The investigation into animal cognition has long wrestled with the boundary between perceptual heuristics and genuine symbolic thought. For centuries, Western philosophical traditions asserted that mathematical reasoning was an exclusively human hallmark, inextricably linked to syntactically complex language and formal instruction. While non-human animals were acknowledged to possess acute sensory faculties capable of discriminating larger fruit clusters or avoiding superior numbers of predators, these behaviors were widely dismissed as sensory approximations driven by low-level perceptual cues rather than true representations of discrete quantity. The modern cognitive revolution, however, catalyzed an empirical reassessment of these assumptions, probing whether non-human minds could move beyond continuous magnitude estimation to represent discrete cardinality.
Among the empirical milestones that dismantled the anthropocentric monopoly on quantitative reasoning, few possess the historical weight and methodological rigor of the research program pioneered by Dr. Sarah T. Boysen at the Ohio State University. Working primarily with a captive female chimpanzee (*Pan troglodytes*) named Sheba, Boysen embarked on a multi-decade enterprise that interrogated the architectural foundations of numerical cognition. Rather than relying solely on operant conditioning paradigms that trained arbitrary stimulus-response associations, Boysen sought to investigate whether an anthropoid ape could acquire a functional, bidirectional symbolic lexicon for numbers, comprehend cardinal and ordinal principles, and deploy these internal representations to solve complex cognitive problems.
The culmination of this research program fundamentally reshaped theoretical models across comparative psychology, evolutionary anthropology, and cognitive neuroscience. Boysen demonstrated not only that chimpanzees could bind abstract Arabic numerals to discrete physical arrays, but that they could spontaneously perform mental arithmetic—specifically, multi-site summation—without explicit reinforcement for addition. Furthermore, her subsequent deployment of the reverse-reward contingency task exposed the delicate tension between visceral perceptual drives and symbolic executive control, revealing how external semiotic tools can liberate cognitive processing from affective interference. This comprehensive treatise examines the theoretical foundations, methodological innovations, empirical breakthroughs, and enduring neurobiological implications of Boysen’s work on chimpanzee numerosity.
1. Historical Context and Foundations of Comparative Numerical Cognition
1.1 Early Inquiries into Animal Mathematical Faculties
The scientific study of numerical competence in animals was deeply influenced by the cautionary tale of Clever Hans at the dawn of the twentieth century. Hans, an Orlov Trotter horse owned by German mathematics teacher Wilhelm von Osten, appeared capable of performing complex arithmetic, calculating fractions, and identifying calendar dates by tapping his hoof. In 1907, psychologist Oskar Pfungst conducted systematic, controlled experiments demonstrating that Hans was not calculating mathematical values; instead, the horse was responding to involuntary, micro-level postural shifts, respiratory alterations, and facial muscle tensions exhibited by his interrogator as the correct number of taps approached. The Pfungst investigation imposed a profound epistemological skepticism across animal psychology, establishing the “Clever Hans phenomenon” as the primary methodological pitfall against which all subsequent comparative cognitive research had to defend itself.
In the wake of this skepticism, German ethologist Otto Koehler initiated pioneering investigations during the 1930s and 1940s into what he termed the “number sense” (Zahlsinn) of non-human species, predominantly using ravens, jackdaws, and parrots. Employing rigorous experimental designs—including simultaneous versus successive item presentation, dynamic visual tracking, and carefully randomized food reward arrays—Koehler demonstrated that birds could match visual patterns based on the number of discrete entities present, independent of spatial geometry, overall area, or pattern configuration. Koehler meticulously distinguished between the capability to act on perceived continuous quantities and the capacity to discriminate “unnamed numbers” (benannte Zahlen), providing early empirical evidence that animals could perceive discrete quantities up to a threshold of approximately seven items without linguistic markers.
Despite Koehler’s early insights, the mid-twentieth-century dominance of radical behaviorism in American psychology effectively suppressed cognitive explanations of quantitative behavior. Under the strictures of operant conditioning models advanced by B.F. Skinner, numerical behaviors were routinely explained away as simple chains of discriminative stimulus-response reflexes reinforced by schedules of food delivery. The resurgence of cognitive ethology in the late twentieth century, propelled by researchers such as Donald Griffin, challenged this mechanistic orthodoxy by asserting that animals maintain internal, mental representations of their ecological environments. This theoretical reorientation catalyzed the critical distinction between proto-numerical estimation—such as evaluating aggregate biomass or visual density—and true symbolic counting, which mandates an invariant internal representation of discrete, countable units.
1.2 Primate Cognition Research Leading Up to the 1980s
By the late 1960s and 1970s, the international scientific landscape was galvanized by great ape language projects that sought to teach symbolic communication systems to chimpanzees, gorillas, and orangutans. Projects involving American Sign Language (ASL) with the chimpanzee Washoe, communicative plastic tokens with Sarah, and keyboard-based lexigrams with Lana and later the bonobo Kanzi, demonstrated that non-human primates possessed previously unrecognized capacities for referential communication, semantic association, and cross-modal concept manipulation. While these linguistic investigations generated substantial controversy regarding syntactic competence, they firmly established that pongid minds could utilize arbitrary physical markers to represent external entities, classes of objects, and qualitative relationships.
Concurrently, David Premack initiated targeted investigations into the abstract conceptual repertoires of chimpanzees, demonstrating that his subject, Sarah, could comprehend proportional relationships and solve proportional analogies involving physical fractions. Premack’s work revealed that chimpanzees could visually evaluate whether a half-filled glass of water bore an equivalence to half an apple, illustrating an underlying cognitive capacity to process abstract ratios and relational similarities without verbal instruction. However, Premack’s paradigms operated predominantly within the realm of continuous analogue quantities, leaving unresolved the question of whether primates could parse the world into discrete, enumerated cardinal sets represented by specific symbolic markers.
Across the Pacific at Kyoto University’s Primate Research Institute, Tetsuro Matsuzawa initiated the “Ai Project” in 1978, training a young female chimpanzee named Ai to recognize and produce Arabic numerals on an automated computer console. Matsuzawa demonstrated that Ai could systematically label sets of objects ranging from zero to nine with corresponding visual keys, exhibiting high accuracy across varied item types. Despite these impressive demonstrations, critics argued that early numeral acquisition studies suffered from potential methodological vulnerabilities. Many tasks utilized static arrays, predictable sequences, or failed to thoroughly decouple item count from perceptual variables such as cumulative surface area, sensory density, and visual duration. Consequently, there remained an urgent methodological imperative to design experimental paradigms capable of proving that an ape understood not just the discriminative association of a digit, but its true cardinal value independent of all lower-level sensory confounders.
1.3 The Epistemological Shift Toward Symbolic Quantification
The transition from documenting simple discriminative conditioning to proving true representational competence required a profound epistemological reorientation within comparative psychology. In a standard discriminative operant paradigm, a subject learns that pressing key “3” when viewing three triangles produces a reward, whereas pressing “2” or “4” does not. This associative paradigm does not guarantee that the animal possesses an internal representation of “three-ness.” The subject might rely on visual heuristics, including spatial dispersion, aggregate perimeter, or the rhythm of saccadic eye movements across the stimuli, entirely bypassing genuine enumeration. To confirm symbolic quantification, experimenters needed to construct tasks where the abstract symbol was detached from immediate visual percepts and utilized dynamically across varying functional contexts.
This methodological challenge highlighted the deep conceptual division between analogue magnitude estimation and discrete cardinality. The Approximate Number System (ANS), shared across human infants and diverse vertebrate lineages, allows organisms to estimate quantities based on continuous ratios governed by the Weber-Fechner law, where discriminatory accuracy is a function of the relative difference between two values rather than their absolute difference. Conversely, true symbolic counting relies on discrete representations: the distinction between 7 and 8 is mathematically identical to the distinction between 1 and 2—namely, a discrete increment of exactly one unit. To empirically validate whether non-human primates could transition across this boundary, research protocols had to isolate the precise mechanisms of ordinal tracking (the serial rank-order of entities) and cardinal correspondence (the absolute numerical magnitude of an enclosed set).
Developing empirical paradigms capable of demonstrating such competence demanded that researchers eliminate continuous perceptual covariations while simultaneously providing non-human subjects with a flexible, bidirectional semiotic system. A chimpanzee had to demonstrate the ability to construct a set of discrete items when presented with a solitary symbol, produce that symbol when encountering an unfamiliar set of varied objects, understand the ordinal position of that symbol within an invariant sequence, and mentally manipulate multiple sets through operational transformations like addition. This epistemological imperative set the stage for Dr. Sarah Boysen’s multi-decade research program.
2. Dr. Sarah Boysen and the Ohio State University Chimpanzee Project
2.1 Establishment of the Comparative Cognition Project
In 1983, Dr. Sarah T. Boysen founded the Comparative Cognition Project at the Ohio State University Department of Psychology, establishing a specialized research center designed to examine the computational and cognitive capacities of chimpanzees. Situated within an academic environment, the laboratory departed from traditional, sterile operant testing facilities by creating an enriched, socially integrated habitat where cognitive inquiry was fused with long-term behavioral welfare. Boysen recognized that the cognitive performance of non-human hominids is profoundly coupled to their affective states, environmental complexity, and social stability. Consequently, the research philosophy at Ohio State departed sharply from automated, impoverished testing boxes in favor of interactive, communicative, and socially supportive experimental paradigms.
The facility was engineered to ensure that cognitive testing sessions were completely voluntary. Chimpanzees were never food-deprived or subjected to aversive training contingencies. Participation in experimental trials was structured around interactive cognitive tasks rewarded with preferred food items such as fruit slices, pellets, juice, and social praise. If an animal chose not to participate, testing was deferred without consequence. This voluntary methodology eliminated the confounding behavioral artifacts induced by stress, captivity neuroses, or depressive states, thereby yielding experimental datasets that reflected the upper boundary of the chimpanzee’s intellectual competence. The core cohort of subjects included Sheba, Kermit, Darrell, and Sarah, each possessing unique behavioral temperaments and cognitive profiles that contributed to the longitudinal investigation of ape cognition.
The physical infrastructure of the Comparative Cognition Project incorporated a complex indoor-outdoor habitat interconnected with dedicated testing rooms. These testing arenas featured specialized experimental apparatuses, including transparent display panels, automated dispensing systems, movable partitions, and computational consoles. This dynamic testing environment enabled the researchers to execute complex spatial foraging paradigms and interactive social tasks that could not be accommodated within the confines of standardized primate cages. The laboratory became a premier center for the empirical exploration of animal minds, providing an ideal scientific crucible for testing the upper limits of non-human numerical cognition.
2.2 Profile of Sheba: The Primary Subject
Sheba, a female chimpanzee (*Pan troglodytes*) born in 1978, served as the primary subject and intellectual catalyst for the Ohio State numerical cognition project. Arriving at the university as an infant after being rejected by her biological mother, Sheba was reared in an enriched human-assisted environment that fostered a high degree of enculturation, communicative trust, and curiosity. Sheba exhibited an exceptional propensity for cross-modal abstraction, behavioral sustained attention, and associative problem-solving. While other chimpanzees in the colony, such as Kermit and Darrell, contributed substantially to various experimental phases, Sheba’s unique cognitive flexibility positioned her as the ideal candidate for tracking the longitudinal acquisition of abstract mathematical systems.
Prior to formal numerical training, Sheba underwent extensive baseline training designed to cultivate semiotic and cognitive readiness. She learned to interact with a broad spectrum of geometric shapes, colors, and objects, mastering complex visual discrimination tasks, match-to-sample paradigms, and associative classification protocols. She demonstrated an intuitive grasp of instrumental tool use and exhibited high social responsiveness toward both human experimenters and her conspecific peers. This cognitive baseline was crucial: it ensured that her subsequent engagement with numerical symbols was not merely an isolated behavioral trick, but rather integrated into a broader conceptual repertoire of symbolic reference and contextual comprehension.
Within the social hierarchy of the captive colony, Sheba occupied an active and socially complex status. Her cognitive testing was carefully balanced alongside daily social dynamics with Kermit, Darrell, and Bobby. These peer interactions proved essential, as later experimental paradigms required testing Sheba within explicit social contexts, such as the reverse-reward contingency task, where cognitive processing intersected with competitive foraging, sharing behaviors, and social impulse control. Sheba’s socio-cognitive maturity provided the empirical foundation for some of the most influential discoveries in the history of comparative psychology.
2.3 Methodological Philosophy and Experimental Rigor
Dr. Boysen maintained that the scientific validity of animal cognition research hinged entirely upon the absolute elimination of unintended human communication. Remembering the legacy of Clever Hans, Boysen integrated rigorous double-blind testing procedures across all phases of the project. During formal testing sessions, the primary experimenter presenting stimuli or evaluating responses was positioned behind physical baffles, opaque screens, or one-way mirrors, or wore specialized reflective sunglasses and neutral facial coverings to obscure micro-saccadic eye movements and subtle postural cues. In many automated tasks, human observers were entirely absent from the immediate testing chamber, with data logged by mechanical input consoles and cross-verified via remote multi-angle videotape systems.
Randomization strategies were applied with mathematical stringency. Spatial configurations of physical items, the positioning of numerical tokens on response trays, and the presentation sequence of target numbers were determined using pseudo-random generation algorithms. This prevented the subjects from using spatial heuristics, such as selecting an item based on its relative left-right placement, visual symmetry, or alternating sequences. Furthermore, stimulus identities were continuously varied: on any given trial, Sheba might be presented with edible items (such as orange segments or grapes) or non-edible objects (such as plastic cups, blocks, or metal hardware), ensuring that her evaluative mechanisms were engaged with the discrete count rather than species-specific consummatory desires.
Crucially, the experimental architecture accounted for the subtle physical attributes that naturally correlate with quantity. In naturally occurring settings, larger numbers of objects typically occupy greater physical volume, create larger retinal projections, weigh more, and generate higher aggregate perimeter contours than smaller sets. To isolate cardinal numerosity from these continuous physical variables, Boysen and her team systematically counterbalanced physical attributes. In controlled transfer tests, a set of three large oranges would possess a cumulative surface area and mass substantially greater than a set of four miniature oranges. If Sheba consistently selected the symbol corresponding to the discrete number of items rather than the overall volume or mass, her performance provided irrefutable empirical evidence of genuine numerical categorization.
3. Theoretical Framework: Numerosity, Counting, and Subitizing
3.1 Defining the Core Principles of True Counting
To differentiate true numerical counting from primitive associative conditioning, comparative cognitive science relies heavily on the theoretical architecture established by developmental psychologists Rochel Gelman and C.R. Gallistel in their seminal 1978 work, The Child’s Understanding of Number. Gelman and Gallistel articulated five fundamental principles that define genuine counting behavior in human ontogeny: the one-to-one correspondence principle, the stable-order principle, the cardinal principle, the abstraction principle, and the order-irrelevance principle. Boysen adopted this structural framework as an empirical diagnostic matrix to determine whether a chimpanzee’s quantitative manipulation met the formal criteria of arithmetic processing.
The one-to-one correspondence principle requires that each item within an enumerated set must be tagged with a single, unique numerical marker, with no item being tagged twice and no item omitted. In chimpanzee experiments, this principle was operationalized by requiring the subject to physically touch, point to, or manipulate each object in an array sequentially while simultaneously coordinating this physical action with a discrete internal or external sequence of symbols. The stable-order principle mandates that the markers used to tag items must be generated in an invariant, repeatable sequence across testing instances (for example, 1, 2, 3, 4, rather than 1, 3, 2, 4). Memorizing an arbitrary sequence of tokens does not fulfill this criterion unless that sequence is maintained consistently as the cardinal continuum scales upward.
The cardinal principle represents the critical cognitive leap: the understanding that the final numerical marker assigned to the last item in a counted array does not merely designate that individual item, but represents the total quantity of the entire set. Without the cardinal principle, an animal might execute a complex sequence of tags simply as a motor routine, without grasping that the terminal marker denotes the property of the entire group. Finally, the abstraction principle states that these counting procedures can be applied across any heterogeneous class of entities—concrete objects, abstract tones, imagined events, or disparate physical shapes—while the order-irrelevance principle establishes that the spatial or temporal order in which the discrete items are enumerated has zero impact on the resulting cardinal value. Boysen’s longitudinal protocols systematically verified Sheba’s adherence to every one of these foundational criteria.
3.2 Subitizing Versus Analogue Magnitude Estimation
The exploration of animal numerosity requires disentangling distinct neuro-cognitive mechanisms that process quantitative information: perceptual subitizing, the Approximate Number System, and symbolic counting. Subitizing, a term coined by E.L. Kaufman et al. in 1949, refers to the rapid, pre-attentive, and remarkably accurate apprehension of small quantities, typically constrained to sets between one and four items. In human subjects, subitizing is characterized by a flat response latency profile, wherein reaction times increase by negligible increments (approximately 40–100 milliseconds per item) within the subitizing range, before abruptly shifting to a steep, linear slope indicative of serial counting for arrays of five or more items.
For quantities exceeding the subitizing threshold, non-symbolic animals and human infants rely primarily on the Approximate Number System (ANS), a phylogenetically ancient cognitive module that yields imprecise, analogue representations of continuous magnitudes. The ANS operates under the mathematical constraints of the Weber-Fechner law: discriminatory precision depends entirely on the ratio between two quantities. For instance, distinguishing between 8 and 16 items (a 1:2 ratio) is cognitively effortless, whereas distinguishing between 15 and 16 items (a 15:16 ratio) yields high error rates and protracted decision latencies. The ANS generates continuous internal distributions with scalar variability—meaning that as the magnitude increases, the internal representation becomes broader and fuzzier.
The central theoretical question confronting Boysen was whether a chimpanzee could transcend the limitations of the ANS and the immediate boundary of subitizing through the acquisition of symbolic notations. When restricted to non-symbolic visual arrays, primates inevitably display scalar variability and ratio-dependent accuracy curves for larger numbers. However, if a subject acquires discrete symbolic labels (such as Arabic digits) that represent discrete values, the cognitive system should exhibit a structural break from the Weber-Fechner law, demonstrating stable, discrete precision regardless of absolute set magnitude. Boysen’s experimental task was to verify whether Sheba was simply using perceptual subitizing and analogue estimation, or whether she had bridged the gap into discrete, symbolic cardinality.
3.3 Symbolic Scaffolding and Semiotic Competence
The conceptual framework underpinning Boysen’s research was deeply informed by the semiotic theories of Charles Sanders Peirce and the developmental psychology of Lev Vygotsky. Peirce posited a tripartite semiotic taxonomy comprising icons (signs that physically resemble their referents), indices (signs that possess a direct, causal, or spatial connection to their referents), and symbols (signs whose relationship to their referents is entirely arbitrary, governed solely by conventionalized rules). Most non-human animal communication relies strictly on iconic or indexical signals. An Arabic numeral, such as the visual glyph “3”, possesses zero iconic resemblance to three round pieces of fruit, nor does it share any indexical or physical spatial connection to the objects. To master Arabic numerals, a chimpanzee must achieve true semiotic competence, mapping an arbitrary visual token onto an abstract mental concept of discrete quantity.
This semiotic transition embodies the Vygotskian principle of psychological tools and symbolic scaffolding. Vygotsky argued that the introduction of external, artificial sign systems fundamentally transforms internal cognitive architecture, mediating thought and liberating the mind from the immediate tyranny of the sensory environment. In human development, words and numerals act as cognitive scaffolding that reorganizes working memory, categorizes perceptual chaos, and enables higher-order executive control. In the context of comparative cognition, Boysen asked whether the provision of external, culturally derived human symbols could function as an analogous cognitive prostheses for a non-human ape.
If an anthropoid ape could internalize arbitrary symbols as representational tags, those symbols would act as cognitive compression mechanisms. Instead of expending working memory resources actively maintaining a transient, spatial representation of scattered physical items, the subject could collapse that perceptual complexity into a single, stable mental token. This symbolic transformation would not only facilitate higher storage capacity in working memory, but also permit the execution of internal mental operations—such as comparison, sequencing, and arithmetic transformation—entirely within an abstract semiotic workspace. The evolutionary implications of this research were profound: if an animal lacking the biological specialization for syntactic language could successfully utilize symbolic tools to reorganize its cognition, then the capacity for symbolic thought must have evolved prior to the phylogenetic divergence of the hominin lineage, long before the emergence of spoken language.
4. Experimental Methodology and Baseline Training Paradigms
4.1 Phase 1: Association of Physical Items with Quantity
The experimental trajectory initiated by Boysen commenced with an intensive, structured baseline phase focused on cultivating physical-item correspondence. Before introducing any abstract graphical symbols, Sheba was required to demonstrate that she could reliably discriminate and match physical sets of concrete entities based solely on discrete item count. Training sessions were conducted daily within a specialized testing enclosure equipped with sliding transparent panels and non-reflective stimulus trays. The target stimuli consisted of highly distinct, familiar physical items, ranging from edible reinforcers (such as orange wedges, whole apples, and monkey chow pellets) to completely inedible items (such as plastic toy figures, wooden cubes, and metal washers).
The initial protocol utilized a classical match-to-sample paradigm. An experimenter presented a sample tray containing a discrete set of items (for example, two oranges) behind a transparent partition. Sheba was presented with two or more comparison trays—one matching the sample array in discrete numerosity (two wooden cubes) and one or more foil arrays containing non-matching quantities (such as one or three objects). To succeed, Sheba was required to touch the matching comparison tray. Crucially, to prevent sensory confounders such as olfactory detection or consummatory distraction, non-edible duplicate stimuli were systematically interspersed throughout the trials, and the spatial positioning of the arrays was counterbalanced using randomized matrix schedules.
Criterion performance metrics were set to exceptionally rigorous statistical thresholds. Sheba was required to achieve a sustained accuracy rate of at least 80% to 90% across multiple consecutive blocks of trials over numerous testing days before the experimental protocol was permitted to progress. Errors were logged and classified, with particular attention paid to whether Sheba’s non-matching selections were distributed randomly or clustered tightly around adjacent numerosities. Once Sheba demonstrated that she could reliably match arrays containing between one and four physical entities based exclusively on discrete count, the researchers advanced to the critical phase of abstract symbol mapping.
4.2 Phase 2: Introduction of Arabic Numerals
The second phase of the empirical protocol initiated the formal mapping of abstract graphical markers—specifically, printed Arabic numerals—onto concrete physical arrays. The laboratory selected standard Western Arabic digits (0 through 4 initially) printed in bold, black sans-serif typeface on uniform, white plastic placards measuring approximately 10 by 10 centimeters. The training regimen avoided iconic cues; the numerals contained no geometric features (such as dots, lines, or hash marks) that could provide sub-symbolic visual clues regarding the quantity represented. The digits functioned as purely arbitrary, symbolic representations of discrete cardinality.
The training protocol was executed through an alternating bi-directional conditioning procedure. In the array-to-numeral condition, Sheba was presented with a physical set of items (e.g., three sliced oranges) and was provided with an array of numeral placards from which she was required to select the correct symbol (“3”). In the reciprocal numeral-to-array condition, Sheba was presented with a single numeral placard (e.g., “2”) and was required to select the corresponding physical tray that contained precisely two items, rejecting trays holding one, three, or four items. Correct responses were reinforced through the delivery of a preferred food reward coupled with enthusiastic vocal reinforcement from the experimenter, while incorrect selections were met with a neutral, brief time-out period accompanied by the immediate withdrawal of the stimulus trays.
As Sheba achieved criterion mastery over the initial set of numerals (0 through 4), Boysen incrementally expanded the numerical range, systematically introducing numerals 5, 6, 7, and eventually 8. Concurrently, the null set—represented by the digit “0”—was introduced into the operational lexicon. The introduction of zero represented a profound cognitive challenge, requiring the chimpanzee to conceptualize the complete absence of physical items not as a non-event or an un-testable condition, but as a discrete, nameable cardinal category. Longitudinal performance data revealed an intriguing error profile: rather than committing random errors across the entire numerical continuum, Sheba’s incorrect selections were overwhelmingly ordinal “near-misses”—selecting “4” instead of “5”—indicating that her internal representations of the symbols were mapped along an organized mental continuum rather than maintained as disparate, isolated associations.
4.3 Transfer Tests and Perceptual Controls
To confirm that Sheba’s performance reflected true adherence to the abstraction principle and was not an artifact of sensory confounders, Boysen instituted an exhaustive battery of transfer tests and perceptual controls. The central vulnerability of animal numerosity research is that physical items inevitably covary with non-numerical continuous dimensions, such as cumulative surface area, visual mass, spatial density, volume, and presentation geometry. If a chimpanzee consistently selects the numeral “4” over “2” simply because the array of four items covers a larger retinal area or projects more visual brightness, the behavior reflects sensory area discrimination rather than mathematical cognition.
To eliminate this possibility, the researchers devised rigorous area-controlled and mass-controlled stimulus sets. In these transfer sessions, the cumulative surface area and aggregate volume of the arrays were inverted or held completely constant across different quantities. For instance, an array of two exceptionally large grapefruit halves was matched against an array of four miniature kumquats or grapes; here, the physical mass and retinal footprint of the two-item array dramatically exceeded that of the four-item array. In other trials, objects of wildly heterogeneous identities, colors, and textures were presented within a single array—such as one plastic block, one orange, and one metal bolt. If Sheba relied on visual uniformity or perceptual density heuristics, her performance on these heterogeneous arrays would deteriorate to chance levels.
The empirical results of these transfer trials were unequivocal. Sheba performed with sustained, statistically significant accuracy on both homogeneous and heterogeneous arrays, regardless of whether the cumulative surface area, physical volume, or spatial configuration favored the smaller or larger set. She successfully applied her acquired numeral vocabulary to completely novel items that she had never previously encountered within the experimental apparatus, including laboratory instruments, gloves, and fabric swatches. These control trials established conclusively that Sheba’s selection behavior was driven strictly by the discrete cardinal numerosity of the items, verifying that her cognitive architecture had successfully decoupled discrete quantity from continuous perceptual dimensions.
5. Arabic Numeral Acquisition and Cardinal Value Attribution
5.1 Demonstration of Bi-Directional Symbolic Comprehension
A foundational benchmark in cognitive semiotics is the demonstration of true bidirectional equivalence. In many lower-animal conditioning studies, an organism can be conditioned to perform an A-to-B mapping (e.g., pecking a triangle when hearing a tone), yet fails completely to execute a B-to-A mapping (selecting the tone when presented with the triangle) without extensive retraining. True symbolic mastery requires that the connection between signifier and signified be fully reciprocal, symmetric, and cognitively accessible in both directions. Boysen’s protocols were meticulously engineered to establish this bi-directional equivalence between Arabic numerals and their corresponding cardinal physical sets.
Sheba demonstrated seamless operational facility across both testing paradigms. When presented with an arbitrary collection of physical objects—whether arranged in organized grids, random spatial clusters, or dropped sequentially into a container—she pointed directly to the corresponding Arabic numeral on a response console. Reciprocally, when presented with a solitary Arabic numeral, Sheba could construct or select an array consisting of that exact number of items. This bi-directional fluidity proved that the Arabic numerals did not function merely as passive discriminative stimuli that triggered an automatic motor response; rather, they served as authentic mental referents that summoned an internal, abstract representation of cardinality.
Mental chronometry data further illuminated the cognitive processes underlying Sheba’s responses. By analyzing response latency profiles across different array sizes, the researchers discovered that for small sets (1 to 3 items), Sheba’s response latencies were relatively brief and uniform, mirroring the human subitizing phenomenon. However, as the number of items extended from 4 to 8, her response latencies exhibited a linear, upward slope. This pattern revealed that Sheba was not relying on an immediate perceptual guess; instead, she was actively scanning, enumerating, and verifying the discrete elements of the set before selecting the corresponding symbolic placard, confirming true serial enumeration in an anthropoid ape.
5.2 The Concept of Zero in Sheba’s Cognitive Repertoire
The mathematical concept of zero represents one of the most intellectually sophisticated achievements in human history, emerging relatively late in human civilizational development compared to the positive natural numbers. Cognitively, the comprehension of zero mandates that an organism recognize the “null set”—the absolute absence of items—not merely as an empty sensory void or an absence of stimulation, but as a discrete, formal quantitative category positioned precisely adjacent to “one” on the ordinal continuum. In comparative cognitive science, demonstrating a true concept of zero requires showing that an animal spontaneously labels an empty space with a specific symbol and recognizes that symbol’s unique mathematical status.
Boysen systematically introduced the Arabic numeral “0” into Sheba’s testing environment by presenting experimental trays that contained no objects. Initial trials presented Sheba with familiar arrays containing items alongside an empty tray; to receive a reward, Sheba was required to match the empty tray with the numeral “0”. Remarkably, Sheba not only mastered this explicit association, but began to exhibit spontaneous, unprompted applications of the zero symbol. When an experimenter accidentally failed to place food items onto a presentation platform during a scheduled trial, Sheba walked to the response console and deliberately selected the “0” placard, accurately reporting the unexpected absence of stimuli.
This behavior mirrors the developmental trajectory observed in human early childhood, wherein children progress from understanding “none” as a linguistic negation to formalizing “zero” as an operational mathematical entity. Sheba understood that “0” was not an indicator of a failed trial or a meaningless token; it signified an explicit state of empty cardinality. Subsequent testing confirmed that Sheba successfully ordered the numeral “0” prior to the numeral “1” in sequential choice paradigms, establishing that her mental number line contained a fully integrated null set anchored at the origin of the positive integers.
5.3 Ordinality and Sequence Comprehension
Cardinality—the understanding of “how many”—is inextricably tied to ordinality—the understanding of “which position” in an invariant, linear progression. While cardinality deals with the aggregate magnitude of an enclosed set, ordinality concerns the formal, sequential relationships between distinct positions: understanding that 3 follows 2, 4 follows 3, and so forth. To evaluate whether Sheba’s acquired Arabic numerals possessed an underlying ordinal structure, Boysen designed sequential ordering tasks in which no physical sets of items were displayed; the trials presented numerals alone, requiring Sheba to make comparative judgments or arrange digits into ascending order.
In these tasks, Sheba was presented with pairs or triads of randomly selected Arabic numerals (e.g., “2” and “6”, or “3”, “5”, and “8”) and was reinforced for touching the numerals in their correct, ascending sequence. Sheba performed these ordinal discrimination tasks with exceptional accuracy, operating far above chance levels from the earliest trial blocks. Notably, her response data exhibited classic numerical distance effects, an empirical phenomenon first documented in human psychophysics by Robert Moyer and Thomas Landauer in 1967. Sheba was substantially faster and more accurate at discriminating between numerals that were far apart on the number line (such as 2 versus 7) than between numerals that were numerically adjacent (such as 5 versus 6).
The manifestation of the numerical distance effect in a non-human primate provides compelling empirical validation for the mental number line hypothesis. It reveals that numerals are mapped onto an internal, spatially organized, analogue mental continuum within the primate brain. When Sheba viewed the visual glyph “2” and the visual glyph “7”, her cognitive apparatus did not treat them as arbitrary geometric categories; it accessed internal analog representations situated at specific spatial-relational coordinates along an internal continuum. The closer two numbers were on this internal continuum, the greater the cognitive interference and the longer the processing latency required to resolve their relative rank, proving that Sheba possessed an authentic, continuous, and organized internal metric of numerical sequence.
6. The Discovery of Spontaneous Addition and Summation Abilities
6.1 Experimental Architecture of the Foraging Summation Task
Having established that Sheba possessed robust, bi-directional comprehension of cardinal numerals and their ordinal relationships, Boysen and her colleague Gary G. Berntson sought to test whether Sheba could use these symbolic representations to perform operations. The researchers designed a foraging paradigm that possessed high ecological validity, mimicking the natural fission-fusion foraging challenges faced by wild chimpanzees while requiring the integration of abstract symbolic markers. The results of this experiment were published in their landmark 1989 paper in the journal Science, entitled “Numerical competence in a chimpanzee (*Pan troglodytes*)”.
The experimental setup was distributed across a large, multi-compartment testing environment. Rather than sitting passively at a testing console, Sheba was permitted to actively explore three distinct, physically separated cache sites (designated as hiding locations A, B, and C) situated within the laboratory. In each trial, an experimenter hid a specific number of food items (such as whole fresh oranges) across the various cache sites while Sheba was positioned behind an opaque visual partition. Importantly, the food items were distributed in varied splits: one cache site might contain two oranges, another might contain one orange, and a third might contain zero oranges. Sheba had never received operational conditioning or reinforcement for the mathematical operation of addition; her prior training had focused on labeling isolated, static arrays.
The protocol dictated that Sheba was released to walk freely between the cache sites, visually inspecting the contents of each location sequentially. She was not permitted to eat the food items during the exploration phase; the oranges were secured inside transparent, locked containers or placed within clear visual receptacles. After visiting and inspecting the disparate cache sites, Sheba was required to return to a central response console located away from all the cache sites. The central console displayed an array of Arabic numerals from 0 to 4 (and subsequently higher). To receive the food as a reward, Sheba was required to select the solitary Arabic numeral that represented the total sum of the food items she had observed distributed across the physical environment.
6.2 Results: Emergence of Spontaneous Addition
The empirical findings from this foraging summation task revealed a remarkable cognitive breakthrough. On the very first block of testing trials, without any preliminary shaping or explicit associative conditioning for combining sets, Sheba spontaneously performed multi-site addition with a high degree of statistical significance. When she found two oranges at the first location and one orange at the second, she returned to the central console and selected the numeral “3”. When she found one orange at each of three cache sites, she selected “3”. When she found zero oranges at one site and two oranges at another, she reliably selected “2”. Sheba’s overall first-trial accuracy rate surpassed 85%, an outcome that conclusively ruled out trial-and-error learning.
Boysen and Berntson systematically varied the distribution of the food items across trials to ensure that Sheba was not relying on spatial heuristics or motor patterns. They introduced diverse numerical partitions for identical sums: for a target sum of 4, the trials varied across splits of 2+2, 1+3, 3+1, 0+4, and 1+1+2. Sheba’s accuracy remained consistently high across all additive combinations, demonstrating that her cognitive architecture was capable of performing an abstract arithmetic integration over time and space. The task imposed substantial demands on working memory: Sheba had to maintain the mental representation of the quantity observed at Cache A while traveling through the physical corridor to Cache B, update her internal tally to incorporate the newly discovered items, and retain that integrated total until she reached the central console to execute her choice.
These findings ignited an intense theoretical debate regarding the precise computational mechanisms driving Sheba’s performance. Two primary hypotheses emerged: the continuous accumulator model and the discrete mental counting model. Under the accumulator model, derived from the work of Russell Church and Warren Meck, each perceived item injects an increment of continuous energy or neural activity into a physiological accumulator, functioning much like an analogue capacitor. When Sheba reached the console, the final accumulated level of activation would be matched against the learned sensory-symbolic thresholds for specific numerals. Under the discrete counting model, Sheba utilized the Arabic numerals as internal tags, performing a covert mental count across the successive encounters. Regardless of which computational model is adopted, Sheba’s performance established that an anthropoid ape can spontaneously execute multi-site numerical summation without explicit training in arithmetic operations.
6.3 Symbolic-to-Symbolic Summation Trials
While the foraging summation task with physical oranges was a breakthrough, critics could argue that the addition was mediated by an analogue, perceptual accumulation of physical biomass: Sheba might have been mentally picturing the physical pile of oranges she would possess if she gathered them all together. To definitively counter this continuous-perceptual critique, Boysen and Berntson escalated the cognitive complexity of the paradigm by replacing the physical food items inside the cache sites with abstract Arabic numerals. In this symbolic-to-symbolic summation paradigm, no food items were visible anywhere in the foraging arena during the exploration phase.
When Sheba arrived at Cache Site A, she did not see physical fruit; she observed a placard bearing an Arabic numeral (such as “2”). When she proceeded to Cache Site B, she observed a second placard bearing an Arabic numeral (such as “1”). She was then required to return to the central testing console and select the numeral that represented the sum of the two observed symbols (in this case, “3”). If correct, she was rewarded with the corresponding number of edible treats. This task required an entirely different level of cognitive abstraction: Sheba could not rely on perceptual accumulation, visual mass, or sensory integration. She had to visually perceive an arbitrary symbol, access its internal cardinal value, hold that value in working memory while navigating to a second arbitrary symbol, decode the second symbol, perform a mental addition of the two abstract quantities, and select the appropriate third symbol to communicate the sum.
The results of the symbolic-to-symbolic summation trials confirmed Sheba’s remarkable arithmetic competence. From the initial sessions, Sheba succeeded on these purely symbolic summation trials at rates significantly exceeding chance, maintaining an overall accuracy rate between 70% and 80%. Her errors remained predominantly ordinal near-misses (e.g., selecting “4” when summing 2+3). This demonstration proved that Sheba possessed a functional arithmetic capacity that operated independently of perceptual continuous variables. The Arabic numerals did not merely signify external objects; they functioned as flexible internal operators within an abstract mental workspace, proving that symbolic mathematical calculation could occur in the absence of human syntactic speech.
7. The Reverse-Reward Contingency Task: Overcoming Perceptual Interference
7.1 The Dilemma of the Reverse-Reward Paradigm
In the mid-1990s, Dr. Boysen, along with colleagues Gary Berntson, Michelle Hannan, and John Cacioppo, utilized Sheba’s numerical competencies to formulate one of the most celebrated experimental paradigms in comparative executive function: the reverse-reward contingency task. Originally developed to investigate the limits of animal associative learning, the reverse-reward paradigm presents an animal with an apparently simple choice between two unequal quantities of food, but imposes a counter-intuitive, inverted reinforcement rule. If the subject points to the larger pile of food, that larger pile is immediately awarded to a conspecific peer (or taken away), and the subject is forced to receive the smaller, unselected pile. Conversely, if the subject points to the smaller pile, the larger pile is delivered to the subject, while the smaller pile is allocated to the partner.
To maximize its own reward, an animal must execute an inhibitory response: it must actively override the visceral, prepotent instinct to reach toward the larger reward, and deliberately point to the smaller array in order to obtain the larger payoff. Despite hundreds of repetitions, nearly every animal species tested—including rhesus macaques, squirrel monkeys, and chimpanzees—systematically and repeatedly fails this task when presented with real, physical food arrays. Even though the animals exhibit obvious signs of distress and frustration when they repeatedly receive the single treat instead of the larger pile, the overwhelming perceptual salience of the larger food treat triggers an automatic, impulsive motor response that completely overrides their rational cognitive evaluation of the contingency.
When Sheba and her conspecific partner, Sarah, were presented with this task using physical food items (e.g., a dish containing four miniature candy-coated chocolates versus a dish containing one), both chimpanzees failed catastrophically. Across hundreds of trials, despite possessing the intellectual capacity to easily discriminate between one and four items, Sheba pointed directly to the larger dish on nearly every single trial. Consequently, her partner received the four treats, leaving Sheba with the single treat. Sheba exhibited visible tantrums, screaming, hitting the console, and expressing acute distress at her outcome; yet, the moment the dishes were reset for the subsequent trial, the visual presence of the four treats exerted an irresistible perceptual pull, and she impulsively pointed directly to the larger dish once again. This striking disconnect between knowledge and behavioral execution exposed a deep vulnerability in the primate executive control system when confronted with biologically salient appetitive stimuli.
7.2 The Symbolic Solution: Introducing Arabic Numerals
Recognizing that the systematic failure on the reverse-reward task was driven by the overpowering perceptual salience and affective pull of physical food, Dr. Boysen devised an ingenious experimental modification. She removed the physical food arrays from the presentation trays and replaced them with the learned Arabic numeral placards. Instead of choosing between a dish of four chocolates and a dish of one chocolate, Sheba was presented with a choice between a placard bearing the numeral “4” and a placard bearing the numeral “1”. The underlying operational rule of the task remained completely unaltered: pointing to the “4” placard resulted in the partner receiving four treats while Sheba received one, whereas pointing to the “1” placard resulted in Sheba receiving the four treats while the partner received one.
The experimental outcome of this symbolic substitution was immediate, dramatic, and profound. The moment the physical food treats were replaced with Arabic numerals, Sheba’s performance shifted instantaneously. Without requiring any additional training trials, Sheba immediately selected the smaller numeral placard (“1”) on her very first trial, thereby securing the larger food payout of four treats for herself. Across subsequent blocks of trials utilizing numerals, Sheba maintained an accuracy rate approaching 90%, flawlessly executing the optimal counter-intuitive strategy across varying pairs of numbers, such as 2 versus 4, 1 versus 3, and 0 versus 3.
Statistical analysis confirmed that this transformation was not a gradual associative reconditioning process, but an immediate cognitive phase shift. To demonstrate that this effect was tied specifically to the abstract nature of the symbols, Boysen subsequently conducted reversal probes, re-introducing the physical food arrays into the presentation trays. The moment real food reappeared, Sheba’s executive control disintegrated: she immediately reverted to pointing impulsively at the larger physical food pile, once again losing the reward to her partner. When the numerals were reinstated, her executive competence instantly returned. The presence of abstract symbols acted as a semiotic buffer, creating psychological distance between the animal’s cognitive processing and her visceral, appetitive drives.
7.3 Cognitive Implications: Inhibition and Executive Control
The findings from the reverse-reward contingency task provided monumental theoretical insights into the evolution of primate executive control, impulse regulation, and the functional role of symbolic mediation. In cognitive neuroscience, executive function comprises the capacity to maintain goal-directed behavioral representations in working memory while actively suppressing competing, prepotent automatic responses. This regulatory network is localized primarily within the prefrontal cortex (PFC), specifically the dorsolateral and ventromedial prefrontal architectures. Boysen’s experiment demonstrated that in non-human primates, as in young children, the prefrontal inhibitory control mechanisms can be easily overwhelmed by sensory and affective inputs originating from evolutionary ancient limbic and ventral striatal circuits.
When physical food is visually present, the immediate, bottom-up sensory salience activates evolutionary subcortical reward mechanisms, triggering an involuntary motor reach toward the appetitive stimulus before top-down prefrontal control can intervene. However, when the physical stimuli are replaced by arbitrary Arabic numerals, the informational input must be routed through higher-order cortical regions for semantic decoding. Because the numeral glyph “1” or “4” does not trigger immediate gustatory or visceral reflexes, the subject experiences a critical temporal and affective delay. This symbolic distance attenuates the immediate impulsive drive, allowing top-down executive control from the prefrontal cortex to successfully modulate the motor output and execute the strategically optimal, counter-intuitive choice.
These empirical findings directly paralleled groundbreaking research in human developmental psychology, most notably Walter Mischel’s iconic “marshmallow test” and delay-of-gratification paradigms. Young children below the age of four systematically fail delay-of-gratification tasks when the physical rewards (marshmallows or pretzels) are left in plain sight; the visual salience provokes impulsive consumption. However, if the rewards are obscured, replaced with abstract representations, or if the children are instructed to mentally reframe the treats as non-edible pictures (e.g., “imagine it is a white, puffy cloud”), their capacity for inhibitory self-regulation increases exponentially. Boysen’s work revealed that the capacity for symbolic representations to act as internal governors of impulsive biological drives is a phylogenetically deep hominoid trait, highlighting the evolutionary roots of self-regulation and executive control.
8. Cognitive Architecture: Symbolic Representation vs. Perceptual Salience
8.1 Dual-System Processing in Primate Cognition
The empirical divergence documented in Sheba’s performance between physical arrays and abstract numerals provides strong cross-species support for dual-process cognitive theories, most famously articulated in modern psychology by Daniel Kahneman and Amos Tversky. Dual-process frameworks posit that cognitive architecture is bifurcated into two distinct modes of information processing: System 1, which is evolutionarily ancient, fast, automatic, heuristic-driven, and highly susceptible to affective and visceral biases; and System 2, which is slower, deliberative, rule-governed, cognitively effortful, and capable of symbolic manipulation and abstract logical reasoning.
In the context of Boysen’s chimpanzee experiments, the presentation of concrete, edible arrays of fruit activates the primate’s System 1 processing architecture. The sight of food automatically recruits evolutionary foraging algorithms that prioritize immediate acquisition and physical capture of the largest detectable resource parcel. In natural ecological environments, pausing to contemplate an inverse reward logic when encountering a food source would lead to immediate starvation or competitive loss to a conspecific. Therefore, System 1 operates through direct perception-action loops that circumvent deliberative reasoning. The physical presence of the food stimulus locks the animal’s cognitive processing into an impulsive, bottom-up sensory feedback loop.
Conversely, when the experimental task presents Arabic numerals, the cognitive processing cannot be completed by System 1’s automatic perceptual heuristics. Because the symbol “4” possesses no intrinsic nutritional or sensory value, it fails to trigger an immediate, subcortical appetitive response. Instead, the input recruits System 2-like processing mechanisms, demanding the retrieval of learned symbolic mappings, the activation of ordinal rank-order representations, and the strategic calculation of rule-governed outcomes. The introduction of symbolic notations effectively shifts the animal’s neural processing from an affective, perception-driven state to a reflective, rule-governed cognitive state, demonstrating that dual-system cognitive architecture does not depend on human linguistic syntax.
8.2 The Semi-permeable Boundary Between Mind and Symbol
The cognitive struggles documented in the reverse-reward task also shed significant light on what developmental psychologist Judy DeLoache termed the dual-representation hypothesis. DeLoache’s research demonstrated that human toddlers struggle to utilize symbolic physical objects—such as a miniature scale model of a room—as informational tools to locate a hidden object in a real room. The child struggles because an object possesses a dual identity: it is simultaneously a concrete, physically attractive thing in itself (a fun miniature toy) and an abstract symbol representing something else. Until children reach a specific developmental threshold (typically around three years of age), the concrete physical salience of the scale model obscures its symbolic, representational function.
Boysen’s work reveals that chimpanzees face an identical cognitive threshold, characterized by a semi-permeable boundary between the physical object and its symbolic representation. When an object is edible and possesses high biological value, its concrete existence completely overwhelms its symbolic capacity. Sheba understood the rule of the reverse-reward contingency at a deep conceptual level, as evidenced by her immediate, flawless execution when numerals were utilized; however, that knowledge was rendered entirely inaccessible in the presence of real food because the physical object could not be cognitively treated as a symbolic token. The food was so intensely an “object in itself” that its abstract informational value was completely obliterated.
By providing external Arabic numerals, Boysen provided Sheba with a purely semiotic medium that was immune to this dual-representation trap. The Arabic numeral was completely devoid of concrete, consummatory value: it could not be eaten, smelled, or played with. Because it was empty of physical appetitive salience, its entire utility resided in its representational function. Boysen’s findings thus provided profound empirical proof that external symbolic media liberate the mind from perceptual entrainment, creating an external semiotic architecture that allows non-human primates to manipulate information about reality without being overwhelmed by reality itself.
8.3 Working Memory Load and Representational Capacities
The multi-stage numerical operations executed by Sheba—including sequential multi-site summation and symbolic comparison—provide valuable insights into the constraints and structural architecture of primate working memory. In cognitive psychology, working memory is conceptualized not merely as a passive storage buffer, but as an active computational workspace responsible for maintaining, updating, and manipulating mental representations in the face of temporal delays, interference, and spatial transitions. In human cognition, working memory capacity is significantly augmented by internal verbal rehearsal (the phonological loop) and the ability to combine discrete units of information into larger, structured representations—a process known as cognitive chunking.
Because non-human hominids lack an internal phonological loop driven by linguistic speech, their working memory must rely on visual-spatial, motoric, and abstract representational mechanisms. During the foraging summation tasks, Sheba demonstrated an impressive capacity to maintain multiple distinct numerical states in active working memory. She had to hold the value of Cache Site A, retain that representation across a physical walking traversal of several meters, update that representation upon encountering Cache Site B, and preserve the final sum while traversing back to the central console. This performance indicates that chimpanzees possess an active, stable executive buffer capable of dynamic updating and maintenance without continuous perceptual support.
Crucially, the Arabic numerals acquired by Sheba functioned as cognitive chunking devices. In classic human cognitive experiments, George A. Miller demonstrated that the capacity of working memory is constrained by the number of discrete “chunks” rather than the absolute quantity of raw data. By binding a complex, multi-item visual array (such as four scattered oranges) into a single, cohesive mental token—the symbol “4”—Sheba was able to compress perceptual data into a single chunk. This symbolic compression drastically reduced the computational load imposed on her working memory, freeing cognitive resources that could then be allocated to spatial navigation, behavioral planning, and arithmetic summation.
9. Neurobiological and Evolutionary Underpinnings of Primate Numerical Competence
9.1 Neural Substrates of Numerical Processing
The sophisticated numerical and symbolic competencies demonstrated by Sheba must be grounded in underlying neuroanatomical structures and electrophysiological networks shared across the hominid lineage. Contemporary cognitive neuroscience, utilizing functional magnetic resonance imaging (fMRI) in humans alongside awake-behaving single-unit electrophysiology in non-human primates, has mapped the core neural substrates of numerical processing to a specialized cortical circuit centered within the intraparietal sulcus (IPS) and the lateral prefrontal cortex (lPFC).
Groundbreaking single-unit recordings conducted by neurobiologists Andreas Nieder and Earl K. Miller in macaques have identified specialized populations of “number neurons” within the fundus of the intraparietal sulcus and adjacent prefrontal regions. These neurons exhibit distinct tuning curves: an individual number neuron discharges maximally in response to a specific preferred numerosity (e.g., three items), with its firing rate decaying systematically as the observed quantity deviates from that preferred value. Crucially, Nieder’s work revealed that while parietal number neurons process raw visual and analogue magnitudes, neurons within the prefrontal cortex are intimately involved in binding arbitrary visual symbols (such as geometric glyphs or Arabic digits) to those parietal magnitude representations, providing an electrophysiological blueprint for the symbolic mapping mastered by Sheba.
The structural homologies between the chimpanzee brain and the human brain provide the anatomical foundation for these cognitive parallels. The chimpanzee neocortex possesses an expanded parietal-frontal computational axis, characterized by well-developed inferior parietal lobules and prominent prefrontal cortices equipped with dense von Economo neurons and extensive reciprocal connectivity. Stanislas Dehaene’s Triple-Code Model of numerical cognition posits that the human brain processes numbers through three interconnected representational formats: an analogue magnitude code situated bilaterally within the intraparietal sulcus, a visual-spatial symbolic code located within the inferior temporal-occipital fusiform areas, and a verbal linguistic code localized in perisylvian language structures. Boysen’s research demonstrates that chimpanzees fully share the first two codes of this neuro-cognitive network; they possess the analogue magnitude and visual-symbolic systems, functioning effectively even in the complete absence of the third, specialized verbal-linguistic code.
9.2 Evolutionary Pressures Driving Numerical Competence
The emergence of quantitative cognition in the primate order was not an evolutionary accident; it evolved in response to intense, multifaceted ecological and social selection pressures. For wild chimpanzees (*Pan troglodytes*), survival within tropical rainforests and savanna-woodlands demands complex spatial foraging strategies. Chimpanzees subsist largely on patchy, ephemeral resources such as ripe fruit groves that vary dramatically across seasons and locations. To optimize their daily caloric budget, wild primates must continuously execute cost-benefit analyses, evaluating food patch yields, estimating depletion rates, and calculating transit distances. An individual that can accurately estimate and compare the relative abundance of disparate foraging patches possesses a decisive selective advantage over a competitor relying on crude trial-and-error.
Beyond ecological foraging, socio-ecological pressures exerted an equally powerful evolutionary push toward numerical competence. Wild chimpanzees inhabit complex fission-fusion societies where social parties continuously fracture and coalesce throughout the day. Inter-community interactions are characterized by lethal territorial aggression and organized border patrols, as documented by Jane Goodall at Gombe and Richard Wrangham at Kibale. During territorial excursions, chimpanzees rely on what primatologists term “numerical assessment”: males evaluate the number of calling individuals in an opposing party against their own coalition size before deciding whether to attack or retreat. Field playback experiments conducted by Michael Wilson and colleagues have shown that male chimpanzees will only advance toward an intruder’s vocalization if their own group outnumbers the rival group by a ratio of at least three to one, demonstrating that numerical evaluation is a vital social survival calculation.
These evolutionary realities provide the adaptive backdrop for Boysen’s findings. The primate brain was already pre-adapted by millions of years of selection to track discrete entities, evaluate coalitional balances, and estimate resource parcels. Dr. Boysen’s laboratory did not install a foreign, artificial cognitive apparatus into Sheba’s mind; rather, she provided an artificial semiotic interface (Arabic numerals) that tapped into, unified, and organized pre-existing evolutionary cognitive adaptations. The latent capacity to mentally combine and compare quantities had long been an evolutionary necessity; Boysen simply gave that cognitive engine a new, symbolic transmission.
9.3 Phylogenetic Continuities and Divergences
Mapping the precise boundaries of numerical cognition across the primate phylogenetic tree illuminates both deep evolutionary continuities and profound evolutionary divergences. The research executed by Boysen demonstrates that the core architectural foundations of mathematical thought—the perception of discrete cardinality, ordinal transitivity, the concept of the null set, and basic arithmetic summation—predate the evolutionary split between the genus Pan and the genus Homo, which occurred approximately six to eight million years ago. These quantitative faculties are not recent cultural inventions, but shared ancestral primate homologies.
Yet, acknowledging these continuities makes the evolutionary divergences between modern human mathematics and non-human primate numerical competence all the more theoretically intriguing. While Sheba demonstrated impressive proficiency with Arabic numerals ranging from 0 to 8, expanding an ape’s numerical lexicon requires extensive, step-by-step training over months or years. A human child, once they master the conceptual logic of the counting routine and the recursive structure of base-10 counting, experiences an intellectual explosion: they realize that every number has an infinite successor ($n+1$), allowing them to count theoretically to infinity without needing individual, discriminative conditioning for every new integer.
This profound evolutionary divergence is rooted in what linguist Noam Chomsky and evolutionary biologists Marc Hauser and W. Tecumseh Fitch identify as discrete infinity and syntactic recursion. Non-human primates lack the biological specialization for open-ended, recursive syntax that allows humans to effortlessly generate infinite, hierarchical cognitive structures from finite sets of elements. While Sheba mastered numbers as discrete symbolic tokens and could combine them operational through addition, she lacked the recursive linguistic machinery required to formulate a self-generating, boundless mathematical continuum. Boysen’s work cleanly demarcates the true evolutionary boundary: great apes possess the fundamental conceptual machinery of number, cardinality, and arithmetic, but the cultural invention of recursive syntax is what enabled the human mind to construct modern mathematics.
10. Comparative Analysis: Chimpanzees versus Other Non-Human Animal Models
10.1 Chimpanzees (Sheba, Ai) versus Rhesus Macaques
To fully contextualize Sheba’s cognitive accomplishments, her performance must be compared against other major non-human primate research programs, specifically Tetsuro Matsuzawa’s work with the chimpanzee Ai at Kyoto University and Elizabeth Brannon and Herbert Terrace’s seminal investigations with rhesus macaques (*Macaca mulatta*) at Columbia University. While Boysen focused on socially embedded, voluntary paradigms and spontaneous arithmetic operations, Matsuzawa’s “Ai Project” utilized an automated, computer-driven training regimen. Ai mastered the numerals 0 through 9, matching them to multi-dot arrays on touch-sensitive monitors with extraordinary speed and accuracy.
Matsuzawa’s longitudinal studies expanded into examinations of working memory and ordinal sequencing with Ai and her son, Ayumu. In the iconic “masked sequence task,” numerals 1 through 9 were displayed briefly on a monitor and then immediately covered with blank white squares the moment the chimpanzee touched the numeral 1. Ayumu demonstrated superhuman spatial working memory, touching the remaining white squares in their correct ascending numerical sequence (2 through 9) with millisecond-level precision, easily outperforming adult human subjects. However, while Matsuzawa’s subjects exhibited astonishing visual-spatial working memory for numerical sequence, Ai was not documented spontaneously performing multi-site foraging summation or utilizing symbols to solve executive-inhibitory dilemmas like the reverse-reward task.
In contrast, research with rhesus macaques conducted by Brannon and Terrace operated strictly within non-symbolic paradigms. In their 1998 and 2000 studies, macaques were presented with visual arrays containing one to four items of varying shapes and sizes, and were reinforced for touching the arrays in ascending ordinal sequence (1 → 2 → 3 → 4). When subsequently presented with completely novel quantities from 5 to 9, the macaques spontaneously transferred the ordinal rule, touching the unfamiliar arrays in ascending order without additional training. This demonstrated that Old World monkeys possess an endogenous ordinal representation of numerosity. However, macaques have not demonstrated the capacity to acquire a bi-directional, arbitrary symbolic lexicon for numbers, nor have they exhibited spontaneous symbolic arithmetic. The divergence between macaques and great apes confirms that while ordinal magnitude estimation is broadly conserved across the primate order, true symbolic semiotic competence and spontaneous mental calculation are cognitive specializations that reach full expression only within the hominoid lineage.
10.2 Avian Numerosity: The Case of Alex the African Grey Parrot
Outside of the mammalian order, the most compelling comparative evidence for sophisticated numerical cognition emerged from the late Dr. Irene Pepperberg’s remarkable longitudinal work with Alex, an African Grey parrot (*Psittacus erithacus*). Working within the “Model/Rival” social-interactive training paradigm, Pepperberg taught Alex to produce recognizable vocalizations of English speech labels to identify objects, shapes, colors, and discrete numerical quantities up to the value of six. Alex’s abilities provided a stunning example of convergent cognitive evolution, demonstrating that complex semiotic and numerical processing could emerge within an avian brain lacking a layered cerebral cortex.
Alex demonstrated genuine cardinal value attribution. When presented with a visually complex, heterogeneous tray containing blue wool balls, red wooden blocks, and green keys, and asked vocal questions such as “How many blue key?”, Alex could visually parse the array, filter out the distractors, and reliably vocalize the correct English number word. Like Sheba, Alex showed clear evidence of the abstraction principle, accurately enumerating completely novel items. Furthermore, Alex spontaneously developed a vocal label for the null set: when asked to identify the color difference between two identical objects, Alex began responding with the word “none,” which Pepperberg subsequently demonstrated he could generalize to represent the absence of a numerical quantity—a conceptual analogue to Sheba’s understanding of zero.
However, an important structural difference distinguishes Alex’s performance from Sheba’s experimental milestones. Alex’s production of numerical labels was intrinsically tied to acoustic speech routines within an interactive conversational context. While Alex could perform basic addition of small sets hidden under cups, his training focused predominantly on vocal labeling of immediate perceptual displays. Boysen’s work with Sheba explored deeper programmatic territories of executive function, specifically demonstrating how the acquisition of physical symbols could restructure executive control, bypass affective interference in reverse-reward tasks, and facilitate spatial foraging calculations. Nonetheless, the cognitive parallels between an African Grey parrot and a chimpanzee demonstrate that complex numerical competence is an evolutionary solution that can be realized through disparate neuroarchitectural designs.
10.3 Canine and Marine Mammal Quantification Paradigms
The exploration of numerical abilities across other mammalian lineages—most notably cetaceans, pinnipeds, and domestic canines—further clarifies the cognitive baseline of non-primate mammals. In marine mammals, Louis Herman and colleagues working with bottlenose dolphins (*Tursiops truncatus*), and Ronald Schusterman working with California sea lions (*Zalophus californianus*), established that these species possess acute analogue magnitude discrimination. Dolphins, utilizing both vision and high-resolution echolocation, can rapidly discriminate between visual or acoustic arrays based on quantity, exhibiting standard Weber-Fechner ratio dependencies.
Similarly, investigations into domestic dogs (*Canis familiaris*) using behavioral looking-time paradigms and non-invasive fMRI have demonstrated that the canine brain contains parieto-temporal circuits capable of primitive, non-symbolic magnitude estimation. For instance, when presented with varying quantities of food treats behind occluders, dogs display prolonged gaze durations when an experimenter creates a mathematically “impossible” outcome (e.g., 1+1=1 or 1+1=3), mirroring the foundational violation-of-expectation experiments conducted by Karen Wynn on human infants. This confirms that canines possess a rudimentary, non-verbal expectation of basic arithmetic transitions.
Yet, despite these impressive foundational capacities, direct cross-species comparisons highlight a sharp cognitive ceiling in non-primate mammals. Neither marine mammals nor domestic canids have ever demonstrated the capacity to acquire a comprehensive, bi-directional symbolic lexicon for numbers. They do not spontaneously combine separate abstract tokens to execute symbolic summation, nor can they use arbitrary graphic markers to override visceral appetitive drives in reverse-reward tasks. While magnitude estimation and perceptual quantity tracking are broadly distributed across vertebrates, the capacity to bind those magnitudes to arbitrary, external symbols and manipulate those symbols in an abstract mental workspace remains a specialized achievement restricted to great apes, humans, and select avian subjects.
11. Methodological Critiques, Replications, and Empirical Debates
11.1 The Clearness of Controls: Addressing Potential Cueing Artifacts
Given the long shadow cast by Clever Hans over comparative psychology, Dr. Boysen’s published findings were inevitably subjected to rigorous scrutiny by critics seeking alternative, non-cognitive explanations for her subjects’ remarkable performance. The primary critique centered on the possibility of involuntary, subtle experimenter cueing. Skeptics questioned whether human handlers, unconsciously anticipating the correct numeral selection or responding to Sheba’s exploratory hand movements, might have provided subtle postural, facial, or auditory cues that guided her choices toward the correct placards during summation and reverse-reward trials.
To definitively address these concerns, Boysen instituted an escalating series of experimental controls that completely insulated her subjects from human behavioral influence. In subsequent replication and control sessions, double-blind methodologies were strictly enforced. In these trials, the human experimenter standing near the chimpanzee wore opaque reflective welding goggles or darkened sensory shields that rendered them entirely blind to the contents of the stimulus trays and the positioning of the response numerals. In other conditions, the experimenter who arranged the stimuli exited the testing room entirely, leaving the chimpanzee to interact with a secondary experimenter who had no knowledge of which numbers or quantities had been presented.
Furthermore, automated data logging systems and blind video coding were deployed to verify trial outcomes. Independent observers, completely naive to the hypotheses and experimental conditions, coded the videotapes with the audio muted and the stimulus trays cropped from the visual frame, scoring only Sheba’s terminal choices. Inter-rater reliability scores consistently exceeded 0.95. The empirical data collected under these completely blinded and automated controls showed zero statistically significant degradation in Sheba’s accuracy rates: she continued to identify cardinal values, perform multi-site additions, and solve the reverse-reward task with identical precision, providing conclusive proof that her performance was completely independent of involuntary human cueing.
11.2 The Enculturation Hypothesis and Generalizability
A more philosophically complex methodological debate emerged surrounding the enculturation hypothesis, advanced most prominently by developmental and comparative psychologists Michael Tomasello and Josep Call. The enculturation hypothesis posits that chimpanzees reared in rich, human-assisted socio-cultural environments—immersed from infancy in linguistic, communicative, and tool-using interactions—do not represent the species-typical cognitive baseline of wild *Pan troglodytes*. Instead, critics argue, these “enculturated” apes develop an artificial, human-scaffolded cognitive phenotype. Through continuous social interactions with humans, their neural circuitry undergoes unique developmental remodeling, effectively granting them cognitive capacities that are completely absent in wild, mother-reared conspecifics.
From this critical perspective, Sheba’s numerical accomplishments cannot be generalized as representative of “chimpanzee psychology” in a broad, naturalistic sense. Instead, Sheba must be viewed as an extraordinary, culturally transformed subject whose biological cognitive potential was artificially unlocked by human semiotic artifacts. Proponents of this view point out that wild chimpanzees, despite their sophisticated tool use and cooperative hunting, have never evolved an indigenous symbolic counting system, an invariant tally stick, or written graphical notations in the wild.
In response to the enculturation critique, Boysen and other evolutionary cognitive researchers argued that enculturation does not create new biological brain structures; it merely provides the environmental input necessary to express latent cognitive potentials already present within the hominid genome. Just as an illiterate human child raised in an environment devoid of formal mathematics would fail to master arithmetic, a chimpanzee raised in an impoverished wild environment cannot express symbolic abilities. The fact that an ape brain can acquire, integrate, and operationalize a symbolic number system without evolutionary genetic alteration proves that the computational architecture required for symbolic mathematical thought was already evolutionarily stable prior to the hominin-pongid split, awaiting only the cultural catalyst of symbolic scaffolding.
11.3 Replication Efforts and Discrepant Findings
As with all landmark scientific breakthroughs, the validity of Boysen’s findings stimulated replication efforts across international comparative laboratories, yielding a nuanced landscape of empirical replications and discrepant findings. The reverse-reward contingency task, in particular, became the subject of extensive experimental replication. Researchers such as Alan Silberberg and David Kearns attempted to replicate the reverse-reward findings with other primate subjects, occasionally reporting that their animals struggled to master the task even when abstract symbols or colored shapes were introduced.
Similarly, comparative studies conducted with rhesus macaques, baboons, and capuchin monkeys often failed to observe the immediate, dramatic release from perceptual interference that Boysen documented when replacing food with numerals. Even within chimpanzee laboratories, researchers such as Murray, Kralik, and Wise documented substantial individual variability: some chimpanzees required dozens of trials to transfer the inhibitory rule to symbols, while others continued to exhibit impulsive biases toward larger quantities even when non-food visual stimuli were utilized. These discrepant findings generated intense scientific dialogue regarding the experimental boundary conditions of symbolic executive control.
Upon systematic meta-analytic review, these empirical discrepancies were reconciled by identifying critical methodological differences between Boysen’s protocols and those of subsequent replication attempts. Many failed replications utilized subjects that lacked Sheba’s extensive, foundational, bi-directional symbolic training: they introduced arbitrary colored tokens or unfamiliar symbols that lacked true cardinal meaning for the animals, treating the task as a simple conditional discrimination rather than a deeply internalized symbolic operation. Furthermore, variations in housing, testing stress, rearing history, and relationship with experimenters profoundly affected the animals’ inhibitory performance. When replication protocols faithfully reproduced the rigorous symbolic baseline training and low-stress voluntary testing conditions established at Ohio State, the core empirical phenomenon held true: abstract symbolic representation consistently provides a cognitive buffer that enhances executive control and overcomes visceral perceptual interference.
12. Enduring Legacy and Implications for Human Cognitive Development
12.1 Impact on Theories of Human Numerical Development
The theoretical reverberations of Sarah Boysen’s research program extended far beyond comparative primatology, fundamentally transforming models of numerical acquisition and developmental cognitive psychology in human children. For decades, developmental theories—dominated by the classic stage-based models of Jean Piaget—asserted that true numerical competence emerged late in human childhood, dependent upon the formal acquisition of conservation of volume and linguistic logical operations. Piaget maintained that young children could not genuinely understand cardinality until they were approximately six or seven years of age.
Boysen’s demonstration that an anthropoid ape could master cardinality, sequence, and addition without linguistic speech provided vital empirical support for alternative developmental models advanced by researchers such as Karen Wynn, Susan Carey, and Rochel Gelman. In 1992, Karen Wynn published her landmark infant cognition study in Nature, showing that five-month-old human infants display looking-time expectations consistent with basic arithmetic calculations (1+1=2 and 2-1=1). Boysen’s cross-species data provided the evolutionary grounding for Wynn’s findings, confirming that the human infant’s numerical capacity is rooted in a phylogenetically ancient, non-verbal cognitive module shared with our nearest living relatives.
Furthermore, Boysen’s work with the reverse-reward paradigm profoundly illuminated the developmental transition termed “conceptual bootstrapping” by Susan Carey. Carey’s framework illustrates how human children utilize external linguistic and cultural place-holders—such as reciting the counting list “one, two, three”—to initially organize perceptual inputs before gradually filling those symbols with rich cardinal meaning. Sheba’s cognitive trajectory demonstrated this exact bootstrapping process in an animal model. Moreover, this research yielded profound clinical applications for human education, informing targeted symbolic scaffolding interventions for children suffering from developmental dyscalculia and attention-deficit/hyperactivity disorder (ADHD), where external semiotic tools are deployed to strengthen fragile internal executive inhibitory circuits.
12.2 Contributions to Animal Ethics, Welfare, and Moral Standing
Beyond its contributions to cognitive theory, Dr. Sarah Boysen’s research program played a pivotal role in transforming international perspectives on animal ethics, the moral standing of great apes, and the legal standards governing captive chimpanzee housing. For the greater part of the twentieth century, biomedical paradigms treated chimpanzees as physiological surrogates for human disease modeling, maintaining them in barren, isolated laboratory cages that completely disregarded their sophisticated emotional, social, and intellectual lives. The dramatic demonstrations emerging from the Ohio State laboratory—revealing that a chimpanzee could conceptualize zero, mentally calculate sums, and utilize abstract symbols to manage impulse control—demolished the Cartesian view of animals as unthinking biological automata.
The empirical proof of high-level abstract reasoning in chimpanzees provided philosophical ammunition for movements such as the Great Ape Project, founded by moral philosophers Peter Singer and Paola Cavalieri in 1993, which campaigned for extending fundamental moral rights—the right to life, individual liberty, and protection from torture—to non-human hominids. Boysen emerged as a prominent, vocal advocate for non-human primate welfare, utilizing her scientific standing to lobby against invasive biomedical research and arguing that species possessing self-awareness, symbolic capacity, and complex mental landscapes suffered severe psychological harm when subjected to confinement and social deprivation.
This scientific and ethical pressure contributed directly to historic regulatory transformations. Over the early decades of the twenty-first century, major scientific institutions, including the National Institutes of Health (NIH) in the United States and research councils across Europe, systematically terminated biomedical testing on chimpanzees. In 2015, the United States Fish and Wildlife Service designated all captive chimpanzees as fully endangered, bringing an end to invasive captive exploitation. Dr. Boysen was directly involved in initiatives to retire laboratory chimpanzees to expansive, naturalistic sanctuaries such as Chimp Haven, ensuring that subjects who had spent decades expanding the boundaries of human scientific knowledge could live out their lives in rich, dignified social environments.
12.3 Directions for Future Research in Comparative Cognition
The empirical horizon opened by Boysen’s pioneering work continues to guide and inspire contemporary frontiers in comparative cognition and evolutionary neuroscience. In the modern era, the methodological toolbox of comparative cognition has expanded far beyond the manual stimulus trays of the 1980s. Modern researchers are deploying non-invasive neuroimaging techniques, such as functional near-infrared spectroscopy (fNIRS), passive functional magnetic resonance imaging (fMRI), and high-density electroencephalography (EEG) with cooperative, awake-behaving primates, mapping the living neural correlates of numerical processing in real time without stress or restraint.
Concurrently, the rapid evolution of computational neuroscience and artificial intelligence has established new theoretical intersections with comparative numerosity. Researchers are currently utilizing deep artificial neural networks (ANNs) to model the spontaneous emergence of number-tuned artificial neurons within hierarchical visual processing layers, comparing the computational trajectories of deep learning networks with the longitudinal learning profiles documented in Sheba and Ai. These computational models are helping to resolve long-standing debates regarding the precise algorithmic transitions that allow a neural network—biological or artificial—to bridge the gap between continuous pixel densities and discrete, symbolic cardinality.
Yet, despite these technological advancements, the core questions articulated by Sarah Boysen remain central to the scientific pursuit of the animal mind. Fundamental inquiries persist regarding the absolute boundary conditions of non-human abstraction: Can non-human primates master multi-digit place-value systems? Can they internalize true multiplicative operations? Can they conceptualize negative quantities or rational fractions within a unified symbolic framework? As contemporary comparative cognitive science continues to interrogate the minds of primates, cetaceans, and corvids, the enduring legacy of Dr. Sarah Boysen and Sheba stands as a foundational monument—an experimental triumph that dismantled human exceptionalism and forever expanded our vision of the evolutionary architecture of intellect.
Conclusion
The numerosity research program directed by Dr. Sarah T. Boysen at the Ohio State University stands as a watershed achievement in the history of comparative psychology. Through an unyielding commitment to experimental rigor, double-blind controls, and ethically sound, socially enriched research paradigms, Boysen dismantled centuries of anthropocentric assumptions regarding the uniqueness of human mathematical cognition. Working alongside her chimpanzee partner Sheba, Boysen demonstrated that the capacity to bind arbitrary visual symbols to discrete cardinal quantities, track invariant ordinal sequences, and conceptualize the null set does not depend on the biological machinery of spoken human language. These cognitive faculties represent deeply conserved ancestral hominid potentials that evolved millions of years before the dawn of modern humanity.
Furthermore, the empirical milestones of spontaneous foraging summation and the symbolic solution to the reverse-reward contingency task illuminated the profound functional significance of symbolic scaffolding. Boysen proved that external signs do not merely label internal thoughts; they actively reorganize the cognitive architecture, providing a semiotic buffer that liberates executive control from the tyrannical grip of immediate, affective sensory inputs. By showing that an abstract numeral allows an ape’s mind to conquer visceral impulse and achieve strategic self-regulation, Boysen’s work united cognitive semiotics, neurobiology, and evolutionary theory into a cohesive scientific framework. Her discoveries fundamentally reshaped models of human child development, catalyzed revolutions in animal welfare legislation, and established a permanent theoretical bridge across the phylogenetic divide, ensuring that our understanding of the animal mind will never again be bound by the limitations of human exceptionalism.
References
- Berntson, G. G., Boysen, S. T., & Cacioppo, J. T. (1993). Neurobehavioral organization and the cardinal number principle in the chimpanzee (Pan troglodytes). Journal of Comparative Psychology, 107(2), 191–201. https://doi.org/10.1037/0735-7036.107.2.191
- Boysen, S. T. (1993). Counting abilities in nonhuman primates: Information parallel to human infant development. Annals of the New York Academy of Sciences, 696(1), 168–183. https://doi.org/10.1111/j.1749-6632.1993.tb32394.x
- Boysen, S. T., & Berntson, G. G. (1989). Numerical competence in a chimpanzee (Pan troglodytes). Science, 244(4911), 1426–1428. https://doi.org/10.1126/science.2736932
- Boysen, S. T., & Berntson, G. G. (1995). Responses to quantity: Perceptual versus cognitive mechanisms in chimpanzees (Pan troglodytes). Journal of Experimental Psychology: Animal Behavior Processes, 21(1), 82–86. https://doi.org/10.1037/0097-7403.21.1.82
- Boysen, S. T., Berntson, G. G., Hannan, M. B., & Cacioppo, J. T. (1996). Quantity-based interference and symbolic representations in chimpanzees (Pan troglodytes). Journal of Experimental Psychology: Animal Behavior Processes, 22(1), 76–86. https://doi.org/10.1037/0097-7403.22.1.76
- Boysen, S. T., & Hallberg, K. I. (2000). Primate numerical competence: Contributions toward understanding the evolution of mathematical thought. Cognitive Science, 24(3), 423–443. https://doi.org/10.1207/s15516709cog2403_4
- Brannon, E. M., & Terrace, H. S. (1998). Ordering of the numerosities 1 to 9 by monkeys. Science, 282(5389), 746–749. https://doi.org/10.1126/science.282.5389.746
- Carey, S. (2009). The Origin of Concepts. Oxford University Press. https://doi.org/10.1093/acprof:oso/9780195367638.001.0001
- Dehaene, S. (2011). The Number Sense: How the Mind Creates Mathematics (Revised and Updated ed.). Oxford University Press.
- DeLoache, J. S. (2000). Dual representation and young children’s use of scale models. Child Development, 71(2), 329–338. https://doi.org/10.1111/1467-8624.00148
- Gelman, R., & Gallistel, C. R. (1978). The Child’s Understanding of Number. Harvard University Press.
- Matsuzawa, T. (1985). Use of numbers by a chimpanzee. Nature, 315(6014), 57–59. https://doi.org/10.1038/315057a0
- Moyer, R. S., & Landauer, T. K. (1967). Time required for judgements of numerical inequality. Nature, 215(5109), 1519–1520. https://doi.org/10.1038/2151519a0
- Nieder, A., & Miller, E. K. (2004). A parieto-frontal network for abstract number representation in the non-human primate. Proceedings of the National Academy of Sciences, 101(19), 7457–7462. https://doi.org/10.1073/pnas.0402239101
- Pepperberg, I. M. (2006). Grey parrot (Psittacus erithacus) numerical abilities: Addition and further experiments on a zero-like concept. Journal of Comparative Psychology, 120(1), 1–11. https://doi.org/10.1037/0735-7036.120.1.1
- Pfungst, O. (1911). Clever Hans (The Horse of Mr. von Osten): A Contribution to Experimental Animal and Human Psychology (C. L. Rahn, Trans.). Henry Holt and Company.
- Premack, D. (1976). Intelligence in Ape and Man. Lawrence Erlbaum Associates.
- Tomasello, M., & Call, J. (1997). Primate Cognition. Oxford University Press.
- Wynn, K. (1992). Addition and subtraction by human infants. Nature, 358(6389), 749–750. https://doi.org/10.1038/358749a0