Psychology by BogdanovaThe Encyclopedia of Psychology for the Artificial Era

RESEARCH DOSSIER

Cognitive Development and Conceptual Change

Abstract. Cognitive development and conceptual change describe the formation, differentiation, integration, and revision of representational systems through which children recognize objects, reason about number, infer kinds, explain causes, coordinate logical relations, and learn scientific concepts. This fundamental research monograph synthesizes historical developmental theories, original infant and child experiments, accounts of core knowledge and conceptual construction, strategy-discovery and dynamic-systems models, and contemporary conceptual-change intervention evidence. It examines conservation and false-belief tasks, infant violation-of-expectation paradigms, categorical induction, biological and astronomical mental models, causal-learning experiments, and microgenetic methods. Competing explanations are evaluated through the distinction between behavioral performance and conceptual competence, with attention to task demand, sampling, measurement validity, independent transfer, replication, and durability. Recent systematic reviews and meta-analyses inform the interpretation of infant looking and classroom interventions. The synthesis distinguishes acquisition, differentiation, integration, and contextual regulation as analytically useful forms of development and identifies open problems concerning representational novelty, knowledge coexistence, cultural generalizability, and causal mechanisms of learning. Human–artificial educational interactions are presented as a future research context requiring independent empirical tests.

Keywords: cognitive development; conceptual change; developmental psychology; object permanence; number concepts; categorization; causal inference; conservation; core knowledge; mental models; scientific reasoning; conceptual change instruction; microgenetic methods; cognitive development theories

Research Object, Definition, and Scope

Cognitive development and conceptual change concern the origins, differentiation, coordination, and revision of knowledge structures as children encounter objects, quantities, living things, minds, causes, and institutionalized forms of explanation. A concept organizes inferences across encounters; conceptual change alters the relations, explanatory commitments, or representational resources through which those inferences are made. Development can involve the acquisition of a new label, refinement of a category, modification of a belief, or more extensive reconstruction of a system of related ideas. The central task is to identify which type of change occurred, at what developmental scale, under what conditions, and through which observable processes Carey (1985) Carey (2009).

The scientific object requires distinctions among perceptual discrimination, action-guiding expectations, verbal knowledge, explicit explanation, and flexible transfer. An infant who looks longer at an impossible event may differentiate its perceptual or physical structure without possessing an adult-like verbal concept of persistence. A child who recites a scientifically correct sentence may lack an explanatory model that supports novel predictions. Accordingly, converging evidence from multiple tasks, ages, and methods is essential when moving from observed performance to proposed conceptual competence. Research on early physical and numerical representations makes this inferential problem particularly clear Baillargeon (1987) Xu & Spelke (2000).

This monograph is a critical integrative synthesis of foundational books, original experiments, theoretical investigations, methodological analyses, and research reviews. Its organizing questions concern early representational resources, mechanisms that transform knowledge, domain-specific patterns of change, and the inferential strength of evidence. Literature is examined for task design, competing explanations, contextual influences, and generalizability. The review does not implement a preregistered systematic search or calculate a meta-analytic pooled effect. Its conclusions are analytic statements about the cited literature, with observed findings separated from causal hypotheses and broader theoretical interpretations Spelke & Kinzler (2007) Gopnik (2012).

Historical Formation of the Developmental Problem

Jean Piaget established a powerful research tradition by connecting children's errors on conservation, classification, seriation, and object-search tasks to the organization of reasoning. In this framework, development involved increasingly coordinated structures of action and thought, with assimilation and accommodation describing how experience enters an existing organization and modifies it. The important historical advance was methodological: children's answers became evidence about forms of reasoning rather than simply incomplete adult knowledge. Subsequent research retained this attention to developmental organization while investigating whether competence depended on domain, task demands, prior experience, or experimental presentation Carey (1985) Siegler (1996).

Research during the later twentieth century altered the interpretation of early ability. Looking-time studies, language-sensitive category experiments, and analyses of cognitive strategies revealed competencies that standard verbal or reaching tasks could obscure. The same infant may exhibit expectations in a gaze-based display while failing an action task, because remembering a hiding location and inhibiting a previous response impose additional requirements. Diamond's repeated assessment of infants on the A-not-B task showed continuously increasing tolerance for a retrieval delay, complicating any claim of a single abrupt transition in object representation Baillargeon (1987) Diamond (1985).

A second tradition studied scientific knowledge as an organized conceptual system. Carey's developmental analysis of biological knowledge, Posner and colleagues' account of accommodation in science learning, and Vosniadou and Brewer's work on children's models of Earth articulated change as a transformation of explanatory commitments. Parallel work on strategy discovery and dynamic systems questioned whether global stages accurately described the uneven, context-sensitive pathways through which new reasoning practices become stable. These traditions remain complementary sources of testable hypotheses, provided that their concepts are tied to appropriate behavioral evidence Carey (1985) Posner et al. (1982) Vosniadou & Brewer (1992) Thelen & Smith (1994).

Theoretical Units of Development

Theories differ in their proposed units of change. Structural developmental accounts describe broad coordinations of operations; core-knowledge proposals identify early domain-specific representational systems; theory-based approaches treat explanatory relations as the significant organization; information-processing models analyze attention, memory, representation, and strategy selection. These are partly different levels of description. A successful theory specifies whether it predicts an individual's response on one trial, a class of inferences within one domain, or the integration of concepts over years. Claims of conceptual discontinuity require evidence that a child's explanatory commitments changed, not merely that a test score improved Spelke & Kinzler (2007) Carey (2009) Siegler (1996).

The idea of core knowledge holds that early cognition is supported by representational systems specialized for such domains as objects, number, actions, and space. Their operations provide constraints and initial resources for learning, but explaining mature scientific concepts demands an account of how representations cross, connect, or exceed those initial systems. Carey describes conceptual development in part through the construction of representational resources with new expressive power. The hypothesis is stronger than the assertion that children learn more facts: it asks how later concepts make distinctions that earlier systems could not formulate Spelke & Kinzler (2007) Carey (2009).

Dynamic-systems and variability approaches locate development in changing relations among perceptual cues, motor activity, task history, available strategies, and environmental scaffolds. Siegler describes multiple strategies competing and changing in frequency within an individual, while Thelen and Smith analyze developmental stability as an emergent property of interacting processes. These accounts warn against treating every transition as a wholesale replacement of one coherent mental theory by another. An adequate synthesis must determine which changes occur at the level of conceptual content and which arise from shifting attention, memory demands, task conditions, or strategy selection Siegler (1996) Thelen & Smith (1994).

Early Representations of Physical Objects

Research on object permanence illustrates the difference between a child's explicit action and an expectation inferred from observation. Baillargeon's 1987 rotating-screen experiments compared infants' looking at physically possible and impossible events involving an occluded box. Longer attention to the impossible display among some young infants suggested expectations concerning object persistence and solidity that were difficult to reconcile with late-emerging object knowledge inferred from manual search alone. The experiment established discrimination under its particular visual conditions; its cognitive interpretation depends on controls for perceptual novelty, familiarization, and event structure Baillargeon (1987).

A-not-B errors provide a distinct measure. In the classic procedure, an infant repeatedly finds a hidden object at location A and later reaches there even after seeing the object hidden at B. Diamond's longitudinal observations related this error to the duration of a delay before response, showing that performance changed gradually as action-guiding recall improved. The result draws attention to coordination of object representations with response inhibition and memory maintenance. The two paradigms ask different questions: gaze-based expectation and successful goal-directed retrieval are overlapping but nonidentical expressions of cognitive organization Diamond (1985) Baillargeon (1987).

The physical-object domain also demonstrates why the apparent onset of knowledge depends on experimental design. Occlusion, containment, support, contact, and continuity can produce different predictions and different informational demands, and infants must interpret a display before any expectation can guide their looking. Treating all measures as interchangeable would hide these differences. A stronger developmental account compares the same cohort across tasks, varies perceptual information systematically, tests out-of-sample predictions, and identifies the circumstances under which an expectation generalizes from familiar demonstrations to novel physical arrangements Spelke & Kinzler (2007) Baillargeon (1987).

Number, Quantity, and the Construction of Exact Concepts

Numerical development includes several distinguishable achievements: approximate comparison of quantities, representations of small sets, counting procedures, the cardinal principle, and eventual understanding of exact integers. Xu and Spelke found that six-month-old infants discriminated large sets when their numerical ratio was sufficiently different, including eight versus sixteen but not eight versus twelve under the tested conditions. Such performance supports early approximate numerosity sensitivity, while remaining different from the capacity to name exact quantities or justify arithmetic operations. Ratio-dependent results also require careful control of area, density, contour length, and other continuous visual properties Xu & Spelke (2000).

Wynn's 1992 infant event experiments invited interpretations of early addition and subtraction, but the strength of a numerical conclusion depends on excluding alternative responses to perceptual configuration or object-tracking events. Later comparisons of small and large sets produced patterns that can be interpreted as evidence for partially distinct representational resources. Carey's account of the origin of concepts treats the relation between early quantities and exact natural-number concepts as a substantive problem of representational construction, where number words and counting practices offer structures unavailable in a purely approximate system Wynn (1992) Xu (2003) Carey (2009).

Counting develops within social and linguistic practices, yet acquisition is more than memorizing a stable verbal sequence. Children learn one-to-one tagging, stable order, and the connection between the final count word and cardinality; they then coordinate these procedures with object classification, comparison, and symbolic operations. Gelman and Gallistel's foundational analysis highlighted principled numerical capacities in the preschool period. A complete developmental explanation must reconcile task-specific competence, the role of teaching and language, and the distinction between understanding a counting procedure and deploying it consistently across unfamiliar sets Gelman & Gallistel (1978) Carey (2009).

Categories, Kinds, and Inductive Generalization

Category development concerns the movement from noticing shared surface features to making inferences about unseen properties, causal relations, and kind membership. Gelman and Markman's experiments pitted category labels against visual similarity, revealing how categorical information can guide young children's generalizations. The paradigm is important because categorization and induction are not identical performances. Recognizing that two items share a name does not automatically demonstrate that the child regards them as sharing a deep causal nature. Researchers must manipulate labels, appearance, and domain knowledge separately if they want to identify the grounds of a child's inference Gelman & Markman (1986).

Contrasting findings illuminate the problem. Sloutsky, Kloos, and Fisher taught four- and five-year-olds category membership for artificial animal-like stimuli and then compared their categorization with induction. Children used the acquired kind information in categorization but relied on perceptual similarity when generalizing new properties. The study demonstrates that the influence of labels and kinds depends on what the task requests, how categories were learned, and how inference is expressed. It also challenges claims that children's category-based generalization is fully stable across situations at a particular age Sloutsky et al. (2007).

A developmental account therefore needs multiple forms of evidence: sorting and labeling, property projection, explanations of why members belong together, transfer to new instances, and longitudinal persistence. Theories of kind-based induction propose expectations about causal or internal commonalities; similarity-based approaches emphasize learned attentional distributions and representational features. Rather than assign one mechanism to every domain, the evidence invites assessment of when a child relies on labels, perceptual features, or specific explanatory knowledge. Domain content and instructional context can moderate all three contributions Gelman & Markman (1986) Sloutsky et al. (2007) Carey (1985).

Developing Biological Concepts

Biological understanding is an especially productive case of conceptual change because children confront distinctions among life, movement, growth, illness, reproduction, and inherited properties. Carey's historical case study investigated development of biological knowledge in childhood and argued that explanatory reorganizations can change how children understand relations among animals, people, and bodily processes. The conceptual issue is whether a child acquires isolated biological facts or constructs a more integrated explanation of what living things are and why they undergo characteristic changes. The distinction can be examined through prediction and explanation tasks rather than definitions alone Carey (1985).

Young children's tendency to understand natural phenomena in functional or purposive terms introduces a further complication. Kelemen analyzed the scope of teleological explanations and proposed that children's interpretation of living things may be shaped by an understanding of goals and designed artifacts. Explaining an animal feature by its function can support useful prediction in some cases while suggesting mistaken explanations in others. A research design must distinguish an answer encouraged by the wording of a question from a stable underlying belief, and separate knowledge of biological function from an assumption about intentional creation Kelemen (1999).

Biological conceptual development connects perception, analogy, testimony, and cultural instruction. Children observe organisms changing, hear adults name kinds and explain invisible processes, and compare familiar cases such as humans or household animals with unfamiliar ones. The emerging organization is tested when they predict properties of novel organisms or revise an explanation after encountering counterexamples. A rigorous study samples different kinds, asks for reasons and transfer, and tracks whether the child continues to deploy the new relation outside the instructional setting. General claims about innate biological theories require evidence beyond one specially framed question Carey (1985) Carey (2009).

Causal Learning and Intervention

Causal development concerns inference about what produces an event and what would happen if a candidate cause were changed. Gopnik, Sobel, Schulz, and Glymour introduced young children to a novel apparatus, the blicket detector, in which some objects activated an outcome. Across three studies, two-, three-, and four-year-olds drew inferences from patterns of variation and covariation. The research showed that young children can reason about causal relationships under experimentally controlled conditions that avoid relying solely on explicit accounts of everyday physical mechanisms Gopnik et al. (2001).

Covariation, intervention, and explanation impose distinct epistemic demands. A repeated association may make one cause likely, yet a child's decision to intervene can supply further information about the causal structure. Research on early scientific thinking describes children as capable of informal experimentation, using statistical evidence, observing other people's actions, and revising causal hypotheses. Such capacities provide plausible learning mechanisms for conceptual change, although precise claims about Bayesian computation require process-level predictions rather than a general analogy between children and scientists Gopnik (2012) Gopnik et al. (2001).

The most informative causal-development experiments contrast alternative generative structures that predict different outcomes. Researchers can compare mere exposure with opportunities for intervention, arrange ambiguous and diagnostic evidence, and test whether new causal beliefs transfer to untrained objects. Children's preferences may also depend on perceived intentionality, prior domain expectations, and how a novel machine is described. These design features matter because task success could reflect local associative learning, inference about hidden kinds, or the reconstruction of a more general causal model. A causal claim becomes stronger when the experiment distinguishes these explanations prospectively Gopnik et al. (2001) Kelemen (1999).

Analogy and Relational Restructuring

Analogical comparison allows a learner to map structured relations from a familiar domain to an unfamiliar one. Gentner's structure-mapping account gives priority to correspondences among relations over superficial similarity of isolated attributes. For conceptual development, this framework helps explain how two concrete examples can support a more abstract relation such as balance, containment, transfer, or causal dependency. A child may initially attend to shared appearances and later use explicit comparison to coordinate a common relational structure across dissimilar objects Gentner (1983).

Analogy becomes a mechanism of conceptual change when comparison reveals a relation that reorganizes subsequent inferences, rather than merely making one problem easier. Evidence of change includes spontaneous use of the relation in a new setting, the ability to explain its applicability, and appropriate rejection of misleading analogies. Instructional demonstrations can also create surface-driven associations that disappear when the examples change. Consequently, experiments must test long-distance transfer and reasoning about the limits of an analogy in addition to performance during instruction Gentner (1983) Posner et al. (1982).

The relation between analogical mapping and other learning mechanisms is a substantive theoretical question. Structured comparison may direct attention toward previously neglected dimensions, provide relational predicates that can be expressed in language, or help learners coordinate competing models. These pathways can operate together with category learning and causal experimentation. Developmental evidence should examine which knowledge is available before the comparison, what is learned during it, and whether a representational change persists. The broader implication is that an effective explanatory model specifies the intermediate operations by which experience becomes a reusable conceptual relation Gentner (1983) Carey (2009).

Language, Symbols, and Explanatory Resources

Language offers children labels for grouping observations and propositions for coordinating relations. Numerical vocabulary, biological terms, causal connectives, and explanatory discourse create resources for representing distinctions that may be difficult to sustain through perception alone. Carey's analysis of exact numerical concepts treats linguistic and symbolic practices as important elements of representational construction. A label, however, can be learned ahead of its intended meaning, and fluent verbal performance may coexist with misconceptions revealed by prediction or intervention tasks Carey (2009) Gelman & Gallistel (1978).

Verbal instruction also changes the conditions under which a concept is measured. Explaining what a word means, choosing a picture, imitating a demonstrator, and generating a prediction involve different amounts of language production, memory, and pragmatic inference. Children's answers about Earth are a case in which everyday language, diagrams, and school explanations can induce apparently compatible sentences within substantially different mental models. Examining verbal claims alongside drawings and novel inference probes helps identify whether the intended scientific meaning has become operational Vosniadou & Brewer (1992) Nobes et al. (2005).

A developmental theory should describe language as a set of interacting practices rather than an isolated causal switch. Dialogic explanation supplies testimony, comparative expressions, requests for reasons, and standards for evaluating evidence. Children integrate these resources with perceptual patterns, prior concepts, and social expectations. Research must evaluate whether the child generalizes beyond the words used during training, whether communication has changed a prediction rather than just a response convention, and whether age-related language demands make one group appear less conceptually sophisticated than another Carey (2009) Gopnik (2012).

Conceptual Change as Representational Reorganization

Conceptual change is most demanding when learning alters a system of mutually supporting assumptions rather than correcting an isolated statement. Carey treated children's developing biological knowledge as a case in which the meanings and explanatory roles of concepts may be reconstructed. In later work she asked how conceptual resources with greater expressive power are formed from earlier representations. Research must therefore specify the prior structure, the candidate new relations, and the tasks that reveal a change in what the learner can infer. Evidence of knowing a new term is insufficient if explanations and predictions still follow an earlier conceptual system Carey (1985) Carey (2009).

Posner, Strike, Hewson, and Gertzog advanced an influential model of scientific conceptual accommodation centered on the learner's appraisal of existing and candidate explanations. Dissatisfaction with an existing account and the intelligibility, plausibility, and fruitfulness of an alternative were proposed as conditions affecting change. This proposal made instruction sensitive to learners' conceptual commitments. It also requires empirical qualification: increasing willingness to repeat an alternative may represent temporary compliance, while conceptual acceptance and durable transfer require independent evidence across contexts and follow-up assessments Posner et al. (1982).

A contemporary synthesis treats conceptual change as an empirical family of outcomes. Some episodes revise confidence in a proposition, some add a distinction within an existing domain, and others change the causal or categorical framework connecting many propositions. These outcomes may occur sequentially and remain incomplete. Classifying change requires observing what persists, which predictions alter, whether errors reappear under pressure, and how an individual handles previously unfamiliar cases. Such a model can accommodate both extended theoretical reconstruction and local adjustments in knowledge activation diSessa (1993) Shtulman & Valcarcel (2012) Carey (2009).

Mental Models of Earth and Astronomical Phenomena

Vosniadou and Brewer investigated elementary school children's explanations of Earth's shape and found responses that appeared inconsistent when examined one sentence at a time. Their analysis proposed that children employed a limited set of organized mental models that reconciled everyday experience with learned claims about a spherical Earth. Responses about edges, people standing on the surface, and what lies beneath the ground were interpreted as parts of synthetic explanations. This research exemplifies the importance of studying relations among answers rather than scoring individual statements as independent pieces of knowledge Vosniadou & Brewer (1992).

Related work on the day-night cycle examined how children coordinated models of Earth, Sun, and Moon with causal explanations for observable events. The results indicated that some children organized an account of day and night around a restricted set of explanatory patterns, including models integrating local visual experience and scientific information. The findings support a developmental account in which new knowledge is assimilated into prior constraints and gradually reorganized. They do not establish that every child follows one fixed intermediate sequence or that a verbally elicited model is a stable theory across tasks Vosniadou & Brewer (1994).

Competing interpretations emerged when researchers changed the available response formats and interview procedures. Nobes, Martin, and Panagiotaki investigated children's scientific knowledge of Earth with methods that prompted renewed scrutiny of the inference from apparently nonstandard answers to coherent alternative mental models. The central methodological question is whether task-specific wording, familiar artifacts such as globes, and the distinction between a local landscape and a planetary object explain some responses. Strong evidence combines open explanation, forced choice, generative prediction, and cross-method consistency Nobes et al. (2005) Vosniadou & Brewer (1992).

Knowledge in Pieces and Contextual Activation

A knowledge-in-pieces approach questions the assumption that learners always draw on an integrated naïve theory. DiSessa's analysis of physics learning describes intuitive knowledge as including relatively local elements that can be activated by particular contexts. Such elements may account for apparently contradictory answers without requiring a child to alternate between complete rival world models. The approach yields a different empirical agenda: identify features that cue an interpretation, trace the coordination of local resources, and examine how instruction reorganizes their use in explanation diSessa (1993).

Coherent mental-model and knowledge-in-pieces accounts direct attention to different units of analysis. A model account expects patterns of responses to cohere across related problems; a contextual-resource account expects variation as task framing and local causal features change. These are discriminable hypotheses when researchers predefine which questions should covary and which contextual manipulations should change answers. Without such predictions, both accounts can be used retrospectively to rationalize the same observations. Evidence must therefore include cross-task reliability, within-person response patterns, and tests beyond the materials used to formulate the theory Vosniadou & Brewer (1992) diSessa (1993).

The developmental synthesis need not assume that coherence is all-or-nothing. A learner may maintain stable relational commitments in one familiar domain while using fragmentary analogies in another; within the same domain, core assumptions may coexist with task-dependent interpretations. Researchers can model degrees of integration, examine how instructional practice changes activation probabilities, and measure whether children recruit a relation spontaneously under unfamiliar demands. This approach links theoretical accounts of conceptual structure to tractable experiments on explanation, memory, and transfer diSessa (1993) Siegler (1996).

Scientific Reasoning and Evidence Evaluation

Conceptual change depends partly on learning to treat evidence as a constraint on one's explanations. Research on children's scientific thinking examines hypothesis generation, controlled comparison, causal inference, and differentiation between a favored idea and observations that might refute it. Gopnik's review of early scientific thinking emphasizes preschoolers' abilities to learn from statistical patterns and informal interventions. Such achievements coexist with demands for guidance when evidence is ambiguous or when alternative hypotheses require coordinated testing. Scientific thinking is therefore a developmental set of practices rather than a single maturity threshold Gopnik (2012).

Kuhn, Amsel, and O'Loughlin examined developing strategies of scientific reasoning, asking how learners reconcile new data with prior belief. Their framework makes epistemic coordination central: a learner must recognize that an explanation and the evidence offered for it are distinguishable objects of thought. This capacity affects which data are sought, how contrary cases are interpreted, and when a belief is revised. Experimental tasks in this tradition can reveal whether a child selectively searches for confirmation, balances alternative explanations, or revises the relevant causal relation Kuhn et al. (1988).

Scientific-reasoning performance is shaped by domain familiarity, vocabulary, opportunities for repeated inquiry, and the structure of the task. One child may identify a controlled comparison with an explicitly presented contrast and fail to invent a suitable comparison independently. A study that only assesses a correct final answer misses the process by which evidence was selected and interpreted. Strong designs retain trial-level records, compare prompted and spontaneous reasoning, and test whether improved reasoning subsequently changes untrained conceptual explanations. Process and outcome evidence belong in the same developmental model Kuhn et al. (1988) Gopnik (2012).

Strategy Discovery and Microgenetic Change

Microgenetic research increases the density of observations during periods when a new procedure or concept is emerging. Siegler and Jenkins examined when children first discovered problem-solving strategies and how those strategies generalized beyond the initial situation. The significance of such work is temporal precision: a new strategy may appear sporadically, coexist with older procedures, and spread gradually across contexts. An age-group comparison would conceal this individual-level process and exaggerate the abruptness of a transition. Repeated trials instead expose variability that provides information about learning mechanisms Siegler & Jenkins (1989) Siegler (1996).

Strategy change and conceptual change should be analyzed together without assuming that they are equivalent. A child can learn a faster way to solve an arithmetic problem while preserving the same numerical understanding; a reconceptualization of cardinality may permit entirely new classes of strategies. To distinguish them, investigators need tasks that demand explanation, transfer, and prediction alongside procedural performance. The sequence of changing strategies can reveal when newly available conceptual relations become useful, although temporal coincidence alone does not identify which process caused which Siegler (1996) Carey (2009).

Dense repeated measurement can itself alter learning because practice, feedback, and investigator attention become part of the child's experience. Microgenetic studies should therefore describe the instructional opportunities created by testing, compare alternative practice schedules when possible, and model between-person variation rather than reporting only the average discovery trial. Their strongest contribution is a fine-grained account of emergence, competition, stabilization, and generalization. They provide a way to connect long-run developmental change to the particular encounters from which new cognitive organization is assembled Siegler & Jenkins (1989) Thelen & Smith (1994).

Conservation, Operations, and Logical Coordination

Conservation tasks ask whether a child recognizes that quantity remains invariant under a transformation that changes appearance without adding or removing the relevant material. In familiar liquid, substance, and number arrangements, a child must identify which dimension is preserved and resist an appealing but misleading perceptual cue. Piaget and Szeminska's account of the genesis of number linked such judgments to the coordination of classification, seriation, and operations, making conservation a test case for explaining the changing structure of thought. These tasks illuminate important relations among measurement, transformation, and justification, but success on any single arrangement cannot by itself establish mastery of a general logical stage Piaget & Szeminska (1941) Siegler (1996).

Task presentation matters. Pouring water into differently shaped containers requires attention to height and width, comprehension of the experimenter's question, memory for the original quantities, and interpretation of why the same question is asked twice. Numerical conservation removes some material properties while introducing spatial configuration and counting opportunities. A child's correct response may come from an experimentally induced counting strategy or a general principle of invariance; an incorrect response may arise from language or attention even when the principle is available in a different form. Cross-task transfer and explanation therefore provide stronger evidence of operational coordination than binary pass–fail classification Piaget & Szeminska (1941) Siegler & Jenkins (1989).

The central developmental problem remains theoretically productive: how do representations of an initial state, an intervening transformation, and a final state become coordinated so that an invariant relation governs inference? Children may arrive at a solution through counting, verbal rehearsal, perceptual comparison, or explicit reversal of the transformation. Researchers can systematically vary these routes and assess whether the resulting competence generalizes to materials, locations, and tasks that do not support the trained procedure. Conservation thus belongs within a broader study of conceptual relations and strategy development, with claims about stages requiring a pattern of independent evidence Piaget & Szeminska (1941) Carey (2009).

Mental-State Concepts and False-Belief Development

Children's understanding of beliefs introduces a particularly clear example of representational change. To predict another person's action when that person has false information, a child must coordinate the object's actual situation with the other person's informational perspective. Wellman, Cross, and Watson's meta-analysis of false-belief studies documented systematic developmental patterns across different task variants and demonstrated why age-related changes must be examined across procedures. The relevant concept of belief has an explanatory role: it permits predictions of actions on the basis of a representation that may diverge from reality Wellman et al. (2001).

Onishi and Baillargeon used a nonverbal violation-of-expectation design to examine whether 15-month-old infants anticipated an actor's behavior according to a true or false belief about an object's location. Their positive findings created an apparent discrepancy with standard verbal false-belief tasks in which many younger preschoolers fail. This discrepancy has more than one interpretation. It may reflect different response demands, partially distinct social-cognitive processes, or alternative readings of what looking-time evidence establishes. An adequate developmental account states which behavioral operations are required to infer representation of another's belief rather than relying solely on a task label Onishi & Baillargeon (2005) Wellman et al. (2001).

Theory-of-mind development also shows that conceptual capacities must be analyzed in relation to communication and social interaction. Understanding that a person can hold false information can support explanations, deception, instruction, and cooperative correction, but these activities involve language, pragmatics, inhibition, and contextual expectations. Assessing belief concepts through multiple action, verbal, and prediction tasks helps reveal how their use becomes more flexible. Links to other research dossiers on social cognition and mental-state understanding belong to the broader scientific map; this monograph examines false belief as one comparative instance of conceptual development Wellman et al. (2001) Onishi & Baillargeon (2005).

Representational Redescription and Reorganization

Karmiloff-Smith proposed a developmental perspective in which knowledge initially embedded in competent performance can later become available in more explicit, manipulable representations. Her representational-redescription account offers a different picture from a simple progression from absence of knowledge to presence of knowledge. A child may solve a task effectively while remaining unable to describe, intentionally modify, or recombine the procedures that support the solution. Reorganization can therefore increase accessibility and flexibility even when immediate accuracy was already high. This distinction helps connect procedural learning with conceptual abstraction Karmiloff-Smith (1992).

A model of redescription must specify its observable signature. If knowledge becomes more explicit, children may transfer it into a new task, explain formerly implicit constraints, deliberately alter a strategy, or coordinate separate domains. Yet explanation itself can add training, and performance on verbal reports is constrained by language ability. Strong research designs distinguish the acquisition of a rule from a change in its accessibility and from the development of skills for reporting it. The proposed mechanism is most informative when these outcome classes show different trajectories rather than being collapsed into a generic measure of understanding Karmiloff-Smith (1992) Siegler (1996).

Developmental change often involves both new information and changing relations among information already available. Redescription, dynamic-systems adjustment, and conceptual restructuring address partly different facets of that process. An individually stable strategy may become selectively recruited in unfamiliar circumstances; a newly articulated relation may in turn change how subsequent experiences are encoded. Distinguishing such pathways requires repeated assessments across time and representational formats. The theoretical aim is an explanatory account of how knowledge grows in generative power, how existing competences are reorganized, and when they become usable beyond their original learning contexts Karmiloff-Smith (1992) Carey (2009) Thelen & Smith (1994).

Methods for Studying Developmental Concepts

Infant looking-time and habituation studies infer expectations from differences in visual attention, whereas reaching, imitation, and choice experiments infer knowledge from structured action. Older children may complete conservation judgments, category induction, causal interventions, explanations, drawings, and prediction tasks. Each method introduces its own response requirements and potential artifacts. A credible developmental claim therefore states the inferential link between the observed measure and the proposed construct, describes the control conditions, and reports how alternative perceptual, motor, linguistic, or social explanations were evaluated Baillargeon (1987) Diamond (1985) Gelman & Markman (1986).

Longitudinal designs follow the same individuals as conceptual capacities emerge, permitting estimates of continuity and the temporal order of related achievements. Repeated testing may change the very process it measures; participation, prior familiarity, and attrition must be documented. Cross-sectional age comparisons are efficient but cannot reveal individual transitions and can confound developmental timing with differences in experience. Microgenetic designs add dense observations during periods of rapid change. Combining these designs with preregistered predictions can help separate gradually increasing competence from abrupt switches produced by arbitrary thresholds Siegler & Jenkins (1989) Siegler (1996) Thelen & Smith (1994).

Measurement validity extends beyond reliability of a test's numerical score. An instrument intended to measure a child's representation of causal structure should discriminate causal reasoning from associative preference and general attention; a test of conceptual revision should measure explanatory transfer rather than only retention of taught phrases. Researchers must consider age-equivalence of task demands, response formats, cultural familiarity of materials, and the possibility that practice changes meaning across time. Strong designs use independent outcome measures, blinded coding where feasible, counterbalancing, explicit missing-data rules, and transparent criteria for judging a response as evidence of understanding Gopnik et al. (2001) Vosniadou & Brewer (1992) Wellman et al. (2001).

Reassessing Infant Looking-Time Evidence

Kunin, Piccolo, Saxe, and Liu conducted a preregistered systematic review and meta-analysis of 76 published and unpublished infant studies, comprising 1,899 participants aged three to twelve months, on early physical and psychological reasoning. They compared attention to unexpected versus expected events with attention to perceptually novel versus familiar displays within the same literature. The authors reported a standardized mean difference of 0.290 for violation-of-expectation contrasts, with a 95% confidence interval of 0.208–0.372, and 0.239 for perceptual novelty contrasts, with a 95% confidence interval of 0.109–0.369. This evidence supports distinguishable contributors to infant looking under the studied conditions Kunin et al. (2024).

The review found that age and habituation related differently to the two kinds of looking effect, consistent with partially independent perceptual and conceptual contributions. Its inclusion of unpublished evidence, preregistration, and direct comparison of predictors strengthen the assessment of an influential paradigm. At the same time, the authors reported mixed evidence concerning publication bias in violation-of-expectation research. A mean contrast across experiments cannot establish that every individual infant reasoned in the same way or that the measured expectation was an explicit concept. These qualifications matter when developmental theories map gaze differences onto enduring knowledge structures Kunin et al. (2024) Baillargeon (1987).

A productive next step is to design studies in which the proposed conceptual interpretation predicts a pattern across different sensory displays, event trajectories, and behavioral responses that a perceptual novelty account would not predict. The relevant developmental question concerns the stability and generalization of the expectation, not simply the existence of longer looking at one event. Multi-laboratory replications, open materials, preregistered coding, and cross-paradigm convergence can clarify which representational claims are warranted. Evidence synthesis should retain distinctions between replication of a behavioral contrast and validation of a theoretical interpretation Kunin et al. (2024) Onishi & Baillargeon (2005).

Developmental Evidence and Competing Explanations

Theories of conceptual development confront a recurring problem of explanatory underdetermination: a successful response may follow from different combinations of knowledge, attention, memory, instruction, and task history. An infant's longer looking can be consistent with an violated expectation, perceptual novelty, or other stimulus-driven processes, depending on the controls. A child who revises a statement after a demonstration may be responding to social authority rather than reconstructing a causal model. Experimental evidence becomes more discriminating when rival accounts are specified before data collection and linked to different predictions for novel cases Kunin et al. (2024) Posner et al. (1982) Gopnik et al. (2001).

Competing models also operate at different levels of explanation. Core-knowledge proposals address initial representational constraints, knowledge-in-pieces accounts emphasize local activation, mental-model theories characterize organized explanatory relations, and microgenetic approaches track transitions in strategy use. Their claims should be compared on matched tasks and developmental observations, not merely by counting demonstrations consistent with each view. Evaluating a developmental account requires attention to whether it predicts an age pattern, within-person change, context sensitivity, or the emergence of new conceptual distinctions. Models can be integrated when their proposed mechanisms genuinely refer to different causal levels Spelke & Kinzler (2007) diSessa (1993) Vosniadou & Brewer (1992) Siegler (1996).

Failure to find a predicted effect should also inform theory. Investigators can test whether an apparently universal transition depends on instruction, whether a presumed domain-general competence appears only with familiar materials, and whether a purportedly new concept disappears under a transfer challenge. Interpreting null findings requires adequate statistical sensitivity and trustworthy measurement; interpreting positive findings requires attention to selective reporting and multiple comparisons. Cumulative research benefits from open stimuli and analysis code, independent replication, and clearly distinguished descriptive, mechanistic, and educational claims Kunin et al. (2024) Aleknavičiūtė et al. (2023).

Conceptual Change Through Instruction

Instructional conceptual change begins with the recognition that learners bring explanatory commitments into educational encounters. These commitments may be useful in familiar circumstances and misleading when extended to unfamiliar scientific problems. Posner and colleagues proposed that learners' evaluation of a current account and of an alternative helps determine whether conceptual accommodation occurs. A productive lesson therefore makes an existing explanation available for examination, provides evidence that matters to the learner, and supports construction of a more explanatory model. The outcome must be judged by reasoning with the new model, rather than by its repetition on an immediate posttest Posner et al. (1982) Vosniadou & Brewer (1992).

Different instructional designs seek change through cognitive conflict, bridging between familiar and unfamiliar situations, refutational texts, structured comparison, or explicit ontological reclassification. Each implies a distinct causal sequence. Conflict makes a mismatch salient; bridging makes a new relation intelligible through accessible intermediate examples; categorical reclassification revises the kind of entity a learner takes the phenomenon to be. These approaches may be combined within actual classroom teaching. Their evaluation requires information about the initial conception, the learning activities, teacher implementation, comparison treatment, timing of assessment, and whether the student transfers to independently constructed examples Pacaci et al. (2024) Aleknavičiūtė et al. (2023).

Teaching also changes how children treat their own explanations as objects of inquiry. Guided prediction, explanation, comparison, and revision can make hidden assumptions explicit and permit discussion of the evidence supporting them. Yet conceptual conflict alone may lead to avoidance, rote substitution, or an additional isolated belief if the learner lacks resources for constructing a useful alternative. The successful educational mechanism is therefore identified through a sequence: a prior explanatory commitment is characterized, a particular learning process is introduced, and a change in independent explanations and transfer is observed over time. Such a sequence makes instructional studies relevant to developmental theory rather than only to test-score improvement Kuhn et al. (1988) Carey (2009) Posner et al. (1982).

Evidence from Intervention Reviews and Meta-Analyses

Pacaci, Ustun, and Ozdemir synthesized 218 primary studies involving 18,051 students in a random-effects meta-analysis of conceptual-change strategies in science education. They reported an overall Hedges' g of 1.10, with a 95% confidence interval from 1.01 to 1.19, on the measures analyzed. Their model-based adjustment for potential publication bias gave an effect estimate of 0.93. These figures support an average advantage for the compared conceptual-change interventions across the included study designs and educational settings; they do not represent one invariant effect or prove that all measured learning involved restructuring of a concept Pacaci et al. (2024).

The same analysis reported a wide prediction interval, from 0.19 to 2.38, indicating substantial dispersion in effects across settings and study characteristics. It also examined cognitive conflict, bridging, and ontological category shift, although the underlying studies and assessment methods differed across these groups. One should therefore avoid treating comparisons of subgroup averages as evidence that one learning mechanism is intrinsically superior for a particular child or concept. Outcome instruments frequently emphasize achievement or misconceptions, which leaves uncertainty about long-term explanatory transfer and the underlying structure of what changed Pacaci et al. (2024).

Aleknavičiūtė, Lehtinen, and Södervik's review and meta-analysis of three decades of biology-oriented conceptual-change interventions likewise identified substantial average intervention effects while emphasizing small-scale studies and frequent absence of randomized designs. Refutational texts were a common instructional format. The evidence supports further investigation of explicit conceptual instruction, alongside a need for stronger experimental contrasts, longer follow-up, and measures that discriminate genuine explanatory reconstruction from improved response selection. Cumulative synthesis should therefore evaluate the construct validity of each outcome before assigning a pooled educational gain the theoretical meaning of conceptual change Aleknavičiūtė et al. (2023) Pacaci et al. (2024).

Persistence, Coexistence, and Delayed Revision

Acquiring a scientific explanation does not guarantee the disappearance of earlier intuitions. Shtulman and Valcarcel asked scientifically educated adults to verify statements that either agreed or conflicted with common naïve intuitions across ten scientific domains. Responses were generally slower and less accurate for statements where the scientific answer competed with an intuitive answer, suggesting continued interference from earlier representations. The experiment directly addresses adult performance; the inference that both kinds of knowledge persist is a plausible explanation of the reported reaction-time and accuracy pattern, while details of the internal representational mechanism require additional convergent evidence Shtulman & Valcarcel (2012).

Coexistence has implications for developmental measurement. A child may give a new correct explanation in a lesson, then rely on an older account when a different representation, time constraint, or concrete example makes that account salient. Apparent relapse can signal context-sensitive activation as well as incomplete learning. To evaluate genuine change, studies should assess spontaneous explanation, unprompted prediction, confidence, transfer, and delayed retention, using alternate items so that answer memory does not masquerade as understanding. Such designs allow a distinction between an added scientific rule and a reorganized system of explanatory commitments Shtulman & Valcarcel (2012) diSessa (1993) Vosniadou & Brewer (1994).

The presence of an earlier intuition is not necessarily a defect in learning; intuitive knowledge can remain a productive basis for analogy and everyday action under appropriate conditions. The developmental question is whether a learner can identify the domain where each explanatory system applies and select an appropriate model for novel evidence. Instruction may increase discrimination between contexts before it produces fully integrated conceptual relations. Longitudinal and process-sensitive studies can trace whether strategic selection becomes faster, whether scientific reasoning generalizes, and whether incompatible assumptions eventually lose influence. These outcomes represent different forms of developmental progress Shtulman & Valcarcel (2012) Siegler (1996) Gentner (1983).

Individual Variation, Culture, and Educational Context

Children develop concepts in environments that differ in materials, language practices, schooling, social relationships, and the kinds of explanations routinely offered by adults. A developmental pattern observed in a narrow group should therefore be presented with its population and task context rather than elevated into an age-universal sequence. Theoretical accounts of early number, biological kinds, and scientific explanation generate testable predictions about how shared cognitive resources interact with local instruction and opportunities for practice. Generalizability depends on comparing the same underlying competence through age-appropriate measures across settings Carey (2009) Gelman & Markman (1986) Gopnik (2012).

Variation within a child is equally informative. Prior familiarity with a scientific domain, comprehension of question wording, interest in an activity, and experience with similar tasks can change observable performance. Differences among learners may reflect opportunities to use a concept or differences in how the learning situation was organized, as well as variation in representational resources. Repeated within-person observations are particularly useful because a competence that appears absent in one interview may be recruited in a later causal or classification task. Researchers should report response variability as data relevant to theory rather than merely measurement noise Siegler (1996) Thelen & Smith (1994).

A culturally and educationally credible research program would examine children across varied linguistic and school settings, document access to relevant experiences, and distinguish measurement adaptation from change in the targeted construct. Interventions should be described sufficiently to permit interpretation across curricula and age groups. Research that compares local explanatory practices can clarify which learning routes depend on particular symbol systems or teacher demonstrations and which generalize broadly. This is an empirical agenda rather than a claim that current evidence establishes the same developmental trajectory for every population Aleknavičiūtė et al. (2023) Pacaci et al. (2024).

Conceptual Development in the Artificial Era

Digital interfaces, conversational artificial systems, adaptive educational software, and computational simulations now provide children with increasingly varied opportunities to receive explanations and test hypotheses. For developmental psychology, the immediate research object is how these interactions alter children's attention, inferential choices, confidence in testimony, and transfer of knowledge. A responsive explanation system might guide comparison of evidence or encourage unexamined answer copying, depending on the learner's task and the system's feedback. Such effects must be measured through controlled learning outcomes rather than inferred from the sophistication or apparent personality of an artificial interlocutor Gopnik (2012) Kuhn et al. (1988).

A coherent study of human–artificial conceptual learning would measure baseline explanations, randomly allocate learners to distinguishable feedback conditions, characterize the interaction actually received, and assess reasoning on unseen problems after an appropriate delay. The relevant mechanisms might include analogy, exposure to counterexamples, guided intervention, or social evaluation of testimony. Researchers should also document age, language proficiency, accessibility, and the quality and factual correctness of generated explanations. These design principles translate established questions about instruction and conceptual change into a setting where explanatory content can vary dynamically across encounters Gentner (1983) Posner et al. (1982) Pacaci et al. (2024).

The Artificial Era · From Homo to Artificial framework opens an additional, explicitly theoretical question: how should concept formation and representational revision be compared across biological learners and artificial cognitive systems? Comparable experiments would require precise operational definitions, matched prediction tasks, longitudinal or repeated training observations, and evidence of transfer to new domains. Apparent structural similarity between human and machine outputs is a starting hypothesis for functional comparison. The literature synthesized here supplies methods for testing representational change in children; conclusions about artificial psychological organization require their own independent empirical program Carey (2009) Spelke & Kinzler (2007).

Integrative Mechanisms and Explanatory Levels

A research synthesis can explain conceptual development through linked processes at several levels. Initial perceptual and representational resources constrain what can be discriminated; attention and action select evidence; categories and causal expectations organize induction; language and social instruction supply new symbols and explanatory relations; practice and feedback change when different strategies are recruited. More extensive conceptual change occurs when these adjustments alter the relational organization supporting prediction, explanation, and transfer. This multilayered account accommodates early core knowledge, dynamic strategy variation, and gradual restructuring without treating any one level as a complete explanation Spelke & Kinzler (2007) Carey (2009) Thelen & Smith (1994).

The evidence base suggests four useful outcome classes. Acquisition adds an item or relation to an existing scheme; differentiation divides a formerly undifferentiated category; integration coordinates several relations into a more general explanatory structure; contextual regulation improves selection among competing interpretations. These are analytical distinctions intended to organize research, not newly discovered cognitive stages. Each is operationalized differently: acquisition by novel property prediction, differentiation by discrimination and justification, integration by cross-context causal transfer, and regulation by appropriate selection under conflicting cues. Longitudinal observations can reveal how these changes interact in a single developing learner Siegler (1996) Carey (2009) Shtulman & Valcarcel (2012).

Mechanistic inference must go beyond correlating age with accuracy. If a proposed process is causal, manipulating that process should change the predicted pattern of later conceptual outcomes while plausible alternatives are controlled. For example, structured comparison should enhance relational transfer beyond equally intensive exposure when the analogy mechanism is decisive; guidance in causal interventions should affect generalization beyond simple associative training when causal inference is implicated. Intervention effects, observational trajectories, and experimentally specified transfer patterns form complementary evidence rather than interchangeable tests. The deepest contribution of developmental research is the ability to explain how a learner becomes capable of new forms of inference Gentner (1983) Gopnik et al. (2001) Pacaci et al. (2024).

Open Scientific Problems and Research Priorities

A major open problem is how to demonstrate genuine representational novelty rather than improved access to an earlier competence. Researchers need instruments that make competing accounts of infant expectation, preschool categorization, and school-age explanations predict different outcomes. This requires well-matched control tasks, convergence across action and verbal measures, and repeated observations. The question of whether knowledge is principally organized as coherent theories, local resources, or flexible mixtures should be tested through within-person patterns and prospective predictions rather than by retrospective labels applied after examining the results Kunin et al. (2024) diSessa (1993) Vosniadou & Brewer (1992).

A second problem concerns mechanisms of durable educational change. Existing intervention syntheses show encouraging average outcomes but considerable variation in designs, materials, and instruments. Future studies should separate immediate learning, delayed retention, explanatory generalization, and metacognitive recognition of conflicting evidence. Randomized comparisons of teaching mechanisms, transparent preregistration, and assessment instruments calibrated for comparable conceptual outcomes would increase causal interpretability. Evidence syntheses could then address whether particular approaches are effective for particular types of revision, rather than aggregating every improvement in science achievement under one label Pacaci et al. (2024) Aleknavičiūtė et al. (2023).

A third problem concerns temporal and population generalizability. Longitudinal cohorts, dense microgenetic observations, and experiments with culturally varied materials can jointly test whether a proposed transition is shared, delayed, or reorganized by experience. Better reporting of demographic recruitment, attrition, language conditions, task familiarity, and measurement reliability would help determine how widely findings travel. Artificial educational environments add new experimental opportunities but also new confounds involving the variability of feedback and learners' assumptions about its authority. The research agenda is consequently both developmental and methodological: specify the concept, model its transformation, and observe its consequences under independent conditions Siegler & Jenkins (1989) Gopnik (2012) Kunin et al. (2024).

Conclusion

Cognitive development and conceptual change form a connected research field concerned with how increasingly powerful systems of knowledge arise through childhood. Findings from object expectation, numerical discrimination, categorization, causal learning, biological concepts, mental-state reasoning, and scientific explanation establish that development is domain-sensitive and method-dependent. Children can display informative early expectations while continuing to acquire the language, strategic flexibility, and inferential coordination required for later explicit understanding. A scientifically adequate account therefore describes both the conceptual content that changes and the task conditions through which that content is observed Baillargeon (1987) Carey (2009) Wellman et al. (2001).

The main theoretical traditions contribute different explanatory resources. Piaget's experimental problems foregrounded organized operations; core-knowledge theories identify early constraints; representational-redescription and theory-based accounts explain new expressive relations; dynamic-systems and microgenetic studies expose gradual, variable pathways; knowledge-in-pieces theories illuminate context-sensitive activation. The resulting synthesis supports a developmental explanation that distinguishes acquisition, differentiation, integration, and contextual regulation as testable forms of change. The empirical question for any proposed mechanism is whether it predicts independent transfer and temporal development beyond the specific task from which it was inferred Piaget & Szeminska (1941) Spelke & Kinzler (2007) Karmiloff-Smith (1992) Siegler (1996).

Current evidence also places limits on sweeping conclusions. A reliable looking-time contrast does not directly measure an explicit theory; accurate answers after teaching do not guarantee restructuring; and older intuitions may still influence reasoning after scientific knowledge is acquired. Original experiments, recent meta-analyses, and methodological comparisons together establish a strong foundation for cumulative research. The most productive future work will connect precisely specified prior representations, experimentally identified learning processes, durable conceptual outcomes, and cross-context replication. In the Artificial Era, the same rigor can guide investigations of how novel computational environments participate in learning while maintaining a clear distinction between human developmental evidence and new cross-system hypotheses Kunin et al. (2024) Shtulman & Valcarcel (2012) Pacaci et al. (2024).

References