Having theory of mind means understanding how others think, including the ability of someone else to have a false belief.
In a famous theory-of-mind experiment that includes false beliefs, children watch scenes involving a character named Maxi, his mother and a chocolate bar. Maxi places the chocolate bar into a blue box and then leaves.
Unbeknownst to Maxi, his mother shows up and moves the chocolate from the blue box into a green box. After Maxi’s mother leaves, Maxi returns and then the child is asked where Maxi will look for the chocolate.
By 4 years old, children can answer correctly: Maxi will look in the blue box.
But do young children really understand that because Maxi did not see his mother move the chocolate, he falsely believes it is still in the blue box?
The answer is no, according to William Fabricius, associate professor of psychology at Arizona State University. For more than a decade, Fabricius and his collaborators have carried out new experiments and have also analyzed previous experiments that collectively show children do not actually understand false beliefs until they are 6 or 7 years old.
This work was published in Monographs of the Society for Research in Child Development on September 21.
“When we overestimate what young children understand about the mind, and thus how others think, we can expect too much from them in terms of social behavior or performance in school,” said Fabricius, who is the lead author of the paper.
Three locations to hide the chocolate bar
One of the first ways the research team tested what children actually understand about Maxi’s false belief was to add a third possible location of the chocolate bar.
When young children are asked where Maxi will look for the chocolate, they answer the blue box 50% of the time and the red box 50% of the time.
“When there are only two locations, 4- and 5-year-old children can answer correctly without truly understanding that Maxi has a false belief about the location of the chocolate bar,” Fabricius said. “Adding a third location results in them guessing at chance between the two empty locations.
The random choices children make when there are three possible locations of the chocolate bar suggest they rely on their rudimentary understanding of seeing and knowing. This research team has named this process “perceptual access reasoning.”
Children use perceptual access reasoning in the following way:
- Seeing leads to knowing
- People who cannot see something do not know about it
- People who do not know will always do the wrong thing
Based on these rules, 4- and 5-year-old children reason that when Maxi returns, he cannot see that the chocolate is in the green box, so he does not know that the chocolate is in the green box. Therefore, children reason that Maxi will make the wrong choice and will look in an empty location.
When there is only one empty location (the blue box), children answer correctly by default. When there are two empty locations (blue and red boxes), they guess.
What happens when Maxi has a true belief, and his mother leaves the chocolate bar alone
Another way the research team tested what young children understand about others’ thoughts was to have the chocolate bar remain where Maxi put it. When Maxi returns, he has a true belief about where the chocolate is.
In this experiment, Maxi again puts the chocolate bar in the blue box and leaves. This time when Maxi’s mother comes in, she leaves the chocolate bar where it is.
Even with just two options – the blue and green boxes – young children fail the true-belief task. They incorrectly answer that Maxi will make the wrong choice and look in the green box.
“Perceptual access reasoning users have an immature concept of knowing as tied to the present situation, and do not yet understand that people have memories that persist across situations. They do not understand that Maxi might remember putting the chocolate bar into the blue box,” Fabricius said.
“The evidence from this series of experiments is consistent that children do not understand mental representation until they are 6 or 7 years old.”
What perceptual access reasoning means for preschoolers
The finding that young children do not understand true or false beliefs and instead rely on perceptual access reasoning is relevant for how they are taught.
“There are strong correlations between theory of mind and a child’s ability to share, be socially appropriate and be able to problem solve and plan,” said Anne Kupfer, director of ASU’s Child Study Lab (CSL) and co-author of the Monograph paper.
The CSL partners with developmental psychology faculty to put research findings into practice and has implemented the findings from the Monograph paper into its preschool curriculum.
“It is important for educators to know at what age a child can finally realize that how they feel, how they think or what they want are not necessarily what everyone else feels, thinks or wants,” Kupfer said.
Sharing a toy is a common situation that requires CSL staff to leverage how young children use perceptual access reasoning. Kupfer described a scenario in which a child wants a toy, but another classmate is playing with it.
The child takes the toy and because they are happy holding the toy, they think everyone is happy. But the child who just lost the toy starts to cry, and the child who took the toy is puzzled.
“That’s where we come in. In this situation we narrate what is happening and role model responses that are based on what the kids understand from perceptual access reasoning,” Kupfer said. “We say to the child who is crying,
‘I can see you are upset and saw that Johnny took the toy away from you. Is that why you are upset?’
We then role model and ask the crying child to tell Johnny why they are upset, because he took their toy.
Then we direct Johnny to look at the sad child’s face and say,
‘She just told you she is upset. Why is she upset?’ Johnny can then answer, ‘Because I took her toy’.”
This example demonstrates how educators can help children learn about others’ mental representations. The child who took the toy begins to understand why they feel happy but the other child does not – a precursor to having theory of mind.
The capacity to recognize that others have minds, and reason about the contents of those minds, is critical for navigating our social world. Often referred to as a ‘Theory of Mind’, this capacity is essential for understanding human action [1], it also provides the foundation for cultural learning [2, 3], reduces prejudice [4], and fosters prosocial behavior [5–8].
The ontogeny of this capacity in humans has intrigued philosophers and researchers for decades. When do children understand that others have minds that represent, and therefore can misrepresent, reality? In some of the earliest work investigating this question researchers used what became known as the Sally-Anne Task (a.k.a., the Classic False Belief task, Maxi Task, or Change-of-Location task; [9]).
In this task, children observe a scenario where a protagonist (e.g., Sally) hides an object in one location and leaves the scene. In Sally’s absence, another character (e.g., Anne) hides the object in a different location. Then, children are asked where Sally will look for the object when she returns. The logic is as follows: if children can predict that Sally will look for the object where she originally hid it—because she will not know what transpired in her absence and thus will hold a false belief—then they understand that a mind can misrepresent reality [9–12].
Across hundreds of replications of this task, and others like it [11–14], 3-year-olds consistently fail by indicating that Sally will first look for the object where it is currently (and where they know it to be), whereas by around 4.5 to 5 years of age, children tend to correctly infer Sally’s false belief (see 14, for meta-analysis).
Notably, Theory of Mind is a multifaceted capacity that is much broader than simply passing the Sally-Anne task, or reasoning about false beliefs. A ‘theory of mind’ is generally believed to involve a suite of abilities like reasoning about emotions, intentions, desires, knowledge, and much more (for review, see [15]).
To examine the developmental trajectory of different Theory of Mind abilities researchers have designed various measures such as the Reading the Mind in the Eyes Test [16], Belief-Desire Reasoning task [17], Strange Stories [18], among many others (e.g., [19–23]). Together, these findings reveal that there is important development in Theory of Mind both before and after children succeed at classic false belief tasks.
Despite the existence of several innovations in Theory of Mind measures, the classic Sally-Anne task is still by far the most widely used measure in the literature, especially for preschool-age children. As of 2020, the original work by Wimmer and Perner has been cited 7,733 times [9], and Wellman and Liu’s Theory of Mind Scale (which includes a variant of the false belief task) has been cited 1,978 times [24].
For better or worse, the Sally-Anne task remains the gold standard in developmental and cognitive science providing the primary foundation for various fields of research relating to the understanding of the mind such as its neuroanatomical loci [25, 26], its role in Autism Spectrum Disorder [27, 28], and ADHD [see, 29], as well as its various cognitive and environmental correlates [30–34], and its ability to predict a whole host of social outcomes such as one’s tendency to exhibit empathy and prosocial behavior [see 8], the degree of one’s peer relationship problems [35–37], and one’s academic success [38, 39].
Given that performance on this task is such a widely used predictor of various social outcomes, and it forms the bedrock on which several fields of research have been built, it is imperative to understand precisely what it is measuring and which aspects of cognition account for the developmental differences.
Unfortunately, there is still widespread disagreement among researchers about what the Sally-Anne task is measuring, and why 3-year-olds fail this classic task [40–42]. One interpretation, sometimes referred to as the ‘Fundamental Change’ view, is that 3-year-olds do not yet comprehend that others have minds that can misrepresent reality [14, 43, 44].
Another interpretation, sometimes referred to as the ‘Processing Demands’ view is that 3-year-olds’ more general cognitive abilities (e.g., working memory, inhibitory control, language) are not sufficiently developed to allow them to succeed. Under this latter interpretation, it is not that 3-year-olds lack the concept of false beliefs [45–47], but that developmental limitations in cognitive processing abilities undermine their performance.
Several variants of this latter interpretation have been put forth emphasizing different cognitive processes but on the whole these researchers argue that if these processing demands were sufficiently reduced, even very young children could successfully reason about false beliefs (for more accounts, see [48–50]).
Several studies have shown that when the Sally-Anne task is modified in ways that clarify the task, increase the motivation of the participant, or minimize task demands, young children’s false belief performance tends to improve. For example, Siegal and Beattie (1991) changed the false belief question to w
here will the protagonist look for the target object first? As opposed to, where will the protagonist look for the target object. They found that this clarification improved children’s false belief performance [51]. Other researchers have shown that when children are more actively involved in the false belief scenario, such as when they are deceiving the protagonist by hiding the target object, young children were more likely to accurately infer the protagonist’s false belief [52–56].
Still other researchers have identified that when specific cognitive demands are reduced, including explicit response generation [e.g., 47], executive functioning [57, 58], attentional biases [e.g., 59], and language comprehension [e.g., 60], young children’s false belief performance tended to improve [42, 51, 52, 61].
Initially, infant research also appeared to provide support for the view that the Classic Sally-Anne task is unnecessarily cognitively demanding for young children [62–65]. Onishi and Baillargeon were first to demonstrate infants’ apparent ability to reason about false belief using an experimental paradigm measuring whether infants looked longer (evidencing surprise) at one of two visual scenes (i.e., one in which an actor behaved in line with a true belief and one with an actor behaving in line with a false belief; [64]). This task placed fewer cognitive demands on participants.
For example, the paradigm did not require any understanding of language, it did not require that the participant make a prediction about one’s future behavior, nor did it require the participant to intentionally choose a response. In the absence of these demands, even 15-month-olds seemed to understand that people can hold false beliefs [66, 67].
Recently, however, the robustness of these findings has been called into question following several failed replication attempts from multiple labs [68–71]. Moreover, even if the initial findings hold true, it is not the changes in infancy that have been shown to predict a host of social outcomes—it is the changes in cognition between three and four years of age that do, and therefore it is those changes that necessitate better understanding.
In the current study, we examine a different cognitive mechanism that has so far received little attention—the effect of the ‘curse of knowledge’ on children’s false belief reasoning. The curse of knowledge is the tendency to be biased by one’s own (current) knowledge when reasoning about a more naïve perspective (a.k.a., the ‘hindsight bias’, ‘creeping determinism’, ‘reality bias’, or the ‘knew-it-all-along effect’; [72–75]). As a classic example of the curse of knowledge bias, adults who learn the outcome of an event (e.g., an election, a battle, a sports game) overestimate the likelihood that others will predict that outcome. In comparison, adults who were not informed of the event outcome tend to be more accurate in their estimates of what others will predict [73, 76, 77].
In the Sally-Anne task, children are made aware of exactly where Anne moved the object even though specific outcome information is not a necessary condition to infer a false belief. Accordingly, to pass the Sally-Anne task, children are required to overcome the curse of knowledge and infer Sally’s naïve perspective. This is especially difficult for younger children, as they tend to be more vulnerable to the curse of knowledge than older children [78–81]. Still, even adults’ false belief reasoning can be impaired by possessing specific outcome information [82].
In their experiment, Birch and Bloom presented participants with a four-container false belief task, where adults were told that a protagonist, named Vicky, was playing her violin and then decided to go outside (Fig 1). Before leaving, she placed her violin inside a blue container. In her absence, participants were told that the locations of the containers were altered, and the violin was moved to another container.
Some of the participants were told exactly where the violin was moved (e.g., red container), and others were simply told that it was moved to a different container, with no specific information as to where. Participants who knew it had been moved to the red container were significantly worse at predicting Vicky’s false belief than participants who did not know where the object had been moved. That is, adults who can unquestionably reason about false beliefs can be biased, or ‘cursed’ so-to-speak, by their own knowledge when reasoning about false beliefs [for replications and extensions of this work, see 83, 84].
It might be tempting to attribute the curse of knowledge to poor inhibitory control, where an individual is biased by their knowledge because they are unable to inhibit the contents of their own knowledge. While there is evidence for the role of inhibitory control on the curse of knowledge [81, 85, 86], inhibitory control does not fully account for the curse of knowledge [87–89]—a point we return to in greater detail in the Discussion.

Stimuli from Birch and Bloom’s (2007) study, where adults are given a plausible reason to believe that Vicki would look for her violin in the current container.
Note, that the red container (current container) in section B is in the same spot as the blue container (original container) in section A. From “The Curse of Knowledge in Reasoning About False Beliefs” by Birch & Bloom (2007), Psychological Science, 18(5), 384.
In the current experiment, we aimed to directly examine the effect of the curse of knowledge on false belief performance by manipulating children’s outcome knowledge, while keeping trials consistent with one another. To test this, we employed the logic detailed in Birch (2005) and used by Birch and Bloom (2007) and created two variants of the Sally-Anne task [see Fig 2; 82, 90], where a protagonist hides an object in one of four containers and leaves the scene. As in the standard task, another character moves the object in the protagonist’s absence. Our critical manipulation was whether the participants were told specific outcome information: For two test trials (Outcome Known, or Cursed, Trials), we kept it consistent with the standard task and told children exactly where the object was moved (e.g., “to the blue box”). In the other two test trials (Outcome Unknown, or Curse-Lifted, Trials), we told them that the object was moved but did not tell them where (e.g., “to one of the other boxes”). After each test trial, children were asked where the protagonist will look for the object. In other words, the manipulation across test trial types allowed us to test whether eliminating, or reducing, the effect of the curse of knowledge (i.e., in the Outcome Unknown Trials) affected children’s false belief reasoning.

A depiction of the four-box Sally-Ann task used in the present experiment.
The researcher presented each child with a visual enactment of four false belief scenarios using colorful stick figures. Each scenario involved a protagonist (a) who put an object in one of the four containers (b). Then, the protagonist leaves (c), and another character placed the object in a different container (d). In one scenario for example, the researcher said to the child: “Sally was playing with her ball, then she got hungry, so Sally put her ball right in here [purple container], and went home. When Sally was gone, Ryan hid Sally’s ball in a different spot! He may have hid it here, or here, or here” [researcher pointed to each of the other containers]. In two scenarios (i.e., Outcome Known Trials), the researcher told the child exactly where the second character placed the object (e.g., “But we know that Ryan hid the ball here.” [blue container]). In the other two scenarios in contrast (i.e., the Outcome Unknown Trials), the researcher told the child: “But we don’t know where he hid it.” Both trial types ended the same way (Panel E, e.g., “Then, Sally came back. Where will she look for her ball?”.
Given previous research suggesting that the curse of knowledge interferes with the ability to reason about false beliefs (e.g., see [76]), we predicted that children would perform more accurately on the false belief task when they are not required to overcome the curse of knowledge. Also, considering that young children are more susceptible to the bias compared to older children and adults [e.g., 79], we hypothesized that the age-related changes in children’s performance on the Sally-Anne task may reflect older preschoolers’ greater ability to overcome the curse of knowledge bias, rather than a conceptual change in their understanding of the mind.
reference link: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7894954/
More information: William V. Fabricius et al, Perceptual Access Reasoning (PAR) in Developing a Representational Theory of Mind, Monographs of the Society for Research in Child Development (2021). DOI: 10.1111/mono.12432