-analysis of theory-of- mind development: The truth about false belief. Child Development, 72(3), 655–684
Wimmer, H., & Perner, J. (1983). Beliefs about beliefs: Representation and constraining function of wrong beliefs in young children’s understanding of deception. Cognition, 13(1), 103–128. doi:10.1016/0010–0277(83)90004–5
Zawidzki, T. (2013). Mindshaping, Cambridge: MIT Press.