Explain what theory of mind is and how it develops, evaluate the main ways of measuring it, and discuss its role in development, autism research and cognition.
Each is in the Psychology glossary, with every other term in the subject.
A lifelong neurodevelopmental condition characterised by differences in social communication and interaction and by restricted or repetitive behaviours and interests, varying widely between individuals.
The proposal that difficulties in understanding between autistic and non-autistic people run in both directions, because each group finds the other’s ways of thinking and communicating hard to read.
The ability to understand and share another person’s feelings; often divided into cognitive empathy (understanding what others feel) and affective empathy (responding emotionally to it).
A set of higher-order cognitive processes, including inhibitory control, working memory and cognitive flexibility, that allow people to plan, control attention and regulate their behaviour.
A belief that does not match reality; understanding that someone can hold a false belief and will act on it is the standard test of theory of mind in children.
A test of adult theory of mind in which participants view photographs of the eye region of faces and choose which of four words best describes the person’s mental state.
The ability to attribute mental states such as beliefs, desires, intentions and emotions to oneself and others, and to understand that they can differ from one’s own and from reality.
A false-belief task in which a familiar container holds something unexpected and the child is asked what another person, or they themselves earlier, would think is inside.
A false-belief task in which an object is moved while a character is absent and the child is asked where the character will look for it, such as the Sally-Anne task.
A method used with infants in which looking times to expected and unexpected events are compared; longer looking at an unexpected event is taken as evidence of a violated expectation.
Explain what theory of mind is and how it develops from infancy to middle childhood.
Describe the main strategies used to measure theory of mind, including unexpected transfer, unexpected contents, implicit looking-time measures and the Reading the Mind in the Eyes Test.
Evaluate strategies used to measure theory of mind with reference to validity, reliability and task demands.
Describe and evaluate Baron-Cohen, Leslie and Frith (1985) and the theory of mind account of autism.
Explain how language, family interaction, culture and brain development contribute to theory of mind.
Discuss the role of theory of mind in understanding development and cognition.
Evaluate research on theory of mind with reference to measurement, bias, causality and perspective.
Theory of mind is the understanding that other people have their own beliefs, desires, intentions and feelings, which may differ from one’s own and from reality, and that these mental states explain what people do. It is used constantly in everyday life: in predicting what a friend will think of a present, in telling a joke, in recognising a lie, and in understanding a story. Its development in early childhood is one of the most studied changes in developmental psychology.
Theory of mind is a cognitive process, so this area draws heavily on the cognitive approach. It also has a biological side, studied through brain imaging of the regions active when people think about others’ thoughts, and a sociocultural side, because conversation with family members and the culture a child grows up in affect when and how it develops. Because mental states cannot be observed, research on theory of mind is above all a study in
With a free account, Augment remembers where you stopped and what you have practiced.
The term was introduced by Premack and Woodruff (1978), who asked whether chimpanzees understand that others have mental states. They used the word “theory” because mental states are not directly observable: a person must infer them from behaviour and context and use them to predict behaviour, as a scientist uses a theory. In psychology the ability is also called mentalising or mind-reading.
The key test of a mature theory of mind is understanding false belief. A belief is a mental representation of the world, and it can be wrong. To predict the behaviour of someone who holds a false belief, a child must set aside what they themselves know to be true and reason from what the other person believes. Passing a false-belief task is therefore taken as evidence that the child understands beliefs as representations, which may be accurate or inaccurate, rather than as copies of reality.
Researchers distinguish several levels. First-order false belief is reasoning about what one person believes (“Sally thinks the marble is in the basket”), which most children in Western samples pass at about 4 to 5 years. Second-order false belief is reasoning about what one person believes about another person’s beliefs (“Anne thinks that Sally thinks…”), passed at about 6 to 7. Theory of mind continues to develop into adolescence and adulthood, as people become better at using it quickly and flexibly in complex social situations.
Theory of mind is closely related to, but different from, empathy. Psychologists often separate cognitive empathy, knowing what someone else is thinking or feeling, which is close to theory of mind, from affective empathy, sharing or responding emotionally to someone else’s feelings. A person can have good cognitive empathy but little affective empathy, as in some manipulative people who understand others’ feelings well but do not care about them. Autistic people are sometimes described as having the opposite profile: difficulty in reading others’ mental states, but ordinary or strong emotional concern for others once they understand.
Because beliefs are invisible, researchers have designed tasks in which a child’s answer is possible only if the child understands someone else’s mental state. Each task operationalises theory of mind in a different way, and each has its own strengths and weaknesses. The development of these tasks is a good example of how psychologists turn an abstract construct into an observable measure, and of the problems this creates.
The first false-belief task was designed by Wimmer and Perner (1983). Children watched a puppet story in which a boy called Maxi puts his chocolate in a cupboard and goes out to play. While he is away, his mother moves the chocolate to a different cupboard. Children were asked where Maxi will look for his chocolate when he comes back. To answer correctly (the first cupboard), the child has to represent Maxi’s false belief, rather than their own knowledge of where the chocolate really is. Most 3- to 4-year-olds pointed to where the chocolate actually was, while most 6-year-olds answered correctly.
Baron-Cohen, Leslie and Frith (1985) simplified the procedure with two dolls, Sally and Anne, and it became the best-known version. Control questions check that a child who fails is not simply confused or forgetful: a reality question (“Where is the marble really?”) and a memory question (“Where was the marble in the beginning?”). A child who answers both correctly but fails the belief question has a specific difficulty with false belief.
In the Smarties task, a child is shown a tube of Smarties sweets and asked what is inside. The child says “Smarties”, and the tube is opened to show that it contains pencils. The child is then asked what another child, who has not seen inside, will think is in the tube. Gopnik and Astington (1988) added a second question, asking children what they themselves had thought was inside when they first saw the tube. Many 3-year-olds said “pencils” to both questions. They failed to attribute a false belief to another child and also misreported their own earlier false belief.
The finding that young children misreport their own past beliefs is important theoretically. It suggests that 3-year-olds’ difficulty is not only about other people but about understanding beliefs as representations that can change and be mistaken. This supports the view that passing false-belief tasks reflects a real conceptual change in children’s understanding of the mind.
A single false-belief task gives a pass-or-fail result and treats theory of mind as one ability. Wellman and Liu (2004) developed a theory of mind scale that tests several steps in order of difficulty: understanding that people can have different desires, different beliefs, that seeing leads to knowing (knowledge access), false belief, and that people can hide their real emotions. In their study of 75 children aged about 3 to 6½, most children who passed a later step also passed all earlier steps. This shows that theory of mind develops in a sequence and allows researchers to measure it more finely.
YouTubeStandard false-belief tasks require children to understand questions, hold a story in memory and inhibit the urge to point to where the object really is. Researchers who suspected that these demands hide earlier understanding designed implicit tasks that measure where infants look instead of what they say. In the violation-of-expectation paradigm, infants watch an event that is consistent or inconsistent with an actor’s belief; longer looking at the inconsistent event is taken as a sign of surprise.
Onishi and Baillargeon (2005) showed 15-month-olds an actor hiding a toy in one of two boxes. In some conditions the toy then moved to the other box while the actor could not see it, so the actor had a false belief. Infants looked longer when the actor reached into the box where the toy really was than when she reached into the box where she falsely believed it to be. The researchers concluded that infants of 15 months already expect people to act on their beliefs, even false ones. Anticipatory looking tasks use a similar logic: an eye-tracker records where a child looks just before the character returns, which shows where the child expects the character to search.
These findings started a major debate. If infants understand false beliefs, why do 3-year-olds fail the verbal tasks? Some researchers argue for two systems: an early, fast, implicit system for tracking others’ beliefs, and a later, flexible, explicit system that depends on language and executive function. Others argue that the infant findings can be explained by simpler processes, such as noticing what the actor last saw.
The replication record of implicit tasks is also mixed: Kulke et al. (2018) attempted to replicate four influential anticipatory-looking tasks in a large preregistered study; only one replicated reliably, and it failed once confounds were removed. Verbal false-belief findings, in contrast, replicate very reliably.
Adults pass simple false-belief tasks, so research with adolescents and adults needs harder measures. The Reading the Mind in the Eyes Test (RMET), revised by Baron-Cohen et al. (2001), shows photographs of the eye region of faces and asks participants to choose which of four words best describes what the person is thinking or feeling (for example “playful”, “comforting”, “irritated”, “bored”). In the revised version, a group of 15 adults with Asperger syndrome or high-functioning autism scored lower than 239 control participants, and scores were negatively correlated with the number of autistic traits measured by the Autism Spectrum Quotient.
The RMET is quick and sensitive to individual differences, but it has been criticised. It relies on a rich English vocabulary of mental-state words, so scores partly reflect verbal ability and education. The photographs show mostly white adults, which may disadvantage participants from other ethnic and cultural groups. It also measures recognition of emotional expressions from static pictures, which is closer to emotion recognition than to reasoning about beliefs. Its construct validity as a measure of theory of mind is therefore debated.
Measure | What the participant does | Strengths | Limitations |
|---|---|---|---|
Unexpected transfer (Maxi, Sally-Anne) | Predicts where a character will look for a moved object | Clear logic; control questions; very reliable age pattern across studies | Heavy demands on language, memory and inhibition; one pass/fail answer |
Unexpected contents (Smarties) | Says what another person, or the child earlier, thinks is in a container | Tests understanding of own as well as others’ beliefs | Same verbal demands; children may be reluctant to admit their own mistake |
Theory of mind scale (Wellman and Liu) | Completes a series of tasks of increasing difficulty | Measures a sequence of steps, not a single ability | Still verbal; designed in the United States, so the order may not be universal |
Violation of expectation and anticipatory looking | Looks at events or locations (infants and young children) | No language needed; can test infants | Looking time has several interpretations; mixed replication record |
Reading the Mind in the Eyes Test | Chooses a word to describe the mental state shown by a pair of eyes (adults) | Quick; detects individual differences in adults | Depends on vocabulary; mostly white faces; closer to emotion recognition |
Simon Baron-Cohen, Alan Leslie and Uta Frith aimed to test whether autistic children have a specific difficulty with theory of mind that is not explained by general intellectual disability. Autism is a lifelong developmental condition involving differences in social communication and interaction, together with restricted or repetitive behaviours and interests. The researchers proposed that a difficulty in representing other people’s mental states could explain the social and communication differences.
The study was a quasi-experiment with three groups of children, compared on the same task. The independent variable was the group, which could not be manipulated: 20 autistic children, 14 children with Down syndrome and 27 typically developing preschool children. The autistic children were older on average and had a higher average verbal mental age than the children with Down syndrome. The Down syndrome group was included as a control for general learning difficulty: if autistic children failed because of low intelligence, the Down syndrome group should fail too.
Each child was tested individually with two dolls, Sally and Anne. Sally put a marble in her basket and left; Anne moved the marble to her box; Sally returned. The child was asked the naming question (which doll is which), the belief question (“Where will Sally look for her marble?”), the reality question (“Where is the marble really?”) and the memory question (“Where was the marble in the beginning?”). The dependent variable was whether the child answered the belief question correctly, having passed the control questions.
Every child answered the naming, reality and memory questions correctly. On the belief question, 23 of the 27 typically developing children (85%) and 12 of the 14 children with Down syndrome (86%) passed. Only 4 of the 20 autistic children (20%) passed. The autistic children who failed pointed to where the marble really was, the box.
Sources: Baron-Cohen, S., Leslie, A. M. and Frith, U. (1985). Does the autistic child have a “theory of mind”? _Cognition, 21(1), 37–46._
Because the autistic children understood the story and remembered where the marble had been, their failure could not be explained by poor memory or misunderstanding. Because children with Down syndrome, with lower verbal mental ages, passed, it could not be explained by general learning difficulty. The researchers concluded that most autistic children have a specific difficulty in attributing beliefs to others, which they later described as mindblindness, and proposed that this could explain many of the social and communication differences in autism.
The study used a single task to measure a broad construct. Using dolls rather than real people may have made the task less meaningful for some children, and the task depends on understanding spoken questions. Later research found that many autistic children with good language pass first-order false-belief tasks, and that autistic adults often pass them easily, although they may solve them by careful reasoning rather than intuitively. The task therefore measures one specific skill in one format, and failing it does not show a complete absence of theory of mind.
The study framed autism as a deficit in the autistic person. Milton (2012), an autistic researcher, proposed the double empathy problem: misunderstanding between autistic and non-autistic people runs in both directions, because each group finds the other’s minds hard to read. Crompton et al. (2020) supported this in a study in which information was passed along chains of eight people: chains of autistic people passed on as much detail as chains of non-autistic people, but mixed chains passed on much less and reported lower rapport. This suggests that measuring theory of mind only from the non-autistic perspective builds a researcher bias into the conclusions.
The groups were not randomly allocated, so other differences between them, such as age, experience of schooling or communication styles, could contribute to the results. The samples were small, which means each child changed the percentages considerably: one more autistic child passing would raise the pass rate from 20% to 25%. Even so, the core finding that autistic children on average find false-belief tasks harder than matched controls has been repeated many times, which supports its reliability, while the explanation of autism as primarily a theory of mind deficit is now considered too narrow.
Theory of mind develops in every culture studied, but its timing varies between children and communities. Research on what explains this variation points to language, social interaction within the family, and brain development, which interact with one another.
Milligan, Astington and Dack (2007) combined 104 studies with a total of 8,891 children under 7. They found a moderate to large relationship between language ability and false-belief understanding, which remained when age was controlled. The relationship was stronger from earlier language to later false belief than the reverse, which suggests that language helps theory of mind to develop rather than simply resulting from it. Language may help in two ways: grammar such as “She thinks that…” gives children a way to represent a belief that differs from reality, and conversation exposes them to other people’s different points of view.
Research with deaf children supports a role for conversation. Peterson and Siegal (1995) tested 26 deaf children aged 8 to 13 in Australia, of normal intelligence, who used sign language but had grown up with hearing parents. 65% failed a simple false-belief task that typically developing children pass at 4 or 5, a rate similar to that of autistic children in earlier research.
Deaf children of hearing parents often have limited access to conversation at home in the early years, while later research found that deaf children of deaf, signing parents perform like hearing children. This natural experiment suggests that early access to conversation, not deafness itself, affects theory of mind.
Perner, Ruffman and Leekam (1994) found that children with more siblings passed false-belief tasks earlier, which they summarised in the title of their paper by describing theory of mind as “contagious”. Siblings provide experience of pretend play, arguments, teasing and deception, all of which require thinking about what another person knows or believes. Later studies have found that parents who often talk about thoughts and feelings with their children, for example explaining that a friend was sad because she thought her toy had been taken, tend to have children who develop theory of mind earlier.
These findings link theory of mind to sociocultural theory: understanding of minds is built partly through social interaction and the cultural tool of language, as Vygotsky proposed for other higher mental functions. The relationships are correlational, though. Families with more conversation about mental states may also differ in other ways, such as parents’ education, and children with better theory of mind may draw more such conversation from their parents.
Brain imaging studies have identified a network of regions active when people think about others’ mental states, including the medial prefrontal cortex and the temporo-parietal junction (TPJ). Saxe and Kanwisher (2003) used fMRI while adults read short stories. A region of the TPJ responded strongly when stories described a character’s beliefs, but not when stories described people’s physical appearance or when they described false representations that were not mental, such as an out-of-date photograph. The researchers concluded that this region is involved specifically in reasoning about the contents of other people’s minds.
YouTubeNeuroimaging shows where theory of mind is processed in adults, but it raises its own questions of measurement and causality. Activation in the TPJ during a task does not show that the TPJ causes theory of mind, and the regions involved continue to develop through adolescence, as brain maturation and synaptic pruning refine social brain networks. The most likely account is an interaction: maturation of brain networks makes theory of mind possible, and social experience, especially conversation, shapes how and when it develops.
If theory of mind is a universal part of human development, the same developmental pattern should appear in every culture. If it is shaped by culture, its timing, sequence or content should vary. Research supports a combination of both.
Wellman, Cross and Watson (2001) combined 178 studies of false-belief tasks in a meta-analysis. A model including age, country and task features explained 55% of the variation in children’s performance. In every country and across task variations, children moved from below-chance performance to above-chance performance during the preschool years. The researchers concluded that false-belief understanding reflects a genuine conceptual change, not an artefact of particular tasks, although the age at which it occurred differed between countries.
Liu et al. (2008) focused on Chinese children. Their meta-analysis compared 196 conditions from mainland China and Hong Kong, with more than 3,000 children, with 155 conditions from the United States and Canada. Chinese and North American children followed parallel trajectories, but the timing differed by as much as two years or more between communities. Callaghan et al. (2005) gave the same standardised false-belief procedure to children in Canada, India, Peru, Samoa and Thailand, and found that children crossed the false-belief milestone at around 5 years in every culture.
Study | Sample | Main finding |
|---|---|---|
Wellman, Cross and Watson (2001) | Meta-analysis of 178 studies in several countries | Same developmental shift everywhere; timing varies by country |
Liu et al. (2008) | Meta-analysis: China and Hong Kong vs USA and Canada | Parallel trajectories; timing differs by up to 2 years or more |
Callaghan et al. (2005) | Canada, India, Peru, Samoa, Thailand; one standard procedure | Children passed at about 5 years in every culture |
Shahaeian et al. (2011) | 77 Australian and 58 Iranian children aged 3–6 | Same overall rate, but a different order of steps |
Sources: Wellman, Cross and Watson (2001), _Child Development; Liu et al. (2008), Developmental Psychology; Callaghan et al. (2005), Psychological Science; Shahaeian et al. (2011), Developmental Psychology._
Shahaeian et al. (2011) tested whether the sequence of steps is universal, using Wellman and Liu’s theory of mind scale. Iranian and Australian children reached theory of mind at similar overall rates, but Iranian children understood knowledge access (that seeing leads to knowing) before diverse beliefs (that people can hold different opinions), while Australian children showed the reverse order. Iranian children followed the same order previously found in Chinese children. The researchers linked this to collectivist values that emphasise respect for elders, avoiding open disagreement and acquiring knowledge, so that children have less experience of people openly expressing different opinions.
Taken together, these studies suggest that the broad milestone of false-belief understanding is universal, while the timing and the order of the steps leading to it are influenced by culture. This illustrates how cultural dimensions such as collectivism and family conversational styles shape cognitive development without changing its basic direction.
Theory of mind matters because so much of social life depends on it. Children with better theory of mind tend to be more popular with peers, better at resolving conflicts and better at cooperative play. Understanding false belief also makes possible deliberate deception, such as lying, keeping secrets and playing hide-and-seek well, as well as understanding irony, sarcasm, persuasion and many jokes. It also helps children understand stories, whose plots often depend on characters who know, believe or want different things.
In cognition, theory of mind is closely linked to executive functions, especially inhibitory control and working memory. To pass a false-belief task, a child must hold two representations in mind (where the object is and where the character thinks it is) and inhibit the urge to answer from their own knowledge. Children with better executive functions tend to pass false-belief tasks earlier. This link raises a measurement question: some failures on false-belief tasks may reflect executive demands rather than a lack of understanding of beliefs.
Theory of mind also contributes to learning. Understanding that other people can have knowledge one lacks makes it possible to learn from teachers and to judge whose information to trust, and understanding that one’s own beliefs can be wrong is part of metacognition, thinking about one’s own thinking. In education, children who understand that others can hold different beliefs are better prepared for discussion and argument, which are central to learning in the zone of proximal development.
Large language models such as ChatGPT have raised a new measurement question. Kosinski (2024) gave 11 language models 40 newly written false-belief tasks, each with true-belief control versions and reversed versions; a model had to solve all eight versions to pass a task. Older models solved none, while ChatGPT-4 (June 2023) solved 75% of the tasks, which Kosinski compared to the performance of 6-year-old children. He suggested that a theory-of-mind-like ability might have emerged as a by-product of language learning.
Ullman (2023), responding to an earlier version of Kosinski’s study, showed that small changes to such tasks often caused a language model to fail, although they would not trouble a child who understood false belief. Examples included making the container transparent and having a trusted friend tell the character what was really inside. This suggests that the models may succeed by matching patterns in text similar to the many false-belief stories in their training data, rather than by representing beliefs. The debate shows that passing a task is not the same as having the ability it was designed to measure, which is also the central problem in measuring theory of mind in children.
Every theory of mind measure infers an unobservable mental ability from behaviour, so validity is the main issue. Verbal false-belief tasks show high reliability, with a consistent age pattern across hundreds of studies, but their demands on language and inhibition may hide earlier understanding. Implicit looking-time tasks remove these demands but are open to simpler interpretations and have a weaker replication record. Adult measures such as the RMET may measure vocabulary and emotion recognition as much as mental-state reasoning. The strongest conclusions come from triangulation across measures with different demands.
The tasks were designed in Western research settings, mostly with middle-class children, and they assume that talking openly about what others think is natural. In some cultures, speculating about another person’s private thoughts is considered inappropriate, which may affect children’s willingness to answer questions about beliefs. Research on autism shows a different form of bias: interpreting differences from non-autistic norms as deficits, as the double empathy problem highlights.
Most evidence about influences on theory of mind is correlational: language, siblings, conversation about mental states and executive functions all correlate with false-belief performance. Longitudinal designs, such as those in the language meta-analysis, give some evidence of direction, and natural comparisons such as deaf children of hearing versus deaf parents isolate one factor. Training studies, in which children are given conversations about mental states or feedback on false-belief tasks, provide the strongest causal evidence, although their effects are usually modest.
Researchers who emphasise biological maturation and a dedicated brain network see theory of mind as a universal human ability that develops on a similar timetable everywhere. Researchers who take a sociocultural perspective emphasise that it is constructed through conversation and culture. The cross-cultural evidence supports an interactionist view: the broad milestone is universal, while timing and sequence vary with language, family and culture.
Concept | How it applies to theory of mind |
|---|---|
Measurement | Verbal tasks vs implicit looking measures vs adult tests; task demands; triangulation |
Bias | Western task design; deficit framing of autism (double empathy problem); RMET faces and vocabulary |
Causality | Language, siblings and conversation correlate with ToM; longitudinal, natural and training studies test direction |
Perspective | Universal brain-based ability vs culturally constructed understanding; cross-cultural studies support both |
Responsibility | Research with young children and autistic participants; how findings about autism are communicated |
Theory of mind can appear in Paper 1 Section B as a scenario about children’s social behaviour, and in Section C as an essay on the role of theory of mind in development or on how it is measured, through a concept such as measurement or bias. False-belief studies are also suitable stimuli for Paper 2 Section B.
Theory of mind is the ability to attribute mental states to oneself and others and to understand that they can differ from one’s own and from reality.
Understanding false belief, measured by unexpected transfer (Maxi, Sally-Anne) and unexpected contents (Smarties) tasks, typically appears at about 4 to 5 years; second-order false belief at about 6 to 7.
Implicit looking-time measures, such as those used by Onishi and Baillargeon (2005), suggest earlier sensitivity to false belief, but their replication record is mixed; the Reading the Mind in the Eyes Test measures adult theory of mind.
Baron-Cohen, Leslie and Frith (1985) found that 85% of typically developing and 86% of Down syndrome children passed the Sally-Anne task, against 20% of autistic children.
The deficit account of autism is challenged by the double empathy problem: misunderstanding between autistic and non-autistic people runs in both directions.
Language, conversation about mental states, siblings and maturation of brain networks including the temporo-parietal junction all contribute to theory of mind.
Across cultures the false-belief milestone is universal, while its timing and the order of earlier steps vary (Liu et al., 2008; Shahaeian et al., 2011).
Theory of mind supports social competence, deception, communication and learning; research on it is above all a lesson in the measurement of unobservable abilities.
I can explain what theory of mind is and describe how it develops.
I can describe unexpected transfer, unexpected contents, implicit and adult measures of theory of mind.
I can evaluate strategies used to measure theory of mind.
I can describe and evaluate Baron-Cohen, Leslie and Frith (1985).
I can explain how language, family, culture and the brain contribute to theory of mind.
I can discuss the role of theory of mind in development and cognition.
I can evaluate theory of mind research with reference to measurement, bias, causality and perspective.