Emerging Trends in The Social and Behavioral Sciences · Two‐Systems View of Children's Theory‐of‐Mind Understanding

Two‐Systems View of Children's Theory‐of‐Mind Understanding

Media

Part of Two‐Systems View of Children's Theory‐of‐Mind Understanding

Title: Two‐Systems View of Children's Theory‐of‐Mind Understanding
extracted text: Two-Systems View of Children’s
Theory-of-Mind Understanding
JASON LOW

Abstract
Theory-of-mind research reveals a puzzling pattern of young children showing
implicit “mindreading” success in indirect false-belief tasks well before they can
pass explicit tasks where they are asked to make direct predictions about the
mistaken agent’s belief or behavior. Relevant theorizing has either boosted indirect
responses (e.g., eye movements) as showcasing infants’ and young children’s innate
psychological reasoning system or scoffed at indirect responses as reflecting only
a shallow causal understanding of behavior. This essay describes new theorizing
suggesting that we have not one but two mindreading systems—an implicit
efficient system (shared by infants, children, and adults) that supports spontaneous
responses such as eye gazing and an explicit flexible system (constructed from
age 4 onward) that supports direct verbal responses. This view has inspired
cutting-edge research documenting signature limits on the kinds of input that
the efficient mindreading system processes. New research shows that the efficient
system is set to help young children and adults minimally track facts relating to
agents and objects, but not relations between agents and propositions. The flexible
system—supporting understanding of belief as such—guides children’s direct
verbal inferences in a wide range of perspective-taking situations that include
ascribing how people interpret a particular object. Future research into the question
of how human beings mindread in fast-moving situations will need insight into
whether there are systematic patterns of limits on implicit understanding that
converge across age groups, multiple paradigms, and diverse populations.

INTRODUCTION
Emma, by Jane Austin, is a celebrated novel that illustrates human beings
trying to ascribe mental states to others and the perils of misinterpretations
and getting people wrong. In a famous episode, a chain of comedic blunders
is set up when Emma persuades her friend Harriet to refuse a marriage proposal from the kind but lowly farmer, Mr. Martin. Emma persuades Harriet
to instead wait for the well-to-do vicar, Mr. Elton, to declare his affections.
Emma’s actions are, however, based on her false belief that Mr. Elton’s tender
Emerging Trends in the Social and Behavioral Sciences. Edited by Robert Scott and Stephen Kosslyn.
© 2015 John Wiley & Sons, Inc. ISBN 978-1-118-90077-2.

1

2

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

attention to her painting of Harriet’s portrait signifies a growing love for Harriet. It turned out that the vicar was more interested in the painting by Emma
rather than the painting of Harriet. Like the characters in Austin’s novel, we
can infer what is on other people’s minds (e.g., what people might believe,
know, desire, hope, and perceive). Predicting, explaining, or changing people’s behaviors on the basis of their mental states underpin many kinds of
social endeavors whether it is to act altruistically, win political campaigns,
negotiate business deals, communicate ideas, or plot romantic schemes.
While the study of how we can know about the mental states of others has
a rich philosophical history, our everyday theory-of-mind understanding
or “mindreading” as it is termed in this essay has broad significance to
researchers in many fields. Among anthropologists, there is attention to how
sociocultural practices can make malleable the ways by which we predict
and interpret others’ actions. Linguists are also interested in the extent to
which attributions of mentality are tied to language users and, perhaps more
fundamentally, provided by mastery of complex linguistic constructions.
There is even significant interest among behavioral ecologists on whether
selection pressures would favor increasingly sophisticated mindreading
abilities and how one might make use of comparative methods to chart
the degree to which nonhuman animals’ mental representations constitute
genuine thoughts and concepts. For developmental scientists, the most
widely recognized reason for studying mindreading is that mental state
concepts such as “belief” constitute a paradigm case of conceptual change
that underscores children’s active theory-like organization of observations,
experiences, and data. Here, the emphasis will be on how mindreading is not
foolproof—we can also misjudge where other people are coming from. The
comedy of errors in Austin’s novel brings into sharp focus that mindreading
is cognitively demanding and might not be at work at every given situation.
This essay will provide a basic overview of how the study of signature limits
on tracking and ascribing others’ actions in mental state terms can serve as a
powerful tool to illuminate and differentiate the multiple cognitive systems
constituting children’s mindreading ability.
FOUNDATIONAL RESEARCH
A widely used measure of children’s everyday mindreading ability is the
false-belief task; this task captures a child’s understanding that people will
act in accordance with their own beliefs even when the child knows that those
beliefs do not match reality. The standard task involves an unexpected transfer of an object’s location to set up a perspective difference between children
and the target actor. Max puts his chocolate in the drawer and goes outside to
play. His mother secretly removes the chocolate from the drawer and hides it

Two-Systems View of Children’s Theory-of-Mind Understanding

3

in the cupboard. Max returns and wants the chocolate. Children are asked to
verbally predict where Max will look for the chocolate: 3-year olds fail systematically answering that Max will look in the current location and 4-year
olds pass predicting that Max will look in the drawer even when he is factually wrong. Rapid conceptual change between the ages of 3 and 4 years
has also been found with respect to children’s appreciation of the relation
between their own beliefs and other people’s beliefs. For example, in the
unexpected contents variant of the false-belief task, children are shown a
Smarties candy tube and asked what is inside the container—children typically answer “Smarties.” The experimenter opens the tube to reveal that it
actually contains pencils. Children are asked to predict what other people
would think if shown the closed tube and told to guess its contents. Four-year
olds predict that others would say “Smarties” just as they did. Three-year
olds, however, report that they always knew the tube contained pencils and
predict that everyone else would think that the tube contained pencils.
Decades of studies support that older preschoolers from age 4 onward show
systematic success in attributing false beliefs in a variety of tasks and, further,
are able to deploy their attributions for different uses or demands. Depending
on cultural differences in sociolinguistic practices and child-rearing patterns,
there can be subtle differences in the average age for passing false-belief tasks
and contrasts in the developmental route that children might take to acquire a
conception of belief. The basic age-related developmental trajectory in direct
appreciation of the effect of false belief on people’s actions appears to be uniform in many countries and has been replicated even among children living
in preliterate hunter-gathering and agrarian communities. Given the consistency of these age-related changes in direct false-belief reasoning, one view
is that there is conceptual discontinuity in the early preschool years when
children’s understanding of mind changes from being nonrepresentational to
representational, and beliefs are appreciated as being representations about
a particular state of the world.
What a child appears to understand as revealed by his or her direct verbal
answers may well be different from what the child intuits as revealed by
his or her indirect or spontaneous responses (e.g., eye gazing). Using the
standard false-belief task, researchers have documented that when children
reach 3 years of age, the location where they verbally predict Max will look
for the chocolate dissociates from the location where anticipate Max will
look for the chocolate: 3-year olds look first at the drawer despite answering
that Max will look in the cupboard. In the absence of similar effects in much
younger children, one might cautiously suggest that 3-year olds’ accurate
gaze responses indicate some form of implicit knowledge that is an early
stage in the well-established discontinuous changes in older preschoolers’
reasoning about false beliefs (Apperly, 2011). However, using a modified

4

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

nonverbal version of the standard false-belief task, researchers have found
that 2-year olds also show accurate gaze anticipations that Max would
search incorrectly for the hidden object. The most well-known challenge
to the view of conceptual change in children’s false-belief understanding,
however, comes from Onishi and Baillargeon’s (2005) landmark “violation
of expectation” study with 15-month olds. Their idea was that if infants
can represent others’ false beliefs, infants should find an event that violated
that concept surprising; infants should stare longer when they see an agent
failing to behave in accordance with his or her false belief compared to
when the agent behaves according to his or her false belief. They found that
infants showed longer looking times when the agent behaved in violation
of her false belief about a toy’s whereabouts to search in the new location
(where only infants knew the toy to be) compared to when the agent acted
in accordance with her false belief to search in the old location where she last
saw the toy. Onishi and Baillargeon suggested that 15-month olds already
possess a representational theory of mind—infants appeal to and reason
about other people’s beliefs when interpreting goal-directed pursuits. Since
then, researchers measuring children’s indirect anticipatory or expectant
gazing have suggested that 13- to 18-month olds can already attribute others’
false beliefs about an object’s location, type, and property. Such findings
raise a startling developmental paradox: How can infants, toddlers, and
young preschoolers, who consistently fail to demonstrate any mindreading
ability on many variations of the standard false-belief task, appear to track
others’ beliefs on indirect measures?
The early mindreading account suggests that infants have an innate
understanding of belief (Baillargeon, Scott, & He, 2010). Direct verbal
predictions in the standard false-belief task makes great demands on general
executive functioning abilities that develop more slowly compared to children’s detailed innate concepts. As such, direct verbal prediction measures
underestimate the depth of children’s early mindreading ability. In the case
of indirect nonverbal responses—as measured in violation-of-expectancy
and anticipatory looking paradigms—only the belief representation process
is involved and sophisticated understanding is revealed at younger ages.
A rival account is the behavior-rule account (Perner & Ruffman, 2005),
which holds that indirect looking responses reflect shallow causal understanding. Infants and children are good at noticing patterns of regularities in
the environment, and can statistically abstract rules that help anticipate and
explain people’s actions. Responses stripped to the level of eye gazing may
reflect learnt (or innate) behavior rules that allow an agent’s future action to
be predicted from current behavior without going through the middle step
of inferring the agent’s mental or informational states. Thus, one alternative
interpretation of Onishi and Baillargeon’s study is that 15-month-old infants’

Two-Systems View of Children’s Theory-of-Mind Understanding

5

looking responses may reflect the shallow behavior rule that people will
search for an object where they last saw it rather than a deep analysis in
terms of people’s beliefs about an object’s whereabouts, which then causes
them to act in the predicted way.
Strictly speaking, even older children’s correct verbal predictions on the
standard false-belief task (that Maxi will search in the cupboard for the chocolate) can be dominated by shallow behavior rules (people look for things
where they last saw or put them). However, many researchers remain satisfied that older children can indeed reason about people’s mental states. First,
from about 4 years of age, individual children pass different varieties of the
standard false-belief task, and also pass other problems that require general
meta-representational thinking. Second, older preschoolers (especially children from about 6 years of age) use mental state vocabulary to explain and
justify their predictions about others’ erroneous actions in false-belief tasks
(e.g., Max will look in the drawer because that is where he thinks the chocolate is).
Skepticism of the strong claim that children represent “belief” in their precocious indirect responses might be partly addressed if it turned out that
individual infants, toddlers, and young preschoolers at least passed tests of
coherence—that is, individual children’s spontaneous responses were systematic across diverse perspective-taking tasks. As it stands, there is evidence
of success among infants in diverse belief-inducing task contexts, but these
are only between age groups and between children. Stronger evidence of
individual infants and preschoolers showing systematic looking responses
across false-belief tasks with different belief-inducing situations combined
with different demands is hard to come by—this makes it difficult to rule out
application of behavior rules when interpreting what indirect measures of
mindreading actually reflect (Low & Wang, 2011). Given the computational
equivalence between the early mindreading and behavior-rule accounts, any
action that can be expected, anticipated, or predicted on the basis of attributing others’ epistemic states can also, in principle, be based on behavior rules.
There is a new theoretical solution: following Apperly and Butterfill (2009),
differences between children’s indirect looking responses and direct verbal
responses may reflect the operation of not one psychological reasoning system but two mindreading systems (henceforth two-systems account).
CUTTING-EDGE RESEARCH
To first bring into focus the two-systems account of human mindreading,
consider examples of mental state reasoning demonstrated by jury members
and competitive sports players. In a court of law, a group of jury members
are evaluating a defendant’s belief that the money he was dealing with

6

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

was not proceeds of drug trafficking: the jury is considering whether the
defendant’s actions were intentional or whether he had a false belief about
the nature of the cash he was accepting. The mindreading exhibited by
the jury members is flexible, informed by interacting inferences about the
defendant’s beliefs, knowledge, desires, and intentions. Such unbounded
inferences about others’ reasons for action characterize a flexible and normative conceptualization of belief as such. Representing belief as such includes
(i) negotiating complex causal structures where the agent’s thoughts and
knowledge cohere and combine in unbounded ways, (ii) argumentative
reasoning where there is no restriction as to which evidence might be relevant to the agent’s belief, and (iii) content that is propositional. Adults can
engage in flexible mindreading in the specific sort of framework described
previously, but such a representation of others’ mental states makes great
demand on language and executive resources. The cognitive demands
of flexible mindreading make it less suitable for handling fast-moving
social situations. Consider the reasoning exhibited by competitive sports
players. A rugby player making a dummy pass quickly anticipates action
without engaging in unbounded inferences about the opponent’s beliefs
that interfere with game playing itself; mindreading is cognitively efficient.
The dual and contradictory demand of mindreading that is flexible on the
one hand and efficient on the other hand motivates a two-systems proposal.
A flexible explicit mindreading system supports direct verbal reasoning;
this system ascribes complex mental states to people (e.g., beliefs) and
has access to one’s general knowledge and store of practical wisdom. The
flexible system is emergent from age 4 as language, executive function, and
general meta-representational abilities develop. An efficient implicit mindreading system supports indirect looking responses; this system is shared
by infants, children, and adults, but only ascribes the simple belief-like state
of registration that approximates belief. An agent is said to register an object
at a location if he or she recently encountered the object at that location and
acts as if the object was still there. A distinct minimal mindreading system
that tracks facts about the correctness of an agent’s registration is sufficient
to help guide young children’s spontaneous responses in certain false-belief
tasks involving relations between agents and objects.
The two-systems view holds that cognitive economy in mindreading is
achieved by sacrificing flexibility over representing relations between agents
and propositions and makes the unique prediction that there are natural
limits on the efficient mindreading system. Apperly and Butterfill (2009)
suggest that one signature limit to representing registration is mistakes
over identity. To make sense of this idea, consider Low and Watts’ (2013)
illustration of substituting coreferential names in belief reports relating to
Lois Lane looking to interview Superman:

Two-Systems View of Children’s Theory-of-Mind Understanding

7

1. Lois believes that Clark Kent is in the newsroom.
2. Clark Kent is Superman.
3. Lois believes that Superman is in the newsroom.
On a belief account, the inference at step 3 is invalid; Lois thinks of Clark
in a particular way and does not necessarily know Clark under the aspect of
Superman. Now consider the inference in the case of registration:
1. Lois registers <Clark, newsroom>.
2. Clark Kent is Superman.
3. Lois registers <Superman, newsroom>.
As registration is a relation to objects and not propositions about them, the
inference is valid without ascribing a belief to Lois. Someone who simply
represents Lois registering <Clark, newsroom> will not understand why she
would search elsewhere for Superman, for registering <Clark, newsroom>
and <Superman, newsroom> are equivalent. A person who represents perspective as such will understand there are different ways of thinking about a
single reference. The discovery of a natural signature limit on children’s (and
adults’) ability to ascribe others’ false beliefs about object identity—when
these ascriptions are automatic or indirect spontaneous responses—makes
a powerful case that more than one mindreading system is required to track
and represent beliefs.
Low and Watts (2013) designed a novel task to probe for signature limits on children’s (and adults’) efficient ability to ascribe false beliefs about
object identity. At the start of the familiarization phase of the task, participants watched two boxes lifted to reveal two toy objects of the same kind
(e.g., a red boat was underneath the left-side box and a blue boat was underneath the right-side box). Both boxes were lowered and an actor entered the
scene. Participants watched the actor observing each object move to occupy
the opposite box (red boat moved from left- to right-side box, then blue boat
moved from right- to left-side box). After lights on the windows leading
to the respective boxes illuminated and a beep sounded, the actor reached
through the window into the box with his preferred colored toy (e.g., blue).
The remaining familiarization trials were similar except the boxes were lifted
to reveal toy cars, ducks, and buggies—the actor always reached for the blue
object (for half of the participants, the actor showed a preference for the red
object).
After establishing to participants the actor’s color preference, the test
phase began. The test phase only involved a single dog-robot toy. The
dog-robot toy had a face and body that was blue when viewed from
one position and a face and body that was red when view from another

8

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

position. Participants watched the two boxes lifted to only show a red robot
underneath the left-side box and then boxes were lowered. The actor entered
and participants watched him observing the robot travel (red aspect facing
participants) from the left- to right-side box. Then the robot emerged from a
recessed viewing chamber inside the right-side box where it was visible only
to participants and not the actor. In the viewing chamber, the robot—with
its red aspect facing participants—spun 180∘ to reveal its blue face–body
aspect on one side and spun 180∘ again to reveal its red aspect. Then the
robot with its red aspect facing participants returned back into the right-side
box. Next, participants watched the actor observing the robot—now with its
blue aspect facing them and red aspect facing him—move from the right- to
left-side box. Then the doors illuminated with a beep sound to indicate the
actor was about to search for a blue object.
If participants are tracking and attributing others’ beliefs as such, they
should appreciate that there is reason for the actor to expect another dog
robot (blue) to be inside the right-side box. Such an appreciation would
entail an understanding of the actor’s false belief about object identity. Only
6% of 3- and 4-year olds and 25% of adults correctly gazed in anticipation of
where the actor believed to be a separate blue robot, while 13, 56, and 95%
of the participants in those age groups provided correct verbal predictions.
Aside from the object-identity-task film, all participants also watched a
standard object-location-task film whereby the actor did not observe a
toy being unexpectedly transferred from one location to another—there
was close to ceiling correct anticipatory gazing shown by the 3-year-old,
4-year-old, and adult age groups alongside the usual developmental increase
in accuracy of their verbal predictions about the actor’s incorrect search
action (31, 75, and 100%, respectively). In sum, all groups showed accurate
eye movements when anticipating an agent’s action in the location task but
showed the same blind spot in spontaneously tracking how an agent viewed
a particular object (as required in the identity task), suggesting that the quick
and efficient mindreading system guiding spontaneous reasoning is limited
in scope. Four-year olds and adults had no trouble making accurate verbal
predictions about agents’ incorrect search actions across the different tasks,
suggesting that effortful verbal reasoning is supported by a distinct flexible
mindreading system.
KEY ISSUES FOR FUTURE RESEARCH
There are challenges in interpreting research in limits on the efficient
mindreading system. For one thing, absence of evidence is not evidence of
absence, and Low and Watts’ (2013) negative findings contrasts with foundational research reviewed previously on violation-of-expectancy paradigms

Two-Systems View of Children’s Theory-of-Mind Understanding

9

where infants’ expectant looking is measured. For example, Scott and
Baillargeon’s (2009) claimed that 18-month olds are able to represent others’
false beliefs about identity; their violation-of-expectancy study involved
two separate objects (a nonstackable 1-piece toy penguin and a stackable
2-piece toy penguin) to measure infants’ successful ability to infer an agent’s
false belief about which toy she was facing. However, given that there were
two toys present in the test scene at all times, critics have suggested it is
possible that infants may be tracking facts about types of objects present
without necessarily understanding that people can interpret or identify a
single object differently depending on their viewing circumstances. Low
and Watts’ design avoided these problems by ensuring that, for their object
identity false-belief task, the spatiotemporal boundaries of the “red robot”
and “blue robot” coincided so as to require children to confront how people
with different viewing angles can interpret a single object differently.
Nonetheless, Low and Watts (2013) only showed limits on perspectivetaking in one paradigm. The single-system early mindreading account could
still push the explanation that a greater degree of executive function skills
(e.g., inhibitory control) might have been needed to solve Low and Watts’
(2013) identity task compared to their location task. Although one might be
seduced to argue that preschoolers’ executive control may not be sufficiently
developed to inhibit inaccurate gazing on the identity task, the temptation
should be resisted; such a move would not easily explain why adults with
mature executive control also turned out to show inaccurate gazing on
the identity task. Such an explanation also does not easily explain why
4-year olds would then apparently have sufficient executive control to give
accurate verbal predictions for the identity task. Moreover, the single-system
early mindreading account contends that it is direct verbal responding on
false-belief tasks, not indirect spontaneous looking, that is burdened with
executive function demands. Following a behavior-rule account, one could
conjure up the rule “if an object is blue, agent will reach for it” to explain
participants’ incorrect looking to the box containing the object in the identity
task. This would not explain why 4-year olds and adults did not follow
such a rule when making direct verbal predictions in the identity task: They
correctly predicted that the agent would reach into the empty box. Overall,
the discovery of blind spots in preschoolers’ spontaneous reasoning of false
beliefs about object identity is consistent with Apperly and Butterfill’s (2009)
two-systems account predictions that the efficient system must be limited.
To help inform theoretical interpretations, one important avenue for future
research is to test for converging patterns of limits on human mindreading
across different paradigms and different age groups. There are allied developments involving other paradigms suggesting that the negative results for
spontaneous higher level perspective-taking documented by Low and Watts

10

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

(2013) is meaningful. Apperly (2011) has examined cognitively efficient mindreading in a simple “Level-1” perspective-taking task (determining what
items are seen by an agent) and in a complex “Level-2” perspective-taking
task (determining how an item is seen by an agent). The clever idea from
Apperly’s team is that if participants automatically calculate others’ perspectives, then participants should do so even when there is no need to do so in
the task. In a Level-1 task, adult participants viewed a scene where an avatar
stood in room with dots on the wall. Sometimes the avatar saw the same
number of dots on the wall as did participants, and other times the avatar
saw fewer dots (some of the dots where positioned behind the avatar that
only participants could see). What was interesting was that when participants had to judge how many dots they themselves could see, participants
responded more slowly and with more errors when the avatar saw a different
number of dots compared to when the avatar saw the same number of dots. It
seems that adults automatically track facts about what others see even when
they do not need to, and even when such computations would interfere with
their own responses. However, such “altercentric interference” effects did not
extend to adults’ (nor older school children’s) handling of complex Level-2
types of perspective-taking tasks. In the latter type of task, participants faced
an avatar seated at a table. The key trials showed certain digits on the table
such as the number “6” where it could look like a “9” to the avatar and a “6”
to participant viewers. This time there was no evidence of participants being
slower or more error prone when judging how the digit appeared to them
even when it appeared differently to the avatar. Experimental paradigms that
provide evidence of automatic mindreading in Level-1 but not Level-2 tasks
lend significant purchase to Low and Watts’ findings that children’s implicit
understanding about false beliefs allows them to track others’ “beliefs” about
object location but not about object identity.
It will be worthwhile for future research to test whether individuals from a wide developmental span (say from the ages of 5 years to
adulthood) systematically show accurate anticipatory looking in the
object-location false-belief task and altercentric interference in the Level-1
perspective-taking task, but systematically show incorrect anticipatory looking in the object-identity false-belief task and no altercentric interference
in the Level-2 perspective-taking task. Evidence of converging patterns of
performance in the very same participants across multiple age groups and
across very different paradigms would refine the notion that the efficient
mindreading system is limited to handling certain kinds of content. Much
more work also needs to be done to trace what is invariant and what changes
across populations when delineating signature limits on efficient compared
to flexible mindreading. The new methods for measuring indirect mindreading as developed by Low and Watts and Apperly and colleagues are suited

Two-Systems View of Children’s Theory-of-Mind Understanding

11

to gauge behavioral phenomena in Westernized and educated children and
adults; they will need to be ecologically adapted for use in more diverse
samples. In so doing, we would be able to conduct cross-cultural studies
to chart whether signature limits on spontaneous mindreading coupled
with age-related changes in direct mindreading are also found, for example,
among populations that show a strong capacity for self-control very early in
development (e.g., in Confucian cultures) or that focus on external powers
rather than internal mental states as a means of explaining human action
(e.g., divinity as a basis for intentionality in Sudanese communities of the
Dinka). If converging patterns of limits were stable across individuals in
very different sociocultural (and linguistic) environments, it would partly
suggest that the structure of the efficient mindreading system has a high
degree of genetic constraint.
As the field goes forward in mapping converging patterns of signature
limits on efficient mindreading within and across age groups (using multiple paradigms and involving diverse populations), a challenging theoretical issue that is likely to emerge is whether the information provided by
the efficient cognitive system intersects with the information provided by
the flexible cognitive system. Following the two-systems account of mindreading outlined here, one suggestion is that while there may be some modest connection between implicit and explicit ascriptions of mental states, the
exchange must be kept to a minimum to preserve the processing efficiency of
tracking belief-like states in fast-moving social interactions. Low and Watts’
(2013) findings support the suggestion that the efficient mindreading system is not likely to be just an implicit homolog of the flexible mindreading system as it will not be able to effectively provide candidate contents
to the later-developing flexible system for a range of situations beyond its
representational powers. Further experiments along such lines could inform
and even revitalize related research on how a complex suite of knowledge,
skills, and experiences—for example, domain general meta-representational
reasoning, objective considerations of action, language, and executive functioning, social relationships, and family dynamics—partly contribute to the
development and construction of a flexible system for mindreading beliefs
and other complex epistemic states.
Mapping of signature limits on efficient mindreading is not a recipe for
avoiding them. Limits on the efficient mindreading system might even
be regarded as being adaptively beneficial, just like how egocentrism can
reduce stereotype threats and improve relationship satisfaction by elevating
the congruence between our own and others’ opinions. A better understanding of the nature of signature limits on the efficient mindreading system
can help inform when and to what extent we should take other people’s
points of view. Thus, it is reassuring that the cognitive trick of managing

12

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

the business of other people’s minds—the dual and conflicting demands in
making abductive inferences to the best explanation about others’ actions in
terms of mental states and how this might be accomplished in fast-moving
social situations—can be achieved by invoking a combination of highand low-level processes, which make complementary tradeoffs between
flexibility and efficiency. Apperly and Butterfill’s (2009) dual-process view
is set to revolutionize research exploring new frontiers in the richness
and complexity of children’s and adults’ social understanding. Vive la
révolution!
REFERENCES
Apperly, I. A. (2011). Mindreaders: The cognitive basis of “Theory of Mind”. Hove, England: Psychology Press.
Apperly, I. A., & Butterfill, S. A. (2009). Do humans have two systems to track beliefs
and belief-like states? Psychological Review, 116(4), 753–970. doi:10.1037/a0016923
Baillargeon, R., Scott, R. M., & He, Z. (2010). False-belief understanding in infants.
Trends in Cognitive Science, 14(3), 110–117. doi:10.1016/j.tics.2009.12.006
Low, J., & Watts, J. (2013). Attributing false beliefs about object identity reveals a signature blind spot in humans’ efficient mind-reading system. Psychological Science,
24(3), 305–311. doi:10.1177/0956797612451469
Low, J., & Wang, B. (2011). On the long road to mentalism in children’s spontaneous
false-belief understanding: Are we there yet? Review of Philosophy and Psychology,
2(3), 411–428. doi:10.1007/s13164-011-0067-y
Onishi, K. H., & Baillargeon, R. (2005). Do 15-month-old infants understand false
beliefs? Science, 308(5719), 255–258. doi:10.1126/science.1107621
Perner, J., & Ruffman, T. (2005). Infants’ insight into the mind: How deep? Science,
308(5719), 214–216. doi:10.1126/science.1111656
Scott, R. M., & Baillargeon, R. (2009). Which penguin is this? Attributing false
beliefs about object identity at 18 months. Child Development, 80(4), 1172–1196.
doi:10.1111/j.1467-8624.2009.01324.x

FURTHER READING
Low, J., & Perner, J. (2012). Implicit and explicit theory of mind: State of the art. British
Journal of Developmental Psychology, 30, 1–13. doi:10.1111/j.2044-835X.2011.02074.x
Perner, J., & Roessler, J. (2012). From infants’ to children’s appreciation of belief.
Trends in Cognitive Sciences, 16(10), 519–525. doi:10.1016/j.tics.2012.08.004

JASON LOW SHORT BIOGRAPHY
Jason Low is a Senior Lecturer in the School of Psychology at Victoria University of Wellington, New Zealand, where he has been a faculty member

Two-Systems View of Children’s Theory-of-Mind Understanding

13

since 1999. Jason completed his PhD and his undergraduate studies at The
University of Western Australia, Perth, Australia. His primary research interests lie in the cognitive underpinnings of human theory-of-mind ability, with
a focus on understanding the development and limits of children’s implicit
and explicit systems to track and ascribe mental states to others. In recent
years, he has also developed research interests in numerical and social cognition in nonhuman animals, focused on the extent to which New Zealand
robins can discriminate between quantities and interpret human behavioral
cues. He has published in Child Development, Psychological Science, and Proceedings of the Royal Society B: Biological Sciences, among other journals. He
is an Associate Editor for the British Journal of Developmental Psychology. Personal webpage: http://www.victoria.ac.nz/psyc/about/staff/jason-low
RELATED ESSAYS
Globalization Backlash (Sociology), Mabel Berezin
Agency as an Explanatory Key: Theoretical Issues (Sociology), Richard
Biernacki and Tad Skotnicki
Empathy Gaps between Helpers and Help-Seekers: Implications for Cooperation (Psychology), Vanessa K. Bohns and Francis J. Flynn
Theory of Mind and Behavior (Psychology), Amanda C. Brandone
Language, Perspective, and Memory (Psychology), Rachel A. Ryskin et al.
Mental Models (Psychology), Ruth M. J. Byrne
Political Ideologies (Political Science), Edward G. Carmines and Nicholas J.
D’Amico
Authenticity: Attribution, Value, and Meaning (Sociology), Glenn R. Carroll
The Inherence Heuristic: Generating Everyday Explanations (Psychology),
Andrei Cimpian
Delusions (Psychology), Max Coltheart
Misinformation and How to Correct It (Psychology), John Cook et al.
Insight (Psychology), Brian Erickson and John Kounios
Micro-Cultures (Sociology), Gary Alan Fine
Cognitive Processes Involved in Stereotyping (Psychology), Susan T. Fiske
and Cydney H. Dupree
Language and Thought (Psychology), Susan Goldin-Meadow
An Evolutionary Perspective on Developmental Plasticity (Psychology),
Sarah Hartman and Jay Belsky
States and Nationalism (Anthropology), Michael Herzfeld
Normal Negative Emotions and Mental Disorders (Sociology), Allan V.
Horwitz
The Development of Social Trust (Psychology), Vikram K. Jaswal and Marissa
B. Drell

14

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

Herd Behavior (Psychology), Tatsuya Kameda and Reid Hastie
Cultural Neuroscience: Connecting Culture, Brain, and Genes (Psychology),
Shinobu Kitayama and Sarah Huff
From Individual Rationality to Socially Embedded Self-Regulation (Sociology), Siegwart Lindenberg
Concepts and Semantic Memory (Psychology), Barbara C. Malt
A Bio-Social-Cultural Approach to Early Cognitive Development: Entering
the Community of Minds (Psychology), Katherine Nelson
Emerging Trends in Culture and Concepts (Psychology), Bethany Ojalehto
and Douglas Medin
Culture as Situated Cognition (Psychology), Daphna Oyserman
Religion (Anthropology), Benjamin Grant Purzycki et al.
Class, Cognition, and Face-to-Face Interaction (Sociology), Lauren A. Rivera
Born This Way: Thinking Sociologically about Essentialism (Sociology),
Kristen Schilt
News Framing Effects and Emotions (Political Science), Andreas R. T. Schuck
and Alina Feinholdt
Theory of Mind and Behavior (Psychology), Amanda C. Brandone
Theory of Mind (Psychology), Henry Wellman
Understanding Biological Motion (Psychology), Jeroen J. A. Van Boxtel and
Hongjing Lu
The Intrinsic Dynamics of Development (Psychology), Paul van Geert and
Marijn van Dijk; Two-Systems View of Children’s
Theory-of-Mind Understanding
JASON LOW

Abstract
Theory-of-mind research reveals a puzzling pattern of young children showing
implicit “mindreading” success in indirect false-belief tasks well before they can
pass explicit tasks where they are asked to make direct predictions about the
mistaken agent’s belief or behavior. Relevant theorizing has either boosted indirect
responses (e.g., eye movements) as showcasing infants’ and young children’s innate
psychological reasoning system or scoffed at indirect responses as reflecting only
a shallow causal understanding of behavior. This essay describes new theorizing
suggesting that we have not one but two mindreading systems—an implicit
efficient system (shared by infants, children, and adults) that supports spontaneous
responses such as eye gazing and an explicit flexible system (constructed from
age 4 onward) that supports direct verbal responses. This view has inspired
cutting-edge research documenting signature limits on the kinds of input that
the efficient mindreading system processes. New research shows that the efficient
system is set to help young children and adults minimally track facts relating to
agents and objects, but not relations between agents and propositions. The flexible
system—supporting understanding of belief as such—guides children’s direct
verbal inferences in a wide range of perspective-taking situations that include
ascribing how people interpret a particular object. Future research into the question
of how human beings mindread in fast-moving situations will need insight into
whether there are systematic patterns of limits on implicit understanding that
converge across age groups, multiple paradigms, and diverse populations.

INTRODUCTION
Emma, by Jane Austin, is a celebrated novel that illustrates human beings
trying to ascribe mental states to others and the perils of misinterpretations
and getting people wrong. In a famous episode, a chain of comedic blunders
is set up when Emma persuades her friend Harriet to refuse a marriage proposal from the kind but lowly farmer, Mr. Martin. Emma persuades Harriet
to instead wait for the well-to-do vicar, Mr. Elton, to declare his affections.
Emma’s actions are, however, based on her false belief that Mr. Elton’s tender
Emerging Trends in the Social and Behavioral Sciences. Edited by Robert Scott and Stephen Kosslyn.
© 2015 John Wiley & Sons, Inc. ISBN 978-1-118-90077-2.

1

2

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

attention to her painting of Harriet’s portrait signifies a growing love for Harriet. It turned out that the vicar was more interested in the painting by Emma
rather than the painting of Harriet. Like the characters in Austin’s novel, we
can infer what is on other people’s minds (e.g., what people might believe,
know, desire, hope, and perceive). Predicting, explaining, or changing people’s behaviors on the basis of their mental states underpin many kinds of
social endeavors whether it is to act altruistically, win political campaigns,
negotiate business deals, communicate ideas, or plot romantic schemes.
While the study of how we can know about the mental states of others has
a rich philosophical history, our everyday theory-of-mind understanding
or “mindreading” as it is termed in this essay has broad significance to
researchers in many fields. Among anthropologists, there is attention to how
sociocultural practices can make malleable the ways by which we predict
and interpret others’ actions. Linguists are also interested in the extent to
which attributions of mentality are tied to language users and, perhaps more
fundamentally, provided by mastery of complex linguistic constructions.
There is even significant interest among behavioral ecologists on whether
selection pressures would favor increasingly sophisticated mindreading
abilities and how one might make use of comparative methods to chart
the degree to which nonhuman animals’ mental representations constitute
genuine thoughts and concepts. For developmental scientists, the most
widely recognized reason for studying mindreading is that mental state
concepts such as “belief” constitute a paradigm case of conceptual change
that underscores children’s active theory-like organization of observations,
experiences, and data. Here, the emphasis will be on how mindreading is not
foolproof—we can also misjudge where other people are coming from. The
comedy of errors in Austin’s novel brings into sharp focus that mindreading
is cognitively demanding and might not be at work at every given situation.
This essay will provide a basic overview of how the study of signature limits
on tracking and ascribing others’ actions in mental state terms can serve as a
powerful tool to illuminate and differentiate the multiple cognitive systems
constituting children’s mindreading ability.
FOUNDATIONAL RESEARCH
A widely used measure of children’s everyday mindreading ability is the
false-belief task; this task captures a child’s understanding that people will
act in accordance with their own beliefs even when the child knows that those
beliefs do not match reality. The standard task involves an unexpected transfer of an object’s location to set up a perspective difference between children
and the target actor. Max puts his chocolate in the drawer and goes outside to
play. His mother secretly removes the chocolate from the drawer and hides it

Two-Systems View of Children’s Theory-of-Mind Understanding

3

in the cupboard. Max returns and wants the chocolate. Children are asked to
verbally predict where Max will look for the chocolate: 3-year olds fail systematically answering that Max will look in the current location and 4-year
olds pass predicting that Max will look in the drawer even when he is factually wrong. Rapid conceptual change between the ages of 3 and 4 years
has also been found with respect to children’s appreciation of the relation
between their own beliefs and other people’s beliefs. For example, in the
unexpected contents variant of the false-belief task, children are shown a
Smarties candy tube and asked what is inside the container—children typically answer “Smarties.” The experimenter opens the tube to reveal that it
actually contains pencils. Children are asked to predict what other people
would think if shown the closed tube and told to guess its contents. Four-year
olds predict that others would say “Smarties” just as they did. Three-year
olds, however, report that they always knew the tube contained pencils and
predict that everyone else would think that the tube contained pencils.
Decades of studies support that older preschoolers from age 4 onward show
systematic success in attributing false beliefs in a variety of tasks and, further,
are able to deploy their attributions for different uses or demands. Depending
on cultural differences in sociolinguistic practices and child-rearing patterns,
there can be subtle differences in the average age for passing false-belief tasks
and contrasts in the developmental route that children might take to acquire a
conception of belief. The basic age-related developmental trajectory in direct
appreciation of the effect of false belief on people’s actions appears to be uniform in many countries and has been replicated even among children living
in preliterate hunter-gathering and agrarian communities. Given the consistency of these age-related changes in direct false-belief reasoning, one view
is that there is conceptual discontinuity in the early preschool years when
children’s understanding of mind changes from being nonrepresentational to
representational, and beliefs are appreciated as being representations about
a particular state of the world.
What a child appears to understand as revealed by his or her direct verbal
answers may well be different from what the child intuits as revealed by
his or her indirect or spontaneous responses (e.g., eye gazing). Using the
standard false-belief task, researchers have documented that when children
reach 3 years of age, the location where they verbally predict Max will look
for the chocolate dissociates from the location where anticipate Max will
look for the chocolate: 3-year olds look first at the drawer despite answering
that Max will look in the cupboard. In the absence of similar effects in much
younger children, one might cautiously suggest that 3-year olds’ accurate
gaze responses indicate some form of implicit knowledge that is an early
stage in the well-established discontinuous changes in older preschoolers’
reasoning about false beliefs (Apperly, 2011). However, using a modified

4

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

nonverbal version of the standard false-belief task, researchers have found
that 2-year olds also show accurate gaze anticipations that Max would
search incorrectly for the hidden object. The most well-known challenge
to the view of conceptual change in children’s false-belief understanding,
however, comes from Onishi and Baillargeon’s (2005) landmark “violation
of expectation” study with 15-month olds. Their idea was that if infants
can represent others’ false beliefs, infants should find an event that violated
that concept surprising; infants should stare longer when they see an agent
failing to behave in accordance with his or her false belief compared to
when the agent behaves according to his or her false belief. They found that
infants showed longer looking times when the agent behaved in violation
of her false belief about a toy’s whereabouts to search in the new location
(where only infants knew the toy to be) compared to when the agent acted
in accordance with her false belief to search in the old location where she last
saw the toy. Onishi and Baillargeon suggested that 15-month olds already
possess a representational theory of mind—infants appeal to and reason
about other people’s beliefs when interpreting goal-directed pursuits. Since
then, researchers measuring children’s indirect anticipatory or expectant
gazing have suggested that 13- to 18-month olds can already attribute others’
false beliefs about an object’s location, type, and property. Such findings
raise a startling developmental paradox: How can infants, toddlers, and
young preschoolers, who consistently fail to demonstrate any mindreading
ability on many variations of the standard false-belief task, appear to track
others’ beliefs on indirect measures?
The early mindreading account suggests that infants have an innate
understanding of belief (Baillargeon, Scott, & He, 2010). Direct verbal
predictions in the standard false-belief task makes great demands on general
executive functioning abilities that develop more slowly compared to children’s detailed innate concepts. As such, direct verbal prediction measures
underestimate the depth of children’s early mindreading ability. In the case
of indirect nonverbal responses—as measured in violation-of-expectancy
and anticipatory looking paradigms—only the belief representation process
is involved and sophisticated understanding is revealed at younger ages.
A rival account is the behavior-rule account (Perner & Ruffman, 2005),
which holds that indirect looking responses reflect shallow causal understanding. Infants and children are good at noticing patterns of regularities in
the environment, and can statistically abstract rules that help anticipate and
explain people’s actions. Responses stripped to the level of eye gazing may
reflect learnt (or innate) behavior rules that allow an agent’s future action to
be predicted from current behavior without going through the middle step
of inferring the agent’s mental or informational states. Thus, one alternative
interpretation of Onishi and Baillargeon’s study is that 15-month-old infants’

Two-Systems View of Children’s Theory-of-Mind Understanding

5

looking responses may reflect the shallow behavior rule that people will
search for an object where they last saw it rather than a deep analysis in
terms of people’s beliefs about an object’s whereabouts, which then causes
them to act in the predicted way.
Strictly speaking, even older children’s correct verbal predictions on the
standard false-belief task (that Maxi will search in the cupboard for the chocolate) can be dominated by shallow behavior rules (people look for things
where they last saw or put them). However, many researchers remain satisfied that older children can indeed reason about people’s mental states. First,
from about 4 years of age, individual children pass different varieties of the
standard false-belief task, and also pass other problems that require general
meta-representational thinking. Second, older preschoolers (especially children from about 6 years of age) use mental state vocabulary to explain and
justify their predictions about others’ erroneous actions in false-belief tasks
(e.g., Max will look in the drawer because that is where he thinks the chocolate is).
Skepticism of the strong claim that children represent “belief” in their precocious indirect responses might be partly addressed if it turned out that
individual infants, toddlers, and young preschoolers at least passed tests of
coherence—that is, individual children’s spontaneous responses were systematic across diverse perspective-taking tasks. As it stands, there is evidence
of success among infants in diverse belief-inducing task contexts, but these
are only between age groups and between children. Stronger evidence of
individual infants and preschoolers showing systematic looking responses
across false-belief tasks with different belief-inducing situations combined
with different demands is hard to come by—this makes it difficult to rule out
application of behavior rules when interpreting what indirect measures of
mindreading actually reflect (Low & Wang, 2011). Given the computational
equivalence between the early mindreading and behavior-rule accounts, any
action that can be expected, anticipated, or predicted on the basis of attributing others’ epistemic states can also, in principle, be based on behavior rules.
There is a new theoretical solution: following Apperly and Butterfill (2009),
differences between children’s indirect looking responses and direct verbal
responses may reflect the operation of not one psychological reasoning system but two mindreading systems (henceforth two-systems account).
CUTTING-EDGE RESEARCH
To first bring into focus the two-systems account of human mindreading,
consider examples of mental state reasoning demonstrated by jury members
and competitive sports players. In a court of law, a group of jury members
are evaluating a defendant’s belief that the money he was dealing with

6

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

was not proceeds of drug trafficking: the jury is considering whether the
defendant’s actions were intentional or whether he had a false belief about
the nature of the cash he was accepting. The mindreading exhibited by
the jury members is flexible, informed by interacting inferences about the
defendant’s beliefs, knowledge, desires, and intentions. Such unbounded
inferences about others’ reasons for action characterize a flexible and normative conceptualization of belief as such. Representing belief as such includes
(i) negotiating complex causal structures where the agent’s thoughts and
knowledge cohere and combine in unbounded ways, (ii) argumentative
reasoning where there is no restriction as to which evidence might be relevant to the agent’s belief, and (iii) content that is propositional. Adults can
engage in flexible mindreading in the specific sort of framework described
previously, but such a representation of others’ mental states makes great
demand on language and executive resources. The cognitive demands
of flexible mindreading make it less suitable for handling fast-moving
social situations. Consider the reasoning exhibited by competitive sports
players. A rugby player making a dummy pass quickly anticipates action
without engaging in unbounded inferences about the opponent’s beliefs
that interfere with game playing itself; mindreading is cognitively efficient.
The dual and contradictory demand of mindreading that is flexible on the
one hand and efficient on the other hand motivates a two-systems proposal.
A flexible explicit mindreading system supports direct verbal reasoning;
this system ascribes complex mental states to people (e.g., beliefs) and
has access to one’s general knowledge and store of practical wisdom. The
flexible system is emergent from age 4 as language, executive function, and
general meta-representational abilities develop. An efficient implicit mindreading system supports indirect looking responses; this system is shared
by infants, children, and adults, but only ascribes the simple belief-like state
of registration that approximates belief. An agent is said to register an object
at a location if he or she recently encountered the object at that location and
acts as if the object was still there. A distinct minimal mindreading system
that tracks facts about the correctness of an agent’s registration is sufficient
to help guide young children’s spontaneous responses in certain false-belief
tasks involving relations between agents and objects.
The two-systems view holds that cognitive economy in mindreading is
achieved by sacrificing flexibility over representing relations between agents
and propositions and makes the unique prediction that there are natural
limits on the efficient mindreading system. Apperly and Butterfill (2009)
suggest that one signature limit to representing registration is mistakes
over identity. To make sense of this idea, consider Low and Watts’ (2013)
illustration of substituting coreferential names in belief reports relating to
Lois Lane looking to interview Superman:

Two-Systems View of Children’s Theory-of-Mind Understanding

7

1. Lois believes that Clark Kent is in the newsroom.
2. Clark Kent is Superman.
3. Lois believes that Superman is in the newsroom.
On a belief account, the inference at step 3 is invalid; Lois thinks of Clark
in a particular way and does not necessarily know Clark under the aspect of
Superman. Now consider the inference in the case of registration:
1. Lois registers <Clark, newsroom>.
2. Clark Kent is Superman.
3. Lois registers <Superman, newsroom>.
As registration is a relation to objects and not propositions about them, the
inference is valid without ascribing a belief to Lois. Someone who simply
represents Lois registering <Clark, newsroom> will not understand why she
would search elsewhere for Superman, for registering <Clark, newsroom>
and <Superman, newsroom> are equivalent. A person who represents perspective as such will understand there are different ways of thinking about a
single reference. The discovery of a natural signature limit on children’s (and
adults’) ability to ascribe others’ false beliefs about object identity—when
these ascriptions are automatic or indirect spontaneous responses—makes
a powerful case that more than one mindreading system is required to track
and represent beliefs.
Low and Watts (2013) designed a novel task to probe for signature limits on children’s (and adults’) efficient ability to ascribe false beliefs about
object identity. At the start of the familiarization phase of the task, participants watched two boxes lifted to reveal two toy objects of the same kind
(e.g., a red boat was underneath the left-side box and a blue boat was underneath the right-side box). Both boxes were lowered and an actor entered the
scene. Participants watched the actor observing each object move to occupy
the opposite box (red boat moved from left- to right-side box, then blue boat
moved from right- to left-side box). After lights on the windows leading
to the respective boxes illuminated and a beep sounded, the actor reached
through the window into the box with his preferred colored toy (e.g., blue).
The remaining familiarization trials were similar except the boxes were lifted
to reveal toy cars, ducks, and buggies—the actor always reached for the blue
object (for half of the participants, the actor showed a preference for the red
object).
After establishing to participants the actor’s color preference, the test
phase began. The test phase only involved a single dog-robot toy. The
dog-robot toy had a face and body that was blue when viewed from
one position and a face and body that was red when view from another

8

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

position. Participants watched the two boxes lifted to only show a red robot
underneath the left-side box and then boxes were lowered. The actor entered
and participants watched him observing the robot travel (red aspect facing
participants) from the left- to right-side box. Then the robot emerged from a
recessed viewing chamber inside the right-side box where it was visible only
to participants and not the actor. In the viewing chamber, the robot—with
its red aspect facing participants—spun 180∘ to reveal its blue face–body
aspect on one side and spun 180∘ again to reveal its red aspect. Then the
robot with its red aspect facing participants returned back into the right-side
box. Next, participants watched the actor observing the robot—now with its
blue aspect facing them and red aspect facing him—move from the right- to
left-side box. Then the doors illuminated with a beep sound to indicate the
actor was about to search for a blue object.
If participants are tracking and attributing others’ beliefs as such, they
should appreciate that there is reason for the actor to expect another dog
robot (blue) to be inside the right-side box. Such an appreciation would
entail an understanding of the actor’s false belief about object identity. Only
6% of 3- and 4-year olds and 25% of adults correctly gazed in anticipation of
where the actor believed to be a separate blue robot, while 13, 56, and 95%
of the participants in those age groups provided correct verbal predictions.
Aside from the object-identity-task film, all participants also watched a
standard object-location-task film whereby the actor did not observe a
toy being unexpectedly transferred from one location to another—there
was close to ceiling correct anticipatory gazing shown by the 3-year-old,
4-year-old, and adult age groups alongside the usual developmental increase
in accuracy of their verbal predictions about the actor’s incorrect search
action (31, 75, and 100%, respectively). In sum, all groups showed accurate
eye movements when anticipating an agent’s action in the location task but
showed the same blind spot in spontaneously tracking how an agent viewed
a particular object (as required in the identity task), suggesting that the quick
and efficient mindreading system guiding spontaneous reasoning is limited
in scope. Four-year olds and adults had no trouble making accurate verbal
predictions about agents’ incorrect search actions across the different tasks,
suggesting that effortful verbal reasoning is supported by a distinct flexible
mindreading system.
KEY ISSUES FOR FUTURE RESEARCH
There are challenges in interpreting research in limits on the efficient
mindreading system. For one thing, absence of evidence is not evidence of
absence, and Low and Watts’ (2013) negative findings contrasts with foundational research reviewed previously on violation-of-expectancy paradigms

Two-Systems View of Children’s Theory-of-Mind Understanding

9

where infants’ expectant looking is measured. For example, Scott and
Baillargeon’s (2009) claimed that 18-month olds are able to represent others’
false beliefs about identity; their violation-of-expectancy study involved
two separate objects (a nonstackable 1-piece toy penguin and a stackable
2-piece toy penguin) to measure infants’ successful ability to infer an agent’s
false belief about which toy she was facing. However, given that there were
two toys present in the test scene at all times, critics have suggested it is
possible that infants may be tracking facts about types of objects present
without necessarily understanding that people can interpret or identify a
single object differently depending on their viewing circumstances. Low
and Watts’ design avoided these problems by ensuring that, for their object
identity false-belief task, the spatiotemporal boundaries of the “red robot”
and “blue robot” coincided so as to require children to confront how people
with different viewing angles can interpret a single object differently.
Nonetheless, Low and Watts (2013) only showed limits on perspectivetaking in one paradigm. The single-system early mindreading account could
still push the explanation that a greater degree of executive function skills
(e.g., inhibitory control) might have been needed to solve Low and Watts’
(2013) identity task compared to their location task. Although one might be
seduced to argue that preschoolers’ executive control may not be sufficiently
developed to inhibit inaccurate gazing on the identity task, the temptation
should be resisted; such a move would not easily explain why adults with
mature executive control also turned out to show inaccurate gazing on
the identity task. Such an explanation also does not easily explain why
4-year olds would then apparently have sufficient executive control to give
accurate verbal predictions for the identity task. Moreover, the single-system
early mindreading account contends that it is direct verbal responding on
false-belief tasks, not indirect spontaneous looking, that is burdened with
executive function demands. Following a behavior-rule account, one could
conjure up the rule “if an object is blue, agent will reach for it” to explain
participants’ incorrect looking to the box containing the object in the identity
task. This would not explain why 4-year olds and adults did not follow
such a rule when making direct verbal predictions in the identity task: They
correctly predicted that the agent would reach into the empty box. Overall,
the discovery of blind spots in preschoolers’ spontaneous reasoning of false
beliefs about object identity is consistent with Apperly and Butterfill’s (2009)
two-systems account predictions that the efficient system must be limited.
To help inform theoretical interpretations, one important avenue for future
research is to test for converging patterns of limits on human mindreading
across different paradigms and different age groups. There are allied developments involving other paradigms suggesting that the negative results for
spontaneous higher level perspective-taking documented by Low and Watts

10

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

(2013) is meaningful. Apperly (2011) has examined cognitively efficient mindreading in a simple “Level-1” perspective-taking task (determining what
items are seen by an agent) and in a complex “Level-2” perspective-taking
task (determining how an item is seen by an agent). The clever idea from
Apperly’s team is that if participants automatically calculate others’ perspectives, then participants should do so even when there is no need to do so in
the task. In a Level-1 task, adult participants viewed a scene where an avatar
stood in room with dots on the wall. Sometimes the avatar saw the same
number of dots on the wall as did participants, and other times the avatar
saw fewer dots (some of the dots where positioned behind the avatar that
only participants could see). What was interesting was that when participants had to judge how many dots they themselves could see, participants
responded more slowly and with more errors when the avatar saw a different
number of dots compared to when the avatar saw the same number of dots. It
seems that adults automatically track facts about what others see even when
they do not need to, and even when such computations would interfere with
their own responses. However, such “altercentric interference” effects did not
extend to adults’ (nor older school children’s) handling of complex Level-2
types of perspective-taking tasks. In the latter type of task, participants faced
an avatar seated at a table. The key trials showed certain digits on the table
such as the number “6” where it could look like a “9” to the avatar and a “6”
to participant viewers. This time there was no evidence of participants being
slower or more error prone when judging how the digit appeared to them
even when it appeared differently to the avatar. Experimental paradigms that
provide evidence of automatic mindreading in Level-1 but not Level-2 tasks
lend significant purchase to Low and Watts’ findings that children’s implicit
understanding about false beliefs allows them to track others’ “beliefs” about
object location but not about object identity.
It will be worthwhile for future research to test whether individuals from a wide developmental span (say from the ages of 5 years to
adulthood) systematically show accurate anticipatory looking in the
object-location false-belief task and altercentric interference in the Level-1
perspective-taking task, but systematically show incorrect anticipatory looking in the object-identity false-belief task and no altercentric interference
in the Level-2 perspective-taking task. Evidence of converging patterns of
performance in the very same participants across multiple age groups and
across very different paradigms would refine the notion that the efficient
mindreading system is limited to handling certain kinds of content. Much
more work also needs to be done to trace what is invariant and what changes
across populations when delineating signature limits on efficient compared
to flexible mindreading. The new methods for measuring indirect mindreading as developed by Low and Watts and Apperly and colleagues are suited

Two-Systems View of Children’s Theory-of-Mind Understanding

11

to gauge behavioral phenomena in Westernized and educated children and
adults; they will need to be ecologically adapted for use in more diverse
samples. In so doing, we would be able to conduct cross-cultural studies
to chart whether signature limits on spontaneous mindreading coupled
with age-related changes in direct mindreading are also found, for example,
among populations that show a strong capacity for self-control very early in
development (e.g., in Confucian cultures) or that focus on external powers
rather than internal mental states as a means of explaining human action
(e.g., divinity as a basis for intentionality in Sudanese communities of the
Dinka). If converging patterns of limits were stable across individuals in
very different sociocultural (and linguistic) environments, it would partly
suggest that the structure of the efficient mindreading system has a high
degree of genetic constraint.
As the field goes forward in mapping converging patterns of signature
limits on efficient mindreading within and across age groups (using multiple paradigms and involving diverse populations), a challenging theoretical issue that is likely to emerge is whether the information provided by
the efficient cognitive system intersects with the information provided by
the flexible cognitive system. Following the two-systems account of mindreading outlined here, one suggestion is that while there may be some modest connection between implicit and explicit ascriptions of mental states, the
exchange must be kept to a minimum to preserve the processing efficiency of
tracking belief-like states in fast-moving social interactions. Low and Watts’
(2013) findings support the suggestion that the efficient mindreading system is not likely to be just an implicit homolog of the flexible mindreading system as it will not be able to effectively provide candidate contents
to the later-developing flexible system for a range of situations beyond its
representational powers. Further experiments along such lines could inform
and even revitalize related research on how a complex suite of knowledge,
skills, and experiences—for example, domain general meta-representational
reasoning, objective considerations of action, language, and executive functioning, social relationships, and family dynamics—partly contribute to the
development and construction of a flexible system for mindreading beliefs
and other complex epistemic states.
Mapping of signature limits on efficient mindreading is not a recipe for
avoiding them. Limits on the efficient mindreading system might even
be regarded as being adaptively beneficial, just like how egocentrism can
reduce stereotype threats and improve relationship satisfaction by elevating
the congruence between our own and others’ opinions. A better understanding of the nature of signature limits on the efficient mindreading system
can help inform when and to what extent we should take other people’s
points of view. Thus, it is reassuring that the cognitive trick of managing

12

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

the business of other people’s minds—the dual and conflicting demands in
making abductive inferences to the best explanation about others’ actions in
terms of mental states and how this might be accomplished in fast-moving
social situations—can be achieved by invoking a combination of highand low-level processes, which make complementary tradeoffs between
flexibility and efficiency. Apperly and Butterfill’s (2009) dual-process view
is set to revolutionize research exploring new frontiers in the richness
and complexity of children’s and adults’ social understanding. Vive la
révolution!
REFERENCES
Apperly, I. A. (2011). Mindreaders: The cognitive basis of “Theory of Mind”. Hove, England: Psychology Press.
Apperly, I. A., & Butterfill, S. A. (2009). Do humans have two systems to track beliefs
and belief-like states? Psychological Review, 116(4), 753–970. doi:10.1037/a0016923
Baillargeon, R., Scott, R. M., & He, Z. (2010). False-belief understanding in infants.
Trends in Cognitive Science, 14(3), 110–117. doi:10.1016/j.tics.2009.12.006
Low, J., & Watts, J. (2013). Attributing false beliefs about object identity reveals a signature blind spot in humans’ efficient mind-reading system. Psychological Science,
24(3), 305–311. doi:10.1177/0956797612451469
Low, J., & Wang, B. (2011). On the long road to mentalism in children’s spontaneous
false-belief understanding: Are we there yet? Review of Philosophy and Psychology,
2(3), 411–428. doi:10.1007/s13164-011-0067-y
Onishi, K. H., & Baillargeon, R. (2005). Do 15-month-old infants understand false
beliefs? Science, 308(5719), 255–258. doi:10.1126/science.1107621
Perner, J., & Ruffman, T. (2005). Infants’ insight into the mind: How deep? Science,
308(5719), 214–216. doi:10.1126/science.1111656
Scott, R. M., & Baillargeon, R. (2009). Which penguin is this? Attributing false
beliefs about object identity at 18 months. Child Development, 80(4), 1172–1196.
doi:10.1111/j.1467-8624.2009.01324.x

FURTHER READING
Low, J., & Perner, J. (2012). Implicit and explicit theory of mind: State of the art. British
Journal of Developmental Psychology, 30, 1–13. doi:10.1111/j.2044-835X.2011.02074.x
Perner, J., & Roessler, J. (2012). From infants’ to children’s appreciation of belief.
Trends in Cognitive Sciences, 16(10), 519–525. doi:10.1016/j.tics.2012.08.004

JASON LOW SHORT BIOGRAPHY
Jason Low is a Senior Lecturer in the School of Psychology at Victoria University of Wellington, New Zealand, where he has been a faculty member

Two-Systems View of Children’s Theory-of-Mind Understanding

13

since 1999. Jason completed his PhD and his undergraduate studies at The
University of Western Australia, Perth, Australia. His primary research interests lie in the cognitive underpinnings of human theory-of-mind ability, with
a focus on understanding the development and limits of children’s implicit
and explicit systems to track and ascribe mental states to others. In recent
years, he has also developed research interests in numerical and social cognition in nonhuman animals, focused on the extent to which New Zealand
robins can discriminate between quantities and interpret human behavioral
cues. He has published in Child Development, Psychological Science, and Proceedings of the Royal Society B: Biological Sciences, among other journals. He
is an Associate Editor for the British Journal of Developmental Psychology. Personal webpage: http://www.victoria.ac.nz/psyc/about/staff/jason-low
RELATED ESSAYS
Globalization Backlash (Sociology), Mabel Berezin
Agency as an Explanatory Key: Theoretical Issues (Sociology), Richard
Biernacki and Tad Skotnicki
Empathy Gaps between Helpers and Help-Seekers: Implications for Cooperation (Psychology), Vanessa K. Bohns and Francis J. Flynn
Theory of Mind and Behavior (Psychology), Amanda C. Brandone
Language, Perspective, and Memory (Psychology), Rachel A. Ryskin et al.
Mental Models (Psychology), Ruth M. J. Byrne
Political Ideologies (Political Science), Edward G. Carmines and Nicholas J.
D’Amico
Authenticity: Attribution, Value, and Meaning (Sociology), Glenn R. Carroll
The Inherence Heuristic: Generating Everyday Explanations (Psychology),
Andrei Cimpian
Delusions (Psychology), Max Coltheart
Misinformation and How to Correct It (Psychology), John Cook et al.
Insight (Psychology), Brian Erickson and John Kounios
Micro-Cultures (Sociology), Gary Alan Fine
Cognitive Processes Involved in Stereotyping (Psychology), Susan T. Fiske
and Cydney H. Dupree
Language and Thought (Psychology), Susan Goldin-Meadow
An Evolutionary Perspective on Developmental Plasticity (Psychology),
Sarah Hartman and Jay Belsky
States and Nationalism (Anthropology), Michael Herzfeld
Normal Negative Emotions and Mental Disorders (Sociology), Allan V.
Horwitz
The Development of Social Trust (Psychology), Vikram K. Jaswal and Marissa
B. Drell

14

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

Herd Behavior (Psychology), Tatsuya Kameda and Reid Hastie
Cultural Neuroscience: Connecting Culture, Brain, and Genes (Psychology),
Shinobu Kitayama and Sarah Huff
From Individual Rationality to Socially Embedded Self-Regulation (Sociology), Siegwart Lindenberg
Concepts and Semantic Memory (Psychology), Barbara C. Malt
A Bio-Social-Cultural Approach to Early Cognitive Development: Entering
the Community of Minds (Psychology), Katherine Nelson
Emerging Trends in Culture and Concepts (Psychology), Bethany Ojalehto
and Douglas Medin
Culture as Situated Cognition (Psychology), Daphna Oyserman
Religion (Anthropology), Benjamin Grant Purzycki et al.
Class, Cognition, and Face-to-Face Interaction (Sociology), Lauren A. Rivera
Born This Way: Thinking Sociologically about Essentialism (Sociology),
Kristen Schilt
News Framing Effects and Emotions (Political Science), Andreas R. T. Schuck
and Alina Feinholdt
Theory of Mind and Behavior (Psychology), Amanda C. Brandone
Theory of Mind (Psychology), Henry Wellman
Understanding Biological Motion (Psychology), Jeroen J. A. Van Boxtel and
Hongjing Lu
The Intrinsic Dynamics of Development (Psychology), Paul van Geert and
Marijn van Dijk