Cognitive PrimatologyComparative PsychologyPhilosophy of Mind

The Theory of Mind in Chimpanzees Experiment (Knower vs. Guesser) – David Premack and Guy Woodruff

A comprehensive academic analysis of Premack and Woodruff’s groundbreaking 1978 study on chimpanzee Theory of Mind and the Knower vs. Guesser paradigm.

memjavad
PUBLISHED
Scientifically Reviewed · Dr. Marwa Abd-Alazim · September 16, 2026
Medically & Scientifically Reviewed Verified: September 16, 2026
Dr. Marwa Abd-Alazim Ph.D.
Professor of Psychology • University of Kerbala
Review Criteria & Clinical Standards

This content undergoes rigorous scientific peer-review and medical editorial standards at Arab Psychology Network to ensure clinical accuracy, validity, and compliance with evidence-based guidelines from leading psychological and healthcare authorities (APA / WHO).

In 1978, cognitive psychologists David Premack and Guy Woodruff published an essay in Behavioral and Brain Sciences that transformed comparative psychology: “Does the chimpanzee have a theory of mind?” Prior to this publication, comparative animal research operated under the dominance of behaviorist mechanics, reducing animal cognition to stimulus-response contingencies, schedule-induced conditioning, and observable motor patterns. Premack and Woodruff directly challenged this paradigm by proposing that an adult chimpanzee (Pan troglodytes) named Sarah could represent the internal, unobservable mental states of others. They suggested that non-human primates could construct cognitive models of intentionality, discerning goals, desires, and knowledge states in conspecifics and human handlers.

The core of this inquiry was the distinction between an organism that merely tracks overt behavior and one that operates as a psychological theorist. To attribute mental states is to possess a Theory of Mind (ToM)—a system of inferences that treats the minds of others as causally efficacious, representational engines. While Premack and Woodruff’s initial 1978 study relied on videotaped problem-solving tasks, the methodological disputes that followed necessitated an experimental design that could isolate epistemic states (knowing versus not knowing) from simple behavioral cues. This led directly to the development of the “Knower versus Guesser” paradigm, a methodology that separated visual perception from informational access to determine whether an animal understands that “seeing leads to knowing.”

Decades later, the Knower versus Guesser paradigm remains a critical reference point in cognitive primatology, developmental psychology, and the philosophy of mind. The experiment sparked debates between mentalistic interpretations of non-human animal behavior and radical behavioral abstraction models, compelling scientists to interrogate the boundaries of animal consciousness. Tracing the trajectory from Sarah’s plastic tokens to contemporary eye-tracking false-belief tasks, the intellectual history of the Knower versus Guesser experiment illuminates the empirical challenges of demonstrating whether non-human animals can read minds or whether they are simply adept at reading behavior.

1. Historical Foundations and the Coining of Theory of Mind

1.1 The Seminal 1978 Paper and Theoretical Foundations

The publication of David Premack and Guy Woodruff’s 1978 paper, “Does the chimpanzee have a theory of mind?” in Behavioral and Brain Sciences, served as an epistemological rupture in comparative psychology. At the time of its release, cognitive ethology was an embryonic discipline, struggling to establish methodological credibility against decades of behaviorist orthodoxy. Premack and Woodruff posited a bold, operationalized definition: an individual possesses a theory of mind if they impute mental states to themselves and others. These mental states, ranging from purpose and intention to knowledge, belief, thinking, doubt, and emotion, are fundamentally unobservable. Consequently, any cognitive system that uses them to predict and interpret another agent’s actions must rely on a representational framework—a “theory.”

The philosophical significance of this framing lay in its break from simple associative learning. Under a strict behaviorist paradigm, an organism’s behavioral trajectory is understood entirely as a function of past reinforcement history, environmental discriminative stimuli, and physiological drives. Premack and Woodruff argued that such low-level accounts were insufficient to explain the adaptive flexibility of primate social interactions. If a chimpanzee could anticipate a human actor’s behavior by inferring an internal representation—such as an unfulfilled goal or a state of knowledge—the animal was not merely reacting to physical cues; it was computing second-order mental representations. The paper inaugurated an ongoing effort to determine whether mentalistic attribution could be proven through empirical testing without falling into anthropomorphism.

1.2 Pre-Experimental Paradigms in Comparative Cognition

Before 1978, experimental paradigms in animal cognition were largely shaped by B.F. Skinner’s operant conditioning and Clark Hull’s drive-reduction theories. While Wolfgang Köhler had demonstrated insight learning in apes at Tenerife decades earlier, mid-twentieth-century laboratory experiments prioritized strict behavioral controls. Primates were primarily tested in Wisconsin General Test Apparatus (WGTA) enclosures, working through repetitive discrimination tasks designed to measure associative strength, stimulus generalization, and reversal learning. Cognitive processes were often treated as an inaccessible “black box,” and attributing internal representations or communicative intent to non-human subjects was widely viewed as unscientific.

Concurrently, human developmental psychology was undergoing a revolution led by Jean Piaget’s genetic epistemology. Piagetian stages demonstrated that human infants progressively develop object permanence, symbolic representation, and the capacity to overcome cognitive egocentrism. Comparative researchers began to ask whether great apes might mirror these early cognitive milestones. Studies exploring cross-modal matching, delayed-response tasks, and symbolic token manipulation emerged in the late 1960s and early 1970s. These lines of research converged to establish cognitive primatology as an independent subfield, shifting the central research question from what animals can be conditioned to do, to how they mentally represent their physical and social environments.

1.3 Defining the Epistemological Framework

To evaluate whether an animal possesses a theory of mind, researchers required a rigorous taxonomy of intentional systems. Philosopher Daniel Dennett provided this architecture with his formulation of the intentional stance. Within Dennett’s framework, a zero-order intentional system possesses no internal beliefs or desires; it operates purely as an algorithmic input-output mechanism. A first-order intentional system possesses internal states such as beliefs and desires about the external world (e.g., “the chimpanzee wants the fruit”). However, a genuine theory of mind requires second-order intentionality: the capacity to hold beliefs about beliefs, or desires about beliefs (e.g., “the chimpanzee believes that the human knows where the fruit is hidden”).

Differentiating second-order intentionality from simpler perceptual awareness became the primary methodological challenge. An animal might track the physical orientation of a competitor’s head without attributing any subjective visual experience to that competitor. Demonstrating knowledge attribution requires proving that the subject differentiates between an agent who has accessed specific information and one who has not, independent of concurrent physical postures. Crucially, the definitive benchmark for full human-like Theory of Mind emerged as the understanding of false belief—recognizing that an agent can hold an internal representation of the world that conflicts with objective reality. This distinction established the baseline criteria required to prove an organism represents another agent’s subjective perspective rather than merely calculating physical behavioral probabilities.

2. Subject Background: Sarah the Chimpanzee

2.1 Sarah’s Rearing History and Language Training

Sarah, an African-born female chimpanzee acquired by the laboratory of David Premack in the late 1960s, served as the primary subject for the initial Theory of Mind experiments. Unlike conspecifics reared in traditional biomedical isolation or standard zoological facilities, Sarah underwent intensive language-training protocols designed to assess artificial symbolic communication. Premack rejected acoustic or gestural systems in favor of a visual-spatial token system. Sarah was taught to manipulate distinctively shaped, colored, and textured pieces of plastic backed with metal, which adhered magnetically to an upright testing board. Each token functioned as an arbitrary symbol representing objects, actions, properties, and logical relationships.

Through systematic training, Sarah learned to interpret and produce multi-token sequences resembling syntactic sentences. She mastered conditional logic (“if-then” constructions), negation, and abstract relational concepts such as “same” versus “different.” However, Sarah’s extensive enculturation introduced complex methodological considerations. Having spent years in close, mediated interaction with human experimenters, she was accustomed to cooperative problem-solving and communicative exchanges. This raised critical questions regarding generalizability: Did Sarah’s performance reflect the latent cognitive capacities of wild Pan troglodytes, or was her mental architecture artificially organized by linguistic scaffolding and cross-species enculturation?

2.2 Cognitive Capabilities Prior to Theory of Mind Testing

Before her involvement in mental state attribution research, Sarah had completed extensive batteries assessing non-verbal and symbolic cognitive capabilities. She demonstrated advanced proficiencies in analogical reasoning tasks, such as solving proportional geometric matrices (A is to B as C is to D) and functional analogies involving everyday tools (e.g., an intact apple is to a cut apple as an unpeeled banana is to a peeled banana, paired with the appropriate cutting instrument). These achievements confirmed that Sarah possessed an ability to identify structural, causal, and relational parallels between disparate physical entities.

Furthermore, Sarah showed competence in causal inference and physical conservation tasks analogous to Piagetian metrics. She could reconstruct disrupted temporal-causal sequences, selecting photographs that correctly represented the intermediate steps between an initial state and a modified end state (e.g., a dry sponge, an intervening action involving water, and a saturated sponge). Her spatial working memory was robust, and she regularly completed complex cross-modal matching tasks that required translating visual representations into tactile choices. Her social interactions with long-term human trainers were equally nuanced; she differentiated between human caretakers based on past interactions, demonstrating an awareness of human behavioral tendencies and contextual routines.

2.3 Ecological Validity Concerns of Enculturated Primates

The reliance on an enculturated subject like Sarah introduced foundational debates regarding the ecological validity of the findings. Critics, including behaviorists and evolutionary ethologists, argued that an ape reared in an artificial human environment, steeped in symbolic tokens and reward structures, represents a cognitive outlier. Such animals might develop heuristic problem-solving strategies that do not occur in wild chimpanzee populations. Wild chimpanzee sociality is defined by intense fission-fusion dynamics, dominance hierarchies, group territorial defense, and intra-specific competition, where misattributing an opponent’s intent carries severe fitness costs.

Conversely, proponents of the enculturation hypothesis, including Premack and later Michael Tomasello, argued that enculturation does not implant non-biological cognitive mechanisms. Instead, human mediation serves as a developmental scaffold that unmasks latent cognitive potential. In this view, wild chimpanzees possess the foundational neurological architecture for complex social cognition, but their competitive ecological environment suppresses the expression of cooperative-communicative competencies. Testing an enculturated ape allowed researchers to investigate the maximum computational capacity of the chimpanzee brain, freed from the survival constraints of the wild.

3. The Initial Videotape Experiments (Premack & Woodruff 1978)

3.1 Experimental Protocol and Videotaped Problem Scenarios

The experimental series that inaugurated Theory of Mind testing in 1978 utilized a novel problem-solving methodology based on 30-second videotaped recordings. Premack and Woodruff presented Sarah with video recordings of an adult human actor situated in a concrete enclosure, confronting various physical impediments to a visible goal. The problems were divided into distinct physical scenarios: an actor attempting to reach a cluster of bananas suspended overhead beyond arm’s reach; an actor trying to grasp bananas resting outside cage bars beyond reach of a short stick; an actor shivering violently before an unlit electric heater with its cord visibly detached; and an actor trapped inside a room unable to open a locked door.

Each video concluded with the human actor actively expressing frustration or remaining suspended in a state of unresolved effort. Immediately following the video clip, the display was terminated, and Sarah was presented with a choice between two or more still photographs. One photograph depicted the functional solution to the actor’s dilemma (e.g., a pile of stepped boxes beneath the hanging fruit, a long pole, a heater cord plugged into a wall outlet, or an intact key inserted into a padlock). The alternative photographs presented non-functional distractors that were visually matched to the scene (such as a bent stick, broken keys, or unattached cords) or depicted actions unrelated to the goal state. Sarah’s task was to select the photograph that represented the appropriate resolution.

3.2 Sarah’s Performance Across Trial Variations

Sarah demonstrated a high degree of accuracy across novel, unreinforced problem scenarios. On her very first trials with completely novel video sequences, she consistently chose the functional solution over the non-functional distractors. Statistical analysis confirmed that her correct selections occurred at levels significantly above chance ($p < 0.005$). Crucially, Sarah did not select alternatives based on low-level perceptual matching or physical visual similarity. For instance, when presented with the unlit electric heater, she did not simply choose a photograph containing any electric cord; she specifically chose the photograph showing the cord plugged into an electrical socket, rather than an unattached or physically severed cord.

Response latencies and behavioral observations revealed that Sarah made these selections rapidly, often requiring only a few seconds of inspection following the termination of the video feed. When novel human actors replaced the familiar experimenters in the video recordings, her performance remained stable, demonstrating that her choices were not dependent on a specific history of reinforced interactions with an individual handler. Her success indicated an ability to extrapolate a functional solution to an unresolved scenario, suggesting that she grasped the structural dynamics of the problems displayed on screen.

3.3 Testing Intentionality: Preferred vs. Disliked Human Actors

To differentiate whether Sarah was solving these tasks via physical causal logic or by attributing mental intentions to the human actors, Premack and Woodruff introduced a social manipulation. They contrasted two human actors who had established distinct social relationships with Sarah: a “beloved” trainer (Keith), who routinely provided food, play, and positive social reinforcement, and an “unfriendly” or disliked trainer, who frequently restricted access to desired items, enforced strict disciplinary boundaries, and exhibited antagonistic social interactions toward her.

When Sarah viewed videotapes of the beloved trainer facing a physical dilemma, she consistently selected photographs depicting successful functional resolutions (e.g., the bananas successfully reached, the heater functioning properly). However, when viewing identical problems faced by the disliked trainer, her choices changed significantly. For the disliked trainer, Sarah systematically selected photographs depicting failure, disaster, or physical misfortune. In one instance involving an unlit heater, she chose a photograph showing the trainer engulfed in flames; in other scenarios, she selected photos of the actor falling, breaking equipment, or failing to retrieve the reward. This finding suggested that Sarah was not simply calculating the mechanical solution to a physical state. Rather, she appeared to evaluate the actor’s internal goal state, modulating her response based on her own emotional attitude toward the individual.

4. Genesis and Architecture of the Knower vs. Guesser Paradigm

4.1 Theoretical Transition from Video Attribution to Informational Access

Despite the findings of the 1978 videotape studies, critics argued that Sarah’s performance could be explained without attributing a Theory of Mind. As commentators in Behavioral and Brain Sciences noted, Sarah’s choice of photographs could be governed by learned causal scripts or associative completion heuristics (e.g., broken objects match negative outcomes; intact objects match positive outcomes). Choosing a picture of a plugged-in cord does not necessarily prove an understanding that the human *wanted* warmth; it may simply reflect an association between heaters and functioning power cords.

Recognizing the need to isolate epistemic states (knowing versus ignorance) from causal problem-solving, Woodruff and Premack adapted their methodology in 1979. They recognized that the most direct way to evaluate mentalistic attribution was to test whether an animal understands the relationship between perceptual access and internal knowledge. This motivated the development of the “Knower versus Guesser” experimental paradigm. The central hypothesis shifted from: “Can an ape perceive a physical goal?” to: “Can an ape determine whether another individual has gained epistemic access to a hidden piece of information based strictly on their visual history?”

4.2 Structural Components of the Knower vs. Guesser Paradigm

The Knower versus Guesser paradigm established an experimental architecture that has since been widely utilized across comparative psychology. The spatial configuration involved a testing room containing a subject animal, an independent human baiter, and two human confederates who served as informants. Crucially, the physical target—a high-value food reward—was hidden inside one of several identical, opaque containers (e.g., inverted cups, covered buckets, or wooden boxes) situated out of the subject’s direct line of sight during the hiding phase.

The roles of the two human informants were carefully manipulated around the event of baiting:

  • The Knower: Positioned so they had direct, unobstructed visual access to the baiter placing the food reward into a specific container.
  • The Guesser: Deprived of perceptual access during the baiting process, achieved by having the Guesser step outside the room, stand behind a physical barrier, or wear an opaque covering over their head.

Once the food was concealed and the visual barriers removed, both the Knower and the Guesser pointed toward the containers. The Knower consistently pointed to the container that held the food, while the Guesser pointed to an empty container. The subject chimpanzee was then allowed to select one container based on these pointing gestures. Because the chimpanzee could not directly see the food, successful choice required discerning which human informant held accurate, valid information.

4.3 Differentiation Between Perceptual Access and Knowledge State

The epistemological challenge of the Knower versus Guesser paradigm centers on operationalizing the principle that “seeing leads to knowing.” In human developmental psychology, children grasp by roughly three to four years of age that visual perception is an informational channel that creates knowledge representations in the mind. If an agent looked inside a box, that agent knows what is inside; if an agent was blindfolded or absent, they are ignorant.

In comparative cognition, demonstrating that an ape understands this connection requires separating visual behavior from mentalistic attribution. An animal might track the physical orientation of human eyes or heads as a simple spatial vector pointing toward a target, without attributing an internal state of “knowledge” to the informant. Alternatively, an animal could use learned associative rules, such as “choose the person whose eyes were open” or “avoid the person who had a bucket on their head,” through reinforcement learning. To rule out these low-level accounts, the paradigm required transfer tasks using novel occluders, testing whether the subject could generalize the concept of informational access without relying on trial-and-error conditioning.

5. Methodology of the Knower vs. Guesser Experiment

5.1 The Physical Apparatus and Spatial Configuration

The standard physical apparatus for the Knower versus Guesser paradigm was engineered to prevent unintentional informational leakage while standardizing lines of sight. The testing arena typically consisted of an ape holding enclosure separated from an adjacent testing space by transparent Plexiglas or heavy wire mesh. Positioned within this testing space was an array of identical, opaque containers—frequently two to four inverted white plastic cups or covered wooden boxes—mounted on a sliding table or fixed platforms beyond the immediate physical reach of the subject.

A primary visual occluder, such as a large movable wooden or opaque fabric screen, was positioned between the subject ape and the baiting table. This screen prevented the chimpanzee from witnessing which container was baited, ensuring the ape had to rely on the informants. The human informants were situated symmetrically on either side of the baiting area. Barrier heights, viewing angles, and the distances between the informants, the baiting containers, and the ape were calibrated so that the line of sight for the Knower was uninhibited, while the sightline for the Guesser was completely blocked. The ape was maintained at a central observation point to prevent side biases from dictating spatial choices.

5.2 The Role Manipulation Protocols

The behavioral scripts executed by the human informants required strict temporal and kinematic standardization. During a trial, the baiter entered the testing arena, drew attention to the food reward, and placed it inside one of the opaque containers while the subject ape was visually occluded. The Knower maintained continuous head and eye orientation toward the baiter, visually tracking the placement of the food. Conversely, the Guesser’s perceptual access was interrupted using one of several standardized role manipulations:

  • Physical Absence: The Guesser left the room entirely before baiting began and re-entered only after the reward was hidden.
  • Head Covering: The Guesser remained seated but placed an opaque bucket or opaque cardboard box entirely over their head.
  • Blindfolding: The Guesser remained seated but wore an opaque cloth blindfold covering their eyes.
  • Back-Turned: The Guesser sat with their back completely turned away from the baiting apparatus.

After the food was hidden, the occluding screen blocking the chimpanzee’s view was retracted. Both the Knower and the Guesser simultaneously extended an arm to point at a container, holding their index finger roughly ten centimeters from their selected box. The informants maintained neutral facial expressions, locked their gaze onto the center of the chosen container, and avoided dynamic micro-movements. The identities of the Knower and Guesser were systematically counterbalanced across trials to ensure that the subject was not simply responding to individual human handlers.

5.3 Trial Structures, Habituation, and Transfer Phases

The experimental trajectory typically proceeded through three successive phases: habituation and baseline pointing comprehension, acquisition trials with the standard Knower versus Guesser configuration, and transfer trials using novel occluders. In the baseline phase, subjects were confirmed to have a basic understanding of human pointing. A single experimenter pointed to a baited container in full view of the subject to confirm that the animal could use a distal pointing cue to retrieve food.

Once this baseline was established, acquisition testing began. The subject was confronted with the binary choice between the Knower and Guesser across blocks of reinforced trials (typically 10 to 20 trials per block). If the subject pointed toward the container indicated by the Knower, they received the hidden reward; if they selected the Guesser’s container, the empty box was revealed, and no food was provided. If an animal required hundreds of reinforced trials to consistently choose the Knower, this could indicate associative learning rather than an understanding of mental states. Consequently, researchers introduced transfer phases with novel occluders (e.g., replacing buckets with paper bags, novel screens, or hand coverings) on trial-one probe configurations to test whether the animal understood the concept of visual occlusion itself.

6. Methodological Controls and Mitigating Clever Hans Effects

6.1 Elimination of Inadvertent Behavioral Cues

A persistent methodological challenge in comparative cognition is the Clever Hans effect, wherein animal subjects respond to subtle, unintentional bodily cues from human handlers rather than the intended experimental stimuli. In the Knower versus Guesser paradigm, human informants might subtly lean toward the baited container, widen their eyes, alter their breathing, or display micro-postural adjustments indicating the food’s location. If an ape detects these cues, successful performance can be achieved without any understanding of the informants’ knowledge states.

To eliminate these inadvertent cues, advanced variations of the paradigm instituted strict kinematic and postural controls. Both the Knower and Guesser were trained to execute identical pointing gestures, standardizing arm angles, extension velocity, and hand positions. Furthermore, the role of the baiter was separated from the informants: the baiter completed the concealment and then retreated behind an occluding blind or left the room, preventing the baiter from cuing the subject. In the most rigorous double-blind iterations, the informants themselves did not know which container held the food; rather, they received auditory cues through headphones directing them where to point, effectively neutralizing unintentional somatic signaling.

6.2 Controlling for Low-Level Perceptual and Spatial Biases

Non-human primates frequently exhibit strong, idiosyncratic behavioral biases in two-choice discrimination tasks. These include side preferences (consistently choosing the container on the left or right regardless of experimental cues), hand biases (preferring to reach with the dominant limb toward the ipsilateral side), and spatial alternation patterns (systematically switching left-right-left-right across consecutive trials). Left unaddressed, these biases can mask cognitive competence or create statistical false positives.

Researchers applied randomized, counterbalanced baiting schedules, ensuring that the food was never placed in the same container for more than two or three consecutive trials. Informant seating positions were similarly counterbalanced so that the Knower was not persistently associated with a preferred side. Researchers monitored response patterns statistically; if an ape exhibited a significant side bias, data analysis evaluated choice accuracy independently for both left- and right-sided presentations. In addition, experimental designs controlled for auditory cues—such as the faint thud of an apple or pellet dropping into a container—by using padded receptacles or introducing white noise to mask sound cues during baiting.

6.3 Addressing Rapid Associative Learning Confounds

The primary critique leveled against the Knower versus Guesser paradigm concerns rapid associative learning. If a chimpanzee receives food reinforcement across multiple trials, it does not need to understand “knowing” or “seeing” to solve the task. The ape needs only to form a low-level associative rule: “Person without bucket = reward; Person with bucket = no reward.” If this association can be acquired in 10 to 30 trials, the animal’s success can be explained by reinforcement learning rather than a Theory of Mind.

To differentiate mentalistic inference from associative conditioning, researchers relied heavily on trial-one analysis during transfer tasks. When an animal that learned the task with a bucket over the Guesser’s head was suddenly presented with a Guesser holding a cardboard shield or wearing an opaque blindfold, researchers recorded the subject’s *immediate* choice on the first trial. If the ape possessed a generalized understanding that visual barriers block perceptual access, it should choose the Knower on trial one of this novel condition. If the animal chose at chance on trial one and only gradually learned to avoid the new occluder over repeated, reinforced trials, this suggested associative rule learning rather than mental state attribution.

7. Empirical Findings and Behavioral Outcomes

7.1 Immediate vs. Longitudinal Choice Accuracy

The empirical outcomes of the original Knower versus Guesser experiments and their immediate follow-ups presented a complex, controversial picture. When chimpanzees were first tested in the standard configuration (Knower present, Guesser absent or wearing a bucket), they did not display immediate, trial-one comprehension. Instead, subjects typically exhibited chance-level performance across their initial test blocks. Success was achieved only longitudinally, after extended exposure to differential reinforcement across dozens, and in some instances hundreds, of trials.

Statistical analysis of these learning curves revealed patterns characteristic of visual discrimination learning rather than insight. Subjects gradually extinguished their initial side biases and learned to avoid the informant who was visually compromised. Individual differences were pronounced: some subjects, like Sarah, demonstrated faster acquisition rates than younger or non-enculturated apes, but immediate insight was absent. The requirement for extensive trial-and-error training left the data vulnerable to behaviorist critiques, which maintained that the primates were simply conditioning to physical markers on the informants’ bodies rather than inferring informational states.

7.2 Performance Across Variable Occlusion Conditions

The limitations of the chimpanzees’ performance became clearer when researchers introduced varied occlusion conditions designed to test the boundaries of their visual comprehension. While subjects successfully learned to differentiate between an informant who was completely absent from the room and one who was present, their performance dropped when the visual barriers were made more subtle. When the Guesser remained in the room but wore an opaque bucket over their head, apes learned to choose the Knower, though this typically required extended training.

However, when confronted with conditions where the Guesser wore a small blindfold covering only the eyes, or held a small cardboard screen directly before the face, chimpanzees consistently failed to discriminate between the informants. They chose the blindfolded informant at chance levels, continuing to follow pointing gestures from humans whose eyes were obscured. Performance patterns suggested that apes were tracking gross bodily orientation: as long as an informant’s body, torso, and face were directed toward the baiting apparatus, the chimpanzee treated that person as an informative communicator, showing little spontaneous sensitivity to whether the eyes themselves were unobstructed.

7.3 The Woodruff and Premack (1979) Deceptive Paradigm Findings

Recognizing the ambiguities inherent in cooperative pointing paradigms, Woodruff and Premack (1979) introduced an alternative framework: intentional deception. They reasoned that if a chimpanzee understands what another agent knows, it should be capable of withholding information from or actively misinforming a competitor. In this study, Sarah was introduced to two types of human trainers: a “cooperative” trainer who shared any hidden food Sarah indicated, and a “competitive” trainer who kept the food for himself whenever Sarah directed him to it.

Over extended trials, Sarah learned to adjust her communicative signaling depending on the trainer’s identity. With the cooperative trainer, she consistently pointed to the baited container. With the competitive trainer, Sarah eventually learned to withhold her pointing gesture entirely or actively point to the *empty* container, leading the competitor away from the food so that she could obtain it later. While Woodruff and Premack interpreted this as evidence of intentional deception and an understanding of the trainer’s beliefs, critics again pointed out the protracted learning curve. Sarah did not spontaneously deceive the competitor; the behavior developed over hundreds of reinforced trials, leaving open the possibility of operant discrimination learning based on food outcomes.

8. Competing Interpretations: Mentalizing vs. Behavioral Abstraction

8.1 The Mentalizing Hypothesis (Premack, Woodruff, de Waal)

The mentalizing hypothesis posits that great apes construct internal, cognitive models of others’ psychological states. David Premack, Guy Woodruff, and later primatologist Frans de Waal argued that great apes navigate complex social systems where predicting conspecific behavior solely via low-level stimulus-response pairings would be computationally inefficient. Social life, they contended, demands a representational framework where visible behaviors (such as head direction, posture, and facial expressions) are translated into inferred mental states (such as “he sees the food,” “he wants the food,” or “he intends to attack”).

Proponents of this view argued that Sarah’s performance across varied tasks—from the photographic problem-solving sequences to the deceptive pointing tasks—reflected an underlying competence in second-order intentionality. Even if learning was required to master specific experimental interfaces, the speed with which apes could generalize rules across social scenarios pointed toward a broader cognitive framework. In this view, mentalizing does not require human-level language; rather, it exists as an evolutionarily ancient, adaptive heuristic that allows primates to predict conspecific actions by treating them as intentional agents.

8.2 The Behavioral Abstraction Hypothesis (Povinelli and Vonk)

In direct opposition to the mentalizing view, comparative psychologist Daniel Povinelli and his collaborator Jennifer Vonk formulated the “behavioral abstraction hypothesis.” Povinelli argued that the mentalizing interpretation fell prey to anthropomorphic projection. According to this critique, chimpanzees are skilled “behavior-readers” rather than “mind-readers.” They possess sophisticated perceptual systems that allow them to extract complex, invariant regularities from the physical behavior of others without ever imputing unobservable mental states.

Under the behavioral abstraction model, an ape does not compute: “This human cannot see the food; therefore, he does not know where it is.” Instead, the ape constructs a direct behavioral rule: “A person whose body and face are oriented toward a container is a reliable predictor of reward.” Because physical postures and behaviors are directly correlated with outcomes in natural environments, an animal can successfully navigate complex social and deceptive interactions relying exclusively on these physical correlations. Povinelli posed the “Logical Problem”: because mental states are always paired with physical behaviors, any experiment that provides behavioral cues can be solved via behavioral abstraction, making mentalizing hypotheses untestable without specific controls.

8.3 The Conceptual Re-Interpretation by Call and Tomasello

In the late 1990s and early 2000s, Josep Call and Michael Tomasello proposed a conceptual re-interpretation of the Knower versus Guesser paradigm. They argued that the early failures of chimpanzees in laboratory pointing tasks were not proof of cognitive absence, but rather artifacts of an ecologically invalid testing paradigm. Chimpanzees do not naturally point to communicate with conspecifics, nor do they naturally share food cooperatively. Expecting an ape to understand a cooperative-communicative pointing gesture from a human informant forced the animal into an evolutionary mismatch.

Tomasello and Call argued that non-human primate social intelligence is primarily adapted for competition. In the wild, chimpanzees use their cognitive capacities to outcompete rivals for food, mates, and dominance. Consequently, when the Knower versus Guesser paradigm was translated into a competitive context—such as competing against a dominant conspecific for a hidden reward—chimpanzees demonstrated visual perspective-taking abilities that had remained undetected in cooperative tests. This perspective synthesized the divide: apes are not Cartesian mind-readers who represent abstract beliefs, but neither are they simple behaviorist machines. They possess an understanding of what others *see* and *know*, but this capacity is bound to specific competitive contexts.

9. Philosophical and Cognitive Implications

9.1 Intentionality and the Philosophy of Mind

The Knower versus Guesser experiment directly impacted contemporary philosophy of mind, particularly discussions surrounding the nature of intentionality. Tracing back to Franz Brentano, intentionality—the capacity of the mind to be directed toward, or represent, objects and states of affairs—was long considered the defining mark of human consciousness. Rene Descartes had posited a radical ontological divide between the mechanistic, non-conscious nature of animals (res extensa) and the thinking, intentional substance of the human soul (res cogitans).

By attempting to operationalize second-order intentionality in a non-human ape, Premack and Woodruff challenged this Cartesian divide. If an animal can form a belief about another agent’s state of information, intentionality can no longer be viewed as a uniquely human, linguistically mediated phenomenon. Instead, it must be understood as an evolved neuro-cognitive adaptation that developed incrementally across primate phylogeny. This shift provided empirical grounding for naturalist philosophies of mind, suggesting that human cognitive architecture is an elaboration of representational systems shared with other great apes.

9.2 Mental Representation and Epistemic States

The distinction between perceptual tracking and epistemic representation is central to the debate over animal mind-reading. Can a non-linguistic organism represent an *epistemic state* (such as knowing, doubting, or believing) as distinct from the immediate sensory reality of its environment? The Knower versus Guesser paradigm evaluated this question by asking whether an ape can treat visual perception as an event that creates an enduring representational state in another mind.

To possess a fully representational Theory of Mind, an organism must understand metarepresentation: the capacity to represent a representation. In the context of the Knower versus Guesser task, the subject must form an internal model of the informant’s internal model of the world. If the chimpanzee only tracks whether an informant’s line of sight intersects a physical object, it is relying on spatial geometric tracking rather than epistemic attribution. Cognitive modularity theories, such as those advanced by Alan Leslie in human developmental research, propose a dedicated “Theory of Mind Mechanism” (ToMM). The primatological debate remains centered on whether non-human primates possess an evolutionary precursor to this module or whether they rely on general-purpose social learning mechanisms.

9.3 The Demarcation Problem Between Mind-Reading and Behavior-Reading

The primary theoretical obstacle emerging from the Knower versus Guesser paradigm is the “Demarcation Problem,” often formalized by philosophers of science such as Susan Hurley, Stephen Stich, and Robert Lurz as the “Logical Problem.” The problem can be framed as follows:

For any proposed mentalistic explanation of animal behavior ($M$), there exists an equivalent behavioral-abstraction explanation ($B$) that predicts the exact same behavioral outcome based on observable cues:

  • Hypothesis M (Mind-reading): The animal infers that the informant saw the food, knows its location, and therefore points to the correct container.
  • Hypothesis B (Behavior-reading): The animal relies on an abstracted behavioral rule: “Follow the pointing gesture of an individual who was facing the target during the concealment event.”

Because mental states (like seeing) naturally co-occur with specific physical orientations (like head direction and unobstructed lines of sight), the animal’s choice can always be explained by the physical cues alone. To break this interpretive stalemate, researchers have had to develop experimental paradigms where mental states and physical behaviors diverge, such as scenarios involving novel visual occluders, mirrors, or deceptive visual cues that render past behavioral associations useless.

10. Replications, Refinements, and Povinelli’s Challenge

10.1 Povinelli, Nelson, and Boysen (1990) Formalization

In 1990, Daniel Povinelli, Keith Nelson, and Sarah Boysen published a direct replication and methodological refinement of the Knower versus Guesser experiment. Working with a cohort of four laboratory-reared chimpanzees, their study sought to address the methodological ambiguities and small sample sizes of earlier research. The experimental apparatus was upgraded to include standardized screens, uniform pointing protocols, and automated baiting mechanisms to reduce inadvertent experimenter cuing.

The initial results appeared to replicate Woodruff and Premack’s findings: over a series of training blocks, the chimpanzees learned to discriminate between the Knower and the Guesser, selecting the container indicated by the Knower at rates significantly above chance. However, Povinelli and colleagues introduced critical transfer probes. When the visual status of the informants was altered using novel, subtle variations—such as having the Guesser hold a small screen over their eyes rather than leaving the room entirely—the subjects’ performance declined. The apes struggled to generalize their performance when the visual impediment was reduced to the eye region, raising early concerns that their initial success had been driven by learning gross behavioral orientations rather than understanding visual attention.

10.2 The 1996 Monograph: ‘What Chimpanzees Know About Seeing’

In 1996, Povinelli and his research team published a landmark monograph titled What Chimpanzees Know About Seeing, detailing five years of testing with seven young chimpanzees across dozens of experimental conditions. Moving beyond pointing paradigms, Povinelli instituted a cooperative-begging paradigm. The chimpanzees, who naturally used an extended-arm begging gesture to request food, were presented with two human experimenters holding food rewards. One experimenter could see the ape, while the other experimenter was visually incapacitated via various occluders.

The test conditions included:

  • Blindfolds: One experimenter wore a blindfold over their eyes; the other wore a blindfold over their mouth.
  • Buckets: One experimenter had an opaque bucket over their head; the other held an identical bucket resting on their shoulder.
  • Screens: One experimenter held a small opaque screen directly in front of their face; the other held the screen behind their head.
  • Back-Turned: One experimenter looked forward toward the ape; the other sat with their back turned.

Across hundreds of trials, the chimpanzees begged from the visually incapacitated experimenter at roughly the same rate as the seeing experimenter on initial transfer trials. They begged from humans wearing blindfolds over their eyes, humans with buckets over their heads, and humans looking away, showing a reliance on whether a human body was present and physically facing them. Povinelli concluded that chimpanzees possess no internal concept of seeing as an attention-directing psychological state; their cognition, he argued, was rooted entirely in behavioral heuristics.

10.3 Critiques of Povinelli’s Negative Findings

Povinelli’s monograph provoked immediate pushback from cognitive primatologists, who identified several potential methodological flaws in his testing regime. First, critics argued that the cooperative begging scenario lacked ecological validity. In nature, chimpanzees do not beg for food from conspecifics using delicate visual assessments; food sharing among adults is rare and typically mediated by harassment or dominance dynamics rather than cooperative visual signaling.

Second, critics pointed to species-specific sensory and morphological biases. Unlike humans, who possess a bright white sclera that highlights gaze direction, chimpanzees have dark sclera and pigmented eyes, making fine ocular tracking difficult. Conspecific communication in wild chimpanzees relies on gross head, neck, and whole-body orientation. Expecting an ape to spontaneously prioritize a small cloth blindfold over the eyes of a human handler ignored the evolutionary morphology of chimpanzee visual communication. Finally, concerns were raised regarding the developmental status of Povinelli’s subjects, who were tested as juveniles and adolescents; human children do not pass comparable Theory of Mind benchmarks until early childhood, leaving open the possibility that the subjects were simply too young to display these cognitive capacities.

11. Comparative Perspectives: Great Apes, Children, and Other Species

11.1 Comparative Developmental Trajectories with Human Children

To contextualize the performance of non-human primates, developmental psychologists adapted the Knower versus Guesser paradigm for human infants and toddlers. When tested on identical two-choice informant tasks, human children exhibit a clear developmental progression. Toddlers between 18 and 24 months of age track gaze vectors and follow pointing gestures to locate hidden objects, but like chimpanzees, they frequently fail to differentiate between informants wearing blindfolds versus those with unobscured vision on initial trials.

By 36 to 48 months of age, a significant cognitive shift occurs. Human children reliably pass the Knower versus Guesser paradigm on trial one across diverse occlusion conditions, explicitly identifying that an individual wearing a blindfold, a bucket, or an eye mask cannot see and therefore does not know the location of the reward. Shortly thereafter, children pass explicit false-belief tasks (such as the Sally-Anne paradigm), demonstrating a fully representational Theory of Mind. This developmental divergence suggests that while basic gaze-following and goal-detection mechanisms are shared across the hominid lineage, the conceptualization of internal epistemic states undergoes specialized, language-scaffolded maturation in humans.

11.2 Paradigms in Other Non-Human Primate Taxa

Comparative research across other non-human primates has mapped the evolutionary distribution of visual perspective-taking abilities. Studies conducted with bonobos (Pan paniscus), gorillas (Gorilla gorilla), and orangutans (Pongo pygmaeus) indicate that all great ape species perform similarly in gaze-following and visual-barrier tasks. Like chimpanzees, bonobos and orangutans can track a human handler’s line of sight around an opaque wall, adjusting their own physical position to inspect what the handler is viewing. However, they encounter the same difficulties when faced with cooperative pointing and subtle eye-masking conditions.

In contrast, Old World monkeys such as rhesus macaques (Macaca mulatta) and baboons (Papio species) fail cooperative Knower versus Guesser paradigms entirely. Yet, when tested in competitive food-theft paradigms, rhesus macaques show a clear sensitivity to human visual access. In experiments designed by Dario Flombaum and Laurie Santos, free-ranging macaques selectively stole food from a human experimenter whose gaze was turned away or whose head was occluded, rather than an experimenter looking directly at the food. This confirmed that sensitivity to visual perspective is not unique to great apes, but its expression is closely tied to competitive behavioral contexts across the primate order.

11.3 Convergent Evolution: Corvids and Domestic Canines

Research into visual perspective-taking extends beyond primates, revealing instances of convergent cognitive evolution in corvids and domestic canines. Corvids, particularly Western scrub-jays (Aphelocoma californica) and common ravens (Corvus corax), rely on food-caching strategies to survive. Studies by Nicola Clayton and Thomas Bugnyar demonstrate that scrub-jays and ravens actively adjust their caching behavior based on whether a competitor can see them. If a rival bird is watching, the caching bird will cache behind opaque barriers, wait for the observer to look away, or return later in private to re-cache the food in a new location.

Domestic dogs (Canis lupus familiaris) present a different comparative profile. Through thousands of years of selective breeding and domestication, dogs have evolved an unusual sensitivity to human cooperative-communicative gestures. Research by Brian Hare and Michael Tomasello shows that dogs outperform chimpanzees in utilizing human pointing, gaze direction, and subtle head turns to locate hidden food on trial one. However, this canine capacity appears to be a specialized communicative adaptation rather than a generalized representational Theory of Mind; dogs struggle in perspective-taking tasks that require abstracting another agent’s internal knowledge state in the absence of communicative cues.

12. Legacy and the Contemporary Scientific Consensus

12.1 The Competitive Paradigm Revolution (Hare, Call, Tomasello 2000s)

The early 2000s brought a resolution to the debate surrounding Premack and Woodruff’s original hypothesis through the “competitive paradigm revolution.” Brian Hare, Josep Call, and Michael Tomasello designed an experimental framework that aligned with the natural socio-ecology of the chimpanzee. Instead of relying on human informants pointing to hidden food, they arranged a competitive scenario involving two conspecific chimpanzees: a dominant individual and a subordinate individual situated on opposite sides of a central testing arena.

Two pieces of food were baited inside the arena: one positioned out in the open (visible to both apes), and the other positioned behind an opaque wooden barrier so that it was visible *only* to the subordinate. Subordinate chimpanzees preferentially retrieved the food hidden behind the barrier, avoiding the open reward where they faced physical retaliation from the dominant. When the opaque barrier was replaced with a transparent plastic screen—which physically occluded the dominant but left the food visibly exposed—the subordinates did not retrieve the food, confirming that they were calculating the dominant’s line of sight rather than relying on physical barriers as hiding shelters. These findings demonstrated that chimpanzees possess an understanding of what others can and cannot see when tested within an ecologically relevant, competitive context.

12.2 False Belief Understanding in Non-Human Apes: Recent Advances

For decades, the capacity to understand false beliefs—recognizing that another individual can hold a mental representation of the world that conflicts with reality—remained the definitive dividing line between human and non-human cognition. Non-human primates consistently failed explicit behavioral false-belief tests. However, in 2016, Christopher Krupenye, Fumihiro Kano, Satoshi Hirata, Josep Call, and Michael Tomasello published an investigation using non-invasive infrared eye-tracking technology, evaluating implicit false-belief understanding in chimpanzees, bonobos, and orangutans.

Using an anticipatory looking paradigm adapted from human infant research, apes watched videotapes of a human actor in a King Kong costume and an experimenter. The human observed King Kong hide an object in one of two boxes. When the human briefly left the room, King Kong relocated the object to the second box and then removed it entirely. When the human returned to retrieve the object, eye-tracking records revealed that the apes looked anticipatorily toward the box where the human *falsely believed* the object remained, rather than the box where it had been last seen. These findings suggest that great apes may possess an implicit understanding of false beliefs, challenging the traditional binary view of Theory of Mind and supporting a tiered evolutionary model of social cognition.

12.3 Enduring Contributions of Premack and Woodruff’s Work

The original 1978 paper by David Premack and Guy Woodruff fundamentally reshaped modern psychology, cognitive neuroscience, and animal behavior. By framing social interaction as an exercise in mental state attribution, they provided the conceptual tools that established Theory of Mind as a cornerstone of cognitive science. Their work led directly to the development of false-belief testing in human developmental psychology by Simon Baron-Cohen, Uta Frith, and Alan Leslie, which transformed clinical approaches to neurodevelopmental conditions such as autism spectrum conditions.

Within primatology, the Knower versus Guesser paradigm established an empirical framework for evaluating internal epistemic states without relying on human language. The methodological debates between mentalistic and behaviorist interpretations compelled comparative psychology to refine its experimental controls, eliminate Clever Hans artifacts, and account for ecological validity in experimental design. While the field has largely moved beyond the polarized debate over whether chimpanzees are full mind-readers or simple stimulus-response machines, Premack and Woodruff’s work provided the foundation for our current understanding of social cognition across the animal kingdom.

Conclusion

The intellectual journey inaugurated by David Premack and Guy Woodruff in 1978 represents one of the most ambitious empirical undertakings in the history of the cognitive sciences. By asking whether a chimpanzee could impute unobservable mental states to others, they challenged the dominant behaviorist paradigms of their time and fundamentally altered our understanding of animal cognition. The Knower versus Guesser paradigm, emerging as a response to the methodological ambiguities of Sarah’s early videotape tests, operationalized the principle of epistemic access, providing an experimental framework designed to determine whether non-human animals understand that visual perception leads to internal knowledge.

The decades of research that followed demonstrated that answering this question was far more challenging than initially anticipated. Daniel Povinelli’s critiques and replications revealed the persistent difficulty of separating mentalistic inference from sophisticated behavioral abstraction. These challenges exposed how easily human researchers can project their own cognitive mechanisms onto animal subjects. Yet, the subsequent competitive revolution led by Brian Hare, Josep Call, and Michael Tomasello demonstrated that when experimental paradigms align with an animal’s evolutionary ecology, non-human primates consistently exhibit cognitive capacities that extend far beyond simple associative conditioning. Chimpanzees actively track the sightlines of competitors, formulate deceptive tactics based on visual occlusion, and, as recent anticipatory eye-tracking paradigms indicate, implicitly track the false beliefs of others.

Ultimately, the legacy of the Knower versus Guesser experiment does not rest on a simplistic binary of whether non-human apes possess a human-equivalent Theory of Mind. Rather, its value lies in demonstrating that social intelligence is a multi-layered, evolutionary adaptation. Basic mechanisms of gaze-following, goal-comprehension, and visual perspective-taking evolved deep within our shared primate lineage, providing the cognitive foundation upon which human language and explicit metarepresentational mind-reading eventually developed. By compelling science to examine the internal worlds of our closest evolutionary relatives, Premack and Woodruff opened a window into the evolution of social awareness, forever changing how we conceptualize the divide between the human and animal mind.

References

★

Rate This Content

0.0 / 5 • 0 votes

Cite This Article

memjavad (2026, September 16). The Theory of Mind in Chimpanzees Experiment (Knower vs. Guesser) – David Premack and Guy Woodruff. PSYCHOLOGICAL DATABASE. https://en.arabpsychology.com/experiments/theory-of-mind-chimpanzees-experiment-premack-woodruff/
memjavad. “The Theory of Mind in Chimpanzees Experiment (Knower vs. Guesser) – David Premack and Guy Woodruff.” PSYCHOLOGICAL DATABASE, 16 September 2026, https://en.arabpsychology.com/experiments/theory-of-mind-chimpanzees-experiment-premack-woodruff/.
memjavad. “The Theory of Mind in Chimpanzees Experiment (Knower vs. Guesser) – David Premack and Guy Woodruff.” PSYCHOLOGICAL DATABASE. September 16, 2026. https://en.arabpsychology.com/experiments/theory-of-mind-chimpanzees-experiment-premack-woodruff/.