The Sally-Anne Test (Theory of Mind) – Simon Baron-Cohen, Alan Leslie, and Uta Frith

The human mind possesses a remarkable, near-automatic capacity to look beyond the surface of physical behavior and perceive the unseen architecture of the internal world. Every human social exchange—from the subtle negotiation of a business contract to the shared laughter of intimate companions—relies upon an unspoken, continuous cognitive calculus: the attribution of mental states. We routinely assign beliefs, desires, intentions, knowledge, and pretenses to those around us, operating on the foundational assumption that other individuals navigate reality not according to how the world objectively exists, but according to their subjective, internal representations of it. In the annals of developmental psychology and cognitive neuropsychology, this foundational social-cognitive capacity is known as Theory of Mind (ToM) or mentalizing.

For decades, developmental psychologists struggled to determine precisely when and how this capacity crystallizes during ontogeny, and whether its disruption could explain the profound social and communicative differences observed in specific neurodevelopmental conditions. While philosophers from René Descartes to Daniel Dennett had long explored the epistemological complexities of intentionality and mental representation, empirical psychology required an elegant, unambiguous behavioral litmus test. This operational breakthrough occurred when developmental researchers realized that the most definitive demonstration of an individual’s grasp of another person’s mind requires showing that they understand that someone else can hold a belief that is factually incorrect—a false belief. If an observer can predict another person’s actions based on a mistaken mental representation rather than on the physical reality known to the observer, that observer has unequivocally decoupled mind from world.

In 1985, a groundbreaking collaborative paper titled “Does the autistic child have a ‘theory of mind’?” transformed the landscape of cognitive science. Authored by Simon Baron-Cohen, Alan M. Leslie, and Uta Frith, the study introduced a deceptively simple puppet scenario known universally today as the Sally-Anne Test. By comparing the performance of autistic children against both typically developing peers and children with Down syndrome, the researchers demonstrated a striking, selective dissociation: despite possessing sufficient chronological and non-verbal mental ages, the overwhelming majority of autistic children failed to appreciate Sally’s false belief, pointing instead to the physical location of the hidden object. This finding revolutionized autism research, replacing mid-twentieth-century psychodynamic speculations with a domain-specific cognitive model and establishing a paradigm that continues to guide developmental psychology, neurobiology, and clinical practice nearly four decades later.

1. Introduction to the Sally-Anne Test and Theory of Mind

1.1 Defining Theory of Mind (ToM) in Cognitive Psychology

The term “Theory of Mind” was formally introduced to the behavioral sciences not through the study of human children, but through comparative primatology. In their seminal 1978 paper, David Premack and Guy Woodruff posed the provocative question: “Does the chimpanzee have a theory of mind?” In their experimental investigations with Sarah, a captive chimpanzee, Premack and Woodruff defined Theory of Mind as the capacity to attribute mental states—such as purpose, intention, knowledge, belief, doubt, and pretense—to oneself and others. They argued that such a system of inferences warrants the label of a “theory” precisely because mental states are not directly observable; they are theoretical constructs deployed by an organism to make systematic, highly accurate predictions about the observable behaviors of conspecifics.

Within cognitive psychology, Theory of Mind represents the bridge between mere sensory perception of bodily kinetics and genuine social comprehension. Crucially, the construct demands a distinct conceptual boundary between first-person self-awareness and second-order interpersonal cognitive modeling. While an organism may experience basic internal drives, affective sensations, or primary perceptual awareness (first-person subjectivity), mentalizing requires constructing a metarepresentational model of another agent’s psychological perspective. This secondary modeling process necessitates that the observer track not only what the other agent perceives, but how that agent evaluates, stores, and updates informational states over time.

From an evolutionary perspective, the emergence of Theory of Mind is widely considered one of the primary drivers of hominin encephalization. As articulated within the Social Brain Hypothesis advanced by Robin Dunbar, the structural complexity of primate neocortical expansion correlates directly with social group size and the corresponding demands of navigating intricate social hierarchies. In ancestral human environments, survival was profoundly tethered to social cohesion, alliance formation, deceptive maneuvers, cooperative foraging, and cultural transmission. Navigating these multidimensional matrices required a cognitive apparatus capable of instantaneous mental state attribution, transforming raw social perception into predictive, strategic social intelligence.

1.2 Historical Context of Autism Research in the Early 1980s

To fully appreciate the paradigm shift inaugurated by the Sally-Anne Test, one must examine the clinical and theoretical landscape of autism research prior to 1985. Throughout the 1950s, 1960s, and into the 1970s, the field remained heavily shadowed by psychodynamic and affective hypotheses. Influenced by psychoanalytic formulations popularized by Bruno Bettelheim, childhood autism was often mischaracterized as an emotional withdrawal or a defensive psychological retreat triggered by cold, unreciprocating parental figures—a toxic conceptualization encapsulated in the stigmatizing “refrigerator mother” myth.

By the late 1970s, pioneering empirical researchers such as Michael Rutter, Marian Hermelin, and Neil O’Connor had begun dismantling these psychodynamic formulations, demonstrating that autism is a biologically based neurodevelopmental condition characterized by prominent cognitive, communicative, and perceptual atypicalities. However, the exact cognitive architecture underlying the social and communicative impairments defined by Leo Kanner in 1943 remained elusive. Most researchers conceptualized the cognitive deficits in autism as either broad, domain-general intellectual disabilities or as primary disturbances in emotional and affective processing, such as an inability to experience fundamental intersubjective relatedness.

The critical inflection point occurred at the Medical Research Council (MRC) Cognitive Development Unit in London, directed by John Morton. The MRC Unit provided an intellectually rigorous incubator that merged cognitive psychology, computational models of mind, and clinical neuropsychology. Researchers at the Unit sought to identify whether the social deficits in autism could be traced to a domain-specific information processing failure—a selective cognitive lesion—rather than a global intellectual impairment. It was within this rich interdisciplinary environment that Uta Frith, along with her doctoral student Simon Baron-Cohen and cognitive theorist Alan Leslie, conceptualized a bold hypothesis: that the quintessential social impairments of autism stemmed from an underlying, circumscribed failure in the cognitive mechanism responsible for computing Theory of Mind.

1.3 The Seminal 1985 Publication: ‘Does the Autistic Child Have a “Theory of Mind”?’

In 1985, Baron-Cohen, Leslie, and Frith published their groundbreaking study, “Does the autistic child have a ‘theory of mind’?”, in the international journal Cognition. The paper opened by drawing a direct conceptual lineage from Premack and Woodruff’s work, transposing the theoretical question from non-human primates to children diagnosed with autism. The authors hypothesized that the core social-communicative pathology observed in autism was neither an affective refusal of contact nor a consequence of general intellectual delay, but rather a specific, innate cognitive deficit in attributing mental states to oneself and other social agents.

The conceptual leap achieved by the authors was the bridge they built between philosophical epistemology and empirical, pediatric behavioral assessment. Drawing upon contemporary philosophical debates regarding how intentionality could be empirically tested without falling into the trap of behavioral mimicry or direct physical cues, they adapted a narrative false-belief experimental framework. By stripping away extraneous linguistic and computational complexities, the authors designed a standardized, visually mediated paradigm—the Sally-Anne scenario—that could be administered directly to children with communicative challenges.

The academic reception of the 1985 paper was immediate and transformative. It catalyzed a profound realignment within developmental psychology, shifting attention from Piagetian domain-general stages of cognitive development toward modular, domain-specific models of social cognition. Within autism studies, the paper provided the first empirically robust, cognitively grounded explanation for the triad of impairments (social interaction, communication, and imaginative flexibility) that characterized the condition. Today, the 1985 paper stands as one of the most cited works in cognitive science, serving as the conceptual foundation for thousands of subsequent investigations into social neuroscience, developmental trajectories, and neurodivergent cognition.

2. Biographical and Academic Profiles: Baron-Cohen, Leslie, and Frith

2.1 Simon Baron-Cohen: Evolutionary Developmental Psychology and Neurodiversity

Simon Baron-Cohen undertook this seminal work as part of his doctoral research under the mentorship of Uta Frith at the MRC Cognitive Development Unit. Following the publication of the 1985 study, Baron-Cohen dedicated his career to refining and expanding the cognitive and neurobiological models of autism. In his landmark 1995 monograph, Mindblindness: An Essay on Autism and Theory of Mind, he codified the conceptual framework that autism is fundamentally marked by a degree of “mindblindness”—a selective difficulty in inferring the intentions, beliefs, feelings, and perspectives of other individuals.

Baron-Cohen later formulated the Empathizing-Systemizing (E-S) Theory, which posits that human cognitive profiles can be understood along two continuous psychological dimensions: empathizing (the drive to identify another person’s emotions and thoughts, and to respond with an appropriate emotion) and systemizing (the drive to analyze or construct systems governed by rules). Under this paradigm, Baron-Cohen argued that the cognitive phenotype of autism often features hyper-developed systemizing skills alongside reduced spontaneous empathizing capacities. Extending this line of inquiry, he proposed the Extreme Male Brain (EMB) theory, linking fetal androgen exposure to specific cognitive profiles.

As the long-time Director of the Autism Research Centre (ARC) at the University of Cambridge, Baron-Cohen spearheaded the development of essential developmental screening instruments, most notably the Checklist for Autism in Toddlers (CHAT) and its quantitative successor, the Q-CHAT. These instruments operationalized key early mentalizing precursors, such as protodeclarative pointing and gaze-following, allowing clinicians to screen for autism in children as young as 18 months. Throughout his later career, Baron-Cohen has actively engaged with the neurodiversity movement, emphasizing that cognitive differences, including heightened sensory sensitivity and superior pattern recognition, must be understood as cognitive variations rather than purely reductive pathological deficits.

2.2 Alan M. Leslie: Modular Cognitive Architectures and Metarepresentation

Alan M. Leslie brought a rigorous computational and cognitive architectural perspective to the 1985 collaboration. Specializing in infant cognition and developmental epistemology, Leslie was deeply influenced by Jerry Fodor’s modularity hypothesis. Leslie sought to map the precise representational mechanisms that permit the human brain to process non-veridical mental representations—that is, thoughts about states of affairs that do not correspond to physical reality—without collapsing into cognitive chaos.

Leslie proposed the existence of an innate, specialized neurocomputational system termed the Theory of Mind Mechanism (ToMM). According to Leslie’s theoretical model, ToMM operates in tandem with a domain-general inhibitory component called the Selection Processor (SP). While ToMM computes possible mental states by processing “metarepresentations,” the Selection Processor actively inhibits salient, competing information regarding actual physical reality (e.g., the true location of an object) to permit the correct selection of a false belief. In his groundbreaking 1987 theoretical paper in Psychological Review, Leslie linked the origins of ToMM to early pretend play, arguing that both pretense and false-belief reasoning rely upon a cognitive process known as “decoupling.”

At Rutgers University, where Leslie established the Center for Cognitive Science (RuCCS), his laboratory continued to dissect the fine-grained computational properties of the infant mind. Leslie’s work provided crucial empirical evidence demonstrating that infants possess an early-developing, core-knowledge system for physical causality and agency, which serves as the developmental substrate for later mentalistic computations. His rigorous theoretical separation between modular representational architecture and domain-general executive resources remains one of the cornerstone contributions to contemporary cognitive science.

2.3 Uta Frith: Cognitive Phenotypes and the Weak Central Coherence Theory

Uta Frith is widely acknowledged as one of the preeminent pioneers of contemporary autism and developmental dyslexia research. Trained at the Institute of Psychiatry in London under Hans Eysenck, Frith established herself as a vital intellectual conduit, notably translating and introducing the clinical observations of Hans Asperger to the English-speaking academic community. Her historical scholarship, combined with rigorous experimental methodologies, elevated autism from an obscure psychiatric diagnostic category to a central model for cognitive neuropsychology.

While championing the Theory of Mind deficit model alongside Baron-Cohen and Leslie, Frith recognized that mentalizing impairments could not fully explain the entire clinical presentation of autism. Specifically, Theory of Mind deficits failed to account for non-social features, such as exceptional attention to fine detail, islets of superior perceptual ability, savant syndrome, and sensory processing differences. To explain these dimensions, Frith formulated the Weak Central Coherence (WCC) Theory. This cognitive model posits that neurotypical cognition is characterized by a drive to pull together disparate information to extract higher-level, contextual, “gist” meaning, whereas autistic cognition demonstrates a local, detail-focused processing bias.

Frith’s leadership at the Institute of Cognitive Neuroscience at University College London (UCL) fostered an entire generation of cognitive scientists. She was among the first researchers to leverage emerging neuroimaging technologies (such as positron emission tomography and functional magnetic resonance imaging) to map the neural substrates of the “social brain” in both neurotypical and autistic populations. Throughout her career, Frith has been an indefatigable advocate for public engagement with science, the advancement of women in academic research, and the absolute necessity of bridging clinical diagnostics with fundamental neuroscience.

3. Theoretical Foundations: The Concept of False-Belief Understanding

3.1 Philosophical Roots: Dennett, Fodor, and Propositional Attitudes

The empirical operationalization of Theory of Mind owes an profound debt to the philosophy of mind that flourished during the 1970s and 1980s. Central to this philosophical infrastructure was Daniel Dennett’s formulation of the intentional stance. Dennett argued that predicting the behavior of complex entities can be achieved at three distinct levels: the physical stance (governed by the laws of physics), the design stance (governed by functional purpose), and the intentional stance. The intentional stance involves treating the entity as a rational agent possessing beliefs, desires, and goals, and calculating its future actions based on the content of those internal states.

Simultaneously, Jerry Fodor’s representational theory of mind established that mental states possess intentionality—they are directed at, or about, states of the world. Fodor conceptualized these states as propositional attitudes: relations between a psychological agent and a proposition (e.g., “Mary [agent] believes [attitude] that ‘the keys are on the kitchen table’ [proposition]”). Crucially, propositional attitudes do not conform to standard truth-conditional logic. They exhibit what philosophers term referential opacity (or intensionality). In referentially opaque contexts, one cannot substitute coreferential terms without potentially altering the truth value of the psychological statement. For example, Mary may believe that Bruce Wayne is a billionaire, yet she may not believe that Batman is a billionaire, despite Bruce Wayne and Batman being the identical physical person.

In 1978, responding to Premack and Woodruff’s chimpanzee experiments in the journal Behavioral and Brain Sciences, Dennett, along with philosopher Jonathan Bennett and cognitive scientist Gilbert Harman, pointed out a critical methodological flaw in previous animal and child experiments: one cannot prove that an organism has a Theory of Mind merely by showing that it can predict an agent’s behavior when that agent holds a true belief. If the agent’s internal belief aligns perfectly with the actual physical state of reality, an external observer could successfully predict the agent’s actions simply by observing physical reality, without ever representing the agent’s internal mental state at all. Therefore, the only definitive proof that an observer relies on a Theory of Mind is the false-belief task: situations where the agent’s internal belief diverges completely from physical truth.

3.2 True Belief versus False Belief: The Crucial Epistemological Threshold

The epistemological threshold between true belief and false belief represents a profound cognitive dividing line. In a true-belief scenario, there is complete isomorphic correspondence between the external world and the internal cognitive map: the object is in container A, and the observed agent believes the object is in container A. A child observing this event does not need to compute an independent model of the agent’s mind; the child can simply rely on direct perceptual realism. The child looks at the physical world, sees that the object is in container A, and assumes that anyone searching for the object will proceed to container A simply because that is where the object physically resides.

In sharp contrast, a false-belief scenario introduces a radical epistemic schism. The external physical reality (the object is currently in container B) stands in direct contradiction to the mental state of the agent (the agent believes the object is in container A). To correctly predict that the agent will search in container A, the observing child must engage in genuine mental representationalism. The child must construct two distinct, simultaneously active cognitive representations:

• Primary Representation (Reality): The physical state of affairs known to the observer through immediate sensory perception (the object is in container B).
• Secondary Representation (Mental State): The counterfactual or subjective state of affairs held in the agent’s mind, which has been rendered inaccurate by an unobserved change in the physical world (the agent thinks the object is in container A).

Holding these two conflicting representations in mind without allowing the dominant, visually verified primary representation to overwrite the unobservable secondary representation imposes a substantial cognitive load. Passing the false-belief threshold confirms that the developing mind has abandoned naive perceptual realism and has embraced an understanding that minds do not provide a direct video feed of reality, but rather generate internal, fallible representations of it.

3.3 Wimmer and Perner’s (1983) Precursor Paradigm: The Maxi Task

The empirical operationalization of Dennett’s false-belief requirement was first achieved by Austrian developmental psychologists Heinz Wimmer and Josef Perner in 1983. Published in Cognition, their study introduced a paradigm designed to track the emergence of belief understanding in typically developing young children: the famous “Maxi” task.

In the Maxi paradigm, children were presented with a detailed narrative enactment using dolls and props. The storyline unfolded systematically:

1. A boy named Maxi helps his mother put away groceries, placing a slab of chocolate into a green cupboard.
2. Maxi leaves the kitchen to play outside in the garden.
3. While Maxi is away, the mother takes some of the chocolate out to bake a cake and subsequently returns the remainder not to the green cupboard, but to a blue cupboard.
4. Maxi returns to the kitchen, feeling hungry, and wants to retrieve his chocolate.

The experimenter then asked the participating child the critical test question: “Where will Maxi look for his chocolate?”

Wimmer and Perner found that typically developing children under four years of age overwhelmingly failed the task, pointing to the blue cupboard where the chocolate was currently located. Between four and five years of age, children transitioned toward the correct answer, pointing to the green cupboard. Despite its brilliance, the Maxi task suffered from notable methodological encumbrances. The narrative was long, linguistically elaborate, and placed heavy demands on working memory, narrative processing, and linguistic decoding. These extraneous processing demands meant that younger children, or clinical populations with language delays, might fail the Maxi task not because they lacked a Theory of Mind, but because they succumbed to the executive and linguistic complexity of the experimental design. This methodological limitation provided the precise catalyst for Baron-Cohen, Leslie, and Frith to streamline the protocol into the Sally-Anne Test.

4. The 1985 Landmark Experiment: Design, Methodology, and Protocol

4.1 Experimental Apparatus, Characters, and Physical Vignette

To eliminate the linguistic and narrative confounders embedded in Wimmer and Perner’s Maxi paradigm, Simon Baron-Cohen, Alan Leslie, and Uta Frith engineered an exceptionally parsimonious, visually transparent enactment. The apparatus consisted of two small character dolls named Sally and Anne, along with two distinct, visually identifiable miniature receptacles: a covered basket and a covered box. The physical object of interest was a standard small red marble.

The vignette was performed live before each child participant, accompanied by an economical, unadorned spoken narrative:

• Sally and Anne are introduced to the child. Sally is holding her covered basket, and Anne is seated beside her with her covered box.
• Sally takes the red marble and visibly places it inside her basket. Sally then closes the basket.
• Sally departs from the scene, walking out of the experimental staging area so that she is explicitly absent from the room and unable to perceive ongoing events.
• While Sally is away, Anne steps forward, opens Sally’s basket, removes the red marble, places it inside her own box, and closes the lid.
• Sally returns to the staging area to retrieve her marble, and the physical enactment halts.

The deliberate choice to use concrete, three-dimensional physical dolls and props enacted in real-time space—rather than verbal stories, two-dimensional line drawings, or complex theatrical roleplay—dramatically reduced verbal processing demands. This design ensured that even young children and those with substantial communicative delays could effortlessly track the spatial trajectory of the object and the physical presence or absence of the social agents.

4.2 Core Test Questions: Belief, Reality, and Memory

The true genius of the Sally-Anne methodology lies in its triangular questioning protocol. Upon Sally’s return to the scene, the experimenter halted the action and presented three distinct verbal probes to the child participant, categorized into the experimental test probe and two mandatory control probes:

• The Critical Belief Question: “Where will Sally look for her marble?” This question probes the child’s predictive capacity regarding Sally’s internal mental representation. To pass, the child must decouple what they personally know to be physically true from what Sally falsely believes, correctly pointing to or naming Sally’s basket.
• The Reality Control Question: “Where is the marble really?” This question evaluates the child’s accurate perception and retention of the current physical state of affairs. The correct response is Anne’s box.
• The Memory Control Question: “Where was the marble in the beginning?” This question evaluates whether the child accurately encodes and recalls the historical chronology of the event. The correct response is Sally’s basket.

The logical necessity of passing both control questions cannot be overstated. In developmental neuropsychology, a failure on an experimental condition is completely uninterpretable without controls. If a child were to fail the Belief Question by pointing to the box, and subsequently fail the Memory Question by forgetting that the marble was ever in the basket, the experimenter could not conclude that the child lacked a Theory of Mind; the failure could simply reflect profound episodic amnesia or attention-deficit decay. Conversely, passing both the Reality and Memory questions acts as a methodological firewall, verifying that failure on the Belief Question represents a specific, isolated inability to compute a mental state.

4.3 Procedural Controls and Methodological Rigor

Baron-Cohen, Leslie, and Frith instituted rigorous procedural controls to eliminate potential sources of experimental bias. To ensure that character bias or spatial preference did not skew results, the positioning of the dolls (left versus right) and the positioning of the receptacles (basket versus box) were meticulously counterbalanced across participants. The experimenters were strictly trained to deliver the questions with neutral, flat prosody, avoiding any unconscious vocal inflections, pitch elevations, or subtle eye-gaze cues that might inadvertently prompt the child toward a specific receptacle.

The scoring criteria were absolute: a child received credit for passing the task if, and only if, they correctly passed the Belief Question and both the Reality and Memory control questions. A failure on either control question led to immediate exclusion of that trial from the mentalizing analysis. Furthermore, the experimental protocol allowed for multimodal response formats. Children who possessed functional verbal expressive language could speak their answers (“In the basket”), while children with limited speech or severe verbal dyspraxia could respond by pointing physically to the target receptacle. By validating unambiguous non-verbal pointing gestures, the researchers ensured that performance bottlenecks would not artificially suppress the measured competence of children with speech differences.

5. Participant Cohorts: Autism, Down Syndrome, and Typical Development

5.1 The Autistic Cohort: Demographic and Diagnostic Characteristics

The core clinical group in the 1985 study comprised twenty children clinically diagnosed with autism spectrum conditions according to established diagnostic standards (such as Michael Rutter’s criteria and the prevailing DSM-III framework). The cohort presented an average chronological age of 11 years and 11 months (ranging from 6 years 1 month to 16 years 6 months), representing an established, clinically stable population.

To establish a rigorous cognitive baseline, each autistic child underwent extensive psychometric evaluation using standardized psychological instruments:

• Receptive Vocabulary: Assessed using the British Picture Vocabulary Scale (BPVS), yielding a mean verbal mental age of 5 years and 5 months.
• Non-verbal Intelligence: Assessed using the Leiter International Performance Scale, yielding a mean non-verbal mental age of 9 years and 3 months.

These psychometric scores were paramount. Cognitive developmental milestones demonstrated that typically developing children cross the false-belief threshold between 4.0 and 4.5 years of age. Because the autistic cohort possessed an average verbal mental age of over 5 years, and a non-verbal mental age exceeding 9 years, their developmental cognitive capacity was well beyond the normative operational threshold required to solve the task, effectively ruling out generalized developmental immaturity as a primary explanation.

5.2 The Down Syndrome Control Cohort: Isolating General Intellectual Disability

The methodological cornerstone of the 1985 paper—and the element that rendered its conclusions so historically definitive—was the deliberate inclusion of a neurodevelopmental control group: fourteen children diagnosed with Down syndrome. This group presented a mean chronological age of 10 years and 11 months (ranging from 6 years 3 months to 17 years 0 months).

Psychometric testing revealed that the Down syndrome cohort had an average verbal mental age of 2 years and 11 months on the BPVS, and an average non-verbal mental age of 5 years and 11 months on the Leiter Scale. Notice that the children with Down syndrome had a significantly lower verbal mental age and a substantially lower non-verbal mental age than the autistic children. Both groups were drawn from specialized educational facilities and shared similar institutionalized or special-education backgrounds, effectively balancing out potential environmental and pedagogical variables.

The theoretical rationale for this cohort was brilliant: if failure on the Sally-Anne Test was simply an artifact of general cognitive delay, low IQ, or the global neurological effects of intellectual disability, then the Down syndrome cohort—having lower intellectual and linguistic scores than the autistic cohort—should perform significantly worse than, or at least as poorly as, the autistic group. Conversely, if Theory of Mind is a modular, domain-specific capacity selectively impaired in autism, the Down syndrome children should pass the test despite their generalized intellectual delays.

5.3 The Neurotypical Control Cohort: Establishing Baseline Developmental Norms

The third group consisted of twenty-seven typically developing, neurotypical children recruited from a standard nursery and primary school environment in London. This cohort possessed a mean chronological age of 4 years and 5 months (ranging from 3 years 5 months to 5 years 9 months).

The neurotypical cohort was purposely selected around the 4.5-year age mark to serve as an exact chronological developmental benchmark. Decades of Piagetian research and contemporary developmental psycholinguistics had suggested that the transition from egocentric perspectives to relativistic social-cognitive processing occurs within the 3.5 to 4.5-year window. Testing this cohort verified that the simplified Sally-Anne apparatus reliably elicited normative false-belief performance in young children, anchoring the experimental task within standard ontogenetic milestones.

6. Detailed Experimental Findings and Statistical Outcomes

6.1 Comparative Pass/Fail Rates Across the Three Experimental Groups

The empirical results of the Sally-Anne Test revealed an unmistakable, statistically decisive dissociation that shocked the developmental psychology community. The pass rates on the critical Belief Question across the three experimental groups were distributed as follows:

• Typically Developing Children: 85% passed (23 out of 27 children correctly identified Sally’s basket).
• Children with Down Syndrome: 86% passed (12 out of 14 children correctly identified Sally’s basket).
• Children with Autism: Only 20% passed (4 out of 20 children correctly identified Sally’s basket).

An astonishing 80% (16 out of 20) of the autistic children failed the Belief Question. When asked where Sally would look for her marble, they pointed not to the basket where Sally had placed it and still believed it to be, but directly to Anne’s box—the physical location where the marble was currently hidden in actual reality. These autistic children, despite having non-verbal cognitive abilities averaging over 9 years of age and verbal abilities well past the 5-year threshold, responded as direct perceptual realists. They predicted Sally’s behavior based on the objective state of the world rather than on the subjective state of Sally’s mind.

6.2 Analysis of Reality and Memory Control Checks

One of the most striking aspects of the 1985 data set was the performance of all three cohorts on the control questions. When presented with the Reality Question (“Where is the marble really?”) and the Memory Question (“Where was the marble in the beginning?”), the performance was uniform:

100% of the participants across all three cohorts passed both control questions without error.

Every single autistic child who failed the Belief Question immediately and accurately answered that the marble was really in the box (Reality Control) and that it had been in the basket in the beginning (Memory Control). This 100% success rate had profound scientific implications:

• It conclusively proved that the autistic children had paid attention throughout the entire live demonstration.
• It demonstrated that they completely understood the verbal semantics of the experimenter’s vocabulary (“really”, “in the beginning”).
• It established that their working memory and episodic memory were fully intact; they had not simply forgotten where Sally had placed the marble.
• It confirmed that their failure on the Belief Question was an isolated, highly specific cognitive failure: the inability to represent what Sally thought.

6.3 Significance of Statistical Disparities

The statistical analysis executed by Baron-Cohen, Leslie, and Frith confirmed the immense divergence between the cohorts. Using Fisher’s exact probability tests, the comparison between the autistic group and the typically developing group was highly significant (p < 0.001). Even more critically, the statistical comparison between the autistic cohort and the Down syndrome cohort was equally robust (p < 0.001), while there was no statistically significant difference whatsoever between the Down syndrome children and the typically developing 4-year-olds (p = 1.0, two-tailed).

This empirical outcome provided an undeniable double dissociation between general intellectual quotient (IQ) and Theory of Mind reasoning. If general intelligence governed mentalizing performance, the autistic children (mean Leiter MA of 9.3 years) should have easily outperformed the Down syndrome children (mean Leiter MA of 5.11 years). Instead, the group with lower intellectual functioning achieved an 86% pass rate, while the group with significantly higher cognitive abilities achieved only a 20% pass rate. This statistical proof provided empirical backing for the modularity hypothesis: the capacity to reason about mental states does not emerge as a mere by-product of general intellectual horsepower, but relies on a specialized neurocognitive architecture that can be selectively disrupted.

7. Cognitive Mechanisms: Metarepresentation and Mindreading Architecture

7.1 Alan Leslie’s Theory of Metarepresentation and Decoupling

To computationally ground the striking failure of autistic children on the Sally-Anne Test, Alan Leslie developed his theoretical model of metarepresentation and cognitive decoupling. In standard cognitive operations, our perceptual machinery generates primary representations. Primary representations are semantic structures that map directly onto the actual physical world: they are strictly bound by veridical reference, true conditions, and real-world existence. If an individual sees a red marble inside a box, their primary representation is simply: [Location: Marble = Box].

However, human social life requires us to entertain hypothetical scenarios, pretenses, and false beliefs that do not match current sensory reality. Leslie argued that to handle these without corrupting our primary knowledge base, the human mind uses a computational architecture that creates a “decoupled” copy of the primary representation. This process yields a metarepresentation, which Leslie formalized as a distinct, tripartite computational relation: an Agent, an Attitude, and a Proposition.

In the Sally-Anne paradigm, this metarepresentation is computed as:

Sally [Agent] — Believes [Attitude] — "The marble is in the basket" [Proposition]

The decoupling mechanism acts as a quarantined buffer. It isolates the proposition “The marble is in the basket” from the primary representational system, preventing it from being overwritten by the visually verified fact that the marble is currently in the box. Leslie argued that the 20% pass rate in the autistic cohort reflects an innate, computational breakdown in the decoupling mechanism. Without functional decoupling, the child’s mind collapses the mental state into the primary perceptual state: the child cannot represent what Sally thinks, and instead reports what the world is.

7.2 The Concept of ‘Mindblindness’ by Simon Baron-Cohen

Building directly upon these empirical findings, Simon Baron-Cohen synthesized his comprehensive model of the human “mindreading system.” In his 1995 framework, Baron-Cohen argued that human social cognition is orchestrated by four distinct, developmentally sequenced cognitive modules:

• The Intentionality Detector (ID): A primitive perceptual mechanism that interprets motion stimuli in terms of basic primitive desires and goals (e.g., approaching an object implies wanting it). Active in early infancy.
• The Eye Direction Detector (EDD): A specialized visual mechanism that detects the presence of eye-like stimuli, tracks the directional vector of gazes, and computes that an agent “sees” what its eyes are directed toward.
• The Shared Attention Mechanism (SAM): A critical computational bridge emerging between 9 and 14 months that builds triadic relationships between the Self, an Other, and an Object. SAM allows for joint visual attention and gaze-monitoring, creating representations of the form: [Self and Other are both attending to the same Object].
• The Theory of Mind Mechanism (ToMM): The ultimate representational system that computes the full repertoire of mental states, converting perceptual inputs into a cohesive framework of beliefs, desires, and intentions, ultimately resolving false beliefs.

Baron-Cohen theorized that in autism, the primitive ID and EDD mechanisms frequently remain intact, allowing autistic individuals to track basic physical goals and eye directions. However, a developmental bottleneck disrupts the functional emergence of the Shared Attention Mechanism (SAM), which in turn compromises the downstream development of ToMM. The resulting condition, mindblindness, leaves the individual navigating a social universe populated by dynamic, unpredictable physical bodies whose underlying psychological drivers remain completely opaque, leading to communicative breakdowns, literal linguistic interpretations, and profound social isolation.

7.3 Interplay Between Theory of Mind and Executive Functioning

While Baron-Cohen and Leslie advanced domain-specific, modular interpretations of the Sally-Anne Test, competing cognitive theorists proposed alternative, domain-general explanations, focusing on the role of executive functioning. Executive functions encompass higher-order cognitive control processes mediated by the prefrontal cortex, including working memory, cognitive flexibility, and, most crucially, inhibitory control.

The Inhibitory Control Hypothesis posits that solving the Sally-Anne Test requires an individual to suppress a prepotent, highly salient piece of knowledge: the actual, physical location of the marble. When asked “Where will Sally look?”, the true location (the box) is intensely active in the child’s working memory. To direct Sally to the basket, the child must actively inhibit this dominant perceptual reality. If an individual possesses severe inhibitory control deficits, they may fail the test not because they lack a conceptual understanding of beliefs, but because they simply cannot suppress their dominant response pointing to the true physical location.

In a famous experiment designed to isolate inhibitory failure from mentalizing deficits, James Russell and colleagues developed the Windows Task. In this non-social task, children had to point to an empty box in order to receive a chocolate treat hidden inside a box with a transparent window. If they pointed to the box showing the chocolate, the experimenter gave the chocolate to an opponent. Autistic children repeatedly failed this non-social task, pointing directly to the visible treat despite knowing it meant losing it. This finding demonstrated that inhibitory control impairments contribute substantially to the performance failures observed in false-belief tasks, fueling an ongoing debate in developmental neuropsychology: is Theory of Mind failure in autism a purely modular socio-cognitive deficit, or is it inextricably linked to prefrontal executive dysregulation?

8. The Sally-Anne Test in Developmental Trajectories

8.1 Typical Milestones of False-Belief Acquisition (Ages 3 to 5)

In typical human ontogeny, the transition from perceptual realism to explicit false-belief understanding represents a universal developmental milestone. Decades of cross-cultural research have shown that between 3.0 and 5.0 years of age, children undergo a fundamental cognitive shift. At 36 months, a typically developing child consistently fails explicit false-belief tasks, predicting that an agent will act in accordance with objective reality. By 48 to 54 months, this error pattern evaporates, and the vast majority of children effortlessly pass the Sally-Anne Test and its variations.

This timeline was validated in a definitive meta-analysis by Henry Wellman, David Cross, and Julanne Watson (2001). Analyzing 178 separate empirical studies covering nearly 600 conditions and thousands of children, the authors confirmed that the developmental progression from failing to passing false-belief tasks is consistent across diverse socio-economic, linguistic, and cultural backgrounds. The meta-analysis established that this shift represents a true conceptual reorganization of the child’s understanding of the mind, rather than a mere artifact of improved linguistic competence or increased task familiarity.

Intriguingly, the dawn of the 21st century brought evidence that the roots of Theory of Mind lie even deeper in infancy. Using non-verbal eye-tracking methodologies, researchers like Kristine Onishi and Renée Baillargeon (2005) demonstrated that 15-month-old infants display implicit false-belief understanding. When tested using violation-of-expectation paradigms, infants looked significantly longer when an actor searched for an object in a location inconsistent with the actor’s false belief, suggesting that the human brain possesses an early-developing, pre-verbal system for tracking beliefs that anticipates the explicit, verbally mediated mastery seen at age four.

8.2 First-Order versus Second-Order False Belief Tasks

The Sally-Anne Test assesses what cognitive psychologists classify as a first-order false belief. A first-order mental state attribution involves modeling a single agent’s direct relationship to a proposition regarding external reality:

"Sally believes [the marble is in the basket]."

First-order Theory of Mind is sufficient to predict simple human actions, understand direct deception, and pass basic developmental screening tasks. However, mature human social interaction frequently operates on nested, recursive levels of intentionality, requiring second-order false belief understanding. Second-order tasks demand that an observer compute a belief about another person’s belief:

"John believes that [Mary thinks [the ice cream van is at the church]]."

Originally operationalized by Josef Perner and Heinz Wimmer (1985) in their classic “Ice Cream Van” scenario, second-order false-belief reasoning represents a significantly more advanced cognitive milestone, typically mastered by neurotypical children between 6 and 8 years of age. In the ice cream van scenario, John and Mary are in the park when they are told the ice cream van will move to the school. Unknown to John, Mary later meets the ice cream vendor, who tells her directly that he has moved his van to the church. John then sees Mary walking to buy ice cream. The child participant is asked: “Where does John think Mary will go to buy ice cream?”

To answer correctly (“To the school”), the child must model John’s mistaken belief about Mary’s knowledge state. In clinical populations, this hierarchy is illuminating: many academically capable, high-functioning autistic adolescents and adults who eventually learn to pass first-order Sally-Anne tasks continue to experience severe breakdowns when confronted with second-order recursive mentalizing tasks. This reveals that structural mentalizing differences persist across developmental trajectories, even when early developmental milestones are eventually reached through compensatory strategies.

8.3 Cross-Cultural Generalizability and Linguistic Influences

A critical question in developmental psychology is whether the milestones exposed by the Sally-Anne Test represent a universal biological maturation or a culturally specific socio-linguistic construct. Cross-cultural research conducted in industrial, rural, and indigenous societies—from urban North America and Western Europe to traditional hunting communities in Cameroon, Samoa, and Peruvian Amazonia—confirms that the developmental shift from failing to passing false-belief tasks universally occurs within the 3.5 to 5.5-year window.

However, specific linguistic environments can systematically accelerate or delay this transition. A compelling factor is the presence of grammatical evidentiality in certain languages, such as Turkish and Korean. In Turkish, verbs must be inflected with specific suffixes depending on whether the speaker directly witnessed the event (-di) or learned of it via hearsay/inference (-miş). Research indicates that children acquiring languages with mandatory evidential markings demonstrate slightly accelerated performance on false-belief tasks, as their native grammar forces them to continually track the perceptual and informational sources of knowledge.

Furthermore, linguist Jill de Villiers formulated the Syntactic Complementation Hypothesis, arguing that the acquisition of specific syntactic structures is the cognitive prerequisite for explicit false-belief reasoning. Specifically, de Villiers pointed to sentential complement structures governed by mental state verbs (e.g., “Sally thinks [that the marble is in the basket]”). In these grammatical configurations, the entire subordinate clause can be semantically false while the overarching matrix sentence remains syntactically true. Children must master this linguistic architecture to reliably decouple mental propositions from external reality, highlighting the intricate dance between language acquisition and social-cognitive development.

9. Methodological Critiques, Limitations, and Replications

9.1 Linguistic and Pragmatic Processing Demands

Despite the historic impact of the 1985 experiment, the Sally-Anne paradigm has faced sustained methodological scrutiny. The most prevalent critique centers on the syntactic and pragmatic processing demands embedded in the critical Belief Question: “Where will Sally look for her marble?”

Developmental psycholinguists, most notably Michael Siegal and colleagues, argued that young children often interpret adult conversational probes through a pragmatic filter of cooperative communication (Gricean maxims). When an adult points to a scene where an object has visibly moved, asking a young child where an agent will look, the child may misinterpret the question not as an inquiry about Sally’s subjective psychological state, but as a practical or moral inquiry about where Sally ought to look if she hopes to actually find the marble. In a series of elegant replications, Siegal and Beattie (1991) modified the phrasing by inserting a single temporal modifier:

“Where will Sally look first for her marble?”

This subtle linguistic adjustment clarified the pragmatic implicature, explicitly directing the child’s attention to the initial behavioral step rather than the ultimate successful outcome. With this single word change, the pass rate among typically developing 3-year-olds increased dramatically, demonstrating that a substantial portion of false-belief task failures stemmed from performance-level pragmatic misalignments rather than a genuine epistemic competence deficit.

9.2 Ecological Validity and Social Context Deficits

A second major critique concerns the ecological validity of the Sally-Anne Test. The scenario takes place in an artificial, sterilized laboratory setting featuring static plastic dolls, sterile boxes, and a tiny marble. Human social life, however, does not occur in frozen tableaus. It unfolds within dynamic, emotionally rich, multisensory environments characterized by continuous streams of facial expressions, vocal prosody, fluctuating power dynamics, body language, and shared cultural histories.

Critics point out that passing or failing a laboratory doll-play vignette has remarkably weak predictive validity when correlated with real-world adaptive social behavior. Many autistic individuals who fail the Sally-Anne Test can successfully navigate certain social interactions using learned scripts, while many high-functioning autistic individuals who comfortably pass the Sally-Anne Test continue to experience profound, exhausting social friction when thrown into messy, ambiguous real-world social environments. The Sally-Anne Test completely lacks an affective component; it tests a cool, intellectualized, deductive form of mentalizing while entirely omitting visceral empathy, somatic resonance, and hot socio-emotional processing.

To overcome these limitations, modern neuropsychology has developed dynamic, naturalistic instruments such as the Movie for the Assessment of Social Cognition (MASC). The MASC presents participants with a 15-minute video depicting real actors engaging in complex social situations involving dating, friendships, and workplace interactions, pausing intermittently to ask questions about the characters’ feelings, intentions, and hidden motives. Such ecologically grounded tools reveal nuanced social processing profiles that binary puppet tasks simply cannot detect.

9.3 Alternative Explanations: Teleology, Simulation, and Weak Central Coherence

Theoretical alternatives to Leslie and Baron-Cohen’s modular nativist account have provided divergent interpretations of the Sally-Anne findings:

• The Teleological Stance: Advanced by Gergely and Csibra, this view argues that young children do not initially reason about subjective internal “beliefs” at all; instead, they adopt a teleological stance that evaluates actions in terms of objective physical goals achieved through the most efficient physical means available. Failure on false-belief tasks simply marks the developmental boundary where teleological calculations of physical efficiency are overridden by mentalistic representational mechanics.
• Simulation Theory: Proposed by philosophers Robert Gordon and Alvin Goldman, Simulation Theory rejects the notion that the mind possesses a specialized “theory” of mind (Theory-Theory). Instead, simulationists argue that we predict other people’s actions by mentally projecting ourselves into their shoes—running our own cognitive and emotional decision-making apparatus “offline” with the other person’s parameters. Under this account, autistic individuals do not suffer from a broken modular engine; rather, they struggle to generate an accurate offline internal simulation due to foundational differences in executive control and sensory processing.
• Weak Central Coherence (WCC): Formulated by Uta Frith, WCC suggests that autistic children fail the Sally-Anne Test because their cognitive style prioritizes local details over global context. Rather than integrating the entire social narrative into a unified event framework, the autistic child may focus obsessively on the most concrete, visually immediate feature of the scene: the actual, physical location of the red marble inside the box.

A crucial line of empirical evidence supporting the sociolinguistic scaffolding view came from studies of deaf children. Candy Peterson and Michael Siegal (1999) compared deaf children of hearing parents with deaf children of deaf parents (“native signers”). Deaf children of hearing parents frequently experience severe early language deprivation due to the communication barrier, whereas native signers have fluent access to sign language from birth. Peterson and Siegal discovered that native signing deaf children passed the Sally-Anne Test at normal developmental rates, whereas deaf children of hearing parents displayed severe delays—matching the performance of autistic children. This proved that extensive early access to shared social conversation about unobservable thoughts and feelings is vital for the timely emergence of Theory of Mind.

10. Neurobiological Correlates and Functional Neuroanatomy

10.1 The Mentalizing Network: Functional Neuroimaging Insights

With the advent of functional neuroimaging technologies in the 1990s and 2000s, cognitive neuroscientists transitioned from behavioral observations to mapping the structural and functional anatomy of the human “social brain.” These investigations revealed that mentalizing does not reside within an isolated cranial structure, but is coordinated by a distributed neural circuit known as the Mentalizing Network (MENT).

The primary nodes of this network include:

• The Temporoparietal Junction (TPJ): Located at the intersection of the posterior superior temporal gyrus and the inferior parietal lobule. Neuroscientist Rebecca Saxe and colleagues demonstrated that the right Temporoparietal Junction (rTPJ) exhibits exceptional functional selectivity for attributing transient mental states, such as thoughts, beliefs, and desires, lighting up specifically during false-belief reasoning tasks while remaining quiet during non-social control tasks.
• The Medial Prefrontal Cortex (mPFC): The anterior hub of the mentalizing system. The dorsal and ventral subregions of the mPFC are heavily engaged in decoupled representation, integrating social information over extended temporal windows, self-other differentiation, and trait attribution.
• The Precuneus and Posterior Cingulate Cortex (PCC): Located within the posteromedial parietal cortex, these areas support the memory retrieval, mental imagery, and spatial perspective-taking required to reconstruct an event from another agent’s physical viewpoint.
• The Superior Temporal Sulcus (STS): Crucial for processing biological motion, gaze shifts, facial expressions, and dynamic social gestures that provide the raw sensory inputs feed into the higher-order mentalizing network.

10.2 Functional and Structural Connectivity in Autism Spectrum Conditions

Neuroimaging paradigms directly adapting the Sally-Anne false-belief design for functional MRI have identified marked functional and structural alterations within the mentalizing network of autistic individuals. Functional neuroimaging studies consistently report atypical activation patterns, characterized by significant hypoactivation of both the right temporoparietal junction and the medial prefrontal cortex during tasks requiring belief attribution.

However, modern social neuroscience views autism less as a focal lesion within a single brain region and more as an alteration of large-scale functional connectivity. During mentalizing challenges, neurotypical brains exhibit tightly coordinated, synchronized low-frequency blood-oxygen-level-dependent (BOLD) oscillations across the entire Default Mode Network (DMN), of which the mentalizing network is a primary subsystem. In autistic cohorts, studies reveal systematic hypoconnectivity between anterior hubs (mPFC) and posterior processing centers (rTPJ/precuneus). The nodes can activate locally, but they fail to integrate their computational outputs across the long-range neural pathways necessary to smoothly resolve dynamic social problems.

Diffusion Tensor Imaging (DTI) and tractography studies have supplemented these functional findings by identifying structural alterations within major white-matter tracts. The arcuate fasciculus and the superior longitudinal fasciculus—fiber bundles responsible for connecting frontal language and executive centers with posterior temporoparietal sensory regions—frequently show altered fractional anisotropy in autistic populations. Electrophysiological investigations utilizing Event-Related Potentials (ERPs) further reveal that the late positive wave (often emerging around 300 to 600 ms post-stimulus, associated with conscious belief processing) is systematically delayed or attenuated in individuals who struggle with explicit false-belief tasks.

10.3 The Mirror Neuron System (MNS) versus the Mentalizing Network (MENT)

A central theoretical challenge in social cognitive neuroscience has been clarifying the functional division between the Mirror Neuron System (MNS) and the Mentalizing Network (MENT). First discovered in the premotor cortex and inferior parietal lobule of macaque monkeys by Giacomo Rizzolatti and colleagues, mirror neurons fire both when an animal executes a motor action and when it observes another animal executing the identical action.

In the late 1990s and early 2000s, researchers like V.S. Ramachandran formulated the “Broken Mirror Theory of Autism,” arguing that a congenital failure of the mirror neuron system was the root cause of the social impairments identified by the Sally-Anne Test. However, widespread scientific debate and subsequent empirical data have largely refuted this simplistic model. The contemporary consensus draws a sharp functional distinction between the two systems:

• The Mirror Neuron System (Sensorimotor / Affective Empathy): Operates as a fast, automatic, bottom-up simulation mechanism. It maps observable bodily kinematics, motor goals, and immediate emotional expressions (such as disgust or pain) directly onto the observer’s own sensorimotor maps.
• The Mentalizing Network (Inferential / Cognitive Empathy): Operates as a top-down, reflective, inferential cognitive architecture. It attributes unobservable propositional attitudes (such as false beliefs, hidden intentions, and abstract knowledge states) that cannot be deduced from motor kinematics alone.

The Sally-Anne Test is structurally unresolvable by the Mirror Neuron System alone. Anne’s physical action of moving the marble cannot provide motor mirror cues that reveal Sally’s absent mental state. Solving the task demands the inferential computational machinery of the mentalizing network. This distinction cleanly illustrates the clinical dissociation frequently seen in autism: many autistic individuals possess deeply resonant affective empathy (feeling profound emotional distress upon witnessing another person suffer), yet simultaneously experience significant challenges with cognitive empathy and Theory of Mind (deducing the complex, non-literal, or mistaken beliefs driving that person’s behavior).

11. Practical Applications in Clinical Assessment and Interventions

11.1 Diagnostic Batteries and Behavioral Screening Tools

The operational framework inaugurated by the Sally-Anne Test fundamentally restructured clinical developmental psychology, inspiring the construction of gold-standard diagnostic instruments used internationally today. The direct conceptual lineage of false-belief testing is embedded throughout the Autism Diagnostic Observation Schedule, Second Edition (ADOS-2). Within Modules 3 and 4 of the ADOS-2—designed for verbally fluent children, adolescents, and adults—examiners specifically probe Theory of Mind through tasks evaluating the participant’s understanding of social relationships, non-literal jokes, nuanced narratives, and the internal mental motives of fictional characters.

Similarly, standardized neuropsychological assessments, such as the NEPSY-II (A Developmental Neuropsychological Assessment), contain dedicated Theory of Mind subtests. These evaluate both verbal and non-verbal mentalizing capacities, moving from basic emotion recognition to complex contextual vignettes that structurally mirror the Sally-Anne false-belief logic. In early pediatric screening, instruments such as the M-CHAT-R/F (Modified Checklist for Autism in Toddlers, Revised with Follow-Up) screen for the functional precursors to Theory of Mind—such as protodeclarative pointing, joint visual attention, and gaze monitoring—identifying developmental divergences years before children reach the chronological age to take an explicit Sally-Anne Test.

For older adolescents and adults with average or above-average intellectual abilities, ceiling effects make the basic Sally-Anne Test clinically useless. To assess mentalizing capacity in these populations, Simon Baron-Cohen and colleagues developed the “Reading the Mind in the Eyes” Task (RMET). In this assessment, participants are shown 36 photographs isolating only the eye region of human faces and must select which of four complex mental state terms (e.g., contemplative, skeptical, insistent) accurately describes the internal state of the person. This advanced assessment continues to be widely deployed across clinical neuropsychiatry, identifying subtle social cognitive differences in conditions ranging from autism and ADHD to schizophrenia and borderline personality disorder.

11.2 Theory of Mind Curriculum and Pedagogical Strategies

Recognizing that Theory of Mind does not develop spontaneously in autistic children led educational psychologists to create structured, explicit instructional interventions. Rather than assuming that children will intuitively deduce the invisible mental states of peers, specialized pedagogical curricula treat mentalizing as an explicit, rule-governed subject that can be systematically taught.

A leading approach, pioneered by Julie Hadwin and colleagues, utilizes explicit visual-spatial scaffolding to make invisible thoughts tangible. These methods leverage tools such as:

• Graphic Organizers and “Thought Bubbles”: By illustrating a cartoon character with a speech bubble (what they say) alongside an explicitly drawn thought bubble (what they privately think or believe), educators provide a concrete visual representation of Leslie’s “decoupling.” Children can visibly see that a person can say one thing while thinking another, or that two characters can hold completely conflicting thoughts about the identical physical object.
• Social Stories™: Developed by Carol Gray, Social Stories use highly structured, individualized narratives to clarify confusing social scenarios. A critical feature of a well-designed Social Story is its deliberate integration of perspective sentences. Perspective sentences explicitly describe the internal mental states, beliefs, feelings, and motivations of other people (e.g., “When the recess bell rings loudly, the teacher thinks it is time to line up. She does not know that the sound hurts my ears.”).

While these didactic interventions reliably teach children to pass formal false-belief tests in laboratory and classroom settings, clinical practitioners frequently encounter the persistent challenge of generalization. An autistic child may learn the explicit algorithmic rules necessary to correctly identify where a character will look in a structured workbook or puppet game, yet struggle to apply this explicit knowledge dynamically during the chaotic, fast-paced exchanges of unstructured playground peer interactions.

11.3 Clinical Implications Beyond Autism: Schizophrenia, TBI, and Dementia

Although the Sally-Anne Test originated as a targeted diagnostic investigation into autism, the concept of Theory of Mind has proven indispensable across the broader landscape of neurology and adult psychiatry. A primary example is the neuropsychology of schizophrenia. Pioneering work by Christopher Frith (1992) demonstrated that the cognitive architecture of schizophrenia features profound, specific disruptions in Theory of Mind. Unlike the developmental delay or reduction in mental state attribution characteristic of autism, individuals experiencing active paranoid or persecutory delusions often exhibit hyper-theory of mind (or hyper-mentalizing). They over-attribute complex intentionality, malevolent motives, and hidden mental conspiracies to random physical events or neutral strangers.

In behavioral neurology, adult patients who sustain Traumatic Brain Injury (TBI)—specifically involving focal contusions to the orbitofrontal cortex, ventromedial prefrontal cortex, or right temporoparietal areas—frequently display dramatic, acquired social-cognitive personality changes. Despite retaining completely intact post-injury intellectual IQ, language fluency, and motor control, these individuals may fail advanced Theory of Mind tasks. They lose the capacity to track social nuances, make inappropriate remarks, commit frequent social faux pas, and exhibit profound interpersonal insensitivity due to the structural disruption of their mentalizing networks.

Theory of Mind paradigms are similarly critical in differential diagnosis across neurodegenerative dementias. In the early stages of Behavioral Variant Frontotemporal Dementia (bvFTD), progressive bilateral atrophy of the anterior temporal lobes, insular cortex, and frontal lobes leads to early, profound collapses in social cognition, moral reasoning, and Theory of Mind performance, often while basic episodic memory remains remarkably preserved. In contrast, patients in the early stages of Alzheimer’s Disease typically demonstrate the inverse cognitive profile: profound episodic memory failure coupled with preserved social facades, emotional reciprocity, and intact basic Theory of Mind capabilities. These clinical distinctions demonstrate that social-cognitive capacity relies upon discrete, dedicated neural systems that can be selectively spared or destroyed by neurodegenerative processes.

12. Evolution of False-Belief Testing and Contemporary Paradigms

12.1 The Shift Toward Implicit and Spontaneous Mentalizing Paradigms

The standard Sally-Anne Test represents an explicit, direct mentalizing paradigm: it halts the action, introduces a foreign experimenter, and verbally commands the child to deliver a deliberate, reasoned judgment (“Where will Sally look?”). This structural format sparked a critical question: what does the human mind do when it is not explicitly interrogated?

This inquiry inspired the development of implicit mentalizing paradigms that measure spontaneous social cognition without requiring linguistic responses. The foundational breakthrough in this domain was achieved by Wendy Clements and Josef Perner (1994) using anticipatory looking protocols. By tracking the spontaneous eye-gaze fixations of children watching a visual enactment, the researchers discovered that children as young as 2 years and 11 months spontaneously shifted their anticipatory gaze toward the correct container (the basket) in expectation of the character’s return, even though those identical children, when immediately asked the explicit verbal question, pointed incorrectly to the box.

This implicit-explicit dissociation reached a dramatic pinnacle in 2007, when Victoria Southgate, Atsushi Senju, and Gergely Csibra leveraged advanced corneal-reflection eye tracking to prove that 25-month-old human infants spontaneously track false beliefs. The infants wore non-invasive sensor caps while watching an actor on a screen. When a chime sounded indicating the actor was about to reach through an aperture to grab an object, the infants’ anticipatory eye movements landed on the location where the actor falsely believed the object to be, before any physical hand movement occurred.

However, the replication crisis within contemporary psychological science has intensely challenged this literature. Large-scale international replication initiatives, such as the ManyBabies Consortium (ManyBabies 2), have struggled to consistently replicate several foundational implicit false-belief eye-tracking effects across multiple independent laboratories. These mixed findings have ignited vigorous academic debate: do these infant gaze patterns reflect genuine, implicit computation of internal propositional mental states, or do they represent lower-level, non-mentalistic perceptual novelties and automated low-level behavioral associations?

12.2 Naturalistic, Continuous, and Advanced Social Cognitive Testing

To overcome the limitations of artificial puppet tasks and binary scoring systems, contemporary cognitive researchers have developed sophisticated, ecologically valid assessment protocols. One widely adopted instrument is the Strange Stories Task, developed by Francesca Happé (1994). The Strange Stories paradigm presents participants with complex, realistic written or auditory vignettes involving subtle social communication: irony, sarcasm, white lies, bluffing, double-bluffing, jokes, and figures of speech. To explain why a character makes an overtly false statement (e.g., saying “Oh, that’s just what I always wanted!” upon receiving a pair of old socks), the participant must recruit flexible, context-sensitive mentalizing skills that go far beyond basic first-order false beliefs.

Another widely validated assessment is the Faux Pas Recognition Test, introduced by Simon Baron-Cohen and colleagues. In this protocol, participants listen to naturalistic everyday stories and must identify instances where an individual unintentionally says something awkward, hurtfully insensitive, or socially inappropriate because they did not know or realize a crucial fact (e.g., someone complaining about a painting without realizing the painter is standing directly behind them). Successfully recognizing a faux pas requires concurrent computation of both a cognitive mental state (the speaker did not know the listener was the artist) and an affective mental state (the listener felt hurt or embarrassed).

The contemporary frontier of social cognitive testing leverages Virtual Reality (VR) and dynamic, computational cognitive modeling. Using high-immersion VR headsets, researchers place participants inside continuous, simulated social environments—such as a crowded café or an academic seminar room—populated by autonomous virtual avatars. By tracking real-time pupillometry, head rotation, eye-gaze fixation, and micro-behavioral choices, researchers can build continuous, Bayesian computational models of belief updating. Rather than categorizing a human mind as simply “having” or “lacking” a Theory of Mind, these advanced paradigms evaluate the continuous probabilistic parameters that govern how brains extract meaning from a chaotic social world.

12.3 Theoretical Refinements: Neurodiversity Paradigm and the Double Empathy Problem

In recent years, the theoretical framework that historically framed the Sally-Anne Test has undergone profound, transformative critique from the neurodiversity movement and contemporary autism scholarship. The foundational premise of the 1985 paper—that autism is characterized by an intrinsic cognitive “deficit,” a broken internal mechanism, or psychological “mindblindness”—has been fundamentally challenged by sociological and cognitive theorists, most notably Damian Milton with his formulation of the Double Empathy Problem.

Milton argues that when social communication breaks down between an autistic person and a neurotypical person, the breakdown is not an isolated cognitive failure residing solely within the autistic brain. Instead, it represents a bidirectional communicative mismatch between two distinct neurotypes, each possessing fundamentally different experiential lifeworlds, communicative norms, sensory processing systems, and expressive habits. Just as the autistic person may struggle to intuit the unspoken intentions and indirect social subtleties of the neurotypical person, the neurotypical person consistently fails to accurately deduce the internal mental states, thoughts, feelings, and intentions of the autistic person—often misinterpreting autistic distress as defiance, or autistic focus as cold indifference.

This theoretical reframing has received compelling empirical validation. Landmark research by Catherine Crompton and colleagues (2020) demonstrated that when autistic individuals interact with other autistic individuals, information transfer is remarkably fluid, accurate, and efficient, displaying high levels of shared rapport, reciprocal communication, and mutual comprehension. Communication breakdowns emerge primarily when interactions cross neurotypes—that is, when autistic individuals interact with neurotypical partners. This empirical proof demonstrates that social friction is an emergent property of relational divergence, rather than an absolute internal pathology.

Re-evaluating the historical legacy of the Sally-Anne Test within 21st-century cognitive science does not require dismissing its empirical brilliance. The 1985 paper successfully dismantled psychodynamic victim-blaming, proved that autism is an organic neurodevelopmental variation, and provided developmental psychology with its most famous experimental paradigm. Yet, as the field matures, cognitive science is transcending the historically narrow “deficit” model. The contemporary view recognizes that human minds populate a rich, heterogeneous cognitive landscape, where mentalizing is not a monolithic, one-way mirror, but an intricate, bidirectional bridge that requires continuous, mutual effort across all neurological variations.

Conclusion

The Sally-Anne Test remains one of the most elegant, consequential, and structurally influential experiments in the history of the behavioral sciences. In a simple vignette lasting less than two minutes, Simon Baron-Cohen, Alan Leslie, and Uta Frith cut through decades of speculative psychoanalytic theories and vague cognitive generalities, landing on the empirical core of social cognition: false-belief understanding. By demonstrating that the vast majority of autistic children struggled to predict an agent’s actions based on a mistaken mental representation—despite possessing the intellectual and chronological capacity to easily do so—the researchers proved that mentalizing is a distinct, specialized dimension of human cognition that does not simply mirror general intellectual quotient.

Over the subsequent four decades, the paradigm initiated by the Sally-Anne vignette served as the catalyst for a sweeping scientific revolution. It gave birth to Leslie’s metarepresentational computational models, Baron-Cohen’s architectural mapping of the mindreading system, and Frith’s holistic explorations of neurodevelopmental phenotypes. The test provided the conceptual key that unlocked the functional neuroanatomy of the social brain, leading directly to the identification of the mentalizing network, the specialized functions of the right temporoparietal junction, and the intricate, large-scale connectivity profiles mapped by modern neuroimaging. Beyond the laboratory, it reshaped the world of clinical practice, transforming early childhood screening, providing the conceptual foundation for diagnostic gold standards like the ADOS-2, and guiding the development of structured educational interventions that continue to benefit neurodivergent individuals globally.

Simultaneously, the evolution of false-belief testing has reflected the broader self-correcting trajectory of cognitive science itself. The methodological critiques of the Sally-Anne protocol—addressing its linguistic demands, its ecological artificiality, and its binary pass/fail limitations—pushed the field toward more nuanced, continuous, and dynamic paradigms, from implicit eye-tracking in infancy to immersive virtual reality assessments. Most profoundly, the rise of the neurodiversity paradigm and the conceptualization of the Double Empathy Problem have recontextualized the historical narrative. We now understand that Theory of Mind is not a static cognitive trophy possessed exclusively by typical brains, but a fragile, bidirectional relational dynamic between diverse human minds.

Ultimately, the story of the Sally-Anne Test is the story of how empirical science can take an invisible, abstract philosophical construct—how one human mind knows what another human mind thinks—and render it completely visible. By tracing the red marble as it travels from the basket to the box, generations of researchers, clinicians, and thinkers have learned to look past the surface of physical actions and peer into the invisible, representational universe that makes human social connection possible. The little dolls on the table in 1985 did not merely test a child’s prediction; they permanently expanded our scientific understanding of the human mind.

References

• Baron-Cohen, S. (1995). Mindblindness: An essay on autism and theory of mind. MIT Press. https://mitpress.mit.edu/9780262522076/mindblindness/
• Baron-Cohen, S., Leslie, A. M., & Frith, U. (1985). Does the autistic child have a “theory of mind”? Cognition, 21(1), 37–46. https://doi.org/10.1016/0010-0277(85)90022-8
• Baron-Cohen, S., Wheelwright, S., Hill, J., Raste, Y., & Plumb, I. (2001). The “Reading the Mind in the Eyes” Test revised version: A study with normal adults, and adults with Asperger syndrome or high-functioning autism. Journal of Child Psychology and Psychiatry, 42(2), 241–251. https://doi.org/10.1111/1469-7610.00715
• Clements, W. A., & Perner, J. (1994). Implicit understanding of belief. Cognitive Development, 9(4), 377–395. https://doi.org/10.1016/0885-2014(94)90012-4
• Crompton, C. J., Ropar, D., Evans-Williams, C. V., Flynn, E. G., & Fletcher-Watson, S. (2020). Autistic peer-to-peer information transfer is highly effective. Autism, 24(7), 1704–1712. https://doi.org/10.1177/1362361320919286
• Dennett, D. C. (1978). Beliefs about beliefs. Behavioral and Brain Sciences, 1(4), 568–570. https://doi.org/10.1017/S0140525X00076664
• de Villiers, J. G., & de Villiers, P. A. (2000). Linguistic determinism and the understanding of false beliefs. In P. Mitchell & K. J. Riggs (Eds.), Children’s reasoning and the mind (pp. 191–228). Psychology Press.
• Dunbar, R. I. (1998). The social brain hypothesis. Evolutionary Anthropology, 6(5), 178–190. https://doi.org/10.7551/mitpress/4737.001.0001
• Frith, C. D. (1992). The cognitive neuropsychology of schizophrenia. Psychology Press.
• Frith, U. (1989). Autism: Explaining the enigma. Basil Blackwell.
• Gray, C. (2010). The new Social Story book (10th anniversary ed.). Future Horizons.
• Happé, F. G. (1994). An advanced test of theory of mind: Understanding of story characters’ thoughts and feelings by able autistic, mentally handicapped, and normal children and adults. Journal of Autism and Developmental Disorders, 24(2), 129–154. https://doi.org/10.1007/BF02172093
• Leslie, A. M. (1987). Pretense and representation: The origins of “theory of mind.” Psychological Review, 94(4), 412–426. https://doi.org/10.1037/0033-295X.94.4.412
• Milton, D. E. (2012). On the ontological status of autism: The ‘double empathy problem’. Disability & Society, 27(6), 883–887. https://doi.org/10.1080/09687599.2012.710008
• Onishi, K. H., & Baillargeon, R. (2005). Do 15-month-old infants understand false beliefs? Science, 308(5719), 255–258. https://doi.org/10.1126/science.1107621
• Perner, J., & Wimmer, H. (1985). “John thinks that Mary thinks that…”: Attribution of second-order beliefs by 5- to 10-year-old children. Journal of Experimental Child Psychology, 39(3), 437–471. https://doi.org/10.1016/0022-0965(85)90051-7
• Peterson, C. C., & Siegal, M. (1999). Representing false beliefs in autism, blindness, and deafness. Cognition, 73(3), 189–214. https://doi.org/10.1016/S0010-0277(99)00057-0
• Premack, D., & Woodruff, G. (1978). Does the chimpanzee have a theory of mind? Behavioral and Brain Sciences, 1(4), 515–526. https://doi.org/10.1017/S0140525X00076512
• Russell, J., Mauthner, N., Sharpe, S., & Tidswell, T. (1991). The ‘windows task’ as a measure of strategic deception in preschoolers and autistic subjects. British Journal of Developmental Psychology, 9(2), 331–349. https://doi.org/10.1111/j.2044-835X.1991.tb00881.x
• Saxe, R., & Kanwisher, N. (2003). People thinking about thinking people: The fMRI investigations of Theory of Mind. NeuroImage, 19(4), 1835–1842. https://doi.org/10.1016/S1053-8119(03)00230-1
• Siegal, M., & Beattie, K. (1991). Natural language, conventional language, and typicality in false belief tasks. Child Development, 62(2), 296–303. https://doi.org/10.1111/j.1467-8624.1991.tb01531.x
• Southgate, V., Senju, A., & Csibra, G. (2007). Action anticipation through attribution of false belief by 2-year-olds. Psychological Science, 18(7), 587–592. https://doi.org/10.1111/j.1467-9280.2007.01944.x
• Wellman, H. M., Cross, D., & Watson, J. (2001). Meta-analysis of theory-of-mind development: The truth about false belief. Child Development, 72(3), 655–684. https://doi.org/10.1111/1467-8624.00304
• Wimmer, H., & Perner, J. (1983). Beliefs about beliefs: Representation and constraining function of wrong beliefs in young children’s understanding of deception. Cognition, 13(1), 103–128. https://doi.org/10.1016/0010-0277(83)90004-5