Core Knowledge Systems in Human Cognition
Historical Paradigms in Cognitive Science vs. Core Knowledge Framework
Traditional Cognitive Science Views:
General-Purpose Learning View: Conceives the human mind as a highly flexible, adaptable mechanism designed to detect regularities in experience through a single, domain-general learning system (Locke, 1689; Hume, 1748). This approach was modernly re-energized by cognitive psychologists and neural network connectionist theorists (Rumelhart & McClelland, 1985; Hinton, 1993).
Massively Modular / Evolutionary Psychology View: Conceives the mind as a vast collection of distinct, special-purpose mechanisms, each selected through evolutionary processes to solve specific adaptive problems (Darwin, 1871; Cosmides & Tooby, 1994; Pinker, 2002; Fodor, 2000).
Societal Framing & Concerns: These opposing perspectives frame debates on mathematical achievement (effort vs. inherited talent), youth violence (learned community aggression vs. evolved intergroup competition predispositions), and the general extent of human agency and self-determination.
The Core Knowledge Alternative:
Empirical findings in developmental science refute both extreme paradigms: humans possess neither a uniform, single general-purpose learning system nor thousands of hyper-specialized evolutionary modules.
Human cognition is instead built upon a small set of separable, early-emerging core knowledge systems.
These core systems serve as the foundational scaffolding upon which domain-general flexible skills, complex belief systems, natural language, and formal education are constructed.
Defining Criteria of Core Systems:
Deep roots in human phylogeny (shared with non-human primates and other animals) and ontogeny (present in early human infancy).
Centered on explicit domain principles that individuate entities and generate inferences about entity behavior.
Bounded by persistent signature limits that identify the operation of the system across tasks, developmental stages, species, and human cultures.
Core System 1: Inanimate Object Representation
Core Principles of Object Representation:
Cohesion: Objects move as connected, bounded wholes (Aguiar & Baillargeon, 1999; Spelke, 1990).
Continuity: Objects move on continuous, unobstructed paths through space and time (Spelke, 1990).
Contact: Objects interact through physical contact and do not act upon one another at a distance (Leslie & Keeble, 1987).
Innate and Early Ontogenetic Origins:
These spatio-temporal principles allow infants and animals to perceive object boundaries, infer full shapes of partially or fully occluded objects, and predict trajectories and resting points.
Present without visual experience: observed in newborn human infants (Valenza, Leo, Gava & Simion, in press) and newly hatched chicks (Regolin & Vallortigara, 1995; Lea, Slater & Ryan, 1996).
Signature Limits of Object Representation:
Set Size Limit: Human infants can track and represent only a small number of distinct objects simultaneously, capped at approximately objects (Feigenson & Carey, 2003).
Lack of Subcategory Specialization: Infants with months of visual experience lack dedicated cognitive subsystems for ecological subcategories like foods or artifacts (Shutts, 2006).
Inapplicability to Non-Objects: The system fails to represent non-cohesive, inanimate non-object entities such as sand piles or liquids as discrete individuals (Huntley-Fenner, Carey & Solimando, 2002; Rosenberg & Carey, 2006; Shutts, 2006).
Phylogenetic Roots and Adult Persistence:
Adult non-human primates possess the identical core system, adhering to continuity and contact constraints (Santos, 2004) and exhibiting a set size limit of objects (Hauser & Carey, 2003).
Visual object-directed attention in human adults is governed by this same system, enabling adults to track separately moving objects provided they follow cohesion and continuity rules (Scholl, 2001; Scholl & Pylyshyn, 1999; vanMarle & Scholl, 2003; Scholl, Pylyshyn & Feldman, 2001; Marino & Scholl, 2005).
When human adults acquire domain-specific knowledge about tools or foods (Keil, Smith, Simons & Levin, 1998; Lavin & Hall, 2001; Santos, Hauser & Spelke, 2001), high attentional load causes finer categorical distinctions to fail while core spatio-temporal object constraints persist (Leslie, Xu, Tremoulet & Scholl, 1998).
Cross-Cultural Universality:
Studies of the Pirahã, an isolated Amazonian population whose language lacks count words beyond 'two', lacks markers for past versus present, and lacks recursive syntactic structures (Everett, 2005), show that members still discriminate objects from non-objects (Everett, 2005) and track objects constrained by the exact same signature set-size limit (Gordon, 2004).
Core System 2: Agents and Goal-Directed Action
Core Principles Governing Agents:
Agent representations bypass spatio-temporal object constraints: agents do not need to be cohesive (Vishton, Stulac & Calhoun, 1998), move on continuous paths (Kuhlmeier, Bloom & Wynn, 2004), or make physical contact to interact with other agents (Spelke, Phillips & Woodward, 1995).
Goal-Directedness: Actions of agents are interpreted as directed toward specific target goals (Woodward, 1999).
Efficiency (Teleological Constraint): Goal attainment is expected to occur via efficient, cost-minimizing means (Gergely & Csibra, 2003).
Contingency & Reciprocity: Agents are identified through contingent interaction (Watson, 1972; Johnson, Booth & O'Hearn, 2001) and reciprocal gestures (Meltzoff & Moore, 1977).
Gaze Direction: Perceptible faces are not mandatory for agent identification (Johnson, Slaughter & Carey, 1998; Gergely & Csibra, 2003); however, when faces are present, gaze direction is utilized to interpret social and non-social actions even by human newborns (Hood, Willen & Driver, 1998; Johnson et al., 1998; Farroni, Massaccesi, Pividori & Johnson, 2004).
Inanimate Distinction:
Infants do not interpret movements of inanimate objects as goal-directed (Woodward, 1998) and do not attempt to mirror or imitate non-agent object motions (Meltzoff, 1995).
Phylogenetic and Neural Foundations:
Newly hatched chicks, rhesus macaques, and chimpanzees infer the visual perspective and line-of-sight of competitors or predators (Agrillo, Regolin & Vallortigara, 2004; Flombaum & Santos, 2005; Hare, Call & Tomasello, 2001).
Supported by premotor mirror neurons observed in captive monkeys, which discharge both when performing an action and when observing another agent perform that action (Rizzolatti, Fogassi & Gallese, 2002).
Mirroring behavior and homologous neural activity occur in human adults (Iacoboni et al., 1999), and representations of goal-directed agency structure intuitive moral judgments regarding harm (Cushman, Young & Hauser, in press).
Core System 3: Numerical Representation
Theoretical Models and Core Properties:
Theoretical formulations include accumulator and neuronal models (Dehaene & Changeux, 1993; Meck & Church, 1983; Church & Broadbent, 1990; Izard & Dehaene, 2006; Gallistel & Gelman, 1992).
Property 1: Imprecision and Ratio Dependence: Core number representations are non-exact, exhibiting scalar variability where imprecision scales linearly with cardinal magnitude. This produces a ratio-limit on set discriminability (Izard, 2006).
Property 2: Domain Abstractness: Representations apply across visual arrays, auditory sequences, and produced motor actions.
Property 3: Combinability: Sets can be compared, added, and subtracted.
Ontogenetic and Phylogenetic Evidence:
Infants discriminate large quantities of objects, actions, and sounds under continuous quantity controls via ratio limits (Xu & Spelke, 2000; Xu, Spelke & Goddard, 2005; Wood & Spelke, 2005; Lipton & Spelke, 2003, 2004; Brannon, Abbott & Lutz, 2004) and perform non-symbolic addition and subtraction (McCrink & Wynn, 2004).
Adult monkeys and humans discriminate large auditory sequences subject to ratio constraints (Hauser, Tsao, Garcia & Spelke, 2003; Barth, Kanwisher & Spelke, 2003) and compute large-number addition/subtraction (Flombaum, Junge & Hauser, 2005).
Cross-modal numerical comparisons in human adults and children are as accurate as intra-modal comparisons (Barth et al., 2003; Barth, La Mont, Lipton & Spelke, 2005).
Developmental Precision Scaling:
Numerical precision sharpens over human development, moving from a required ratio of in \text{-}month-old infants to a ratio of in human adults (van Oeffelen & Vos, 1982; Izard, 2006).
Cross-Cultural Universality:
The Mundurukú (Amazonian indigene group lacking formal education, verbal counting routines, or exact number words beyond 'three') discriminate large approximate numerosities with a ratio limit equal to educated French adults (Pica, Lemer, Izard & Dehaene, 2004).
Mundurukú adults and uninstructed US preschool children perform non-symbolic approximate addition and subtraction on visual object arrays and auditory sequences (Pica et al., 2004; Barth et al., 2005; Barth, LaMont, Lipton, Dehaene, Kanwisher & Spelke, 2006).
Core System 4: Spatial Layout and Geometry
Core Principles:
Represents distance, angle, and sense (left/right) relationships among extended environmental surface layouts.
Fails to capture surface features (color, texture, odor) and fails to represent spatial geometry among movable arrays of discrete objects.
Reorientation Behavior:
When disoriented, young children and animals reorient using layout geometry exclusively (Hermer & Spelke, 1996; Cheng, 1986; Cheng & Newcombe, 2005).
Children fail to reorient using the geometry of a discrete object array (Gouteux & Spelke, 2001) and fail to use array geometry to locate objects when oriented while the array moves (Lourenco, Huttenlocher & Vasilyeva, 2005).
Landmark Memory Dual-Process Model:
Under disorientation, animals and young children frequently fail to use non-geometric landmarks (e.g., a single colored wall) for reorientation (Margules & Gallistel, 1988; Wang, Hermer & Spelke, 1999; Lee, Shusterman & Spelke, in press).
Landmark utilization relies on two independent systems: an encapsulation reorientation module sensitive purely to geometry, and an associative memory process linking localized areas to explicit objects (Cheng, 1986; Lee et al., in press).
Adult and Cross-Cultural Evidence:
Human adults fall back on geometric reorientation when performing spatial tasks under verbal or spatial shadowing interference (Hermer-Vazquez, Spelke & Katsnelson, 1999; Newcombe, 2005).
Uneducated Mundurukú indigenes spontaneously extract and utilize geometric relationships in 2D graphic maps and 3D extended environments (Dehaene, Izard, Pica & Spelke, 2006).
Core System 5: Candidate System for Social Partners and Group Membership
Theoretical Rationale:
Humans exhibit universal predispositions to form coalitions, categorize individuals into in-groups vs. out-groups, and display automatic preferences for the in-group ('us' vs. 'them') (Cosmides & Tooby, 2003; Baron & Banaji, 2006).
Visual Race and Gender Preferences in Infancy:
\text{-}month-old infants display visual looking preferences for faces belonging to their own race (Kelly et al., 2005; Bar-Haim, Ziv, Lamy & Hodes, 2006).
This preference is shaped by social exposure: Caucasian Israeli infants prefer Caucasian faces; African Ethiopian infants prefer African faces; African Ethiopian infants living in predominantly Caucasian Israeli society show no race preference (Bar-Haim et al., 2006).
Infants prefer looking at faces matching the gender of their primary caregiver (Quinn, Yahr, Kuhn, Slater & Pascalis, 2002).
Language and Accent as the Primary Evolutionary In-Group Cue:
In human evolutionary environments, contact with distinct racial groups was rare, and both genders coexisted in every community; however, spoken dialect and accent varied significantly between adjacent social groups (Braudel, 1988; Cosmides & Tooby, 2003).
Newborn infants prefer listening to their native language over foreign languages (Mehler, Jusczyk, Lambertz & Halsted, 1988; Moon, Cooper & Fifer, 1993).
\text{-}month-old infants presented with bilingual English/Spanish speakers look preferentially at the person who previously spoke their native language (Kinzler & Spelke, 2005).
\text{-}month-old infants in Boston presented with native English and French speakers offering different foods preferentially choose the food offered by the native English speaker (McKee, 2006).
Cognitive Extensions, Limitations, and Conceptual Change
Role of Core Knowledge in Uniquely Human Capabilities:
Language Acquisition: Word learning builds directly upon pre-existing core concepts (Bloom, 2000).
Map Navigation: Symbol and map understanding is built upon core spatial geometry representations (Dehaene et al., 2006).
Mathematics: Verbal counting and formal symbolic arithmetic build upon core approximate number representations (Wynn, 1990; Dehaene, 1997; Carey, 2001; Spelke, 2003; Feigenson, Dehaene & Spelke, 2004).
Cultural Transmission: Social partner representations enable cultural learning, transmitting specialized skills within cooperative human networks (Tomasello, 1999).
Cognitive Limitations and Social Maladaptations:
Core principles do not hold at extreme micro or macro scales: physical objects are not cohesive or continuous at quantum scales, space is non-Euclidean, and advanced mathematical concepts exceed core limits (McCloskey, 1983; Gelman, 1991; Hawking, 2002; Randall, 2005).
The evolved social partner system ('us' vs. 'them') produces modern intergroup prejudice, xenophobia, and violent conflict. Historical examples of language-based oppression include severing tongues of non-English speaking slaves in US history, executing Russian speakers post-Alaska purchase, and outlawing public German speech during World War II (Shell, 2001).
Malleability and Conceptual Change:
Core intuitions are resilient but can be overturned through conceptual change.
Scientific and Developmental Progress: Physicists construct models of higher-dimensional non-Euclidean space and massless particles (Hawking, 2002; Randall, 2005). Preschoolers alter their understanding of number when learning to count (Spelke, 2000) and alter agent concepts when learning biological processes like digestion and respiration (Carey, 1985, 2001).
Social Bias Moderation: Infant race preferences alter with environmental diversity exposure (Bar-Haim et al., 2006), and adult/child intergroup biases diminish through structured contact (Pettigrew & Tropp, 2006).