RESEARCH DOSSIER
Prior Knowledge, Misconceptions, and Conceptual Change
Abstract. Prior knowledge shapes the interpretation of new information and the explanatory relations through which learners understand scientific and disciplinary problems. Misconceptions identify context-dependent patterns of inference that conflict with a specified scientific account, while conceptual change concerns a durable transformation in the organization, application, or coordination of knowledge. This fundamental research monograph examines the historical development of misconception research, original studies in mechanics, astronomy, biology, chemistry, and mathematics, and the major theories of accommodation, knowledge-in-pieces, framework models, ontological reassignment, and conceptual coexistence. It evaluates diagnostic interviews, two-tier probes, concept inventories, process measures, delayed transfer, refutation texts, contrasting cases, interactive engagement, and expertise-sensitive instruction. Reviews and meta-analyses are interpreted in light of causal identification, measurement validity, heterogeneity, and publication bias. The evidence supports benefits of targeted conceptual-change instruction for measured learning but requires stronger designs to establish the mechanisms and durability of representational reconstruction. The synthesis identifies priority questions concerning competing cognitive explanations, generalizability, and knowledge revision in human–artificial instructional interactions. This work is an integrative critical scientific monograph, not a preregistered systematic review or original experiment.
Keywords: prior knowledge; misconceptions; conceptual change; educational psychology; scientific reasoning; alternative conceptions; cognitive conflict; accommodation; knowledge-in-pieces; framework theories; ontological shift; refutation texts; concept inventories; transfer of learning; metacognition; expertise reversal; science education; conceptual change assessment
Research Object, Definition, and Scope
Prior knowledge is the collection of representations, procedures, expectations, explanatory commitments, and familiar situations that a learner can recruit when encountering a new problem. A misconception, as the term is used in educational psychology, is an identifiable pattern of interpretation or inference that conflicts with a relevant disciplinary account under specified conditions. Conceptual change is a durable alteration in the organization, application, or explanatory relations of knowledge that changes what a learner can predict, explain, and transfer. These constructs are connected because new information is interpreted through existing understanding, and the same knowledge that enables interpretation can selectively resist revision. The scientific question concerns which elements of prior understanding are productive, which inferences fail, and what processes generate more discriminating explanations National Research Council (2000) Smith et al. (1994).
A learner may remember a correct definition while continuing to predict events by an earlier intuitive rule. Conversely, a person may perform a task accurately without articulating the underlying mechanism. The distinction between possession, accessibility, use, and explanatory organization of knowledge prevents a high posttest score from serving as sufficient evidence for reconstruction. Conceptual change can include differentiation of previously merged ideas, coordination of representations across contexts, reclassification of a concept, or the addition of an explanatory constraint. Its success should be indexed by discriminating predictions, novel cases, uncertainty calibration, and delayed retrieval, according to the domain investigated diSessa & Sherin (1998) Shtulman & Valcarcel (2012).
This monograph critically synthesizes original studies of scientific intuitions, developmental mental models, diagnostic measurement, theory revision, reading-based knowledge revision, and interventions. It uses an integrative, question-guided narrative design rather than a preregistered systematic review or new pooled meta-analysis. Primary studies establish what was observed in particular tasks; review articles and quantitative syntheses describe the distribution of findings; theoretical models furnish competing explanations. The claims are graded by methodological relevance and external validity, and the distinction between conceptual understanding and conventional achievement remains central. This dossier concentrates on instruction-induced interaction between existing knowledge and disciplinary explanation, leaving broad developmental origins and domain-general memory acquisition to adjacent research programs Posner et al. (1982) Pacaci et al. (2024).
The Historical Discovery of the Learner's Starting Point
Research on knowledge-dependent learning has several intellectual lineages. Developmental investigators, prominently Jean Piaget, treated children's explanations as evidence about the organization of reasoning. Subsequent work on scientific learning increasingly documented the rich commitments students bring into formal classrooms. The move from scoring answers as right or wrong to examining reasons, intermediate models, and patterns of inference transformed a pedagogical concern into a psychological research program. Carey examined the reconstruction of biological knowledge in childhood, while studies in mechanics revealed stable predictions acquired through ordinary encounters with motion Carey (1985) Clement (1982).
By the late 1970s and early 1980s, science education researchers had begun to describe students' alternative frameworks as systematic intellectual resources rather than incidental classroom mistakes. Posner, Strike, Hewson, and Gertzog formalized conditions under which a scientific conception might be accommodated: dissatisfaction with an existing conception and the intelligibility, plausibility, and fruitfulness of an alternative. Their account, published in 1982, brought the language of philosophy of science into the investigation of student learning. Its significance lies in specifying a mechanism to investigate, although later empirical work does not imply that all conceptual revision follows one uniform sequence Posner et al. (1982).
A parallel measurement tradition made students' starting points visible in large classes. Halloun and Hestenes analyzed common-sense ideas about force and motion in 1985; Hestenes, Wells, and Swackhamer later developed the Force Concept Inventory as a diagnostic instrument. Their work encouraged evaluations of conceptual reasoning independently of routine formula manipulation. The resulting literature connected qualitative interview evidence to scalable tests, but also created enduring questions about whether any single answer pattern uniquely identifies a coherent underlying theory Halloun & Hestenes (1985) Hestenes et al. (1992) Treagust (1988).
Core Constructs and Levels of Explanation
A schema is an organized structure of knowledge that facilitates interpretation by relating current input to a familiar pattern. A mental model is a task-sensitive representation used to simulate a system or generate an expectation. An explanatory framework comprises commitments about what exists, how relations are organized, and what counts as a cause or sufficient reason. Knowledge elements can also be smaller, locally activated interpretive resources. These units operate at distinct analytical scales, so the same learner may exhibit a stable domain-level belief across interviews and variable responses across particular situations Vosniadou & Brewer (1992) diSessa (1993).
Conceptual understanding must be separated from declarative familiarity, procedural competence, and the ability to reproduce a textbook formulation. One can compute an answer using a remembered equation while still expecting the physical event to unfold differently. One can also express a plausible causal analogy without being able to transfer it to novel boundary conditions. Accordingly, experimental designs benefit from pairing verbal reports with prediction tasks, explanation prompts, and counterexamples. Strong inference requires researchers to specify the representational construct and to demonstrate that chosen indicators distinguish it from literacy, reading demand, memory, and test sophistication Clement (1982) Treagust (1988).
Misconception labels carry explanatory assumptions. A wrong response may reflect an organized alternative account, a useful heuristic applied outside its normal domain, a vocabulary ambiguity, a locally constructed guess, or an inaccessible correct representation. Smith, diSessa, and Roschelle argued that students' existing knowledge also supplies the materials from which expertise develops. In this perspective, productive and problematic knowledge can coexist within the same conceptual network. Operational clarity requires investigators to describe the precise prediction, explanation, context, and criterion of disciplinary correctness before labeling a pattern as a misconception Smith et al. (1994) diSessa & Sherin (1998).
Prior Knowledge in Comprehension and Meaning Construction
Prior knowledge shapes what a person notices, what inferences appear available, and how a text or demonstration is understood. A technical sentence is interpreted through relationships retrieved from earlier experience, not as an isolated sequence of words. Research on reading scientific texts shows that both the knowledge readers possess and the structure of the explanation influence their ongoing comprehension processes. Kendeou and van den Broek used think-aloud and reading-time methodologies to investigate how such processes unfold during reading rather than inferring them exclusively from final test scores Kendeou & van den Broek (2007).
The value of prior knowledge depends on its alignment with the current task. Familiar causal patterns can support predictions, reduce the need for repeated explanation, and make an unfamiliar system intelligible. The same processes can lead readers to fill explanatory gaps with an intuitive relation that a scientific account rejects. Retaining material in memory and constructing a scientifically coherent situation model therefore demand different kinds of assessment. A useful research design asks whether readers connect claims through warranted causal relations and can tell when an earlier connection no longer applies National Research Council (2000) Kendeou et al. (2014).
Knowledge-sensitive instruction has an important asymmetry. For novices, explanation and demonstrations may supply relationships that are otherwise unavailable; for experts, repetitive guidance may compete with existing organization and add unnecessary processing demands. Kalyuga described the expertise-reversal effect as a reversal in the relative effectiveness of instructional arrangements as learners' prior knowledge changes. This finding makes prior knowledge a moderator of instructional effectiveness and cautions against a universal recipe for concept revision independent of the learner's starting point Kalyuga (2007).
Productive Knowledge, Analogy, and Transfer
Everyday interpretations often contain relations that can be refined into disciplinary explanations. A child who expects unsupported objects to fall, or a beginner who links a push with a visible change, already possesses constraints that can guide carefully designed comparisons. The theoretical challenge is to preserve useful relational structures while making their limits explicit. Gentner's structure-mapping account of analogy explains why correspondence between relations can be more informative than superficial resemblance between objects. An analogy is pedagogically useful when the mapped relations preserve the mechanism relevant to the target phenomenon Gentner (1983) Smith et al. (1994).
Transfer provides a stronger test of conceptual development than repeating a practiced example. Near transfer may involve changing surface features while preserving familiar relational structure; farther transfer requires selecting the relevant structure without obvious cues. Investigators should ask what cues prompted retrieval and whether performance survives changes in wording, representation, and context. The same intervention can improve practiced responses while leaving learners dependent on the instructional analogy, especially if its limits were never examined Gentner (1983) Schwartz et al. (2011).
Constructive use of prior knowledge includes contrasts as well as analogies. Contrasting cases make invariant relationships easier to discover because learners must explain both similarity and difference. Schwartz and colleagues compared sequences in which adolescents were told a principle before practicing with sequences in which learners first attempted to invent formulations from contrasting cases and were then instructed. Such experiments test the timing of explanatory support, not an unrestricted claim that discovery should replace teaching. The resulting question is how preparation, guidance, and explanation can be sequenced to strengthen subsequent transfer Schwartz et al. (2011) Mayer (2004).
Misconceptions as Empirical Objects
A misconception becomes an empirical object when it is specified sufficiently to be elicited, compared across contexts, and distinguished from rival accounts. In mechanics, the expectation that continuing motion requires a continuing force can yield characteristic answers under frictionless idealizations. In astronomy, the coexistence of a statement that Earth is round with an imagined edge can reveal an attempt to coordinate incompatible representations. In biological explanation, purpose-based reasoning may intrude on causal accounts of evolution or physiology. Each example demands a distinct instrument and theoretical interpretation Clement (1982) Vosniadou & Brewer (1992) Carey (1985).
Stability is not the same as coherence. An answer that recurs across several differently worded tasks suggests a robust pattern; it does not alone establish an internally consistent theory. A single student may draw on several ways of interpreting the same event, and a test item may unintentionally cue one of them. Interviews, confidence ratings, and designed follow-up questions offer a richer basis for distinguishing competing explanations. Instruction that treats every incorrect response as the same fixed belief risks targeting an interpretation that was never stable in the first place diSessa (1993) Smith et al. (1994).
Scientific correctness is itself domain- and condition-dependent. Statements about objects, inheritance, heat, or mathematical infinity can be appropriate under one explicitly defined model and inappropriate under another. A defensible diagnostic practice specifies the domain, assumptions, levels of idealization, and expected causal mechanism before classifying a student answer. This practice also separates misconception research from generalized judgments of intelligence or ability. The research target is a revisable relationship between prior representation, task conditions, and the scientific explanation against which it is evaluated Treagust (1988) National Research Council (2000).
The Experimental Tradition in Mechanics
Mechanics became a major laboratory for investigating prior knowledge because everyday motion provides abundant experience while Newtonian explanations require distinctions that ordinary experience can obscure. Clement's 1982 study used written tests and videotaped problem-solving interviews to investigate students' beliefs about force and acceleration. The work described a recurring interpretation that linked continuous motion to continued force. Its methodological importance lies in showing qualitative explanatory difficulty among learners who could encounter formal formulas but retain a competing intuition about the underlying event Clement (1982).
Halloun and Hestenes subsequently analyzed common-sense concepts about motion and documented an initial knowledge state that was often little altered by conventional instruction. The Force Concept Inventory, introduced by Hestenes, Wells, and Swackhamer in 1992, operationalized central distinctions in Newtonian mechanics through conceptual alternatives. These instruments changed the evidentiary scale of physics education research, permitting comparisons across classes and instructional conditions. Their items nevertheless assess specified interpretive judgments; whether a response reveals a stable explanatory framework requires corroborating tasks or explanations Halloun & Hestenes (1985) Hestenes et al. (1992).
The familiar prediction that a released object carried by a moving person falls directly downward provides another experimental example. McCloskey, Washburn, and Felch examined the straight-down belief through paper-and-pencil judgments and tasks involving perceptual presentation and action. Their evidence linked some intuitive physics responses to how motion is perceived and mentally represented. The study cautions against inferring that classroom error is only a verbal failure: a mistaken prediction may survive dynamic presentation or personal experience. Equally, it cautions against assuming the same source for every learner who supplies the same wrong answer McCloskey et al. (1983).
Astronomy, Earth Models, and Explanatory Consistency
Vosniadou and Brewer's 1992 study of first-, third-, and fifth-grade students investigated children's models of Earth's shape. Some responses appeared contradictory when scored as isolated propositions: a child could affirm that Earth is round while also imagining an edge from which someone might fall. Their analysis proposed that such patterns could be organized through synthetic mental models combining instructed information with prior presuppositions grounded in everyday spatial experience. The theoretical claim is that learning may involve intermediate organizations of knowledge rather than immediate replacement of one sentence by another Vosniadou & Brewer (1992).
A mental-model explanation is compelling when it predicts relations among independently elicited answers and makes falsifiable expectations about novel questions. Its interpretation is weakened when the interview itself supplies cues, when children interpret terms differently from the investigator, or when a proposed model is inferred mainly from the same few responses it was designed to fit. Nobes, Martin, and Panagiotaki revisited the development of scientific Earth knowledge, illustrating how differences in questioning and interpretation may alter the portrait of children's concepts. The resulting debate concerns how strongly individual responses warrant attribution of a unified internal model Nobes et al. (2005) Vosniadou & Brewer (1992).
Astronomy also illustrates a broader problem of scale. A learner may know that a planet is spherical and still reason locally using the geometry of a flat ground surface. Success requires coordinating egocentric visual experience with a representation of Earth as an astronomical body, and the relevant relation may be demanded differently in drawing, verbal explanation, or prediction tasks. Educational progress should therefore be assessed across questions about shape, gravity, location, movement, and perspective rather than by recognition of the word sphere alone Vosniadou & Brewer (1992) Nobes et al. (2005).
Biological Kinds, Evolution, and Causal Explanation
Biological reasoning tests the limits of familiar concepts such as living, growing, needing, having a function, and belonging to a kind. Carey examined the reconstruction of biological understanding as children differentiated organisms from other entities and coordinated new relations among inheritance, bodily function, and causal processes. Knowledge revision in biology thus extends beyond learning specialized vocabulary: it may change which features count as essential to an explanation, what causes are attributed to organisms, and how categories support inference Carey (1985).
Students may rely on purpose-oriented or essentialist explanations because these support efficient everyday judgments. An explanation that an organ exists in order to fulfill a need can be conversationally helpful yet insufficient as a historical evolutionary explanation. The problem is not solved by forbidding purpose-language in every context; it requires teaching the causal standards appropriate to the question. Domain-specific intervention studies should assess whether learners can distinguish proximate physiological function, evolutionary history, and intentional design when contexts are varied Carey (1985) Talanquer (2006).
The 2023 review and meta-analysis by Aleknavičiūtė, Lehtinen, and Södervik analyzed conceptual-change interventions in biology. Evolution and photosynthesis were prominent research topics. The review reported substantial gains for interventions compared with conventional teaching while identifying publication bias, sample-size effects, and a widespread tendency to measure knowledge enrichment rather than genuine restructuring. These findings make biological instruction an important test case for the distinction between adding correct statements and revising relations among concepts Aleknavičiūtė et al. (2023).
Chemistry and the Organization of Everyday Causal Heuristics
Chemical processes ask learners to interpret particles, interactions, conservation, and transformations that are seldom directly visible. Everyday reasoning about substances can privilege perceptible properties, simple one-to-one causes, and an assumption that visible changes imply disappearance or creation. Talanquer's account of commonsense chemistry sought patterns within a large diversity of alternative conceptions, thereby making it possible to study underlying heuristics rather than compile disconnected lists of errors. The proposal is explanatory: a shared inference pattern may produce several misconceptions across chemical topics Talanquer (2006).
The instructional challenge is to coordinate macroscopic observations, particulate models, and symbolic representations without confusing movement between them with proof of scientific understanding. A learner may balance an equation accurately while retaining a material-image account that predicts conservation incorrectly in an unfamiliar setting. Diagnostic instruments should make the relationships between representation and causal model observable. They should also be sensitive to language and diagram conventions, since these can determine whether a question measures chemistry reasoning or the interpretation of a representation Treagust (1988) Talanquer (2006).
Curricular efforts that emphasize fewer, more connected explanatory ideas reflect this challenge. Talanquer and Pollard described a redesign of introductory chemistry around ways of thinking and core conceptual relationships rather than encyclopedic topic coverage. Such work illustrates a principled response to misconception research: define the structure that learners must coordinate and then construct opportunities to reason across examples. Evaluation still requires comparative evidence that the redesign improves concept use beyond the taught cases Talanquer & Pollard (2010).
Concepts Across Mathematics and Other Disciplines
The transformation of a concept can be more complicated than the correction of a proposition. Mathematics requires learners to differentiate quantity from notation, coordinate part–whole relations, and recognize when a procedure that works for integers fails with fractions or rational numbers. Conceptual growth can involve replacing an inappropriate ordering rule, redefining the unit of comparison, or building a relation that applies across representations. The theory of conceptual change accommodates these cases when the unit of change is explicitly stated and learning outcomes test the intended mathematical structure diSessa & Sherin (1998) Smith et al. (1994).
An educational difficulty recurs when learners overgeneralize productive knowledge. A heuristic such as 'multiplication makes a quantity larger' may fit many experiences with whole numbers and then fail for fractions between zero and one. The failure reveals both prior learning and a missing boundary condition. Effective conceptual instruction uses cases that expose the boundary and requires explanation of why the original rule worked in its familiar domain. The cognitive target is a conditionalized representation capable of selecting the correct operation by relation and context rather than by surface pattern Smith et al. (1994) Gentner (1983).
Studies across physics, astronomy, biology, and mathematics also challenge an overly unified model of misconception. A conceptual conflict can concern perceptual prediction, a category assignment, a network of explanatory commitments, or a formal rule. Therefore, differences in measurement and intervention effectiveness between disciplines may arise from the phenomenon itself, rather than solely from research quality. Comparative claims require common definitions of transfer and durability while allowing domain-specific interpretations of what counts as reconstruction Pacaci et al. (2024) Carey (1985).
The Persistence of Earlier Intuitions
Conceptual learning does not guarantee that earlier intuition disappears. Shtulman and Valcarcel asked adults with years of science education to rapidly verify statements either congruent or incongruent with commonsense interpretations. Responses to incongruent statements were slower and less accurate across ten science domains, including astronomy, evolution, genetics, matter, mechanics, and thermodynamics. These patterns support a coexistence account in which a scientifically accepted representation can be available while an earlier intuition continues to exert influence during rapid judgment Shtulman & Valcarcel (2012).
The coexistence account does not by itself show that every prior idea remains intact, or that a response delay identifies a single stored naive theory. A latency difference can be affected by familiarity, sentence complexity, accessibility, and response selection. The strongest evidence combines carefully matched stimulus conditions with tests of multiple predictions and change over time. Nevertheless, speeded reasoning demonstrates an outcome that standard delayed achievement tests may miss: scientific competence can include managing competition among representations, not simply possessing a newly learned statement Shtulman & Valcarcel (2012) diSessa (1993).
Instructional claims should therefore distinguish the fate of an old representation from the observable success of a new one. A student might revise a causal network, add an exception, learn to inhibit a default response, or acquire context cues that select between explanations. These mechanisms make different predictions for response latency, confidence, interference, and transfer. Longitudinal evidence is needed to determine whether apparently stable postinstruction understanding reflects durable integration or context-dependent suppression of an earlier explanation Shtulman & Valcarcel (2012) Kendeou et al. (2014).
Accommodation and the Conceptual Ecology Model
The accommodation account developed by Posner and colleagues remains a foundational model of instruction-induced conceptual change. Its four well-known conditions describe a learner becoming dissatisfied with an existing conception and recognizing an alternative as intelligible, plausible, and fruitful. Intelligibility concerns whether an alternative can be grasped; plausibility concerns whether it appears compatible with warranted background commitments; fruitfulness concerns whether it offers new explanations or solutions. These are epistemic relations among conceptions and evidence, not guaranteed psychological stages through which every student must pass Posner et al. (1982).
The model is especially useful when research distinguishes a failure to understand a new explanation from failure to accept or deploy it. A learner may comprehend a model and regard it as inconsistent with direct experience; another may accept a formula yet see no reason to use it outside an examination. Such differences suggest experiments that vary explanation, counterevidence, and opportunities to apply a model. They also reveal the difficulty of measuring subjective dissatisfaction reliably: a learner may verbally acknowledge a contradiction while continuing to use the old rule in an unprompted prediction Posner et al. (1982) Chinn & Brewer (1993).
Accommodation is often discussed as a replacement of one conception by another. Subsequent evidence instead supports a family of outcomes, including reorganization, differentiation, contextual selection, and coexistence. The valuable legacy of Posner's model is its insistence that learning depends on a conceptual ecology of previously held commitments. Its limitation as a complete psychological theory is that the posited conditions alone do not specify the fine-grained causal operations by which representations change, nor do they establish durable use after the supportive instructional context disappears Smith et al. (1994) Shtulman & Valcarcel (2012).
Knowledge-in-Pieces and Resource-Based Explanations
diSessa's knowledge-in-pieces perspective treats intuitive reasoning as drawing on numerous small, experience-derived elements rather than a single fully articulated naive theory. A phenomenological primitive, or p-prim, is an intuitive interpretive resource such as the expectation that an imposed effort produces a result. Such resources may be activated differently across physical situations. Learning can involve changing which elements are triggered, how they are coordinated, and which explanatory inferences they support. The perspective predicts contextual variability as a feature of cognition rather than merely a sign of unreliable measurement diSessa (1993).
In resource-oriented research, students' existing ideas are evaluated for the roles they can play in more sophisticated understanding. A familiar observation may be inadequate as a global law but useful as a starting analogy when its relevant constraints are exposed. Smith, diSessa, and Roschelle argued that a uniformly deficit-based interpretation of misconceptions overlooks productive continuities between novice and expert knowledge. Their position makes it possible to ask which learner resources support a scientific account and which integrations or boundary distinctions remain to be constructed Smith et al. (1994).
diSessa and Sherin later introduced coordination classes to describe organized strategies for extracting particular forms of information from the world. This account specifies a type of conceptual change in terms of both representational structure and performance across contexts. It proposes that successful instruction enables learners to identify the same relevant quantity or relation across varied presentations. The explanatory benefit is a direct link between conceptual organization and a measurable pattern of flexibility; the challenge is to identify the appropriate scope and verify the hypothesized coordination through multiple tasks diSessa & Sherin (1998).
Framework Theories and Synthetic Mental Models
The framework-theory perspective emphasizes broad presuppositions through which learners interpret new knowledge. When scientific statements conflict with familiar spatial or causal assumptions, learners may combine elements into synthetic models that preserve parts of both. Vosniadou and Brewer's Earth-shape study illustrates how statements that appear inconsistent can become intelligible within a child's provisional representation. Conceptual change, on this account, requires recognizing and revising assumptions that organize a range of connected inferences Vosniadou & Brewer (1992).
This view generates strong empirical demands. If a synthetic model is genuinely explanatory, it should predict answers to new questions about falling, location, and the relationship between local experience and planetary form. Researchers must also examine whether alternative interview wording elicits the same model and whether apparent consistency arises from how answers are categorized. Work revisiting children's Earth knowledge demonstrates that the classification of intermediate conceptions is sensitive to method. A framework theory gains support from convergent cross-task evidence, not from the attractiveness of a coherent narrative alone Nobes et al. (2005).
Framework-based and knowledge-in-pieces accounts may describe different organizational levels of the same episode. A learner can retain broad expectations about terrestrial space and nevertheless switch between locally activated explanations depending on the question. Their disagreements become experimentally productive when formulated as predictions: one account may expect a stable cluster across tasks, another stronger sensitivity to situation cues. Longitudinal work can determine whether instructional change proceeds through wholesale revision, selective local reconfiguration, or a sequence that includes both processes diSessa (1993) Vosniadou & Brewer (1992).
Ontological Category Reassignment
Chi, Slotta, and de Leeuw proposed that some conceptual learning difficulties arise because a scientific phenomenon belongs to an ontological category different from the one learners initially assume. Their 1994 account contrasted interpretations of phenomena as matter-like things with interpretations as processes or constraint-based interactions. In this approach, the difficulty of revision partly depends on whether a scientific concept requires crossing a category boundary. The theory seeks to explain why certain topics resist straightforward teaching even when learners can repeat the vocabulary Chi et al. (1994).
The category-shift hypothesis is valuable because it yields distinct instructional predictions. Rather than merely correcting an incorrect statement, instruction may need to help learners represent a phenomenon's mode of existence differently. For example, discussion of heat or light may require examining transfer and interaction rather than imagining a conserved material substance in a container. Such explanations must be evaluated against the possibility that difficulty also arises from mathematical demands, language, perceptual cues, or insufficient time for practice. Demonstrating a category shift requires evidence about explanation and transfer, not just improved recognition of terminology Chi et al. (1994) Talanquer (2006).
Modern reviews compare ontological strategies with cognitive conflict and bridging approaches, but any ranking depends on which domains and outcome measures enter the evidence base. The 2024 meta-analysis by Pacaci and colleagues grouped interventions by these three orientations and found positive mean effects across categories while reporting substantial heterogeneity. Because the review's primary studies often used researcher-developed assessment instruments, comparative effectiveness should be interpreted together with measurement design and the strength of the underlying causal comparisons Pacaci et al. (2024).
Anomalous Evidence, Conflict, and Theory Protection
Counterevidence can create an opportunity for revision, but it has no fixed psychological effect outside the interpretive system of the learner. Chinn and Brewer analyzed how students and scientists may respond to anomalous data by ignoring, rejecting, reinterpreting, setting aside, or otherwise accommodating an observation without altering a central theory. Their account identified several possible reactions, with theory change only one outcome. This framework established that presenting a surprising demonstration cannot be equated with producing conceptual change Chinn & Brewer (1993).
The evidentiary problem is recursive: whether an observation counts as anomalous depends on assumptions about measurement, experimental reliability, background conditions, and the expected event. A student who doubts an apparatus may rationally discount a conflicting result; another may correctly identify a missing factor that makes apparent contradiction disappear. Productive instruction therefore must help learners evaluate the provenance and precision of evidence and compare explanations capable of generating it. When conflict is presented without explanatory support, learners may remember the spectacle while preserving the original causal account Chinn & Brewer (1993) Posner et al. (1982).
The strongest instructional use of conflict combines eliciting a prediction, showing trustworthy data, exploring why the expectation failed, and supporting construction of a replacement explanation that predicts subsequent cases. Even this sequence cannot guarantee stable revision. A well-designed experiment distinguishes the effect of an anomaly from the added explanatory guidance and checks whether participants understand the evidentiary relation. Causal claims about conflict require randomization or suitable controls and outcomes that measure conceptual organization after instructional cues have been removed Pacaci et al. (2024) Mayer (2004).
Conceptual Coexistence, Selection, and Inhibition
The coexistence of intuitive and disciplinary knowledge challenges accounts of learning framed solely as deletion. In speeded-reasoning research, participants may give scientifically correct answers more slowly when those answers contradict familiar intuitions. This pattern can be explained by continued activation of a competing response, greater control requirements, or both. It makes inhibition and context-sensitive selection plausible mechanisms of competent scientific reasoning, although the behavior alone cannot distinguish all possible representational architectures Shtulman & Valcarcel (2012).
An important consequence is that apparent resistance after instruction may have more than one cause. A learner might lack a scientific representation, fail to recognize when it applies, or lack time to inhibit an immediate alternative. These causes invite different interventions and measurements. Timed judgments probe competition; unconstrained explanations probe explicit organization; delayed far-transfer tasks test whether the new relation can be generated under changed circumstances. Evidence from any one of these indicators should be interpreted as evidence for the particular capacity it measures Shtulman & Valcarcel (2012) diSessa & Sherin (1998).
This view also clarifies why conceptual mastery can remain effortful. If scientific reasoning repeatedly recruits a newer, less automatic relation, success may depend on deliberate checking even long after formal instruction. Expertise may partly alter speed, cue selection, and the integration of representations while leaving some intuitions accessible. The research agenda should therefore follow individual learners across increasing experience, separating stable accuracy from automaticity and investigating which features of a context still provoke the earlier pattern Kalyuga (2007) Shtulman & Valcarcel (2012).
Refutation Texts and Knowledge Revision During Reading
Refutation text is an instructional genre that identifies an alternative belief, states why it fails under specified conditions, and explains a more adequate account. Its psychological rationale is that correcting a misconception requires activation of relevant prior knowledge and integration of an alternative causal explanation. The design differs from simply presenting the correct fact: it explicitly addresses the relation between competing interpretations. Reviews of science reading research report that such texts can improve learning compared with conventional expository presentations, although effectiveness depends on the content, reader, and outcome Tippett (2010) Sinatra & Broughton (2011).
Kendeou, Walsh, Smith, and O'Brien experimentally investigated knowledge-revision processes during reading. Their work studied how refutation and explanation influence the disruption caused by a previously held commonsense belief, offering a process-level complement to posttest comparisons. The key interpretation is that a text can help readers activate and revise an otherwise interfering representation while comprehension is unfolding. Yet reduced reading disruption and accurate posttest answers are different indicators; whether revision endures and transfers must be determined independently Kendeou et al. (2014).
Refutation works through meaning, not through the mere presence of a sentence denying an error. Readers need a comprehensible alternative with explanatory reach, a reason the earlier expectation was attractive, and evidence that discriminates the accounts. Poorly matched refutations may repeat an unfamiliar false claim or fail to supply the background needed to understand the correction. A strong evaluation compares explanations of equal quality and reading demands, probes delayed misconceptions, and records the baseline knowledge from which any revision occurs Sinatra & Broughton (2011) Kendeou & van den Broek (2007).
Explanation, Self-Explanation, and Concept Mapping
Explanation can change learning because it forces a person to identify which relations make an observation intelligible. Lombrozo described explanatory reasoning as a process that connects new information with prior beliefs and influences causal inference and generalization. In instruction, asking learners to predict, justify, and contrast explanations can reveal missing mechanisms. The value of explanation depends on accuracy and feedback, because a confidently generated narrative can also consolidate an unsupported causal relation Lombrozo (2006) National Research Council (2000).
Concept maps externalize propositions and links among terms, offering a way to inspect how learners organize a topic. Schroeder and colleagues synthesized studies of studying and constructing maps and reported a moderate average benefit for learning, with larger observed effects for constructing maps than for studying supplied maps under their respective comparisons. The meta-analysis does not imply that every map demonstrates conceptual change. An elaborate map may reproduce taught associations while preserving the same causal misunderstanding in applied reasoning Schroeder et al. (2018).
A defensible conceptual-change use of mapping examines change in relationships: whether a learner distinguishes cause from correlation, condition from consequence, or related processes that were previously conflated. Preinstruction maps can diagnose starting relations; subsequent maps can show revision; targeted transfer tasks determine whether those revisions guide new decisions. Researchers should assess interrater agreement for coding links and protect against a misleading gain due simply to increased familiarity with mapping conventions Schroeder et al. (2018) Treagust (1988).
Bridging Analogies and Contrasting Cases
A bridging analogy links an intuitive example that learners already understand to a more difficult target through intermediate cases that preserve relevant relational structure. Its aim is to make a new scientific inference plausible by reusing a productive precursor. Structure-mapping theory provides a rationale: a successful transfer maps explanatory relations rather than superficial features. The instructional test is whether learners can state what remains invariant across examples and where the analogy breaks down Gentner (1983) Smith et al. (1994).
Contrasting cases provide a different but compatible route. By varying features systematically, they invite learners to compare structures and generate rules that explain both successful and failed examples. Schwartz and colleagues' research on practicing versus inventing with contrasting cases demonstrated the importance of when telling occurs relative to student exploration. The evidence concerns particular experimental comparisons in adolescent physics learning; it does not license treating unguided discovery as preferable across contexts. Effective use requires feedback, carefully chosen contrasts, and a subsequent explanation that formalizes what students have noticed Schwartz et al. (2011) Mayer (2004).
A common weakness of analogical instruction is analogical overextension. The analogy can introduce its own misconception if an irrelevant source feature is mistaken for a law of the target domain. A more complete design elicits the initial analogy, uses it to predict a shared relation, then supplies a contrasting instance where the analogy fails. Transfer evaluation should include cases that reward recognizing the boundary, not merely cases that repeat the successful original correspondence Gentner (1983) Talanquer (2006).
Interactive Engagement and Guided Conceptual Inquiry
Interactive engagement combines predictions, discussion, feedback, demonstrations, and opportunities to revise explanations. Its educational significance derives from the way participation makes prior knowledge observable and gives instructors multiple occasions to test emerging concepts. In Hake's survey of introductory mechanics classes, 62 courses included 6,542 students. The reported average normalized gain was substantially higher for the interactive-engagement group than for traditional courses. Because this evidence aggregates nonrandom course comparisons, it supports a broad instructional association more directly than it isolates the specific causal effect of any single conceptual-change component Hake (1998).
Intervention design should isolate whether gains arise from immediate response feedback, time on task, peer explanation, instructor scaffolding, better alignment between teaching and assessment, or a combination. A well-designed classroom trial can randomize sections or instructional units where feasible, record implementation fidelity, and measure delayed conceptual transfer. Qualitative observations can then illuminate how different learners interpret the same demonstration or discussion, while the quantitative outcomes determine how reliably the intervention changes learning under that design Pacaci et al. (2024) Hake (1998).
Mayer's examination of discovery learning challenged a simple equation between active cognition and minimal instructional guidance. The strongest conceptual-change activities combine learner interpretation with structured explanations and appropriately timed support. Opportunities to be wrong can be productive when the task allows learners to understand why a prediction failed, while unsupported exposure to complex phenomena may overwhelm learners who lack essential background. The appropriate balance depends on concept structure, learner expertise, and assessment goals Mayer (2004) Kalyuga (2007).
Expertise-Sensitive Scaffolding and Cognitive Demand
Prior knowledge changes the functional difficulty of instruction. A novice may need worked examples, explicit definitions, and guidance about which features of a display are relevant; an experienced student may be slowed by the same redundant explanations. Kalyuga's review of expertise reversal integrates evidence that instructional supports can change their relative value as knowledge increases. This interaction creates a practical problem for concept-change interventions delivered uniformly to students who possess different initial models Kalyuga (2007).
Adaptation begins with diagnosis and should continue through formative probes. For an unfamiliar principle, the instructor can supply a clear explanation and a closely guided prediction; for an intermediate learner, a contrasting case can require reconstructing a relation; for a proficient learner, counterexamples and transfer tasks can expose hidden limits. This is an instructional design hypothesis that must be tested rather than a verified schedule for all disciplines. Its merit is that it makes support contingent on observed performance instead of nominal grade or age Kalyuga (2007) National Research Council (2000).
Cognitive demand complicates interventions intended to create conflict. If a surprising experiment requires substantial procedural, mathematical, or linguistic processing, learners may struggle with those demands while the conceptual claim remains opaque. A fair test of a mechanism therefore equates extraneous demands across conditions when possible and measures relevant baseline fluency. The intended outcome is the representation of a scientific relationship, so designers should simplify incidental features while preserving the inferential conflict central to that relationship Mayer (2004) Pacaci et al. (2024).
Evidence-Based Sequencing of Conceptual Instruction
An instructional sequence should make the learner's initial explanation explicit before introducing incompatible evidence. A prediction prompt identifies what the learner expects and why; a carefully chosen observation tests that expectation; a scientific explanation supplies an alternative relation; a contrasting case tests the limits of both accounts; and a delayed task evaluates whether the resulting understanding survives withdrawal of cues. This sequence is a synthesis of mechanisms described in the accommodation, anomalous-data, and refutation literatures, not a universally validated five-step treatment Posner et al. (1982) Chinn & Brewer (1993) Sinatra & Broughton (2011).
Sequence comparisons are more informative than declarations that one activity is intrinsically effective. Refutation may precede a demonstration or follow it; an analogy may prepare an explanation or create a false expectation if introduced without limits. Similarly, guided invention may aid later learning when contrasting cases have a useful underlying structure. In each instance, instructional order changes which prior knowledge is recruited and when a learner has enough support to integrate a discrepancy. Trials should predefine the target concept, transfer criterion, instructional dose, and plausible competing explanations Schwartz et al. (2011) Pacaci et al. (2024).
A cumulative program of research would compare components within the same topic while tracking individual starting points. Factorial designs could vary prediction elicitation, refutation, explanation, and feedback; process measures could identify interpretation during learning; long-delay tests could examine reconstruction and retention. Such designs would move the field from broad intervention labels toward an account of why specific learners revise specific concepts under specific circumstances Kendeou et al. (2014) Treagust (1988).
Diagnostic Interviews and Misconception Probes
Conceptual-change assessment begins with an explicit account of the target inference. A diagnostic interview can ask a learner to predict an outcome, explain the causal relation, compare a discrepant event, and reconsider the prediction after seeing evidence. Its strength is the opportunity to distinguish the reasons behind superficially identical answers. Its weakness is that the interviewer may introduce cues, demand characteristics, or terms that alter the explanation being elicited. Researchers can address these threats by using a standardized core protocol, counterbalanced prompts, independent coding of responses, and prompts testing consequences the interviewer has not mentioned Treagust (1988) Vosniadou & Brewer (1992).
Two-tier diagnostic items extend selected-response assessment by pairing a choice of outcome with a choice of reason. Treagust's framework for developing diagnostic tests emphasizes eliciting known alternative conceptions, translating them into response options, and checking the resulting instrument empirically. A correct prediction with an incorrect reason, or a correct reason paired with an incorrect prediction, can be especially informative. Yet such pairings remain constrained by the options offered; a learner may recognize a plausible answer without generating the explanation independently. Confidence ratings and open-ended follow-ups improve interpretability Treagust (1988) Hestenes et al. (1992).
The unit of diagnosis should match the hypothesis. A single misleading item may assess recognition of vocabulary; a set of varied cases can assess whether an explanation transfers; an interview over time can detect changes in a learner's conceptual connections. When instruments are used with culturally or linguistically diverse samples, checking the meaning of terms and the equivalence of response processes is essential. The quality of a conceptual-change claim depends on a demonstrable connection between the construct, the elicitation method, and the inference drawn from the resulting answer National Research Council (2000) Treagust (1988).
Construct Validity, Response Processes, and Measurement
Conceptual change is a latent construct inferred from observable performances. Researchers commonly measure it using concept inventories, explanation quality, drawing or model-construction tasks, classification judgments, reading comprehension, and applied problem solving. These measures overlap imperfectly. Improvement in an inventory score might reflect retrieval of a taught rule, while improvement in an explanation task might reflect verbal fluency or familiarity with the rubric. Construct validity consequently requires an articulated measurement model and evidence that indicators are sensitive to the intended conceptual relations Hestenes et al. (1992) diSessa & Sherin (1998).
Response-process evidence can discriminate competing explanations for a score. Verbal protocols reveal what a learner considered; interviews clarify the meaning assigned to a question; timed judgments probe the availability or competition of intuitive and disciplinary responses. These methods also create different demands. A speeded task increases the role of response selection and language processing, whereas an unhurried essay recruits reflective explanation and writing ability. A multi-method design permits investigators to ask whether the same change appears under conditions that tap different capacities Shtulman & Valcarcel (2012) Kendeou et al. (2014).
Measurement requires attention to item difficulty, ceiling and floor effects, and changes in task interpretation between pretest and posttest. Repeating the same items can generate practice and memory benefits independent of knowledge restructuring. Parallel forms, novel predictions, and blind scoring reduce selected threats. When learners are nested within classes, analyses should account for shared instructional conditions, and comparisons across groups should avoid equating gains calculated from different baseline score distributions. A strong report therefore presents the scoring rubric, assessment timing, baseline differences, reliability evidence, and examples of response categories Treagust (1988) Hake (1998).
Longitudinal Trajectories, Delayed Transfer, and Durability
Immediate postinstruction success answers an important but restricted question: whether learners can give targeted responses after recent exposure. Durable conceptual change requires additional evidence. A delayed assessment can examine whether the new explanation remains accessible after instructional reminders disappear, while transfer items can test whether it governs unfamiliar cases. An intervention that boosts the immediate test without changing delayed prediction may have increased surface familiarity rather than modified a stable explanatory organization Pacaci et al. (2024) Shtulman & Valcarcel (2012).
Longitudinal assessment also makes coexistence visible. A learner may initially alternate between a scientific explanation and a familiar heuristic, later select the scientific one with deliberation, and eventually apply it more fluently. Repeated assessments should measure both accuracy and the conditions that elicit a response. Whenever possible, the same conceptual relation should be probed through different representational formats so that a gain is not equated with memorizing one diagram or wording. These trajectories remain hypotheses until individual-level repeated measures support them diSessa & Sherin (1998) Shtulman & Valcarcel (2012).
The interpretation of delay depends on retention interval, intervening instruction, and opportunities to use the knowledge outside a study. A follow-up after several weeks has a different evidentiary meaning from a follow-up conducted the next day, and a student exposed to further classroom work is undergoing a different intervention history. Studies should report these conditions and consider measurement attrition. The central evidence for durable understanding is flexible use of the relevant relation at meaningful intervals under conditions that are not identical to training National Research Council (2000) Kalyuga (2007).
Experimental Identification and Competing Causes
Claims about the cause of conceptual change require a design capable of isolating instructional components. Random assignment of learners or classes can reduce selection bias, but classroom interventions also raise practical issues of teacher effects, contamination, fidelity, and dependence among students sharing the same lesson. Observational comparisons can reveal associations between activity types and learning outcomes, yet they cannot by themselves rule out prior achievement, voluntary participation, or instructor expertise. The instructive history of interactive-engagement research illustrates the difference between large-scale educational patterns and experimentally identified mechanisms Hake (1998).
An explanation-plus-conflict intervention differs from a conflict-only intervention in several ways: it may provide more instructional time, higher-quality examples, and more direct alignment with the evaluation items. A factorial design could vary anomaly exposure and explanatory support separately while holding duration and test conditions constant. Such a study would assess whether apparent conflict effects depend on the availability of a constructive explanatory alternative. Similarly, comparison of refutation texts with matched expository texts should equate readability and the quality of the scientific explanation Chinn & Brewer (1993) Kendeou et al. (2014).
Participants' prior understanding is both the starting condition and a likely effect modifier. A student already using a scientific representation can respond differently from one whose reasoning is dominated by an everyday heuristic. Preregistered subgroup analyses, adequately measured baseline knowledge, and independent transfer outcomes would enable more precise causal estimates. At the same time, researchers should avoid dividing continuous knowledge into arbitrary novice and expert classes after inspecting results. The logic of expertise reversal makes moderation theoretically plausible, but identifying it depends on transparent comparative evidence Kalyuga (2007) Pacaci et al. (2024).
Quantitative Syntheses and Their Evidentiary Scope
Pacaci, Ustun, and Ozdemir's 2024 meta-analysis assembled 218 independent studies of conceptual-change strategies in science education, representing 18,051 participants. It reported a positive overall standardized effect, approximately Hedges' g = 1.10 with a 95% confidence interval of 1.01 to 1.19, while detecting substantial heterogeneity. The size of this pooled estimate should not be read as a universal effect on deep conceptual restructuring: studies differed in methods, subjects, comparisons, and outcome instruments. The authors grouped studies into cognitive-conflict, bridging-analogy, and ontological-shift approaches, enabling a broader comparison but not eliminating the diversity of mechanisms and measurement practices Pacaci et al. (2024).
A separate review by Aleknavičiūtė, Lehtinen, and Södervik focused on intervention research in biology over roughly three decades. Their analysis reported advantages for conceptual-change interventions while examining methodological differences, including publication bias and the prevalence of outcomes capturing knowledge enrichment rather than conceptual reconstruction. This distinction matters because an educational intervention may successfully expand what a learner can state and still leave an incompatible earlier explanation available for later use. Pooling both outcomes without a construct-level distinction obscures what has changed Aleknavičiūtė et al. (2023).
Research syntheses are particularly informative when they report effect distributions, comparisons and assessment quality rather than solely an aggregate benefit. Researcher-developed instruments can closely reproduce the language and scenarios of an intervention, inflating apparent transfer. Small-study effects, selective publication, and clustered assignment can further influence estimates. A responsible synthesis treats meta-analytic gains as evidence that interventions can improve measured scientific learning and separately asks which primary studies demonstrate structural and durable conceptual revision under rigorous follow-up Pacaci et al. (2024) Aleknavičiūtė et al. (2023).
Replication, Heterogeneity, and Evidence Quality
The research record demonstrates broad plausibility for instructional strategies that engage prior knowledge, elicit prediction, and support scientific explanation. It also demonstrates heterogeneity. Different learner groups, assessment instruments, subject matters, and instructional comparisons can yield different outcomes even under the same intervention label. Cross-study variation invites mechanistic explanation: a cognitive-conflict activity may be effective when it makes a causal difference observable and ineffective when it depends on background knowledge that students lack Pacaci et al. (2024) Sinatra & Broughton (2011).
Independent replication should reproduce the critical manipulation and the measurement of the proposed conceptual mechanism in different classrooms and cultural settings. Replication of a score increase is useful, but claims of conceptual reconstruction demand replication of transfer, reasoning patterns, or the inferred change in explanatory relations. Reporting materials, coding rules, attrition, missing data, and analyses would allow future investigators to distinguish an idiosyncratic successful lesson from a generalizable process. Transparent evidence review also requires attention to published null findings and to interventions that produce shallow or short-lived changes Aleknavičiūtė et al. (2023) Schroeder et al. (2018).
The field benefits from treating theory and evidence as reciprocally constraining. A coherent theoretical account supplies predictions about what should change, for whom, and in which situations. High-quality primary studies test those predictions with instruments that differentiate conceptual mechanisms. Evidence synthesis then evaluates whether the predicted pattern recurs beyond a single research group or educational context. No intervention category resolves these questions automatically; explanatory progress requires increasingly discriminating designs and cumulative replication diSessa & Sherin (1998) Pacaci et al. (2024).
Individual Differences, Expertise, and Epistemic Engagement
Learners bring different combinations of subject knowledge, confidence, explanatory habits, and educational experience to the same scientific problem. These differences matter because a contradiction is noticed only against an expectation, and a new model becomes plausible through relationships the learner is able to evaluate. What appears to one student as an explanatory breakthrough may appear to another as an unfamiliar collection of terms. Effective research records specific baseline knowledge and the mechanisms of engagement instead of treating chronological age or grade as an adequate substitute for conceptual starting point Posner et al. (1982) Kalyuga (2007).
Epistemic engagement refers here to evaluating why an explanation should be accepted, how observations discriminate among alternatives, and what counts as a justified inference. Its contribution to conceptual change can be tested by manipulating opportunities to compare competing accounts or justify predictions while measuring whether these activities improve explanation and transfer. A learner's willingness to express uncertainty or revise a belief is context-sensitive; interview prompts, teacher feedback, and assessment stakes may all influence it. Strong evidence therefore separates measured reasoning from an assumed stable personal disposition Chinn & Brewer (1993) National Research Council (2000).
Expertise-sensitive designs can respond to differences without building a permanent taxonomy of students. Periodic probes reveal which relationships are accessible and which conditions trigger failure. Learners with little background may benefit from clear explanatory examples; those with partial understanding may benefit from discriminating contrasts; those with high fluency may require subtle transfer cases that expose unrecognized limits. These are adaptable instructional proposals supported by a theory of moderation, not a single proven sequence for every individual or discipline Kalyuga (2007) Schwartz et al. (2011).
Language, Culture, and the Ecology of Classroom Explanation
Scientific knowledge enters learning environments through language, diagrams, material practices, and the institutions that authorize explanatory claims. A term such as force, energy, living, or theory may carry an everyday meaning that partially overlaps with its disciplinary use. Consequently, an apparent misconception sometimes reflects a conflict among meanings or pragmatic interpretations rather than a fully articulated countertheory. A rigorous interview examines what learners understand the question to ask before treating their answer as evidence about a scientific representation Treagust (1988) Talanquer (2006).
Cultural experience influences which situations are familiar, which analogies are intelligible, and which forms of evidence are persuasive. A classroom example drawn from one kind of home, tool, or environment may recruit different starting knowledge in another community. Research should test whether an assessment item retains the same meaning after translation and whether a proposed mechanism survives culturally varied materials. These are methodological requirements inferred from the social and linguistic setting of learning; evidence from one national curriculum should be generalized only within the limits supported by its design National Research Council (2000) Kendeou & van den Broek (2007).
Classroom dialogue creates opportunities to make relations visible through questions, competing predictions, and the reformulation of explanations. Discussion, however, is not intrinsically a mechanism of scientific revision. Its effects depend on the evidence considered, the quality of conceptual guidance, and whether learners must justify their eventual position. The instructional ideal is a setting in which revising an interpretation is normal intellectual practice, and progress is judged by explanatory adequacy across situations rather than by the willingness to repeat an authoritative sentence Hake (1998) Mayer (2004).
Teacher Knowledge, Curriculum, and Institutional Conditions
Teachers translate research constructs into choices about demonstrations, diagnostic questions, explanations, and assessments. Their work begins with selecting an inferential target narrow enough to investigate and important enough to organize related knowledge. A curriculum that lists numerous terms without relationships may improve verbal familiarity while leaving causal organization weak. A concept-centered sequence makes relationships explicit, revisits boundary cases, and uses evaluation questions that require applying rather than naming the organizing principle. This principle is compatible with reforms in chemistry education organized around ways of reasoning Talanquer & Pollard (2010).
Institutional pressures shape what can be measured. A high-stakes short examination rewards fast responses to familiar formats; a longitudinal investigation requires time for explanation, repeated observation, and feedback. The resulting mismatch can lead educators to infer conceptual mastery from an increase in rehearsed performance. Valid evaluation should include both practical achievement and concept-specific indicators of reasoning, and it should disclose how far instructional materials resemble test materials Hestenes et al. (1992) Pacaci et al. (2024).
Teacher professional development can draw on diagnostic examples showing how the same apparently wrong answer arises through different cognitive routes. Such examples invite teachers to ask whether a learner is struggling with a representation, a term, an inferential connection, or the selection of relevant prior knowledge. The effectiveness of professional development itself requires direct evidence, ideally studies tracing changes in teacher decisions to changes in student understanding. The present evidence provides a rationale for diagnosis-led teaching without establishing that one professional-development format universally improves conceptual change Treagust (1988) National Research Council (2000).
Human–Artificial Instructional Systems in the Artificial Era
Artificially generated explanations, adaptive tutoring, and dialogue systems introduce new instructional conditions for prior-knowledge research. An artificial tutor can present varied examples, elicit predictions, and adapt feedback to responses. In the Artificial Era, the substantive psychological question is whether these interactions improve the learner's causal explanation, transfer, and discrimination among competing models. This dossier treats human–artificial education as a research frontier and proposes no established effect specific to generative tutoring without intervention evidence designed for that outcome Kalyuga (2007) National Research Council (2000).
An adaptive system requires a model of the learner's current interpretation. A wrong answer alone is ambiguous: it may reflect a robust alternative conception, a misleading prompt, missing background knowledge, or accidental response selection. Useful adaptation would therefore probe reasons, ask follow-up predictions, and track whether the learner's explanation generalizes to unseen cases. Experiments can compare scripted or human-guided instruction with adaptive artificial dialogue while equating instructional dose, content quality, and teacher involvement. Delayed conceptual transfer, rather than subjective satisfaction with the tutor, should be a central outcome Treagust (1988) diSessa & Sherin (1998).
Artificial explanation also creates a distinct epistemic problem: fluent language may make a scientifically mistaken causal account appear plausible. Systems should support inspection of evidence and acknowledge uncertainty, with source verification independent of rhetorical coherence. Future comparative studies can test whether exposing learners to contrasting explanations, asking them to challenge model claims, and supplying verifiable original sources enhances conceptual discrimination. The broader research opportunity is to study how humans organize and revise knowledge when explanatory partners can generate unlimited examples while possessing variable reliability Chinn & Brewer (1993) Sinatra & Broughton (2011).
Critical Synthesis of the Scientific Explanations
The literature supports an integrated account in which prior knowledge acts as both interpretive resource and constraint on scientific learning. Accommodation theory identifies epistemic conditions for accepting a new account; knowledge-in-pieces theories illuminate local interpretive resources and their coordination; framework models emphasize broader commitments; ontological accounts locate certain difficulties in category assignment; coexistence research documents continuing competition after instruction. These accounts propose partly different units of explanation and may each be informative at the scale for which they were formulated. Their empirical value depends on distinct, testable predictions rather than on verbal compatibility alone Posner et al. (1982) diSessa (1993) Vosniadou & Brewer (1992) Chi et al. (1994) Shtulman & Valcarcel (2012).
The intervention record demonstrates that conceptual-change approaches can improve measured science learning, with unusually large average effects in some syntheses. That conclusion has an essential measurement qualification. Many studies assess targeted content soon after instruction using instruments close to the material taught, and designs vary in their control of competing causes. Conceptual reconstruction is a stronger claim than an achievement gain: it requires evidence about the structure, range, durability, and explanatory adequacy of the learner's new representation. Meta-analysis improves the precision of observed outcome estimates without resolving ambiguity in what each outcome measures Pacaci et al. (2024) Aleknavičiūtė et al. (2023).
A productive synthesis treats conceptual change as an empirical family of processes. Acquisition adds an explanatory relation; differentiation separates concepts previously fused; reorganization changes how ideas support one another; conditionalization defines the situations in which an inference applies; inhibitory selection permits scientific reasoning despite an accessible competing intuition. These categories are analytic distinctions drawn from the reviewed traditions, not a newly validated taxonomy. They direct investigators toward tasks that can determine what changed and avoid equating the disappearance of an incorrect response with deletion of the underlying knowledge diSessa & Sherin (1998) Shtulman & Valcarcel (2012) Smith et al. (1994).
Open Research Problems and an Empirical Agenda
The first open question concerns representational identification: when does a response pattern warrant inference to a coherent belief, a local heuristic, an organized model, or a competition between representations? Resolving this problem requires multiple tasks selected to generate diverging predictions for rival theories. Researchers should preregister how responses will be interpreted and retain room for novel, unanticipated explanations. Cross-method convergence would improve the interpretability of observed change, while transparent cases of disagreement would refine the boundaries of existing constructs diSessa (1993) Vosniadou & Brewer (1992).
The second problem concerns causal mechanism and long-term development. Comparative experiments should isolate how expectation elicitation, evidence quality, explanatory support, analogy, feedback, and learner expertise interact. The same individuals should be followed beyond the immediate posttest, with delayed transfer and alternative contexts that can expose residual intuitions. The analytic goal is to determine whether successful learning requires representational replacement, reorganization, condition-sensitive retrieval, or some combination, and which designs can distinguish those outcomes Chinn & Brewer (1993) Kalyuga (2007) Shtulman & Valcarcel (2012).
The third problem concerns generalizability and evidence integrity. Studies would benefit from independent replications, culturally and linguistically validated assessments, accessible reporting of measures and materials, and clear separation between general achievement and conceptual reconstruction. Synthetic reviews should distinguish observed effect sizes from construct-level evidence and examine differences in follow-up, instructional intensity, and instrument proximity. Research on human–artificial teaching belongs within the same methodological standard, with a particular need to test explanation reliability and knowledge transfer rather than mere apparent conversational success Pacaci et al. (2024) Aleknavičiūtė et al. (2023) Treagust (1988).
Conclusion
Prior knowledge, misconceptions, and conceptual change form a coherent research object at the intersection of educational, cognitive, and developmental psychology. The central phenomenon is the transformation of understanding when new disciplinary explanations meet representations already acquired through experience and instruction. Those earlier representations organize interpretation, support useful inference, and sometimes generate expectations that conflict with scientific models. The study of learning consequently requires diagnosis of existing relations and observation of how they are revised, coordinated, selected, or retained National Research Council (2000) Smith et al. (1994).
The historical progression from developmental interviews and alternative-conception studies to concept inventories, competing psychological theories, process measures, and large-scale syntheses has produced a substantial body of evidence. Researchers have identified recurring intuitive patterns in mechanics, astronomy, biology, and chemistry, demonstrated that scientific knowledge may coexist with earlier intuitions, and tested instructional strategies designed to promote more adequate explanation. The strongest interpretation of the intervention literature is that targeted conceptual-change instruction can improve measured learning; whether a particular study demonstrates durable restructuring depends on its outcomes and design Clement (1982) Vosniadou & Brewer (1992) Shtulman & Valcarcel (2012) Pacaci et al. (2024).
The continuing scientific task is precise identification of what changes. A correct definition, a successful calculation, a revised explanation, and flexible delayed transfer supply different forms of evidence. Mature conceptual-change research connects the knowledge a learner begins with to the mechanisms used during instruction and the conditions under which new understanding governs future reasoning. This approach establishes conceptual change as a testable problem of psychological organization and provides a rigorous basis for investigating educational practice, explanatory technologies, and knowledge development across the Artificial Era diSessa & Sherin (1998) Kendeou et al. (2014) Aleknavičiūtė et al. (2023).
References
- https://doi.org/10.1016/j.edurev.2023.100556
- Carey, S. (1985). Conceptual change in childhood. MIT Press. https://mitpress.mit.edu/9780262031103/conceptual-change-in-childhood/
- Chi, M. T. H., Slotta, J. D., & de Leeuw, N. (1994). From things to processes: A theory of conceptual change for learning science concepts. Learning and Instruction, 4(1), 27–43. https://doi.org/10.1016/0959-4752(94)90017-5
- Chinn, C. A., & Brewer, W. F. (1993). The role of anomalous data in knowledge acquisition: A theoretical framework and implications for science instruction. Review of Educational Research, 63(1), 1–49. https://doi.org/10.3102/00346543063001001
- Clement, J. (1982). Students' preconceptions in introductory mechanics. American Journal of Physics, 50(1), 66–71. https://doi.org/10.1119/1.12989
- diSessa, A. A. (1993). Toward an epistemology of physics. Cognition and Instruction, 10(2–3), 105–225. https://doi.org/10.1080/07370008.1985.9649008
- diSessa, A. A., & Sherin, B. L. (1998). What changes in conceptual change? International Journal of Science Education, 20(10), 1155–1191. https://doi.org/10.1080/0950069980201002
- Gentner, D. (1983). Structure-mapping: A theoretical framework for analogy. Cognitive Science, 7(2), 155–170. https://doi.org/10.1207/s15516709cog0702_3
- Hake, R. R. (1998). Interactive-engagement versus traditional methods: A six-thousand-student survey of mechanics test data for introductory physics courses. American Journal of Physics, 66(1), 64–74. https://doi.org/10.1119/1.18809
- Halloun, I. A., & Hestenes, D. (1985). Common sense concepts about motion. American Journal of Physics, 53(11), 1056–1065. https://doi.org/10.1119/1.14031
- Hestenes, D., Wells, M., & Swackhamer, G. (1992). Force concept inventory. The Physics Teacher, 30(3), 141–158. https://doi.org/10.1119/1.2343497
- Kalyuga, S. (2007). Expertise reversal effect and its implications for learner-tailored instruction. Educational Psychology Review, 19(4), 509–539. https://doi.org/10.1007/s10648-007-9054-3
- Kendeou, P., & van den Broek, P. (2007). The effects of prior knowledge and text structure on comprehension processes during reading of scientific texts. Memory & Cognition, 35(7), 1567–1577. https://doi.org/10.3758/BF03193491
- Kendeou, P., Walsh, E. K., Smith, E. R., & O'Brien, E. J. (2014). Knowledge revision processes in refutation texts. Discourse Processes, 51(5–6), 374–397. https://doi.org/10.1080/0163853X.2014.913961
- Lombrozo, T. (2006). The structure and function of explanations. Trends in Cognitive Sciences, 10(10), 464–470. https://doi.org/10.1016/j.tics.2006.08.004
- Mayer, R. E. (2004). Should there be a three-strikes rule against pure discovery learning? The case for guided methods of instruction. American Psychologist, 59(1), 14–19. https://doi.org/10.1037/0003-066X.59.1.14
- McCloskey, M., Washburn, A., & Felch, L. (1983). Intuitive physics: The straight-down belief and its origin. Journal of Experimental Psychology: Learning, Memory, and Cognition, 9(4), 636–649. https://doi.org/10.1037/0278-7393.9.4.636
- National Research Council. (2000). How people learn: Brain, mind, experience, and school (expanded ed.). National Academies Press. https://doi.org/10.17226/9853
- Nobes, G., Martin, A. E., & Panagiotaki, G. (2005). The development of scientific knowledge of the Earth. British Journal of Developmental Psychology, 23(1), 47–66. https://doi.org/10.1348/026151004X20649
- Pacaci, C., Ustun, U., & Ozdemir, O. F. (2024). Effectiveness of conceptual change strategies in science education: A meta-analysis. Journal of Research in Science Teaching, 61(6), 1263–1325. https://doi.org/10.1002/tea.21887
- Posner, G. J., Strike, K. A., Hewson, P. W., & Gertzog, W. A. (1982). Accommodation of a scientific conception: Toward a theory of conceptual change. Science Education, 66(2), 211–227. https://doi.org/10.1002/sce.3730660207
- Schroeder, N. L., Nesbit, J. C., Anguiano, C. J., & Adesope, O. O. (2018). Studying and constructing concept maps: A meta-analysis. Educational Psychology Review, 30(2), 431–455. https://doi.org/10.1007/s10648-017-9403-9
- Schwartz, D. L., Chase, C. C., Oppezzo, M. A., & Chin, D. B. (2011). Practicing versus inventing with contrasting cases: The effects of telling first on learning and transfer. Journal of Educational Psychology, 103(4), 759–775. https://doi.org/10.1037/a0025140
- Shtulman, A., & Valcarcel, J. (2012). Scientific knowledge suppresses but does not supplant earlier intuitions. Cognition, 124(2), 209–215. https://doi.org/10.1016/j.cognition.2012.04.005
- Sinatra, G. M., & Broughton, S. H. (2011). Bridging reading comprehension and conceptual change in science education: The promise of refutation text. Reading Research Quarterly, 46(4), 374–393. https://doi.org/10.1002/RRQ.005
- Smith, J. P., III, diSessa, A. A., & Roschelle, J. (1994). Misconceptions reconceived: A constructivist analysis of knowledge in transition. Journal of the Learning Sciences, 3(2), 115–163. https://doi.org/10.1207/s15327809jls0302_1
- Talanquer, V. (2006). Commonsense chemistry: A model for understanding students' alternative conceptions. Journal of Chemical Education, 83(5), 811–816. https://doi.org/10.1021/ed083p811
- Talanquer, V., & Pollard, J. (2010). Let's teach how we think instead of what we know. Chemistry Education Research and Practice, 11, 74–83. https://doi.org/10.1039/C005349J
- Tippett, C. D. (2010). Refutation text in science education: A review of two decades of research. International Journal of Science and Mathematics Education, 8(6), 951–970. https://eric.ed.gov/?id=EJ905216
- Treagust, D. F. (1988). Development and use of diagnostic tests to evaluate students' misconceptions in science. International Journal of Science Education, 10(2), 159–169. https://doi.org/10.1080/0950069880100204
- Vosniadou, S., & Brewer, W. F. (1992). Mental models of the earth: A study of conceptual change in childhood. Cognitive Psychology, 24(4), 535–585. https://doi.org/10.1016/0010-0285(92)90018-W