Elizabeth Spelke – 1949 Present

Biography

For more than half a century, the nature of the human mind has been debated through the lens of developmental psychology: Does the infant arrive in the world as an unformed tabula rasa, assembling coherent perceptions solely through the slow accretion of sensorimotor associations, or is the newborn endowed with an innate, evolutionarily ancient conceptual architecture? Few cognitive scientists have reshaped this fundamental epistemological landscape as profoundly as Elizabeth S. Spelke. Born in 1949, Spelke has spearheaded an intellectual revolution that fundamentally dismantled the twentieth-century constructivist consensus, demonstrating through rigorous experimental paradigms that human infants possess sophisticated, highly structured systems of foundational knowledge long before they can crawl, speak, or manipulate physical objects with deliberate manual dexterity.

Working at the intersection of developmental psychology, cognitive science, evolutionary biology, and philosophy of mind, Spelke formulated the landmark Core Knowledge framework. Her central thesis posits that the human mind does not begin as a general-purpose, undifferentiated learning engine. Instead, human cognition is underwritten by a discrete set of phylogenetically ancient, domain-specific representational systems. These systems—encompassing inanimate physical objects, numerical magnitudes, geometric spatial layouts, intentional agents, and social group categories—provide the bedrock upon which all subsequent cultural, linguistic, and scientific competencies are constructed. By shifting the experimental focus from motor execution to visual attention, Spelke provided empirical substance to ideas that had lingered as philosophical conjectures since René Descartes, Immanuel Kant, and Noam Chomsky.

This comprehensive monograph traces the intellectual trajectory, empirical breakthroughs, methodological innovations, theoretical debates, and translational applications that define Elizabeth Spelke’s ongoing career. From her formative graduate training under Eleanor Gibson to her long-standing directorship of the Laboratory for Developmental Studies at Harvard University, Spelke’s empirical discoveries have challenged classic Piagetian developmental stages, resisted radical behaviorist and connectionist reductionism, and provided a powerful defense of nativist cognitive architecture. Today, her framework not only informs early childhood educational policy and developmental diagnostics across the globe, but also provides the foundational benchmarks against which modern artificial intelligence and computational neuroscience measure their models of human commonsense reasoning.

1. Introduction to Elizabeth Spelke and the Revolution in Developmental Psychology

1.1 Biographical Trajectory and Academic Formative Years

Elizabeth Shilin Spelke was born on May 28, 1949, entering academia during an era when behaviorism was yielding to the early stirrings of the cognitive revolution. She pursued her undergraduate education at Radcliffe College and Harvard University, graduating in 1972. During these formative undergraduate years, psychology was grappling with the profound implications of Noam Chomsky’s critique of B.F. Skinner. Chomsky had proposed that the human capacity for natural language relies on an innate, rule-governed Universal Grammar that cannot be accounted for by operant conditioning. Inspired by these debates, Spelke began to question whether a parallel innate structure might govern human concepts beyond the linguistic domain, extending to the physical world of solid objects, spatial geometries, and living entities.

Following her undergraduate studies, Spelke entered the doctoral program in psychology at Cornell University, where she worked under the direct mentorship of the pioneering perceptual psychologist Eleanor J. Gibson. Gibson, renowned for her creation of the visual cliff paradigm and her ecological approach to perceptual learning, exerted a foundational influence on Spelke’s empirical approach. While Gibson emphasized the direct extraction of perceptual affordances from the ecological environment, Spelke pushed Gibsonian experimental methodology toward abstract mental representations. Spelke realized that infant looking behavior could reveal not only what infants see, but what they infer about unseen properties of the world. She earned her Ph.D. from Cornell in 1978, presenting a dissertation that investigated intermodal perception in young infants.

Spelke launched her independent academic career at the University of Pennsylvania, where she held a faculty post from 1979 to 1986. It was during her time at Penn that she designed the earliest iterations of the Violation-of-Expectation paradigm, publishing seminal papers on infant physical reasoning that sent shockwaves through developmental psychology. In 1986, she moved to the Department of Brain and Cognitive Sciences at the Massachusetts Institute of Technology, an intellectual powerhouse home to figures such as Jerry Fodor, Noam Chomsky, and Steven Pinker. At MIT, Spelke honed the modular and computational aspects of her theoretical framework. In 1996, she returned to Harvard University as a Professor of Psychology, eventually co-founding and directing the Laboratory for Developmental Studies. At Harvard, Spelke expanded her research program into cross-species comparative cognition, social categorization, and global educational interventions.

1.2 The Historical Paradigm Shift from Piagetian Constructivism

To appreciate the magnitude of Spelke’s empirical contribution, one must understand the prevailing orthodoxy that preceded her work: the constructivist developmental framework articulated by the Swiss psychologist Jean Piaget. Piaget’s stage theory asserted that infants spend their first eighteen to twenty-four months in the “sensorimotor stage.” Under the Piagetian view, neonates possess only primitive neonatal reflexes (such as sucking, grasping, and visual tracking). According to this classical model, neonates have no concept of an objective physical reality existing independently of their subjective sensory impressions and motor routines. Piaget claimed that object permanence—the understanding that an object continues to exist when out of sight—develops gradually through motor interactions, such as reaching, grasping, and manual un-hiding, reaching true maturity only around eight to twelve months of age.

Spelke identified a fatal methodological confound at the heart of Piaget’s experimental architecture: the conflation of cognitive competence with motor execution. An infant of four months lacks the manual dexterity, neuromuscular control, and motor planning required to reach out, grasp a fabric cloth, and pull it away to reveal a hidden toy. Piaget interpreted the infant’s failure to manually retrieve the occluded toy as evidence that the infant believed the object had ceased to exist—summarized in the aphorism “out of sight, out of mind.” Spelke recognized that demanding coordinated, multistep manual search tasks from motorically immature infants drastically underestimated their conceptual understanding. If one wished to measure the infant mind, one had to devise experimental paradigms that bypassed the physical limits of motor manipulation entirely.

By shifting the methodological medium from manual action to ocular tracking and preferential looking time, Spelke inaugurated a major paradigm shift. When experimental tasks measured looking durations rather than manual reaching, infants as young as three to four months demonstrated an unmistakable understanding of object permanence, physical solidity, spatio-temporal continuity, and mechanical support. They looked significantly longer at physically impossible events—such as an object passing cleanly through a solid barrier—than at physically possible control events. This shift reframed the infant mind from an unorganized sensorimotor reflex loop into an endowed cognitive agent that generates hypotheses, makes structured inferences, and is guided by an innate, abstract ontology of the physical universe.

1.3 Core Tenets of Spelke’s Nativist and Evolutionary Perspective

Spelke’s theoretical framework is fundamentally nativist and deeply informed by evolutionary biology. Central to her argument is the realization that human infants face an impossible computational problem if forced to learn the basic structure of the physical and social world through unguided statistical association alone. The sensory input striking an infant’s retina is fundamentally fragmented, noisy, dynamic, and ambiguous; two-dimensional retinal projections undergo radical shifts in luminance, scale, and perspective with every head movement. If infants possessed no pre-existing interpretative machinery, they could never infer that two distinct retinal patches of color belong to a single, continuous, three-dimensional physical object that remains stable across occlusions.

To solve this poverty-of-the-stimulus problem, Spelke posited that the human genome provides a set of innate, evolved inductive biases. These biases are not learned through environmental interaction; rather, they are phylogenetically ancient adaptations sculpted by millions of years of natural selection. In ancestral environments, organisms that possessed innate mechanisms for tracking predators through occlusion, calculating spatial trajectories across uneven terrain, and discriminating quantities of food or conspecifics survived and reproduced at vastly higher rates than organisms that had to learn the basic laws of physics through trial-and-error. For Spelke, infant cognition is not an unformed prototype of adult intellect, but an evolved survival toolkit operating with exceptional domain-specific elegance.

Philosophically, Spelke’s empirical nativism revitalized the tradition of Cartesian rationalism. Where René Descartes postulated innate ideas of geometry, God, and extension, and Immanuel Kant proposed a priori categories of space, time, and causality as prerequisites for perceptual experience, Spelke converted these philosophical speculations into empirically testable hypotheses in developmental laboratories. By demonstrating that three-month-old infants expect objects to obey continuity, cohesion, and contact mechanics, Spelke provided empirical support for the claim that the human mind comes equipped with an innate conceptual framework that structures perception long before the advent of formal schooling or linguistic mastery.

2. Theoretical Foundations: The Core Knowledge Framework

2.1 Definition and Architecture of Core Knowledge Systems

The Core Knowledge framework stands as Elizabeth Spelke’s most influential theoretical synthesis. At its foundation, the framework posits that human cognition is anchored by a small number of innate, domain-specific, and phylogenetically ancient conceptual systems. Each core system is dedicated to a distinct ontological domain: inanimate physical objects (mechanics), numerical magnitudes and sets (arithmetic), geometric environmental layouts (navigation), intentional agents (psychology), and social group categories (sociology). These systems are universal across the human species, manifesting in identical structural forms across diverse geographic locations, cultures, and socioeconomic environments.

The architecture of a core knowledge system is characterized by four defining criteria:

• Domain-specificity: The principles governing one core system apply exclusively to its specific ontological target. For instance, the mechanical principles of cohesion and contact govern physical inanimate objects, but they do not apply to the internal mental states or social allegiances of intentional agents.
• Task-specific encapsulation: Each core system operates with relative autonomy from top-down conceptual penetration, functioning automatically, rapidly, and independently of conscious deliberation or cultural instruction.
• Phylogenetic antiquity: Core systems are not uniquely human; their foundational computational mechanisms are widely shared with non-human animals, including non-human primates, domestic chicks, and rodents.
• Ontogenetic stability: Core systems do not vanish or transform fundamentally as a child matures; instead, they remain operational throughout the entire human lifespan, serving as the foundational computational substrate over which adult reasoning operates.

Crucially, Spelke makes a sharp architectural distinction between initial core representations and acquired explicit cultural knowledge. Core systems do not represent all possible facts about a domain; they are sparse, skeletal, and limited in scope. For example, the core object system represents an object’s spatial boundary, continuous trajectory, and solidity, but it remains blind to an object’s color, texture, functional utility, or monetary value. These initial representations do not represent the totality of human understanding, but rather serve as the cognitive foundation upon which culturally invented symbol systems, formal educational frameworks, and scientific theories build.

2.2 Domain-Specificity Versus Domain-General Learning Mechanisms

The Core Knowledge framework stands in stark contrast to theoretical models that rely exclusively on domain-general learning mechanisms, such as classical behaviorism, early connectionism, and contemporary radical empiricist artificial intelligence models. Domain-general theories postulate that the brain is a general-purpose optimization machine governed by uniform statistical association engines—such as Hebbian synaptic plasticity, backpropagation, or general associative conditioning. In these frameworks, the same computational algorithm is assumed to learn language, recognize human faces, calculate arithmetic sums, and navigate complex spatial topography, with all domain-specific structure emerging entirely from the statistical distributions present in the environmental input.

Spelke has consistently contested this domain-general architecture using empirical evidence. She demonstrates that infant behavior exhibits distinct cognitive signatures across different domains that cannot be accounted for by a unified statistical learner. For example, infants apply strict spatio-temporal continuity constraints to inanimate artifacts, but they abandon these constraints when evaluating the goal-directed trajectories of self-propelled, intentional agents. An infant expects a wooden block to move continuously along a smooth trajectory when struck, but readily accepts that an adult human can halt, pivot, and reverse direction without physical contact. If a domain-general statistical learner were operating over raw pixel arrays, both visual inputs would be processed with identical spatio-temporal filtering.

Rather than replacing learning, core knowledge systems channel and constrain subsequent experiential learning. By pre-segmenting the perceived world into distinct ontological categories, core systems reduce the hypothesis space an infant must navigate. When an infant encounters a novel visual scene, core knowledge provides an immediate baseline parsing: the scene is already populated with discrete solid objects that cannot pass through one another, bounded geometric surfaces that define navigable space, and intentional agents acting toward goals. Learning does not begin in a void; it unfolds through the directed elaboration of these core conceptual categories.

2.3 The Epistemic Persistence of Core Systems Throughout Life

A central tenet of Spelke’s thesis—one that sharply differentiates her from developmental constructivists like Susan Carey—is the principle of conceptual persistence. In many traditional developmental frameworks, early infant notions are thought to be fundamentally rewritten, replaced, or radically restructured as children undergo cognitive development and acquire scientific concepts. Piaget posited radical conceptual restructuring between developmental stages; Carey argued for “Quinian bootstrapping,” wherein initial conceptual structures are abandoned in favor of incommensurate, newly acquired concepts.

Spelke, by contrast, argues that core knowledge systems are never overwritten, extinguished, or discarded. They persist intact across adulthood, functioning beneath reflective consciousness and often in direct tension with explicit, scientifically acquired beliefs. For example, an adult who has mastered Newtonian mechanics and can solve complex calculus equations regarding gravitational vectors still possesses an intuitive core object mechanics system that operates on simple principles of direct contact and impetus. When asked to predict the trajectory of a ball exiting a curved spiral tube, even college-educated adults routinely demonstrate intuitive physical errors, expecting the ball to continue along a curved trajectory rather than exiting along a straight tangent. This core intuition operates automatically in split-second real-world judgments, bypassing formal academic training.

Empirical evidence for this epistemic persistence is corroborated by cross-cultural studies demonstrating that fundamental core representations are invariant across human populations, regardless of formal schooling, literacy, or symbolic numeracy. Whether testing infants in modern Boston, adults in industrialized cities, or members of the indigenous Munduruku and Pirahã populations living in the Amazonian basin, core intuitions of basic geometry, approximate numerical ratios, and object mechanics remain identical. While cultural exposure equips humans with distinct explicit tools—such as counting words or Euclidean compasses—the underlying core knowledge machinery remains universal and unchanged.

3. Methodological Innovations in Infant Cognitive Science

3.1 The Violation-of-Expectation (VoE) Paradigm

Because preverbal infants cannot speak, follow complex verbal instructions, or coordinate intricate manual motor tasks, understanding their cognitive architecture required a radical methodological innovation. Spelke, working alongside and building upon the early looking-time paradigms of Robert Fantz and the habituation methods of Eleanor Gibson, perfected what is now recognized as the gold standard of developmental cognitive neuroscience: the Violation-of-Expectation (VoE) paradigm. The operational logic of the VoE paradigm is straightforward yet methodologically powerful: it leverages the evolutionary impulse of human infants to look longer at unexpected, anomalous, or novel events than at expected, familiar, or predictable ones.

The standard VoE architecture unfolds in two successive phases:

1. Habituation or Familiarization Phase: The infant is presented with an event until their visual attention wanes to a predetermined criterion (habituation), or they are exposed to a fixed number of familiarization trials displaying a standard physical interaction. This establishes an experimental baseline.
2. Test Phase: The infant is presented with two alternative test events: a physically possible event and a physically impossible (violation) event. Crucially, the impossible event is experimentally designed to either share higher superficial perceptual novelty with the habituation event or to appear visually similar to the possible event, isolating the physical or logical violation as the primary causal variable.

If the infant’s looking behavior were driven solely by low-level perceptual features (such as retinal contrast, color, or raw motion velocity), their visual fixation times between the two conditions would either show no significant difference or favor the more perceptually novel baseline. If, however, the infant has formed an abstract mental representation of the unseen physical properties of the scene (such as object permanence, solidity, or continuous trajectory), the physically impossible event will violate their internal mental model. This violation elicits cognitive surprise, manifesting as significantly prolonged looking duration. Through the VoE paradigm, Spelke turned infant visual gaze into an objective index of the human mind’s innate ontological expectations.

3.2 Intermodal Perception and Cross-Modal Transfer

Long before her famous object mechanics studies, Spelke established herself as a leader in perceptual development through experiments investigating intermodal perception—the capacity to synthesize and coordinate sensory information across distinct perceptual modalities, including vision, audition, and haptic touch. During the 1970s, prevailing associationist frameworks, descended from the empiricist philosophy of George Berkeley, argued that sensory modalities are entirely isolated at birth. Under this classical view, an infant must slowly learn to associate the visual image of a mother’s face with the auditory sound of her voice through thousands of repeated co-occurrences.

Spelke overturned this assumption in a landmark series of experiments demonstrating that infants possess an innate sensitivity to amodal perceptual invariants—properties such as rhythm, temporal synchrony, spatial co-location, and intensity that are shared across sensory systems. In her classic 1976 experiment, Spelke presented four-month-old infants with two side-by-side films: one depicting a woman playing a game of peek-a-boo, and the other showing a hand bouncing a baton against a surface. Simultaneously, a single auditory soundtrack matching the rhythm and temporal tempo of only one of the films was played from a speaker located midway between the screens. Without any prior training or reinforcement, the infants directed their visual attention to the video that matched the temporal cadence of the audio track.

Spelke extended this methodology to investigate cross-modal transfer between tactile and visual modalities. In experiments with one-month-old and four-month-old infants, objects with distinct structural textures (for example, smooth spherical shapes versus knobby, nubbed surfaces) were placed inside the infants’ mouths or hands out of visual sight. After the infants explored these objects tactually, the items were removed, and the infants were presented with high-contrast visual displays of both shapes. The infants looked significantly longer at the shape they had previously explored only through touch. This proved that infants do not experience William James’s hypothesized “blooming, buzzing confusion” of disconnected sensory inputs; rather, they map amodal geometric and temporal structures across distinct sensory channels from the earliest weeks of life.

3.3 Addressing Methodological Debates and Skepticism

The remarkable findings generated by the Violation-of-Expectation paradigm ignited significant methodological and philosophical controversy within cognitive science. Prominent critics, such as Marshall Haith, Richard Bogartz, and Charles Clifton, argued that Spelke and her nativist colleagues were guilty of “rich interpretations” of “lean data.” These critics argued that an infant looking longer at an impossible event does not constitute proof of abstract conceptual knowledge, deductive reasoning, or an understanding of object permanence. Instead, they claimed, looking times could be explained by low-level perceptual biases, such as differences in retinal movement patterns, peripheral flickers, habituation asymmetries, or visual interest in novel physical configurations.

Spelke responded to these critiques with a sustained program of empirical controls. She pointed out that if infant looking times were driven by low-level perceptual features, subtle changes in display lighting, velocity, or surface texture should systematically alter or erase the looking-time differences. In systematic replications, Spelke and her collaborators matched the impossible and possible test events for every observable low-level perceptual variable: luminance, overall motion paths, surface area, and total displacement time. When impossible events were rendered physically possible by altering unseen background occluders (which did not alter the foreground perceptual motion paths), the infants’ increased looking time disappeared. The looking time tracked the *physical impossibility* of the scene, not its superficial perceptual kinematics.

Furthermore, Spelke addressed replication challenges by establishing standardized visual habituation criteria and automated corneal-reflection eye-tracking systems, eliminating human observer bias through strict double-blind coding protocols. To ensure that results were not idiosyncratic artifacts of specific experimental displays, Spelke conducted cross-laboratory replications across diverse populations. By showing that infants’ looking patterns generalized across three-dimensional physical stages, two-dimensional video projections, computer animations, and varied real-world objects, Spelke demonstrated that the infant visual response is an index of deep, abstract, and robust conceptual models of the physical environment.

4. Core System 1: Inanimate Object Mechanics and Spelke’s Principles

4.1 The Principle of Cohesion

The first core knowledge system identified by Spelke is the system of inanimate object mechanics, often referred to as intuitive physics. At the center of this system is the **Principle of Cohesion**. This principle defines what an infant considers to be a physical object: an entity that moves as a bounded, internally connected, and coherent whole, preserving its structural boundaries over space and time. Objects maintain their internal unity; their constituent parts move together, and they do not spontaneously shear, fracture, or dissolve into independent pieces without the application of external force.

To evaluate this principle empirically, Spelke presented three-month-old infants with displays in which an experimenter’s hand grasped the top half of an object and lifted it upward. In the *possible event*, the entire bounded object moved as a cohesive unit, rising smoothly into the air. In the *impossible event*, the object sheared in two: the top half lifted away with the experimenter’s hand while the bottom half remained resting on the floor. Infants looked significantly longer at the shearing event than at the cohesive movement event. The infants had never observed this specific physical object before, yet they maintained an a priori expectation that the visible boundaries of the object defined an internally cohesive, connected whole.

This principle of cohesion operates as an essential computational constraint on the infant visual parsing system. When confronted with an environment containing partially overlapping, stacked, or contiguous surfaces, an infant cannot rely solely on color or texture similarity to segment the scene into discrete objects; surfaces of identical color often belong to different objects, while single objects often possess varied patterns and colors. Instead, Spelke showed that infants rely primarily on **common motion**: surfaces that move together are immediately inferred to be parts of a single, cohesive, bounded object. Surfaces that move independently, even if identical in visual appearance, are parsed as distinct physical entities.

4.2 The Principle of Continuity

The second foundational pillar of Spelke’s intuitive physics is the **Principle of Continuity**. This principle states that objects move only on continuous, uninterrupted paths through space and time. A physical object cannot spontaneously disappear from one spatial location and reappear in another without tracing a continuous trajectory through the intervening space; furthermore, two solid objects cannot occupy the same region of space and time simultaneously (the property of solidity).

Spelke demonstrated this principle in a series of occlusion experiments. In these studies, an object moved behind two spatially separated, vertical occluders. Between the two occluders was an empty, visible gap. In the *possible event*, as the object traversed from the left of the stage to the right, it moved behind the first occluder, became visibly apparent as it crossed the central gap, moved behind the second occluder, and finally emerged on the right side. In the *impossible event*, the object traveled behind the first occluder, did *not* appear in the visible gap between the screens, and then emerged from behind the second occluder. Despite the short duration of the disappearance, infants as young as two-and-a-half to three-and-a-half months of age looked significantly longer at the discontinuous event.

These findings carried monumental implications for the classical Piagetian view of object permanence. If infants lacked a concept of an object persisting through space and time while out of sight, the impossible event would not have violated their expectations; the infant would have simply seen one object vanish and another object appear, accepting each as an isolated perceptual event. The infant’s cognitive surprise proved that they had constructed a mental representation of an enduring, spatially localized object that continued to trace a continuous spatiotemporal path behind the occluders. The infant computed the object’s hidden trajectory, generated an explicit expectation that the object would traverse the visible gap, and experienced cognitive conflict when this spatiotemporal expectation was violated.

4.3 The Principle of Contact

The third core principle governing inanimate object mechanics is the **Principle of Contact**. This principle dictates that inanimate objects act upon one another only through direct physical contact; they do not interact via action at a distance. An inert physical object cannot spontaneously launch itself into motion, nor can it induce motion in another inert object without physical collision or mechanical connection. This principle serves as the foundation for human causal reasoning about the mechanical world.

To demonstrate this principle, Spelke adapted the classic launching paradigms pioneered by the Belgian psychologist Albert Michotte for use with preverbal infants. In these experiments, infants observed two objects: Object A and Object B. In the *contact condition*, Object A moved across the display, made direct physical contact with Object B, and Object B immediately began moving along the same trajectory—a classic mechanical collision. In the *action-at-a-distance condition*, Object A moved toward Object B but stopped short by several centimeters, whereupon Object B spontaneously began moving along the trajectory without any physical contact occurring.

Infants as young as six months habituated readily to the direct contact collisions, interpreting them as standard mechanical events. However, when exposed to the action-at-a-distance condition, infants exhibited sharp increases in looking time, indicating surprise at the violation of mechanical causality. Crucially, Spelke and her colleagues demonstrated that this expectation of contact mechanics is strictly **domain-specific**. When the identical action-at-a-distance trajectories were enacted by two human beings or animated shapes possessing biological motion cues (such as self-propulsion and flexible change of shape), infants showed no surprise whatsoever when one person began walking after another person stopped short. The Principle of Contact is strictly dedicated to inanimate, mechanical objects, proving that infant cognition possesses an innate categorical distinction between physical objects and animate agents.

5. Core System 2: Numerical Representations and Infant Mathematics

5.1 The Approximate Number System (ANS) in Infancy

The second domain within the Core Knowledge framework is the system for representing numerical quantities. For decades, traditional cognitive psychology assumed that true mathematical competence is entirely dependent upon the acquisition of language and symbolic counting systems. Without words for numbers (“one,” “two,” “three”), it was believed, the human mind could not calculate, compare, or track numerical quantities. Spelke, in collaborative work with Fei Xu, Stanislas Dehaene, and Justin Halberda, challenged this assumption by proving that preverbal infants, non-human animals, and adults across all cultures share an innate, non-symbolic system for representing approximate numerical magnitudes: the **Approximate Number System (ANS)**.

The ANS represents quantities not as discrete symbols, but as noisy, continuous mental magnitudes mapped onto an internal, analog mental number line. Because it operates on analog representations, the ANS is constrained by the mathematical principles of the **Weber-Fechner Law**: the discriminability of two numerical quantities depends not on their absolute arithmetic difference, but on the **ratio** between them. For instance, distinguishing 8 items from 16 items is computationally identical to distinguishing 16 items from 32 items, as both comparisons share an identical 1:2 ratio, even though the absolute numerical difference between the second pair is twice as large.

In a seminal paper, Xu and Spelke (2000) habituated six-month-old infants to visual displays containing arrays of dots representing a specific quantity (for example, 8 dots). To eliminate all potential perceptual confounds, the researchers varied the continuous physical variables across trials: total dot surface area, dot density, individual dot size, and overall perimeter length were systematically balanced. During the test phase, infants were presented with alternating displays containing either the familiar quantity (8 dots) or a novel quantity (16 dots). The six-month-old infants looked significantly longer at the novel numerical displays, successfully discriminating 8 from 16 dots. However, when tested on a 2:3 ratio (such as 8 versus 12 dots), six-month-olds failed to discriminate the difference. By nine months of age, this ratio acuity sharpens to 2:3 (e.g., discriminating 8 from 12, or 16 from 24), and continues to refine across childhood until reaching adult levels of approximate acuity, which allows discrimination of ratios as fine as 9:10 or 10:11.

5.2 The Object Tracking System for Small Exact Sets

While the Approximate Number System provides an innate representation of large quantities governed by ratio limits, infants also possess a second, independent core system for processing small collections of items: the **Object Tracking System (OTS)**, often associated with the perceptual phenomenon of subitizing. The OTS does not represent quantity as an abstract analog magnitude; rather, it is anchored in the perceptual mechanisms of visual attention and visual working memory. Drawing on Zenon Pylyshyn’s model of visual indexes (FINSTs, or “fingers of instantiation”), the OTS operates by assigning discrete, spatio-temporal attentional pointers to individual objects in a visual array.

Because the OTS relies on visual working memory, it is constrained by a strict capacity limit: it can simultaneously track a maximum of **three to four** individual items. Within this small range, the OTS provides precise, exact numerical representations. In experiments measuring infant manual searching or looking times at hidden objects (such as crackers placed into an opaque bucket), infants easily track differences between 1 versus 2 crackers, and between 2 versus 3 crackers, consistently crawling toward or reaching into the container holding the larger quantity.

The definitive empirical signature of the OTS is its **catastrophic failure** when set sizes exceed its working memory limit. While an infant can successfully discriminate 2 crackers from 3 crackers, their performance completely collapses when presented with a choice between 2 crackers versus 4 crackers, or 1 cracker versus 4 crackers. If infants were evaluating these small sets using the Approximate Number System, a comparison of 1 versus 4 (a 1:4 ratio) would be vastly easier to discriminate than a comparison of 2 versus 3 (a 2:3 ratio). Instead, the presence of 4 objects overwhelms the available attentional pointers of the OTS, causing the working memory representation to crash entirely. This dissociation confirms that infant arithmetic is underwritten by two functionally and neurologically distinct core systems: one for tracking small exact sets through visual attention, and another for computing large approximate magnitudes through analog ratios.

5.3 Bridging Core Numerical Systems to Symbolic Mathematics

One of the central questions addressed by Spelke’s research is how children transition from these two non-symbolic, core systems to the uniquely human domain of formal, symbolic mathematics. Neither core system on its own is sufficient to represent modern mathematics: the OTS provides exact precision but is capped at three or four items, while the ANS can represent quantities up to infinity but is fundamentally imprecise and noisy. How does the human child arrive at the concept of exact large numbers, such as “exactly forty-seven”?

Spelke argues that the breakthrough occurs through the developmental synthesis of these two systems, mediated by the acquisition of natural language. Around the age of two to four years, children in numerate cultures memorize the verbal counting list (“one, two, three, four…”). Initially, this list is recited as an ungrounded string of words, lacking numerical meaning. Through systematic linguistic interaction, the child slowly maps the first words of the list (“one,” “two,” “three”) directly onto the discrete representations provided by the Object Tracking System.

Once the child reaches the limits of the OTS around the word “four,” a profound conceptual leap occurs: the discovery of the **successor function**. The child realizes that the structural property of the verbal count list—moving forward by one word in the sequence—corresponds to the physical addition of exactly one unit to any set, a realization that maps onto the continuous scale of the Approximate Number System. Language acts as the combinatorial bridge, binding the exact discrete indexing of the OTS with the unbounded scope of the ANS. Longitudinal studies conducted by Spelke and her colleagues have demonstrated that the early acuity of an infant’s ANS directly predicts their future mathematical competence in formal elementary school arithmetic, confirming that symbolic mathematical achievement is built upon the foundation of core numerical systems.

6. Core System 3: Spatial Navigation and Geometric Foundations

6.1 The Geometry of the Surrounding Layout

The third system of core knowledge represents the continuous geometric layout of the spatial environment. Spatial navigation is an existential requirement for any mobile organism; a foraging animal must venture into the environment and return to its nest or safety, calculating its position relative to the physical landscape. Spelke, in landmark collaborative research with Linda Hermer, discovered that when human infants and young children become disoriented, their reorientation relies on a specialized, encapsulated cognitive module dedicated strictly to the **macroscopic geometric shape** of the surrounding environment.

To demonstrate this phenomenon, Hermer and Spelke (1994, 1996) designed the classic **reorientation paradigm**. An infant or toddler was placed inside a controlled, rectangular room. In the corner of the room, a desirable toy was hidden inside a concealed box in full view of the child. The child was then blindfolded, gently picked up, and spun around until completely disoriented with respect to absolute egocentric coordinates. Once unblindfolded, the child was encouraged to find the hidden toy. Because a rectangular enclosure has two-fold rotational symmetry, two diagonal corners are geometrically indistinguishable: both have a long wall to the left and a short wall to the right. If children use the geometric shape of the layout, they should search equally at the correct corner and the diagonally opposite corner (the rotational error corner), while systematically ignoring the other two corners.

Children searched at the correct corner and the rotational opposite corner with equal frequency, ignoring the non-congruent corners. The critical test occurred when the researchers added high-contrast, non-geometric features: they painted one of the short walls a bright, salient blue, providing a unique landmark that broke the room’s rotational symmetry. Despite the presence of the blue wall, disoriented toddlers under the age of four continued to make the rotational error at high rates, searching the incorrect corner half the time. Even though the children could easily see and identify the blue wall, their spatial reorientation system was encapsulated: it operated strictly over geometric variables—metric distance and angle—while ignoring non-geometric featural landmarks.

6.2 Comparative and Neurobiological Underpinnings of Spatial Core Knowledge

The discovery of the encapsulated geometric reorientation module in human toddlers led to a broader question: Is this spatial geometry system a uniquely human adaptation, or is it a phylogenetically ancient system shared across the animal kingdom? Spelke, collaborating with comparative psychologists and neuroscientists such as Giorgio Vallortigara and Russell Phillips, conducted parallel reorientation experiments across multiple non-human vertebrate species, including domestic chicks, rodents, and non-human primates.

The results confirmed the evolutionary antiquity of core spatial geometry. Day-old domestic chicks, reared in controlled visual environments without prior navigational experience, demonstrated the exact same pattern of reorientation behavior when placed in rectangular enclosures. When disoriented, the chicks used the metric properties of the enclosure’s walls to navigate to hidden food caches, making the identical two-fold rotational errors and failing to integrate non-geometric color cues. The same geometric encapsulation has been replicated in rats, mice, and rhesus macaques, proving that the geometric reorientation engine is an evolutionary adaptation conserved across hundreds of millions of years of vertebrate evolution.

Neurobiologically, this core geometric system is tied to specialized neural structures within the hippocampal formation and the parahippocampal cortex. The discovery of **place cells** in the hippocampus by John O’Keefe and **grid cells** in the medial entorhinal cortex by Edvard and May-Britt Moser provided the physiological basis for the computations Spelke observed behaviorally. Grid cells fire at regular, tessellated geometric intervals across an environment, constructing an internal, metric coordinate map of physical space that functions independently of visual surface features. Spelke’s behavioral paradigms captured the functional output of this ancient neuro-computational mapping system in preverbal human infants.

6.3 Integration of Geometric Maps with Local Landmarks

While toddlers and non-human animals are constrained by geometric modularity during reorientation, normal adult humans are not: an adult disoriented in a rectangular room with a single blue wall immediately navigates to the correct corner, using the phrase “to the left of the blue wall” to break the geometric symmetry. This divergence raised a key developmental question: How, and when, does the developing human mind bridge the gap between the encapsulated geometric core map and local visual landmarks?

Spelke, along with Linda Hermer-Vazquez, demonstrated that this developmental transition occurs around the age of five to six years, coinciding with the acquisition of specific spatial language. Specifically, children become capable of using landmarks to reorient at the exact point in development when they acquire the productive use of complex spatial prepositions linked to left-right distinction—such as “to the left of the long green wall” or “to the right of the blue door.” Language provides the compositional syntax necessary to bind two previously isolated cognitive modules: the geometric engine that calculates metric distance and direction, and the visual object-recognition engine that identifies surface color, texture, and landmark identity.

To verify that natural language serves as the real-time combinatorial medium for this integration, Hermer-Vazquez, Spelke, and Katsnelson (1999) tested adult humans under conditions of **verbal shadowing**. Adult participants were placed in the reorientation room with a colored wall while simultaneously performing a demanding verbal task: listening to continuous spoken speech through headphones and repeating every word back immediately. Under verbal shadowing, which preoccupied the linguistic processing system, adult participants regressed to the performance of two-year-old toddlers and non-human chicks: they were unable to use the blue wall to reorient, making the classic two-fold rotational geometric error. Conversely, when adults performed a non-linguistic shadowing task of equal cognitive difficulty (such as shadowing a complex rhythmic drumbeat), their ability to use the landmark remained intact. This demonstrated that the flexible integration of core geometric knowledge with local landmarks relies on the real-time engagement of natural language.

7. Core System 4: Agents, Actions, and Early Social Understanding

7.1 Identifying Animate Agents and Teleological Action

The fourth core knowledge domain is the system dedicated to **intentional agents and their goal-directed actions** (intuitive psychology). In the sensory world, an infant encounters both passive physical objects and active, living beings. If an infant applied the mechanics of Core System 1 to human beings, they would expect people to move only when pushed and to remain motionless until struck. Conversely, if an infant applied psychological reasoning to rocks or blocks, they would expect inert items to seek out food and communicate. Spelke, working alongside developmental researchers like György Gergely, Amanda Woodward, and Gergely Csibra, proved that infants possess a core system for parsing animate agents based on distinct motion cues and intentional principles.

The primary perceptual triggers that activate Core System 4 are **self-propulsion**, **biological motion**, and **contingent responsiveness**. When an entity demonstrates that it can initiate its own movement without external physical impact, or when it moves in a non-rigid, articulate manner, infants immediately assign it to the category of an agent. Once an entity is categorized as an agent, the infant interprets its behavior using the **teleological stance**—the assumption that the agent’s movements are directed toward specific goals, rather than being meaningless physical trajectories.

Furthermore, Spelke demonstrated that infants expect agents to adhere to the **Principle of Rational Efficiency**. Infants expect an agent to pursue its goals via the most direct, physically efficient path available, given the constraints of the environment. In a classic paradigm, infants were habituated to a small circle jumping over a physical barrier to reach a larger circle. In the test phase, the physical barrier was removed. If the circle continued to jump through the air along its old path (which was now an inefficient trajectory through empty space), infants looked significantly longer than if the circle moved along a new, straight line directly toward the goal. Even though the straight-line path was visually novel, it was recognized as rationally efficient. Infants do not simply track kinematics; they infer an agent’s underlying goals and evaluate whether its actions are computationally optimal.

7.2 Social Evaluation and Moral Intuitions in Early Life

Building upon Spelke’s insights regarding the core psychology of intentional agents, developmental cognitive scientists—including Spelke’s former collaborator and student Karen Wynn, alongside Kiley Hamlin and Paul Bloom—investigated whether this core agent system contains the foundations of social evaluation and moral intuition. These studies explored whether preverbal infants evaluate agents based on the **valence** of their interactions with others.

In these experiments, infants observed animated, three-dimensional puppet shows featuring geometric characters with googly eyes. In the baseline event, a “Climber” character struggled to move up a steep hill. In alternating trials, two distinct social interactions occurred:

• The Helper Event: A second character (e.g., a yellow triangle) entered the stage and pushed the Climber gently up the hill, aiding it in reaching its goal.
• The Hinderer Event: A third character (e.g., a blue square) entered the stage and pushed the Climber down the hill, preventing it from reaching its goal.

Following these presentations, infants were presented with a tray holding the Helper and the Hinderer characters to measure physical reaching preferences. Infants as young as three to six months of age overwhelmingly reached for the prosocial Helper over the antisocial Hinderer. Furthermore, infants exhibited surprise when the Climber subsequently approached the Hinderer rather than the Helper, demonstrating that infants expect agents to affiliate with prosocial actors. These findings demonstrated that human morality is not an entirely learned cultural construct; its earliest building blocks—tracking goals, evaluating cooperative versus competitive behavior, and favoring prosocial actors—are embedded within the core psychology system.

7.3 Gaze Following and Shared Intentionality

An essential component of the core agent system is the recognition of eyes as perceptual organs that extract information from the environment. Very young infants do not view the eyes simply as high-contrast physical patches; they recognize eye gaze as an intentional vector connecting an agent to a specific target in external space. Spelke’s laboratory has mapped the precise developmental trajectory of **gaze following** and its role in establishing **triadic interactions**—the shared communicative focus between an infant, an adult, and an external object.

In experiments evaluating visual tracking, infants observed an adult turn their head and eyes toward one of two identical objects placed on either side of a room. Infants systematically followed the adult’s directional line of sight, shifting their own visual attention to focus on the target object. Crucially, Spelke and her colleagues showed that this gaze following is not triggered by mere head movement: if an adult turns their head toward an object while their eyes remain closed or covered by an opaque blindfold, infants do not follow the head turn. The infant’s looking system specifically tracks the visibility of the agent’s open eyes.

This early sensitivity to gaze forms the basis of what Michael Tomasello and Spelke identify as shared intentionality. Gaze tracking allows the infant to infer communicative intent: when an adult looks at an object, points to it, and produces an unfamiliar vocalization, the infant infers that the word refers to the object within the adult’s visual focus. This capacity to treat eye gaze as an informational cue pointing to unseen epistemic states (attention, interest, reference) bridges primitive core agency with the cultural transmission of language and tool use.

8. Core System 5: Social Categorization and In-Group Biases

8.1 Linguistic and Accent Cues as Social Markers

In her later work, Spelke, in collaboration with Katherine Kinzler and Kristin Shutts, identified a fifth domain of core knowledge: the system for **social categorization and in-group identification**. Humans are an obligately social species, and survival has historically depended upon navigating complex social groups, identifying potential allies, and avoiding hostile coalitions. Spelke investigated the innate cognitive biases that guide an infant’s initial parsing of the human social landscape, yielding surprising discoveries regarding how the infant brain categorizes social partners.

In contemporary society, race is widely assumed to be the most primary, visually salient axis of human social categorization. However, through an evolutionary lens, Spelke and Kinzler hypothesized that racial categorization could not be an innate, evolutionarily ancient adaptation. Throughout Pleistocene human evolution, ancestral hunter-gatherers would never have encountered individuals of different races; human travel was limited to foot locomotion, meaning neighboring bands were morphologically and racially similar. The primary social variations encountered between ancestral groups were differences in dialect, language, and cultural accent. Therefore, natural selection would have favored mechanisms that prioritized **linguistic and vocal cues** over physical morphology.

To test this hypothesis, Kinzler, Dupoux, and Spelke (2007) conducted experiments testing five-to-six-month-old infants and ten-month-old infants from monocultural settings. Infants were shown alternating films of two women who smiled silently. Then, the women spoke: one spoke the infant’s native language (e.g., English), while the other spoke a foreign language (e.g., French), or one spoke the native language with a native accent while the other spoke with a foreign accent. When both women subsequently appeared side-by-side in silence, infants looked significantly longer at the native speaker. At ten months, when offered identical toys from both women, infants selectively reached out to accept toys offered by the native-accented speaker. Crucially, when physical race (e.g., White versus Black faces) was directly pitted against linguistic accent, infants and young children prioritized **accent over race**, selectively choosing friends and learning partners who shared their native accent, even when their physical race differed.

8.2 Development of Social Group Preferences and Resource Distribution

The core system for social categorization quickly broadens from visual attention into selective behavior and resource distribution during early childhood. Children do not merely watch social groups; they actively align their behaviors, selective imitation, and cultural choices with members of their perceived in-group.

Spelke, Kinzler, and Shutts demonstrated this dynamic across multiple functional domains:

• Selective Learning and Epistemic Trust: When toddlers observe two adults demonstrating how to operate a novel physical artifact, they systematically imitate the actions of the adult who speaks with a native accent, dismissing the demonstration of the foreign-accented speaker. The child treats the in-group speaker as a reliable, culturally authoritative source of information.
• Food and Artifact Preferences: In studies evaluating food selection, toddlers are significantly more likely to taste and accept novel, unfamiliar foods if they have previously witnessed an in-group, native-speaking adult eating that food with enjoyment, whereas endorsements by foreign-accented individuals produce little effect.
• Expectations of Fair Distribution: By preschool age, children expect resources (such as stickers or treats) to be distributed fairly among members of a neutral group. However, when resources are scarce, their expectations shift: they expect and approve of an in-group distributor giving resources to an in-group member, revealing early intuitions of group loyalty.

These findings demonstrate that the child’s mind is equipped with early social heuristics designed to accelerate cultural learning. Because culture is transmitted primarily through language and social conventions, the core social categorization system directs the infant’s attention toward those community members who share their linguistic dialect, ensuring the transmission of local survival skills, culinary habits, and tool proficiencies.

8.3 Implications for Combating Bias and Prejudice

The realization that human social categorization is rooted in core evolutionary mechanisms carries profound implications for understanding, and mitigating, prejudice, stereotyping, and systemic discrimination. Spelke’s empirical research deconstructs the assumption that racial prejudice is an inevitable, biological reflex of the human visual system. Instead, her work proves that racial bias is a culturally learned, socially constructed phenomenon, mapped onto an underlying evolutionary architecture that is natively sensitive to language and accent rather than skin pigmentation.

Because children do not possess an innate, evolutionary “race detector,” racial biases emerge secondarily when a culture structures race as a salient social boundary. When children observe that racial categories track differences in status, wealth, geography, and social grouping, their core social categorization system latches onto these physical features, treating them as markers of group identity. Racial essentialism is not an inevitable biological fate, but an emergent consequence of applying core categorization systems to a racially stratified environment.

This insight provides clear paths for educational, psychological, and policy interventions. If the child’s brain inherently prioritizes linguistic, communicative, and behavioral commonalities over visual morphology, educational environments can counteract racial prejudice by emphasizing shared communicative experiences, multilingual exposure, and diverse cooperative goals early in life. By decoupling the core social categorization module from superficial physical traits, society can leverage children’s innate drive for social affiliation to build inclusive communities rather than out-group antagonisms.

9. Natural Language as the Combinatorial Medium of Unique Human Thought

9.1 The Compositional Role of Natural Language

Throughout Elizabeth Spelke’s empirical career, a foundational paradox emerged across her data: If human infants share these core knowledge systems—for mechanics, numbers, geometry, agency, and social categories—with other vertebrate species, what explains the immense evolutionary divergence of human intellect? Why can human beings write poetry, invent symbolic algebra, construct telescopes, formulate quantum mechanics, and chart democratic governments, while chimpanzees, dogs, and chicks remain bound to basic survival routines within their evolutionary niches?

Spelke’s answer to this central question is the **Compositional Language Hypothesis**. While non-human animals and preverbal infants possess domain-specific core systems, these systems operate as isolated, encapsulated computational modules. Non-human animals can represent geometric space (Core System 3) and can track small sets of objects (Core System 2), but they cannot combine representations across these distinct domains into a unified, integrated thought. They cannot form the compound conceptual representation: *”the three small rocks to the left of the tall tree.”* Each core system remains an isolated computational silo.

The evolutionary breakthrough of the human species, Spelke argues, was the emergence of **natural language**. Language is not merely a communicative tool for broadcasting internal thoughts to other conspecifics; it is an internal combinatorial computational system. With its recursive syntax, compositional grammar, and open lexical architecture, natural language serves as a universal medium—a cognitive bridge capable of taking representations from distinct core systems and binding them into entirely novel, highly flexible conceptual structures that do not exist within any single core system alone.

9.2 Syntactic Structures and Complex Concept Formation

The mechanism through which natural language synthesizes core knowledge domains is illustrated by the acquisition of syntactic spatial and numerical structures. Consider the navigational challenge of reorientation: Core System 3 provides a geometric frame of reference (distinguishing “short wall” from “long wall”), while the visual feature recognition system identifies landmarks (“blue wall”). In isolation, the geometric module cannot process color, and the landmark system cannot determine metric orientation. Natural language solves this by providing closed-class syntactic elements—specifically spatial prepositions (“to the left of,” “inside,” “behind”)—that allow the brain to merge these representations into a single sentence: *”The treasure is to the left of the blue wall.”*

This compositional power is equally evident in the creation of exact large numbers. As discussed in Section 5, Core System 2 provides approximate analog magnitudes without exact bounds, and the Object Tracking System provides exact boundaries capped at three items. Natural language introduces the counting sequence, providing an infinite recursive structure: each number word ($N$) possesses an exact successor ($N+1$). By binding the discrete object indexes of the OTS with the continuous magnitude scale of the ANS using verbal syntax, language creates the formal concept of exact natural numbers.

Spelke found striking empirical support for this thesis in studies of deaf children raised in language-isolated environments—specifically deaf children of hearing parents who lacked early access to either spoken language or conventional sign language, communicating instead via idiosyncratic “home-sign” systems. While these children demonstrated intact core knowledge (they understood object mechanics, approximate numbers, and geometry normally), they struggled with tasks requiring the cross-domain combinatorial synthesis of geometry and landmarks, or the execution of exact large-number arithmetic. Once these individuals acquired a formal, structured sign language (such as American Sign Language or Nicaraguan Sign Language) complete with spatial prepositions and count words, their performance on cross-domain tasks improved rapidly. Syntax provides the representational glue that turns core modules into integrated human thought.

9.3 Comparative Cognitive Limits: Why Non-Human Primates Lack Flexibility

This linguistic bridging hypothesis provides a clear explanation for the cognitive limits observed in non-human primates. In extensive comparative studies testing chimpanzees, bonobos, and rhesus macaques on tasks derived from infant research, Spelke and her colleagues discovered that non-human primates consistently demonstrate performance on par with preverbal human infants within any isolated core domain. A chimpanzee understands object cohesion and continuity; it can estimate approximate numerical ratios using its ANS; it reorients using geometric enclosures; and it tracks the goals and eye gaze of intentional agents.

Where the non-human primate reaches an absolute cognitive ceiling is the point where a task requires the productive, compositional integration of two distinct core domains. A chimpanzee can use the geometric properties of an enclosure to reorient, and it can learn that a red box contains food, but it cannot flexibly combine these two representational structures into an integrated representation to find food hidden “at the corner with the short wall that is red.” The non-human primate mind lacks the recursive, compositional syntax required to conjoin domain-specific representations across encapsulated modules.

Despite decades of primate language experiments attempting to train chimpanzees and bonobos to use lexigrams or manual signs, non-human primates do not develop true recursive grammar or compositional syntax. Their use of signs remains primarily instrumental, associative, and ungrammatical. Without an internal compositional syntax, the primate conceptual landscape remains a collection of isolated core domains. The cognitive flexibility of our species—our ability to generate an infinite variety of novel scientific, poetic, and mathematical thoughts—is not due to superior general intelligence, but to natural language acting as a combinatorial engine across ancient core knowledge systems.

10. Theoretical Clashes: Spelke in Discourse with Contemporary Cognitive Science

10.1 Spelke Versus Connectionism and Radical Empiricism

Throughout the cognitive revolution and into the modern era of deep learning and generative artificial intelligence, Spelke has remained a prominent critic of radical empiricism and unconstrained connectionist architectures. From the parallel distributed processing (PDP) models of the 1980s to modern transformer neural networks containing hundreds of billions of parameters, radical empiricists have maintained that human-level intelligence can emerge from homogeneous, domain-general architectures exposed to massive volumes of sensory data via statistical optimization algorithms.

Spelke has countered these claims through both theoretical critiques and empirical demonstrations centered on the **poverty of the stimulus**. She points out that human infants achieve their core conceptual competence after mere months of visual exposure, operating on noisy, ambient, real-world sensory inputs without the millions of labeled trials or internet-scale text corpora required by contemporary deep learning models. A three-month-old infant needs only minimal visual experience to understand that solid objects do not pass through one another. When unconstrained, domain-general statistical learners are tested on identical physical reasoning problems, they routinely struggle with basic physical properties, hallucinating impossible collisions, discontinuous teleports, and incoherent physical transformations.

Spelke argues that deep learning models fail to build robust commonsense representations because they lack the innate inductive constraints of the human cognitive architecture. When an artificial neural network is trained without domain-specific priors for object cohesion, continuity, and agentive teleology, it learns superficial statistical correlations across pixel manifolds rather than underlying causal models of the physical world. For Spelke, statistical association and machine learning are powerful cognitive mechanisms, but they can only generate robust, human-like intelligence when operating over the foundation of an innate, core conceptual architecture.

10.2 The Debate with Susan Carey: Quinian Bootstrapping

One of the most nuanced and productive intellectual debates in modern developmental psychology occurred between Elizabeth Spelke and the eminent cognitive developmentalist Susan Carey. While both scholars are prominent defenders of nativism who reject Piaget’s sensorimotor framework, they diverge fundamentally on the question of **conceptual continuity versus conceptual change** in human development.

Carey argues for a process she terms **Quinian bootstrapping**. According to Carey, early infant core representations are fundamentally limited, sensorimotor, and incommensurate with adult concepts. Through development, children undergo true conceptual revolutions—analogous to paradigm shifts in the history of science—wherein old concepts are abandoned and entirely new representational systems with greater expressive power are constructed. For instance, Carey argues that the concept of natural number ($1, 2, 3…$) is not merely a combination of the ANS and the OTS; it is a genuinely new conceptual format that cannot be reduced to its predecessor systems. In Carey’s model, linguistic placeholder structures allow the child to slowly build new concepts that transcend the initial core architecture.

Spelke, by contrast, defends the **Conceptual Continuity Thesis**. She contends that core conceptual systems are never replaced, abandoned, or fundamentally rewritten. Rather than undergoing incommensurate conceptual revolutions, human cognitive development proceeds through the **enrichment, extension, and compositional combination** of permanently enduring core representations. For Spelke, the human concept of number does not require an ungrounded semantic placeholder to bootstrap itself into existence; it is constructed through the literal syntactic composition of the OTS and the ANS, orchestrated by natural language. Spelke emphasizes that adult cognitive errors across intuitive physics, spatial navigation, and mathematics reflect the persistent, unaltered operation of these foundational core systems, proving that the human conceptual foundation remains continuous throughout our lives.

10.3 Dialogue with Developmental Robotics and Modern AI

In recent years, Spelke’s empirical paradigms have expanded beyond developmental psychology into artificial intelligence, machine learning, and developmental robotics. As contemporary artificial intelligence has advanced through large language models (LLMs) and computer vision systems, researchers have encountered a persistent problem: while these systems can write fluent essays and generate photorealistic images, they frequently lack basic **physical common sense**. An LLM may pass a standardized university bar examination, yet fail to predict whether a stack of unbalanced blocks will topple or understand that a cat cannot pass through a closed door.

Leading figures in artificial intelligence, including Yann LeCun, Josh Tenenbaum, and external research teams at organizations like DeepMind and the Allen Institute for AI, have turned directly to Spelke’s Core Knowledge framework to address these vulnerabilities. Researchers have developed benchmark suites—such as the IntPhys (Intuitive Physics) benchmark and the BabyLM challenge—designed explicitly around Spelke’s Violation-of-Expectation tests. These platforms evaluate whether an artificial vision model exhibits surprise (measured via reconstruction error or loss spikes) when shown videos violating cohesion, continuity, or contact mechanics.

Spelke has engaged in this dialogue by advocating for a hybrid computational architecture in AI that integrates deep learning with explicit core knowledge priors. She argues that purely autoregressive, text-based LLMs will never achieve artificial general intelligence (AGI) through text prediction alone, because natural language in humans evolved as an external combinatorial medium operating over pre-existing core knowledge representations of physical objects, space, and agency. Without anchoring artificial neural systems in physical and social commonsense engines modeled on the human infant mind, artificial agents will remain brittle pattern matchers devoid of true understanding.

11. Translational Science: Educational Interventions and Policy

11.1 Applying Core Knowledge to Early Childhood Mathematics

Elizabeth Spelke’s scholarship has extended beyond theoretical debates and laboratory experiments into translational science, generating targeted pedagogical interventions designed to improve early childhood education. Recognizing that formal elementary school mathematics is built upon the dual foundations of the Approximate Number System and the spatial geometry engine, Spelke sought to determine whether targeted cognitive training of these core systems could improve children’s formal mathematical achievement.

Collaborating with developmental psychologists, educational designers, and software engineers, Spelke designed evidence-based preschool games that target intuitive, non-symbolic numerical and geometric representations. These games do not drill flashcards, recite multiplication tables, or teach rote algorithms; instead, they immerse preschool children in playful, intuitive tasks:

• Non-symbolic ANS Games: Children compare large arrays of moving dots without counting, rapidly deciding which of two energetic characters has gathered more food, thereby exercising their ratio acuity on the analog mental number line.
• Intuitive Geometric Games: Children explore geometric shapes, identifying parallelism, collinearity, metric angles, and rotational symmetry across various visual mazes, tapping directly into the core navigation and layout system.

The objective of these interventions is to build an intuitive, non-symbolic cognitive foundation before children are introduced to formal mathematical symbols and equations. When an abstract symbol like “8” is introduced to a child whose ANS acuity and spatial representations have been enriched through practice, the symbol immediately attaches to a robust, intuitive mental magnitude, rather than being memorized as an ungrounded, arbitrary mark on a page.

11.2 Randomized Controlled Trials in Global Educational Settings

Rather than confining these educational interventions to university laboratory preschools, Spelke and her collaborators submitted their core-knowledge pedagogy to testing via large-scale randomized controlled trials (RCTs). In a landmark, multi-year global initiative, Spelke partnered with the Nobel laureate economists Esther Duflo and Abhijit Banerjee, alongside the Abdul Latif Jameel Poverty Action Lab (J-PAL) and the Pratham education initiative, to evaluate these programs across low-income communities in Delhi, India.

The intervention involved over 1,500 preschool children enrolled in urban slums, randomized into three distinct groups:

1. An experimental group that played core-knowledge games exercising non-symbolic numbers and spatial geometry for several hours a week.
2. An active control group that played language and social games of identical duration and interactive engagement.
3. A passive control group that followed the standard, existing preschool curriculum.

The results, published in high-profile policy and economic journals, demonstrated that children in the core-knowledge intervention group made substantial gains in non-symbolic numerical estimation and spatial geometric reasoning compared to their peers. Crucially, when these children transitioned into formal primary schools months later, their gains generalized to **formal, symbolic mathematics**: they were significantly more proficient at counting, identifying printed digits, comparing symbolic numbers, and solving basic arithmetic addition and subtraction problems. By translating basic developmental science into evidence-based educational tools, Spelke demonstrated that interventions targeting core knowledge can help narrow socioeconomic readiness gaps before primary education begins.

11.3 Rethinking STEM Pedagogy in Light of Cognitive Architecture

Spelke’s empirical findings have contributed to a fundamental reassessment of STEM (Science, Technology, Engineering, and Mathematics) pedagogy throughout primary, secondary, and undergraduate curricula. Traditional science education has often operated on the assumption that students enter the classroom as blank slates ready to absorb scientific definitions, or that their conceptual frameworks must be dismantled through aggressive confrontation with counter-evidence. Spelke’s work proves that students enter every science classroom equipped with an active, deeply entrenched system of core physical and mechanical intuitions that operates automatically beneath reflective awareness.

This insight explains why physics students encounter persistent difficulties when learning Newtonian mechanics. A student’s intuitive Core System 1 operates on principles of contact and impetus: when an object moves, the core system expects that it must have been pushed, and that its motion will naturally decay as its internal “impetus” wanes. In direct contrast, Newton’s First Law of Motion states that an object in motion will continue in motion along a straight line indefinitely unless acted upon by an unbalanced external force. Because the core system’s intuitions conflict with Newtonian mechanics, students frequently revert to core impetus models when solving physics problems outside of memorized algorithmic exercises.

To address this challenge, Spelke advocates for pedagogical designs that explicitly align abstract scientific concepts with the cognitive constraints of the human mind:

• Educators must recognize that core intuitions cannot simply be erased or overwritten; they persist throughout the human lifespan and must be consciously managed.
• Pedagogy should help students build explicit **metacognitive awareness** of when their automatic core physical intuitions are operating, teaching them to contrast these intuitive shortcuts with formal mathematical and scientific models.
• Curricula should leverage core spatial and geometric intuitions through visual diagrams, physical analogies, and spatial representations, transforming the core systems into cognitive assets that support, rather than hinder, the acquisition of abstract STEM concepts.

12. Honors, Legacy, and the Future of Developmental Cognitive Science

12.1 Institutional Leadership and Scholarly Mentorship

Beyond her empirical and theoretical contributions, Elizabeth Spelke has played an enduring role in shaping the institutional and academic landscape of modern cognitive science. As the director of the Laboratory for Developmental Studies at Harvard University, Spelke created a vibrant research center that has served as a premier training ground for multiple generations of developmental psychologists, comparative cognitive scientists, and cognitive neuroscientists. Her laboratory has modeled interdisciplinary collaboration, bringing together developmentalists, developmental roboticists, evolutionary biologists, and philosophers of mind.

Spelke’s legacy is reflected in the achievements of her students, postdoctoral fellows, and collaborators, many of whom have become leaders across the discipline of psychology. Scholars such as Katherine Kinzler, Kristin Shutts, Moira Dillon, Shari Liu, and Lindsey Powell have advanced Spelke’s framework, extending core knowledge research into the domains of social cognition, geometric mapping, intuitive psychology, and neuroimaging. Her mentorship has been characterized by empirical rigor, philosophical depth, and a commitment to testing theoretical hypotheses through controlled experimental designs.

Furthermore, Spelke has served as an intellectual bridge between developmental psychology and adjacent academic disciplines. Her collaborations with the linguist Noam Chomsky, the computational neuroscientist Josh Tenenbaum, the philosophers Jerry Fodor and Ned Block, and the developmentalist Susan Carey have helped integrate developmental science into the broader cognitive revolution. She demonstrated that studying preverbal infants is not an isolated specialty within applied psychology, but an indispensable methodology for solving fundamental problems regarding the nature, structure, and evolution of the human mind.

12.2 Major Honors and International Recognition

Throughout her career, Elizabeth Spelke has received the highest honors and accolades awarded within the behavioral, psychological, and cognitive sciences. Her groundbreaking work has earned her election to the most prestigious scientific bodies in the United States and internationally:

• Election to the **National Academy of Sciences** (2001), recognizing her role in establishing the empirical foundations of infant cognitive architecture.
• Election to the **American Academy of Arts and Sciences** (1997).
• Named one of *Time Magazine’s* “America’s Best” scientists in 2001, highlighting her role in the cognitive revolution.
• Awarded the **Jean Nicod Prize** (2009) in Paris, France—one of the world’s most prestigious honors in the philosophy of mind and cognitive science—delivering a historic lecture series titled *Natural Number and Natural Geometry*.
• Recipient of the **C.L. de Carvalho-Heineken Prize for Cognitive Science** (2016), awarded by the Royal Netherlands Academy of Arts and Sciences for her foundational work on the origins of human thought.
• Recipient of the Association for Psychological Science (APS) **William James Fellow Award**, honoring a lifetime of transformative contributions to the fundamental principles of psychology.

Her theoretical writings and empirical studies have accrued tens of thousands of academic citations, altering introductory psychology textbooks, graduate seminars, and developmental curricula worldwide. By transforming our understanding of the infant mind, Spelke established that human infants enter the world equipped with an elegant set of foundational conceptual tools.

12.3 Unsolved Problems and Future Frontiers in Core Knowledge Research

As Elizabeth Spelke’s framework enters its fifth decade of scholarly influence, developmental cognitive science looks toward emerging research frontiers to address remaining questions within the Core Knowledge paradigm. The first major frontier is the precise mapping of the **genetic, epigenetic, and neurobiological substrates** that generate core systems. While functional neuroimaging (fMRI, fNIRS) and high-density EEG have confirmed that infants recruit localized cortical networks when processing objects, numbers, and faces, the developmental genetics that construct these neural circuits during prenatal and neonatal development remain a critical area of investigation.

A second major frontier is the integration of Core Knowledge with **computational cognitive science and artificial general intelligence**. Ongoing work in probabilistic programming, Bayesian modeling, and neuro-symbolic artificial intelligence seeks to translate Spelke’s core principles—cohesion, continuity, contact, rational efficiency—into explicit mathematical and algorithmic architectures. By building artificial systems that integrate these innate, domain-specific priors with flexible learning mechanisms, researchers aim to overcome the vulnerabilities of modern deep learning and create artificial agents that match the commonsense reasoning of a human child.

Ultimately, Elizabeth Spelke’s legacy rests on her transformation of how we conceptualize the human mind. By bringing empirical precision to the classical debates of philosophy, she demonstrated that the human infant is neither an empty slate waiting to be inscribed by experience, nor a passive reflex loop trapped within a sensorimotor stage. Instead, the human newborn arrives as an endowed, active, and rational thinker, equipped with an evolutionary inheritance of core knowledge systems that anchor our understanding of the world, empower our creativity, and unite humanity through a shared cognitive architecture.

Conclusion

Elizabeth Spelke’s research has redefined our understanding of human development and the origins of thought. Over five decades of empirical research, she has moved the debate on cognitive architecture from philosophical speculation into controlled, replicable science. By creating innovative looking-time and cross-modal paradigms, Spelke proved that preverbal infants understand the basic structure of the physical and social world—evaluating the continuous trajectories of physical objects, estimating approximate numerical quantities, navigating metric spatial layouts, interpreting the intentional goals of agents, and categorizing social partners based on communicative cues long before the onset of speech or coordinated movement.

The Core Knowledge framework stands as a lasting contribution to modern cognitive science, resolving historic dichotomies between radical empiricism and abstract rationalism. Spelke has demonstrated that human cognition relies on an evolved architecture composed of phylogenetically ancient, domain-specific modules. These innate core systems do not vanish as we grow, nor do they restrict our intellectual development; instead, they serve as the foundational computational substrate over which our cultural tools, scientific theories, and symbolic languages operate. When natural language weaves these domain-specific modules together, the human mind achieves the cognitive flexibility that defines our species.

From laboratory cribs in Massachusetts to early childhood classrooms in Delhi, Elizabeth Spelke’s work continues to reshape developmental diagnostics, educational curricula, and artificial intelligence design. Her legacy reminds us that to understand the heights of human intellect—our science, mathematics, technology, and art—we must look to its origins in the infant mind. In the gaze of a three-month-old infant, Elizabeth Spelke revealed the enduring foundations of human thought.

References

• Carey, S. (2009). The Origin of Concepts. Oxford University Press. https://doi.org/10.1093/acprof:oso/9780195367638.001.0001
• Dillon, M. R., Kannan, H., Dean, J. T., Spelke, E. S., & Duflo, E. (2017). Cognitive science in the field: A preschool intervention to train geometric and numerical abilities in rural India. Science, 357(6346), 47–55. https://doi.org/10.1126/science.aal4724
• Gergely, G., Nádasdy, Z., Csibra, G., & Bíró, S. (1995). Taking the intentional stance at 12 months of age. Cognition, 56(2), 165–193. https://doi.org/10.1016/0010-0277(95)00661-H
• Hamlin, J. K., Wynn, K., & Bloom, P. (2007). Social evaluation by preverbal infants. Nature, 450(7169), 557–559. https://doi.org/10.1038/nature06288
• Hermer, L., & Spelke, E. S. (1994). A geometric module for surface layout properties in children. Nature, 370(6484), 57–59. https://doi.org/10.1038/370057a0
• Hermer, L., & Spelke, E. S. (1996). Modularity and development: The case of spatial reorientation. Cognition, 61(3), 195–232. https://doi.org/10.1016/S0010-0277(96)00714-7
• Hermer-Vazquez, L., Spelke, E. S., & Katsnelson, A. S. (1999). Sources of flexibility in human cognition: Dual-task studies of space and language. Cognitive Psychology, 39(1), 3–36. https://doi.org/10.1006/cogp.1998.0713
• Kinzler, K. D., Dupoux, E., & Spelke, E. S. (2007). The native language of social cognition. Proceedings of the National Academy of Sciences, 104(30), 12577–12580. https://doi.org/10.1073/pnas.0705345104
• Piaget, J. (1954). The Construction of Reality in the Child. Basic Books. https://doi.org/10.1037/11168-000
• Shutts, K., Kinzler, K. D., McKee, C. B., & Spelke, E. S. (2009). Social information guides infants’ selection of foods. Journal of Cognition and Development, 10(1–2), 1–17. https://doi.org/10.1080/15248370902966636
• Spelke, E. S. (1976). Infants’ intermodal perception of events. Cognitive Psychology, 8(4), 553–560. https://doi.org/10.1016/0010-0285(76)90018-9
• Spelke, E. S. (1990). Principles of object perception. Cognitive Science, 14(1), 29–56. https://doi.org/10.1207/s15516709cog1401_3
• Spelke, E. S. (2000). Core knowledge. American Psychologist, 55(11), 1233–1243. https://doi.org/10.1037/0003-066X.55.11.1233
• Spelke, E. S. (2003). What makes us smart? Core knowledge and natural language. In D. Gentner & S. Goldin-Meadow (Eds.), Language in Mind: Advances in the Study of Language and Thought (pp. 277–311). MIT Press. https://doi.org/10.7551/mitpress/4117.003.0016
• Spelke, E. S., Breinlinger, K., Macomber, J., & Jacobson, K. (1992). Origins of knowledge. Psychological Review, 99(4), 605–632. https://doi.org/10.1037/0033-295X.99.4.605
• Spelke, E. S., & Kinzler, K. D. (2007). Core knowledge. Developmental Science, 10(1), 89–96. https://doi.org/10.1111/j.1467-7687.2007.00569.x
• Spelke, E. S. (2022). What Babies Know: Knowledge and Wonder in the Natural Child. Oxford University Press. https://doi.org/10.1093/oso/9780190051600.001.0001
• Xu, F., & Spelke, E. S. (2000). Large number discrimination in 6-month-old infants. Cognition, 74(1), B1–B11. https://doi.org/10.1016/S0010-0277(99)00066-9