Theory of Mind

Abstract

Theory of mind is the mental process of attributing mental states—beliefs, desires, intentions, and knowledge—to oneself and to others, and of using those attributions to explain and predict behaviour. Coined in 1978 to ask whether chimpanzees represent the minds of others, it became central to developmental psychology once the false-belief task supplied a decisive test: understanding that a person can act on a belief the observer knows to be false. Typically developing children pass explicit false-belief tasks around the age of four, yet infants show sensitivity to others' beliefs far earlier, a discrepancy that drives much current debate. Competing accounts frame the capacity as an intuitive theory, a simulation of other minds, or a maturing cognitive module, while its neural basis centres on a dedicated social-brain network. Theory of mind now bridges developmental, comparative, and cognitive neuroscience.

Keywords: theory of mind, false belief, mentalizing, mental-state attribution, social cognition

A child watches another child hide a chocolate in a cupboard and leave the room; while the first is away, someone moves the chocolate to a drawer. Asked where the returning child will look, a three-year-old typically points to the drawer, where the chocolate actually is, while a five-year-old points to the cupboard, where the other child still believes it to be. That small shift marks one of the most studied transitions in cognitive development: the arrival of an explicit understanding that other people act on their beliefs about the world rather than on the world itself. The capacity to represent such beliefs, desires, and intentions—and to treat them as causes of behaviour—is what psychologists call theory of mind, a phrase that has organised research across development, autism, primate cognition, and social neuroscience since the late 1970s (Frith & Frith, 2005).

What Theory of Mind Is

MeSH defines theory of mind as “the ability to attribute mental states, e.g., beliefs, desires, or intentions to others, and to understand that they may differ from one's own.” The definition captures two components that the research literature keeps distinct. The first is mental-state attribution: representing that a person holds an internal state—a belief, a desire, an intention, an emotion—that is directed at the world. The second is decoupling: grasping that this state need not match reality or one's own perspective, so that a person can want what is bad for them, intend what they fail to do, or believe what is false. It is the second component that makes the capacity powerful and that the classic tasks are designed to probe, because attributing a state that happens to be true can be mimicked by simply reading the world rather than the mind.

The term entered psychology through a question about animals. David Premack and Guy Woodruff asked whether a chimpanzee has a theory of mind, meaning whether it imputes mental states to others rather than merely responding to their behaviour, and named the whole system a theory because these states are unobservable posits used to predict and explain what agents do (Premack & Woodruff, 1978). The commentaries on that paper proposed the test that would dominate the field: if a subject truly represents another's belief, it should be able to predict action driven by a belief the subject knows to be false. Three decades of comparative work have since refined the original question, showing that chimpanzees understand a good deal about what others see, know, and intend while stopping short of the human capacity to represent false beliefs (Call & Tomasello, 2008).

The False-Belief Test

The false-belief task turned a philosophical question into an experiment. Heinz Wimmer and Josef Perner built the first version around a story character, Maxi, who puts chocolate in one location and leaves; the chocolate is moved in his absence, and the child is asked where Maxi will look for it. Answering correctly requires the child to set aside their own knowledge of the chocolate's real location and reason from Maxi's outdated belief. Wimmer and Perner found a sharp developmental shift: children below about four answered from reality, while older children answered from Maxi's false belief (Wimmer & Perner, 1983). The result was striking because the younger children were not merely less accurate—they failed systematically, as though the very possibility of a belief diverging from the facts were unavailable to them.

Simon Baron-Cohen, Alan Leslie, and Uta Frith adapted the paradigm into the Sally-Anne task and used it to probe autism. Sally hides a marble in a basket and leaves; Anne moves it to a box; the child is asked where Sally will look. Most typically developing four-year-olds and most children with Down syndrome—a group with intellectual disability included as a control to show any deficit was not a by-product of low general ability—passed, but the large majority of autistic children in the sample failed, pointing to the returning character's knowledge of the marble's true location rather than her false belief. The authors argued that autism involves a specific impairment in the capacity to represent mental states, decoupled from general intelligence (Baron-Cohen, Leslie, & Frith, 1985). The finding gave developmental psychology one of its most influential links between a cognitive construct and a clinical condition, and made the false-belief task a standard instrument. The strong form of that hypothesis—that a theory-of-mind deficit is the core explanation of autism—was later qualified: the deficit is neither universal across autistic individuals nor unique to autism, and it leaves the condition's non-social features unexplained (Tager-Flusberg, 2007). A later meta-analysis of 178 conditions confirmed that the age-four transition is robust across many variations of the task—different objects, motives, and ways of asking the question—while showing that specific manipulations reliably shift the age of passing, evidence that the change reflects a genuine conceptual development rather than an artefact of any one procedure (Wellman, Cross, & Watson, 2001). The demonstration below runs the unexpected-transfer task and lets its critical variable—whether the character witnesses the move—be toggled, so that the difference between a true and a false belief can be seen to drive the predicted search.

Development of Theory of Mind

The false-belief transition is one milestone in a longer developmental sequence. Before it, toddlers already understand a great deal about desires and intentions: that people feel bad when they do not get what they want, that a person looking at one object intends to act on it, and that seeing leads to knowing. Henry Wellman and colleagues showed that these understandings emerge in a reliable order, forming a scale on which desires are grasped before beliefs and diverse beliefs before false beliefs, so that theory of mind develops as an ordered progression rather than a single switch (Wellman et al., 2001). The concept that reorganises around age four is representation itself: the child comes to understand that the mind does not simply mirror the world but represents it, and that representations can be partial, outdated, or wrong.

Alison Gopnik and Janet Astington sharpened this point by testing the false-belief transition alongside two related tasks: the appearance-reality distinction, in which a sponge painted to look like a rock is both a sponge and rock-like, and the representational-change task, in which children must recall their own earlier false belief about a deceptive container. Performance on all three improved together around the same age, which the authors took as evidence for a single underlying acquisition—an understanding of mental representation—rather than a set of unrelated skills (Gopnik & Astington, 1988). This convergence is a central reason the age-four shift is read as conceptual: it is not confined to one story format but appears wherever a task requires treating a mental state as a representation that can misfire. The transition is not purely conceptual, however: individual differences in false-belief performance are strongly predicted by inhibitory control, the capacity to suppress a prepotent response, suggesting that passing the task also requires the executive resources to hold one's own knowledge in check while reasoning about another's (Carlson & Moses, 2001). Recent work extends the timeline downward, arguing that the foundations are laid in infancy and that the later verbal milestone reflects the added demands of the task rather than the first appearance of the concept (Scott & Baillargeon, 2017). The demonstration lets the age of a hypothetical child be varied and plots an approximate probability of passing a standard false-belief task, tracing the characteristic rise through the preschool years.

Competing Accounts

What kind of cognitive achievement is theory of mind? Three families of answer have structured the debate. The theory-theory holds that children construct an intuitive theory of mind much as scientists construct theories, revising it in the face of evidence, so that the age-four shift is a genuine conceptual change from a desire-based to a belief-based psychology. The simulation theory holds instead that people understand others not by consulting a theory but by using their own mind as a model, imagining themselves in the other's situation and reading off the mental states that result. The modularity view holds that mental-state representation is the work of a specialised, largely innate cognitive system that matures on a biological timetable, its outputs fast and mandatory rather than reasoned.

Alan Leslie's work anchors the modular account. He argued that the capacity to represent mental states is a species of metarepresentation—the ability to hold a representation of a representation—and that its earliest expression is not false belief but pretend play, in which a two-year-old treats a banana as a telephone while keeping the real and pretended meanings apart. On this view a dedicated Theory-of-Mind Mechanism supplies the format for representing attitudes such as believes and pretends, and the developmental change reflects the maturing of processes that select the right content rather than the acquisition of the concept itself (Leslie, 1987). A later synthesis proposed that these positions need not compete outright: humans may possess two systems, an early-developing, efficient, and inflexible one that tracks belief-like states from infancy, and a later, cognitively demanding one that supports explicit, flexible reasoning about beliefs—a division that reconciles infants' precocious performance with preschoolers' failures on verbal tasks (Apperly & Butterfill, 2009). Others question whether the ability is a biological given at all, arguing that mentalizing concepts are acquired through cultural learning and language much as literacy is (Heyes & Frith, 2014). Table 1 sets the accounts side by side, and the demonstration that follows illustrates the recursive structure—beliefs about beliefs—that any account must explain.

Implicit Theory of Mind in Infancy

The clean age-four story was complicated by evidence from infants far too young to speak. Kristine Onishi and Renée Baillargeon adapted the false-belief scenario for 15-month-olds using looking time: infants watched an actor hide a toy and then reach for it after the toy had, or had not, been moved in her absence. The infants looked longer when the actor reached to the location inconsistent with her belief—as though they expected her to act on where she thought the toy was, not on where it actually was (Onishi & Baillargeon, 2005). Because the measure requires no verbal response and no explicit prediction, the result suggested that some form of belief tracking is present more than two years before children pass the standard task. Related studies reported that even seven-month-olds and adults are influenced by another agent's belief, their responses shifting with a bystander's perspective even when it was irrelevant to the task—though whether this reflects genuine belief tracking or a lower-level attentional effect remains contested (Kovács, Téglás, & Endress, 2010).

These findings provoked the field's central current controversy. One reading, advanced by Scott and Baillargeon, is that infants possess a genuine, if implicit, representation of belief, and that the later verbal task simply adds executive and linguistic demands that mask an early competence (Scott & Baillargeon, 2017). A more deflationary reading holds that infants deploy simpler, belief-like processes—tracking what an agent has registered about an object—that fall short of full belief attribution, consistent with a two-systems architecture (Apperly & Butterfill, 2009). The debate has been sharpened by replication concerns: several high-profile implicit false-belief effects have proven difficult to reproduce, and a large-scale reinvestigation of the anticipatory-looking paradigm failed to find the expected pattern, cautioning against strong claims built on a single measure (Kulke et al., 2018). The question of what infants really represent remains open and is among the most active in developmental science.

The Social Brain

Functional neuroimaging has localised mentalizing to a consistent set of regions. Rebecca Saxe and Nancy Kanwisher showed that the temporoparietal junction responds far more strongly when people read stories that require inferring a character's thoughts than when they read closely matched stories about physical states of affairs or a person's appearance, isolating a region engaged specifically by the attribution of beliefs rather than by social stimuli in general (Saxe & Kanwisher, 2003). Convergent work identified a broader mentalizing network—the right and left temporoparietal junction, the medial prefrontal cortex, the posterior superior temporal sulcus, and the temporal poles—that is recruited when people think about the contents of other minds, and that is partly dissociable from the systems supporting the perception of faces and biological motion (Frith & Frith, 2005). The right temporoparietal junction in particular behaves less like a general social hub and more like a region tuned to the abstract representation of beliefs. Figure 1 shows the principal nodes of this network and their approximate cortical locations.

Figure 1

The Core Mentalizing Network on the Lateral and Medial Cortical Surface

Worked Example

Consider a second-order false-belief problem, the kind used to probe theory of mind beyond the preschool years. In the ice-cream-van story, John and Mary both see an ice-cream van in the park. Mary goes home. The van then moves to the church, which John sees but Mary does not; unknown to John, Mary later learns of the move from the driver. The question is where John thinks Mary will go for ice cream. Answering requires holding two embedded mental states at once: a belief about a belief. John believes that Mary believes the van is still in the park, because as far as John knows, Mary never learned it had moved. The correct answer is the park.

The reasoning can be laid out as levels of embedding. A first-order attribution represents one person's state about the world: Mary believes the van is in the park. A second-order attribution represents one person's state about another's state: John believes that Mary believes the van is in the park. First-order false belief is typically mastered around age four; second-order false belief, which requires tracking the nesting and the separate information each character received, is usually not passed until about six or seven, the finding that the ice-cream-van task was originally devised to establish (Perner & Wimmer, 1985). The jump illustrates that theory of mind is not a single hurdle but a recursive capacity: each added level of embedding—A thinks that B thinks that C thinks—imposes a further load, and the number of levels a person can reliably hold is limited. The worked case makes concrete why the false-belief test is diagnostic: its difficulty comes not from the facts of the world, which are fully available, but from the requirement to represent a mind whose contents diverge from those facts.

Discussion

Theory of mind has proved unusually productive as a construct because a single task made an abstract capacity measurable and then travelled across fields. The false-belief test connected developmental psychology to clinical questions about autism, to comparative questions about primate cognition, and to neuroscientific questions about the social brain, and each connection fed back to refine the construct (Baron-Cohen et al., 1985; Call & Tomasello, 2008; Saxe & Kanwisher, 2003). The robustness of the age-four transition across 178 conditions gave the field a rare fixed point, and the convergence of false belief with the appearance-reality and representational-change tasks pinned that transition to an understanding of mental representation rather than to any one story (Wellman et al., 2001; Gopnik & Astington, 1988).

The central tension in the current literature is between this orderly picture and the evidence that infants track beliefs long before they can speak. If the implicit findings are taken at face value, the age-four milestone marks not the birth of the concept but the point at which an early competence becomes usable in explicit, verbal reasoning; if they are treated cautiously, they may reflect simpler processes that only resemble belief attribution (Onishi & Baillargeon, 2005; Apperly & Butterfill, 2009). The two-systems proposal offers a principled way to hold both facts together, and the modular and constructivist accounts continue to divide over how much of the capacity is given and how much is built (Leslie, 1987; Heyes & Frith, 2014). What is no longer in doubt is that representing other minds is a distinct achievement with an identifiable developmental course and neural basis, even as its architecture remains contested.

Current Directions

Two concerns dominate contemporary work. The first is measurement. The replication difficulties that beset implicit false-belief paradigms have prompted scrutiny of the whole toolkit: many widely used theory-of-mind tasks turn out to have poor test-retest reliability and to correlate weakly with one another, which undermines their use as measures of a single stable trait and complicates individual-differences research (Warnell & Redcay, 2019). A related critique argues that the field has stretched the label theory of mind to cover tasks that make quite different demands—visual perspective-taking, emotion recognition, belief reasoning—and calls for sharper operational definitions before results are pooled (Quesque & Rossetti, 2020). Advanced adult instruments that ask observers to infer complex mental states from photographs of the eye region alone are part of this scrutiny, prized for sensitivity in autism research yet debated as to exactly which component of mentalizing they index.

The second is integration. A large meta-analysis of neuroimaging studies has begun to fractionate the mentalizing network by the kind of inference required, distinguishing the cognitive routes engaged by different theory-of-mind tasks and mapping them onto partly separable neural systems, a step toward replacing a single undifferentiated construct with a structured model (Schurz et al., 2021). Alongside this, the cultural-learning perspective reframes the developmental question, asking how much of mind-reading is transmitted through language and social practice rather than maturing on a fixed schedule, and pointing toward cross-cultural and training studies as the decisive evidence (Heyes & Frith, 2014). Together these strands are moving the field from asking whether an agent has a theory of mind toward specifying which processes, at which ages, in which contexts, do the work.

Glossary

Appearance-reality distinction.

The understanding that an object can look like one thing while being another, such as a sponge painted to resemble a rock; it matures alongside false-belief understanding.

Decoupling.

Holding a represented state apart from reality, so that a belief or pretence can be entertained without being taken as true; a prerequisite for representing false beliefs.

False-belief task.

A test in which a character holds a belief the observer knows to be false; predicting the character's action requires reasoning from the belief rather than from reality.

First-order belief.

A mental state one person holds about the world, such as Sally believing the marble is in the basket; the level tested by the standard false-belief task.

Implicit theory of mind.

Sensitivity to others' mental states expressed in spontaneous measures such as looking time, present in infancy before children can pass explicit verbal tasks.

Mental-state attribution.

The act of ascribing an internal state—a belief, desire, intention, or emotion—to oneself or another and treating it as a cause of behaviour.

Mentalizing.

A near-synonym for theory of mind in cognitive neuroscience, denoting the process of inferring the contents of other minds; associated with a dedicated brain network.

Metarepresentation.

A representation of a representation, such as representing that someone believes something; on modular accounts it is the computational format underlying theory of mind.

Second-order belief.

A belief about a belief, such as John thinking that Mary thinks the van is in the park; typically mastered around age six or seven.

Simulation theory.

The proposal that people understand others by using their own mind as a model, imagining the other's situation and reading off the resulting mental states.

Temporoparietal junction.

A cortical region at the border of the temporal and parietal lobes; its right side responds selectively when people reason about others' beliefs.

Theory of mind.

The capacity to attribute mental states to oneself and others and to understand that they may differ from one's own and from reality.

Theory-theory.

The proposal that everyday understanding of mind is an intuitive theory, built and revised from evidence, whose revision produces the age-four conceptual change.

Two-systems account.

The view that theory of mind comprises an early, efficient, inflexible system for tracking belief-like states and a later, flexible system for explicit belief reasoning.

Unexpected-transfer task.

The original false-belief format in which an object is moved while a character is absent, so the character holds a false belief about its location; the Maxi and Sally-Anne tasks are versions.

Key Researchers

Renée Baillargeon (contemporary). Developmental psychologist at the University of Illinois Urbana-Champaign; with Kristine Onishi showed that 15-month-old infants track an actor's false belief in a looking-time task, reframing the developmental timetable of theory of mind. Faculty Page - Wikipedia - ORCID

Simon Baron-Cohen (contemporary). Developmental psychopathologist at the University of Cambridge; co-authored the 1985 Sally-Anne study linking autism to a theory-of-mind deficit and devised the Reading the Mind in the Eyes test. Faculty Page - Wikipedia - ORCID

Uta Frith (contemporary). Cognitive-development pioneer at University College London; co-authored the 1985 autism study and, with Chris Frith, helped found the cognitive neuroscience of mentalizing and the mapping of the social brain. Faculty Page - Wikipedia - ORCID

Alan M. Leslie (contemporary). Cognitive scientist at Rutgers University; argued that theory of mind rests on a domain-specific metarepresentational mechanism first expressed in pretend play, the leading modular account. Faculty Page - Wikipedia - ORCID

Josef Perner (contemporary). Developmental psychologist at the University of Salzburg; with Heinz Wimmer introduced the unexpected-transfer false-belief task, the paradigm that made explicit belief attribution testable in children. Faculty Page - Wikipedia - ORCID

David Premack (1925-2015). Comparative psychologist at the University of Pennsylvania; with Guy Woodruff coined the phrase “theory of mind” in 1978, launching the research programme and the false-belief test that grew from its commentaries. Wikipedia

Rebecca Saxe (contemporary). Cognitive neuroscientist at the Massachusetts Institute of Technology; with Nancy Kanwisher identified the right temporoparietal junction as a region selectively engaged when people reason about others' beliefs. Faculty Page - Wikipedia - ORCID

Henry M. Wellman (contemporary). Developmental psychologist at the University of Michigan; with David Cross and Julanne Watson produced the definitive meta-analysis of false-belief performance and mapped the ordered sequence of theory-of-mind development. Faculty Page - Wikipedia - ORCID

Frequently Asked Questions

What is theory of mind?
Theory of mind is the capacity to attribute mental states such as beliefs, desires, and intentions to oneself and others, and to understand that these states can differ from one's own and from reality. It is what allows people to explain and predict behaviour in terms of what agents think and want rather than only what they do (Premack & Woodruff, 1978).

Why is the false-belief task so important?
Because passing it requires representing a belief that conflicts with the facts, which cannot be done by simply reading the world. A child who predicts that a character will search where the character wrongly thinks an object is, rather than where it really is, demonstrates that they represent the mind as distinct from reality (Wimmer & Perner, 1983).

At what age do children develop theory of mind?
Children typically pass explicit false-belief tasks around the age of four, a transition confirmed across many task variations. Earlier understandings of desire and intention appear before this, and second-order beliefs—thoughts about thoughts—are usually mastered by about six or seven (Wellman et al., 2001).

Do infants have a theory of mind?
Infants as young as 15 months show sensitivity to others' beliefs in looking-time studies, well before they can pass verbal tasks. Whether this reflects genuine belief attribution or a simpler belief-like process is debated, and some of the effects have been hard to replicate (Onishi & Baillargeon, 2005; Kulke et al., 2018).

How is theory of mind related to autism?
A classic 1985 study found that most autistic children failed the Sally-Anne false-belief task despite adequate general ability, suggesting a specific difficulty in representing mental states. The finding has been influential, though later work shows the picture is more nuanced than a single deficit (Baron-Cohen et al., 1985).

What are the main theories of how it works?
Three families dominate: the theory-theory, on which understanding of mind is an intuitive theory revised from evidence; the simulation theory, on which people model others using their own minds; and the modular view, on which a specialised mechanism represents mental states. A two-systems account seeks to combine the strengths of these positions (Leslie, 1987; Apperly & Butterfill, 2009).

Which parts of the brain support theory of mind?
Mentalizing engages a consistent network including the temporoparietal junction, medial prefrontal cortex, posterior superior temporal sulcus, and temporal poles. The right temporoparietal junction is especially tied to reasoning about beliefs rather than to social processing in general (Saxe & Kanwisher, 2003; Frith & Frith, 2005).

Do animals have a theory of mind?
The original question was asked of chimpanzees. Three decades of research indicate that chimpanzees understand a good deal about what others see, know, and intend, but do not show the human capacity to represent false beliefs, so they possess some components of theory of mind without the full system (Call & Tomasello, 2008).

References