Concepts and Conceptual Structure

The Philosophical Problem of Concepts

Concepts are implicated whenever a thinker recognizes an animal as an animal, judges that one proposition follows from another, considers a possible object, or learns that an apparently familiar category has an unexpected structure. These operations have different immediate purposes, yet each appears to involve repeatable elements of thought. A concept is, in a preliminary philosophical sense, an element or capacity whose recurrence helps explain how different thoughts can concern the same kind, property, relation, or object. Calling concepts elements of thought does not settle their ontology. A recurring element might be a mental representation, a publicly shareable sense, a practical ability, a functional role, or a theoretically heterogeneous family. The central philosophical task is to explain their recurrence without presupposing the answer to that ontological question (Margolis & Laurence, 2019).

Four phenomena motivate an account. First, thought is general: one can think about cats without representing every cat individually. Second, thought is recombinable: the same conceptual elements appear in judgments about a cat sleeping, a cat hunting, and a cat being larger than another animal. Third, thought is normatively assessable: identifying a whale as a fish may be comprehensible while being taxonomically incorrect. Fourth, thought is learnable: a person's classifications and inferences can change while inquiry continues to concern the same world. Each phenomenon supplies a constraint, and no single constraint is sufficient to identify the nature of concepts.

The investigation therefore proceeds through the interdependence of individuation, content, structure, possession, acquisition, and application. Individuation asks when a concept remains the same across thinkers or occasions. Content asks what the concept represents and what determines its extension. Structure asks which properties, associated representations, rules, or inferential connections belong to the concept itself. Possession asks what a thinker must be able to do to count as having it. Acquisition asks how that competence arises and changes. Application asks how a concept guides categorization, reasoning, and linguistic conduct. Treating these questions separately makes it possible to test theories by their explanatory achievements rather than by their rhetorical generality.

Definitions, Objects, and Levels of Explanation

A concept can be analyzed along at least three established ontological models. On a representational account, concepts are constituents of token mental states, perhaps structured vehicles in a cognitive system. On an ability account, to possess a concept is to have a capacity for discrimination, reasoning, or action under suitable conditions. On an abstract-object account, a concept is a shareable content or sense, instantiated or grasped by individual thinkers. These accounts answer different explanatory questions. A psychological vehicle may instantiate a meaning; a competence may involve a vehicle; and a public meaning may be grasped by several agents. Their compatibility depends on whether 'concept' designates one level or is systematically used at several levels (Margolis & Laurence, 2019; Shea, 2024).

A concept must also be distinguished from a word, a perceptual image, a category, and a proposition. Words can express concepts, while synonymous words may express a common content and ambiguous words may express different contents. A mental image presents determinate features and often lacks the generality of the concept used to interpret it. A category is a class of objects under some organization; a concept is whatever enables the class to be represented, discriminated, or reasoned about. A proposition is typically the content of a complete truth-evaluable judgment, while concepts function as constituents or capacities involved in expressing such judgments. These distinctions are substantive: a theory that identifies concepts with word meanings must account for prelinguistic thought, and a theory that identifies them with categories must distinguish the extension of a representation from its cognitive significance.

A useful methodological test asks what varies while the alleged concept remains constant. Two thinkers may have different stereotypes for the same chemical element and nevertheless refer to that element. One thinker may use one expression with two distinct modes of presentation of the same person. An individual may preserve a reference while revising a causal theory of its bearer. Conversely, two thinkers might exhibit similar typicality judgments yet mean different things by a term in scientific or institutional practice. These cases show why extension, cognitive significance, and classificatory profile cannot simply be equated.

The analysis adopted in this monograph is explanatory rather than stipulative. It uses 'concept' for a repeatably deployable component of content-bearing thought while investigating what realizes that component. This working characterization preserves the subject of inquiry and places the burden of explanation on competing theories.

Plato: Generality, Definition, and the Unity of a Kind

Plato's dialogues establish a question that later theories of concepts repeatedly reformulate: what enables many heterogeneous instances to be understood under one intelligible account? In the Meno, Socrates challenges Meno to give an account of virtue itself instead of naming the different excellences assigned to men, women, children, and other positions (72a–74b). An enumeration is informative about the extension of a term, but a definition is intended to explain why all its instances belong together. The challenge exposes the gap between collecting examples and specifying a unity that can be applied to further cases. This is a methodological beginning for the theory of concepts, even though Plato's own primary concern includes the nature of the virtue to be known and the possibility of acquiring that knowledge (Plato, Meno, 72a–74b).

The transition from several instances to one account contains a philosophical demand of considerable strength. If the learner can distinguish virtue from its apparent instances, that ability must exhibit some stability beyond the accidental features of the examples already observed. But the demand leaves open the form of this stability: it might be a definition, the grasp of a real nature, an ability to reason under a norm, or another organized capacity. Plato's questions thus establish an explanatory target without establishing a modern psychological theory of stored concepts. The subsequent history is largely an investigation into what the relevant unity requires.

In the Parmenides, Plato probes difficulties in the relationship between Forms and particular things, including participation and the problem of how a separate Form relates to the many instances allegedly characterized by it (130e–135c). The challenge matters because an ontological solution to common predication must explain how universal intelligibility arises for finite thinkers. Positing a shared Form does not automatically show how the thinker represents, learns, or applies its significance. In a later idiom, an account of the objects of general thought and an account of the structure of general thinking must be jointly intelligible, but their explanatory tasks remain distinct.

This distinction clarifies a recurring historical error. The claim that many objects are instances of a common property belongs to metaphysics; the claim that thinkers organize objects under a common representation belongs to philosophical psychology and semantics. Plato brings the questions into intimate relation without resolving them by simple identification. Their separation makes it possible to examine, for example, whether a realist about properties must endorse abstract concepts, or whether a nominalist can nonetheless maintain that conceptual contents are publicly shared. Both possibilities require further argument.

The Platonic search for an account also reveals why conceptual inquiry is normative. A poor definition can be rejected although it successfully recalls familiar examples. Adequacy concerns extension to new cases, explanatory coherence, and the ability to justify inclusion or exclusion. These standards survive transformations in the ontology of universals. Later theories inherit the question of how a thinker moves from a set of representations to an intelligible and defensible general judgment.

Ancient Foundations: Predication, Kinds, and Definition

Aristotle's Categories distinguishes homonymous from synonymous cases at 1a1–15 by reference to a shared name and the account corresponding to that name. The distinction already raises the problem of whether sameness of expression coincides with sameness of what is meant. An animal and a painted animal may share a name under an equivocal use, whereas a human being and an ox are animals under a common account. Aristotle's analysis concerns forms of predication and being, not a modern computational inventory of mental representations. Nevertheless, it provides an enduring philosophical problem: how does an expression remain applicable to many particular subjects while its account preserves determinate significance? The Greek term logos carries the force of an account, explanation, or articulated description, and a translation limited to 'word' would obscure this role (Aristotle, Categories, 1a1–15).

Aristotle's classification of what is said and what exists also prompts a distinction between a universal predicable and the individual substance to which it is attributed. A general account can organize knowledge of numerous instances while each individual remains a particular. This division is historically foundational for later questions about universals and properties. The metaphysical issue asks whether common natures are real and how they exist; the theory of concepts asks how a thinker can grasp and apply general content. They interact but have different explanatory targets. A nominalist and a realist about universals can agree that human beings categorize and reason with repeatable representations while disagreeing about the ontology of the properties represented.

Ancient theories of definition supply an additional constraint. A definition seeks an account that makes the kind intelligible and supports correct inferences about it. The classical idea of genus and difference can organize a taxonomic explanation, but its success for some scientific and logical classifications gives no general guarantee that ordinary concepts have stored necessary and sufficient conditions. The historical legacy of definition is thus an explanatory ambition: make the unity of classification intelligible through a determinate account. Contemporary experimental work tests the psychological claim that actual categorization operates according to such accounts, rather than denying the value of definitions in logic and systematic inquiry.

Late Antiquity and Medieval Accounts of Universal Signification

The transmission of Aristotle's Categories through Porphyry's Isagoge made the distinction among genus, species, difference, property, and accident foundational for medieval logical education. Porphyry famously suspended judgment on whether genera and species subsist independently or only in thought, whether they are corporeal or incorporeal, and whether they exist apart from or within sensible particulars. The philosophical importance of these questions lies in their separation of three domains that can otherwise be confused: the structure of what exists, the generality of what is conceived, and the signification of general terms. Later discussions of the problem of universals repeatedly examine the relations among these domains (Klima, 2022).

Boethius's commentary on Porphyry gave the Latin tradition a developed account of the problem. If a universal were one numerically identical object wholly shared by numerous distinct things, a metaphysical difficulty would emerge concerning the manner of such sharing. Boethian approaches explored how universal understanding could arise from particular things without requiring a separately existing universal thing to be present in each instance in precisely the way an individual exists. The resulting idea of a universal as an object of general understanding highlights an epistemological question: how can a finite singular act of understanding concern indefinitely many things? A history of concepts must recognize that this question received medieval treatments long before the vocabulary of cognitive science appeared.

Peter Abelard's Logica Ingredientibus, in discussing the problem inherited from Porphyry, makes universal signification central. He examines how a common term applies to many individuals and what explanatory relation grounds that common predication. A general term, a mental understanding, and the agreement or resemblance among particular things can be analytically distinguished. His account thus anticipates the need to explain how linguistic universality and conceptual generality relate without assuming that a common word names an independently existing universal entity. These are interpretive themes of the medieval scholarship surveyed by Klima (2022); this monograph does not present paraphrases of modern analysis as direct quotations from Abelard's Latin.

Medieval accounts associated with Thomas Aquinas further develop the cognitive role of abstraction from sensible particulars and the relation between intelligible forms and acts of understanding. Questions of species intelligibilis, intellectual reception, and judgment have their own technical meaning and should not be assimilated wholesale to twentieth-century internal representations. Nonetheless, these accounts ask how intellectual apprehension can be universal in scope despite its origin in particular experience. Medieval nominalist approaches, including the later semantic theory of William of Ockham, assign a central role to mental and spoken terms, supplying another route for investigating the unity of predication and the individuation of meanings.

A simple sequence from medieval realism to medieval nominalism conceals genuine variations. What one philosopher counts as a universal in things may be different from what another calls a universal concept or a common sign. Nor does disagreement about the existence of extra-mental universals straightforwardly predict agreement or disagreement about the psychological structure of a concept. These distinctions constrain modern historical comparisons. The central inheritance is a structured problem: one individual act can be about many things, one sign can be predicated of many things, and one general judgment can be correct or incorrect. A philosophical account must explain how these relations are possible and how far each depends upon the others.

Abstraction and Generality in Early Modern Philosophy

John Locke's Essay Concerning Human Understanding, Book III, chapter 3, addresses a striking asymmetry: existing things are particular, while most words are general. His explanation locates generality in abstract ideas that omit circumstances restricting an idea to one specific existence. A word becomes general by signifying an idea capable of representing several individuals. Locke thereby connects an account of concept acquisition, a theory of linguistic generality, and a nominal-essence approach to species. In §§6–9, abstraction makes a representation more extensive by leaving aside differentiating information; in §11, universality pertains to the signifying role of ideas and words rather than to the separate physical existence of universal words or images (Locke, 1690, III.iii).

The philosophical strength of Locke's approach lies in its explanation of economy. A finite thinker can discuss indefinite numbers of individuals without memorizing a unique name and description for every one. Its difficulty lies in the conditions of omission. If an image of a triangle necessarily has some particular geometry, what could it be for that image to have none of the properties that distinguish kinds of triangles? A general sign might work without the mind constructing a wholly indeterminate picture. The distinction between an act of selective attention and the production of a literally featureless representation becomes decisive.

George Berkeley's Principles of Human Knowledge, Introduction §§6–20, challenges the doctrine of abstract ideas when it demands a representation separated from all determinate sensible qualities. Berkeley's criticism allows particular ideas to be employed generally through their representative function. The key insight is that generality may concern how a representation is used rather than the internal indeterminacy of an image. This does not by itself provide a complete theory of semantic reference, but it removes the assumption that successful general thought must contain a special picture of an indefinite object (Berkeley, 1710).

David Hume further elaborates the account in A Treatise of Human Nature, I.i.7. His claim that ideas are particular in their character while general in their representative employment locates general thinking in the relation among particular ideas, customary association, and general terms. A given image can guide thought about many other instances because use and association extend its role. The result is an early articulation of the separation between the intrinsic qualities of a vehicle and its representational function. It anticipates the enduring question whether the cognitive structure of a concept is internal to a stored representation or distributed through dispositions to use it (Hume, 1739, I.i.7).

These three positions should not be collapsed into an empiricist doctrine with a single psychology. Locke foregrounds the construction of abstract general ideas; Berkeley contests certain abstractions; Hume develops generality through the customary employment of particular ideas. Their disagreement remains relevant to modern debates over imagery, perceptual grounding, and the distinction between a mental vehicle and the generality of its content.

Kant: Concepts, Judgment, and Discursive Understanding

Kant places concepts within a systematic account of cognition. In the Critique of Pure Reason, the distinction between intuition and concept separates the immediate and singular presentation of an object from discursive representation through a general mark. At A68/B93–A69/B94, Kant analyzes judgment through the unifying function of concepts, while A320/B376–377 distinguishes conceptual and intuitive representations. Concepts are implicated in rules of judgment and in the organization of possible cognition, not merely in the storage of names or pictures. The logical functions of judgment also inform the table of categories and the transcendental account of objective experience (Kant, 1781/1787).

Kant thereby changes the governing question. The problem is no longer solely how a general idea is abstracted from particular observations. It is also how representations can be synthesized and taken as belonging to an objectively assessable judgment. For Kant, conceptual understanding contributes the rule-governed unity through which an object can be determined in thought. Concept possession is therefore connected to an ability to judge, but the relation between empirical concepts and pure categories must be preserved. Empirical concepts have histories of comparison, reflection, and abstraction; the categories play a distinct transcendental role in Kant's account of the possibility of experience.

The schematism problem introduces a further dimension. A general conceptual rule must be applicable to the temporal and sensible particularity of experience. The schema mediates the relation between concept and intuition, and the discussion indicates that possessing a general rule does not automatically explain successful recognition of its instances. Modern distinctions between semantic content and the processing mechanism of classification can be read as responses to an analogous explanatory gap, although that modern terminology should not be attributed to Kant himself.

Kant's theory is therefore indispensable to an account of conceptual structure in two respects. It connects concepts to the inferential unity of judgments, and it requires an explanation of the application of general content to particular instances. A model that treats concepts solely as lists of familiar perceptual features cannot, without supplementation, explain the normative significance of their role in judgment.

Frege: Concept, Sense, Reference, and Thought

Frege's logical and semantic work reorganizes the analysis of content. In Über Sinn und Bedeutung (1892), he distinguishes Sinn from Bedeutung. Two expressions may have the same reference while presenting it differently, and identity statements can carry cognitive significance that cannot be accounted for solely by the objects designated. The referential result of a complete judgment and the thought it expresses belong to different explanatory levels. The distinction becomes especially important in attitude contexts, where a substitution of co-referential expressions can change what a person is reported to believe (Frege, 1892a).

An essential terminological precaution concerns the word 'concept'. Frege uses Begriff technically for a function of a certain logical kind, often understood in relation to the reference of a predicate; his Sinn is not simply what Frege himself calls a concept. Some contemporary philosophers identify concepts with Fregean senses or modes of presentation, but this is a subsequent theoretical development. Frege's Über Begriff und Gegenstand (1892) also distinguishes concepts from objects in his logical account. The distinction prevents a historical error in which twentieth-century philosophical usage is projected back onto Frege's technical vocabulary.

Frege's distinction nevertheless yields a strong contemporary argument for fine-grained conceptual individuation. Suppose an agent believes that the first person to arrive is a skilled musician but does not know that the first person to arrive is the current director. The object is identical, yet the agent's reasoning differs according to the mode under which that object is conceived. If a theory individuates concepts only by referent, it must introduce an independent explanation of these differences in cognitive significance. Fregean theories locate that explanation in differences of sense. Atomist theories can respond by recognizing distinct representations or relations without treating descriptive senses as the determinants of reference. The dispute therefore concerns whether the mode of presentation constitutes conceptual identity, rather than whether differences in modes of presentation matter.

Frege also raises the problem of compositional significance. Complete thoughts contain structured components, and the meaning of a complex expression must be explained with reference to constituent meanings and their combination. An account of concepts inherits a demand to explain systematic recombination while preserving the different cognitive roles of distinct contents.

Wittgenstein and the Open Texture of Conceptual Practice

Wittgenstein's Philosophical Investigations, particularly §§65–71, investigates the demand for one feature common to every instance covered by a word. His example of games brings forward overlapping patterns of resemblance across diverse activities. The word 'game' remains intelligible and practically applicable even when the attempt to isolate a single defining characteristic encounters persistent counterexamples. This is a challenge to the universal requirement that ordinary concepts possess classical definitions. It is not a demonstration that no concept can have rigorous boundaries; mathematical and stipulated concepts may be precisely defined. Rather, it questions the assumption that every concept must be explained through one hidden essence.

The surrounding account of language games makes application central. A concept is learned and exercised in practices where distinctions are taught, contested, corrected, and extended. Mastery may depend on recognitional capacities, participation in social forms of life, and the ability to respond appropriately under changing circumstances. The meaning of a term can be clarified by studying its uses, and a concept's apparent unity need not be represented by one list of shared qualities. The importance of this inquiry is methodological: philosophical explanation must attend to the diversity of practices in which terms are intelligible before imposing a uniform conceptual architecture.

Wittgenstein's discussion subsequently influenced prototype-oriented accounts of categorization, but the connection should be interpreted carefully. Family resemblance in philosophical analysis is not itself an experimentally specified similarity metric. A psychological model that computes similarity across features adds a theory of mental processing, memory, and decision. The difference matters because semantic openness and graded typicality are separate phenomena. A class can have determinate membership while its members vary in typicality; conversely, a vague classification can arise for reasons other than prototype comparison.

Analyticity, Explication, and the Limits of Conceptual Analysis

A major twentieth-century controversy investigated whether conceptual relations can ground truths independently of empirical facts. In a familiar analysis, a statement is analytic when its truth is explained by meaning or by the relevant semantic rules, while a synthetic statement requires further facts about the world. If philosophical problems could be resolved through the articulation of the meanings of their terms, conceptual analysis would enjoy a distinctive method of justification. That ambition appears in different forms in logical empiricism, ordinary-language philosophy, and later discussions of a priori knowledge.

Rudolf Carnap developed precise approaches to linguistic frameworks in which logical and semantic rules can be explicitly fixed. A constructed language can determine which statements count as valid or analytic under its specified rules. Carnap's notion of explication also permits the development of a more exact concept that fulfills specified scientific or philosophical purposes while preserving relevant connections with a less precise earlier use. The point of explication is methodological: conceptual improvement can be assessed by clarity, formal utility, and contribution to systematic knowledge, rather than by a requirement to recover one hidden psychological structure shared by every ordinary speaker (Carnap, 1950).

W. V. O. Quine's 'Two Dogmas of Empiricism' (1951) presses a different line of inquiry. Definitions and claims about synonymy often appear to rely on the very notions of meaning and analyticity they are supposed to explain. Quine questions the independence of analyticity as a fundamental explanation of truths and emphasizes the revisability of commitments within interconnected bodies of theory. The challenge to conceptual structure is significant: if every supposedly essential connection is insulated from counterevidence by a mere label, it becomes hard to tell when an analysis discovers an existing content and when it stipulates a new use (Quine, 1951).

The contrast between Carnap and Quine must be represented at the correct level. Carnap's specification of semantic rules for a particular constructed framework does not depend upon the empirical assertion that each natural-language speaker mentally stores those same rules. Quine's critique is particularly forceful against unsupported claims about synonymy in unrestricted natural language. In carefully specified formal systems, rules and interpretations can be stipulated and studied for their consequences. The philosophical question becomes the relation between those systems and the practices they are intended to clarify (Carnap, 1950; Quine, 1951).

This debate supplies a precise warning for theories of concepts. Failure to find a simple definition in ordinary practice does not show that precise concepts cannot be constructed; success in constructing a formal definition does not show that the same definition is cognitively realized in ordinary language. Conceptual analysis can have a reconstructive, critical, or revisionary role. The kind of success being claimed must be specified. A philosophical theory should distinguish discovery about existing semantic practice from a proposal for improved terminology, and it should state the standards by which each is assessed.

Concept Possession and Normative Competence

The possession question asks what must be true of a thinker for the attribution 'has the concept C' to be justified. A minimal answer appeals to stable capacities for deploying C across relevant contexts. Yet mere behavioral success is underdetermining: a person may repeat a word fluently while systematically misunderstanding its inferential implications, and another may competently identify instances without knowing the conventional word for the kind. Possession is therefore assessed against a network of recognitional, inferential, explanatory, and corrective abilities.

Christopher Peacocke's A Study of Concepts (1992) offers a systematic possession-condition approach that treats concepts in relation to the conditions under which thinkers grasp them. Such accounts seek an explanatory connection between a concept's identity and the capacities constitutive of possessing it. Their challenge is to explain competence without requiring every possessor to share all of the same beliefs or to execute identical methods of recognition. If possession conditions are too demanding, ordinary concept users appear to lack concepts they plainly employ. If they are too permissive, possession fails to explain the difference between understanding and imitation.

A distinction between constitutive and enabling capacities is useful. The capacity to draw an elementary consequence from a concept might be constitutive of its mastery on an inferential account; memory, attention, and sensory acuity might enable its use without belonging to its conceptual identity. The difficulty is to justify this partition. A theory may specify a normative role that an agent must be answerable to, while allowing different psychological procedures to implement that role. This leaves the issue of degrees of possession open. Child learners, novices, and experts may participate in a common classificatory practice while displaying radically different mastery.

The question becomes especially pressing when concepts are socially shared. An individual may defer to expert practice in using a chemical or legal term. This supports a distinction between full theoretical understanding, adequate participation, and bare verbal repetition. The appropriate possession condition may therefore be indexed to the kind of concept and the level of competence being attributed.

Definitional Structure and Its Explanatory Limits

The classical theory of conceptual structure proposes that a concept is constituted by a definition containing features individually necessary and jointly sufficient for category membership. The promise is explanatory economy: if a thinker possesses the defining conditions, correct categorization and many elementary inferences become available. For certain explicitly constructed mathematical concepts, definitions genuinely determine an exact extension relative to an interpretation. The difficulty is extending the same model to every ordinary kind, artifact, social role, and evaluative term.

Definitions encounter two characteristic problems. A proposed definition of an ordinary category may be contradicted by a clear instance, or it may include an object that speakers reasonably reject. Moreover, knowing a list of features does not yet explain the speed, typicality, context sensitivity, and flexibility of actual categorization. The classical theory must show both that the definitions are available to thinkers and that the relevant cognitive performance is governed by them. A philosophically adequate intensional definition might nevertheless be psychologically inaccessible as an algorithm of recognition.

The relationship between correctness and classification must be analyzed precisely. A person can classify a robin as a bird quickly because it resembles familiar birds, while classifying an unusual bird correctly after receiving genealogical information. This difference suggests that the cognitive path to an answer may differ from the standard that makes it correct. Definitional theories may model one standard for a constructed kind while prototype and theory-based mechanisms model different routes of application. This methodological distinction prevents a premature inference from the failure of a categorization mechanism to the failure of all definitional explanation.

Prototypes, Typicality, and Graded Organization

The prototype approach models a category through the distribution of characteristic features or proximity to central examples. Eleanor Rosch and Carolyn Mervis's 1975 experiments investigated relations between perceived typicality and shared category attributes, supplying empirical evidence that category members can be more or less representative. A robin is commonly judged more typical of the bird category than a penguin. This judgment does not entail that the penguin has a diminished biological status as a bird. Typicality is a graded psychological property; category membership may be evaluated through different classificatory standards (Rosch & Mervis, 1975).

Prototype models explain efficient recognition and the clustering of familiar instances. They also explain why some associations are activated more readily than others. Their conceptual challenge arises when familiar feature clusters conflict with theory-guided judgment. An animal can be visually altered to resemble another kind while retaining its biological identity. A concept's application may therefore depend on causal origin, underlying constitution, or institutional practice rather than surface similarity. Prototype information alone cannot settle the relevant kind membership in every reflective context.

Compositionality presents a second challenge. A typical pet and a typical fish do not directly determine the typical pet fish; the combination may evoke an especially colorful aquarium animal. Complex concepts may have features that are not inherited from the ordinary prototypes of their parts. This is not a decisive argument against psychologically real prototypes. It is an argument against treating their mere combination as the complete semantics of complex thought. Prototype theory becomes stronger when its scope is specified: it explains some dimensions of cognitive organization and rapid categorization, while additional mechanisms explain reference, inferential revision, and complex content (Margolis & Laurence, 2019).

Exemplars, Contextual Categories, and Cognitive Flexibility

Exemplar accounts explain categorization through the storage and retrieval of encountered individual instances. A new object can be classified by comparison with remembered cases rather than by comparison with one averaged representation. Such models capture sensitivity to unusual subgroups, experience-dependent variation, and the influence of individual learning histories. A person may recognize an uncommon species because it resembles a particular previously encountered specimen, even when it diverges markedly from the category prototype.

A prototype and an exemplar model need not disagree about the final judgment; they disagree about the representation and computation that explain it. Where an averaged prototype compresses a distribution of features, an exemplar model preserves information about encountered instances. Their comparative success depends on the task, category, learning conditions, and response measure. The philosophy of concepts must therefore resist turning a model's experimental performance in one domain into a metaphysical definition of every concept.

Lawrence Barsalou's work on ad hoc categories further complicates any inventory of fixed stored kinds. People can organize objects under temporary and goal-dependent descriptions such as items to take on a camping trip or things to rescue from a flooded room. These categories can display coherent organization without corresponding to stable lexical entries learned through repeated exposure (Barsalou, 1983). They indicate that conceptual activity includes construction, not solely retrieval. The explanatory question shifts toward the capacities that make rapid creation, maintenance, and reuse of such categories possible. A theory that identifies each temporary constructed category with a stored primitive concept needs independent evidence for the corresponding stored representation.

Theory-Based Concepts, Causal Expectations, and Explanation

According to the theory theory, concepts are embedded in bodies of explanatory expectations concerning how their referents behave, originate, and relate to other kinds. Scientific and ordinary classifications can depend on presumed causal structure. A living animal remains the animal it is even when outward appearance changes, because classifications track generation and biological continuity rather than merely visible features. An artifact may be categorized by intended function, causal history, or design. The theory theory models how background explanatory commitments influence both reflective classification and concept acquisition (Carey, 2009; Margolis & Laurence, 2019).

Its strength is evident when a surprising discovery changes what a thinker predicts while leaving a category available for further investigation. A child who learns that a whale is a mammal revises a network of assumptions about reproduction, respiration, and biological relations. This illustrates the role of theoretical connections in organizing concepts. The problem is to specify which beliefs are constitutive of the concept and which are changeable information about its referent. If every belief belongs to conceptual identity, ordinary learning threatens to replace the concept whenever a belief changes. If no belief matters, theory structure loses the role of determining conceptual content.

The distinction between a concept's inferential connections and a theory's complete body of commitments is consequently essential. Local explanatory commitments may be required for some kinds of competence, while peripheral empirical beliefs can change without destroying conceptual continuity. A complete theory theory must articulate this partition and explain how concept sharing remains possible among people with widely different scientific and everyday beliefs.

Natural Kinds, Semantic Externalism, and Shared Reference

Questions of concept identity become especially sharp in the semantics of natural-kind terms. Two speakers may associate a term with substantially different descriptions and yet participate in a common practice of referring to a substance or biological kind. If the shared reference is determined by historical use, environmental relations, and communal practices, an exclusively private list of descriptive features cannot fully explain it. Hilary Putnam's discussion of the meaning of 'meaning' combines the argument that meaning involves the environment with the division of linguistic labor: ordinary speakers can competently defer to specialists whose practices help determine the relevant extension (Putnam, 1975).

Putnam's Twin Earth thought experiment considers counterparts using superficially similar terms in environments where the relevant underlying substances differ. Its purpose is to challenge the assumption that sameness of internally characterized psychological state necessarily guarantees sameness of reference. The methodological lesson for concept theory is that psychologically indistinguishable recognitional procedures may participate in different referential relations. The lesson does not imply that cognitive architecture is irrelevant. It requires the theory to distinguish between the psychological means by which a thinker identifies something and the worldly conditions that fix the meaning or extension of the thought.

Social dependence yields another tension. If some competence consists in deference to external expertise, the possession of a concept cannot always require the thinker to know its complete scientific definition. At the same time, mere ability to repeat a technical term does not suffice for every attribution of understanding. A differentiated account must distinguish being a competent participant in a community's referential practice from being able to explain the scientific principles that govern a category. The possibility of partial understanding makes this distinction indispensable.

Externalism also complicates conceptual change. When scientific discovery alters the prevailing characterization of a substance while maintaining the same historical chain of reference, a theory that equates the concept with the entire stereotype would diagnose wholesale conceptual replacement. A theory emphasizing reference can allow substantial theoretical revision while maintaining referential continuity. Neither result is automatic: historical continuity of vocabulary does not guarantee continuity of referent, and the discovery that two terms co-refer does not erase differences in cognitive significance. Conceptual identity requires a reasoned choice of individuation conditions appropriate to the inquiry.

The lesson for the subsequent analysis of atomism is therefore twofold. World-involving relations can supply explanations that internal descriptions leave incomplete. Yet a theory of reference must explain which relations are meaning-determining, how a learner establishes them, and how reference can be sustained through errors of recognition. Social and environmental relations are candidate explanatory determinants that need further specification, not labels that by themselves solve the problem of concept individuation.

Conceptual Atomism, Reference, and External Determination

Jerry Fodor's conceptual atomism disputes the thesis that the content of most lexical concepts is determined by internal definitions, prototypes, or inferential relations. On this account, a primitive concept can possess reference in virtue of a suitable relation between a mental symbol and the world. Its conceptual identity is not constituted by a descriptive analysis of that symbol into other concepts. The attraction is that shared reference may survive differences in associated beliefs and recognition strategies. The account also supports a strongly compositional architecture, since atomic constituents can be recombined without inheriting an unstable mass of background associations (Fodor, 1998).

This approach belongs in the historical context of anti-descriptivist semantics, including arguments associated with Saul Kripke and Hilary Putnam. The explanatory point is that what a term designates may be fixed through causal, historical, environmental, and social relations not exhausted by what an ordinary speaker believes about the referent. A person can use the term for a substance while possessing an inaccurate theory of it. Yet a concept theory must still explain why thought is rationally assessable, why certain inferences are intelligible, and how a thinker comes to possess the relevant symbol in the first place.

Atomism therefore faces the acquisition and reference-fixing problems. If a primitive concept is associated with an object in the world, which causal relations are meaning-conferring rather than accidental? Encountering a dog while looking at its shadow does not by itself settle whether the resulting representation refers to dogs, shadows, moving shapes, or another property. A complete account needs a principled account of reference selection. Separating referential individuation from the mechanisms of recognition is promising; supplying the relation that makes the separation determinate remains the theory's central task.

Inferential Roles, Holism, and Social Normativity

Inferential-role approaches analyze conceptual significance through a representation's place within patterns of reasoning. The concept 'and' is related to rules governing the movement between a conjunction and its conjuncts; other concepts may be linked to explanatory, material, or practical inferences. On stronger versions, a concept's content is constituted by some or all of these roles. The motivation is that understanding a concept appears inseparable from appreciating at least some of what follows from its correct application.

The challenge is the scope of the content-constituting role. Global holism threatens conceptual stability: if every inferential connection contributes to content, any revision of a belief may change the content of the concepts employed. A more selective molecularism identifies a privileged set of constitutive connections, but then must explain why some inferences are essential and others are revisable. A social-pragmatic account may locate correctness in norms governing public assertion, challenge, and justification. Robert Brandom's Making It Explicit develops the significance of inferential articulation and discursive commitments for semantic content, emphasizing the social practices through which claims are assessed (Brandom, 1994).

The central analytical result is that two roles require separate examination. An inference may be psychologically habitual, or it may be normatively licensed. Habitual association explains why one thought triggers another; normative licensing explains why a transition is correct or incorrect. A concept theory that reduces inferential structure to mere association loses the distinction between error and justification. Conversely, a normative semantics that ignores computational and developmental constraints leaves the actual acquisition and execution of competence unexplained.

The relationship between reference and inferential role remains open to hybrid treatment. Reference may constrain which objects a concept concerns, while inferential norms determine aspects of its cognitive significance. The hybrid view becomes informative only when it specifies how disagreements between inferential role and worldly reference are resolved and what preserves concept identity through correction.

Compositionality, Productivity, and the Structure of Thought

Human thinkers understand and form complex thoughts whose exact combinations they have never previously entertained. This productivity motivates a compositional analysis: the content of a complex representation is determined by the contents of its parts together with their mode of combination. Compositionality supports the explanation of systematicity, the related capacity to understand recombinations such as one animal pursuing another and their roles being reversed. The principle is central in philosophy of language and has been extended into theories of mental representation (Szabó, 2025).

A precise claim must identify the semantic property being composed. Reference, truth-conditional content, cognitive significance, and pragmatically enriched interpretation may obey different constraints. Compositionality also does not entail that every component is a classical definition. Atomic semantic values can compose, and structurally organized representations can interact with contextual mechanisms. Its explanatory force lies in the relation between stable components, combination rules, and indefinitely varied complex contents.

Prototype effects expose the need for this precision. A psychologically retrieved image of a typical pet fish may have emergent associations; such emergence does not by itself show that the truth conditions of 'pet fish' fail to be compositionally determined. An account must distinguish conceptual content from typicality profiles and the results of imaginative elaboration. Conversely, merely defining complex contents in a compositional formalism does not explain the observed richness of psychological combination.

The problem also concerns cognitive architecture. Classical language-of-thought approaches explain productive thinking through structured symbols and operations. Alternatives investigate distributed and situated representations whose content-sensitive operations may be realized through other architectures. The philosophical issue is whether the structured dependencies that explain recombination are constitutive of thought, and how their realization constrains concept identity.

Concept Acquisition, Core Cognition, and Conceptual Change

A theory of concepts should explain both the origin of a repertoire and its transformation. Some categories are acquired through repeated perceptual discrimination; others involve the construction of higher-level explanatory relations, mathematical representations, and social instruction. A person can learn a new word without acquiring a new concept, or acquire a useful distinction without immediately learning the established word for it. The acquisition of a concept must therefore be distinguished from lexical learning, the memorization of examples, and the acquisition of isolated true beliefs.

Susan Carey's The Origin of Concepts (2009) places conceptual development in relation to early cognitive systems and the construction of later representational resources. The significance of this approach is that mature concepts may support representational capacities whose elements are not simply copies of a child's earlier concepts. Learning number, biological kinds, or abstract scientific notions can involve reorganizing the system in which new judgments become possible. Conceptual change may preserve practical continuity with earlier classifications while changing the structural resources by which the subject is understood.

This raises a philosophical identity problem. If the meaning of a scientific concept is determined by its surrounding theory, a major theoretical revolution might replace that concept. Yet scientists can disagree about the same phenomenon over time and can communicate across conceptual revisions. Complete semantic invariance is implausible in all cases; complete discontinuity cannot explain the continuity of inquiry. A defensible account should identify specific continuities in reference, measurement, explanatory task, and practices of correction while acknowledging that some conceptual transformations genuinely change what can be thought and asserted.

The same issue concerns comparison across persons and communities. Shared language permits communication despite different acquired associations and levels of mastery. The correct theoretical question is therefore not whether all individuals contain psychologically identical representations, but which relations suffice for sufficiently stable common reference and reciprocal assessment.

Embodiment, Perceptual Grounding, and Nonconceptual Representation

Embodied and perceptually grounded theories emphasize that conceptual activity is deeply connected with perceptual, motor, and situated resources. Jesse Prinz's Furnishing the Mind develops an empiricist account of concepts and their perceptual basis. Lawrence Barsalou's perceptual-symbol-systems proposal argues for conceptual representations related to perceptual simulation rather than treating conceptual thought as wholly amodal symbolic manipulation (Prinz, 2002; Barsalou, 1999). The central philosophical significance is an alternative explanation of how representations acquire cognitive accessibility and practical relevance.

These proposals must be distinguished from the simple claim that every concept is a mental picture. A simulation or a flexible perceptual representation can involve distributed features and context-dependent activation. Even so, abstract, mathematical, logical, and institutional concepts raise questions about how perceptual grounding can supply the required determinacy and inferential organization. Research in perceptual cognition may show how concepts are realized or accessed without independently settling their semantic ontology.

The debate over nonconceptual content reinforces this distinction. The issue is whether a perceptual state can represent features of the world without requiring its bearer to possess the concepts used by a theorist to specify that content. The conceptually articulated judgment that an object is red and the visual capacity to discriminate an array of shades may impose different possession requirements. The debate matters to infants, animals, perception, and the architecture of concept acquisition. Gareth Evans's work is an important source for the contemporary framing of nonconceptual content; later discussions distinguish several nonequivalent ways of drawing the conceptual/nonconceptual division (Bermúdez & Cahen, 2024).

A complete account should permit empirical models to constrain theories of implementation while allowing semantic and normative questions to remain explicit. The existence of a perceptually implemented classification mechanism alone does not establish that its representation is a concept in every philosophical sense of the term.

Heterogeneity, Pluralism, and Eliminativism

Modern research has produced multiple successful explanatory models: prototypes capture graded typicality; exemplar models capture learning from remembered cases; theory-based models capture causal expectations; atomic accounts emphasize referential stability; inferential-role theories highlight rational competence. This heterogeneity gives rise to conceptual pluralism, the proposal that concepts involve more than one kind of structure, or that different representational kinds underwrite different tasks. A pluralist theory becomes successful when it explains how these structures are coordinated and specifies whether they are components of one concept or distinct concepts linked by a common function (Margolis & Laurence, 2015).

Edouard Machery's Doing without Concepts (2009) advances a sharper challenge. The heterogeneity hypothesis suggests that representations grouped under the name 'concept' may not constitute the uniform kind required by some scientific theories. On an eliminativist response, cognitive science should study prototypes, exemplars, theories, and other specific representations without assuming that all form one homogeneous psychological natural kind. The argument is methodological: an overly broad category can conceal differences in causal mechanism and experimental prediction.

The objection to eliminativism is that a higher-level category need not be homogeneous in implementation to be scientifically or philosophically explanatory. Different mechanisms might realize a shared functional role, just as different physical devices can implement the same logical operation. Yet simply asserting that 'concept' names a function does not solve the problem. The functional role must be defined with explanatory precision, and the account must show when distinct representations genuinely instantiate that role.

The strongest conclusion from heterogeneity is therefore neither unrestricted unification nor automatic elimination. The evidence calls for an articulated division of explanatory labor: distinguish mental vehicles, semantic contents, stored information, inferential norms, and the processes that deploy them. A higher-level concept theory can survive that division if it specifies the coordination that makes the different layers jointly relevant to thinking.

Concepts, Public Language, and the Possibility of Correction

The study of concepts reaches beyond isolated individual minds because conceptual identities are stabilized and revised in linguistic communities. Scientific disciplines, legal institutions, mathematical practices, and everyday social interactions develop public criteria of use. A single speaker may incompletely understand the concept at issue while participating intelligibly in a communal practice. This introduces social dependence into questions of reference and possession. Disagreement is often evidence of an ongoing common subject rather than proof of conceptually isolated speakers.

Public correction requires a distinction between what an agent takes to be true and what the use of a concept licenses under the governing standards. A student can misuse a term and receive an explanation that changes the student's future reasoning. This process connects semantic competence, epistemic responsibility, and education. The possibility of learning from correction presupposes that there is sufficient continuity between the earlier and later uses for the mistake to be recognized as a mistake about the same subject.

The conceptual significance of public practices does not settle whether concepts themselves are abstract shared objects or individually realized capacities. It instead constrains both positions. An abstract-content view must explain how finite agents grasp and deploy a common content. A capacity view must explain what makes different individual performances instances of the same publicly recognizable competence. A representational view must relate internal states to external practices and referents. These are complementary demands generated by the same phenomenon of mutually assessable thought.

The account also raises questions about artificial cognitive systems. A system may manipulate compositional representations, construct categories, and participate in public correction. Whether this suffices for concept possession depends on the theory of possession and the evidential standards for attributing semantic and normative competence. This is a general question about realizations of thought; its investigation should be pursued by assessing architectures and practices instead of assuming that one biological implementation fixes the definition of every conceptual role.

A Systematic Synthesis of Conceptual Structure

The preceding analysis establishes that several relations commonly described as 'conceptual structure' have distinct logical and explanatory positions. Semantic structure concerns the contribution a component makes to the content of a thought. Inferential structure concerns transitions among judgments and representations, including their normative standing. Cognitive structure concerns stored associations, typicality patterns, explanatory expectations, and access procedures. Deployment structure concerns the processes by which a thinker applies and combines conceptual resources on particular occasions. Social structure concerns publicly maintained standards, linguistic coordination, and correction. These relations interact in ordinary cognition, but they are not interchangeable definitions.

This analysis can be expressed as an explanatory constraint. Suppose a theory T asserts that a specific structure S constitutes concept C. T must show that S individuates C under relevant changes of context, explains the core competence attributed to possessors, and coheres with the content that C contributes to complex thoughts. If S predicts typicality judgments but neither individuates reference nor explains compositional behavior, S is an important psychological model whose status as constitutive conceptual structure remains to be established. If S explains reference but leaves recognition unexplained, it needs an account of deployment. If S explains a public norm but cannot accommodate learning and variation in mastery, it needs an account of possession and development.

This yields a substantive distinction between constitutive and accompanying structure. Constitutive structure fixes which concept is present in the relevant theoretical account. Accompanying structure contributes to its availability, use, learning, and elaboration. The distinction does not predetermine which theory succeeds. Atomists, inferentialists, and pluralists disagree about which structures are constitutive; the method requires each to state the relevant individuation conditions and defend them with appropriate evidence.

A second result concerns the relation between generality and determinacy. Generality allows a representation to apply across distinct cases. Determinacy supplies standards for when that application is appropriate. Classical abstraction, referential semantics, social use, and causal theories provide different explanations of these features. The history of concepts is partly the history of discovering that the same observable classificatory behavior can arise from different representational and normative organizations.

A third result concerns continuity through learning. Scientific understanding can develop while reference is maintained, and some radical transformations can create genuinely new conceptual resources. Conceptual identity is therefore a question to be argued for in a particular case, not inferred simply from stable vocabulary or from changed associated beliefs. The most promising comprehensive theories offer a disciplined explanation of continuity that integrates semantic, psychological, historical, and social evidence.

Open Philosophical Questions

Several problems remain fundamental. The first concerns the individuation of concepts across thinkers. What exact degree of shared reference, inferential competence, and public participation is sufficient for common possession? The answer may vary with the kinds of concept at issue, but a theory must explain the relevance of that variation.

The second concerns the origin of primitive concepts. An atomist needs an account of reference-fixing that distinguishes meaningful worldly relations from accidental correlations. A structured account needs to explain how finite thinkers acquire constitutive inferential networks without presupposing the concepts those networks are intended to generate. Acquisition theories must also describe the boundary between developing a new concept and improving one's knowledge under an existing one.

The third concerns compositionality and context. The semantic contribution of a concept to a complex thought must remain systematic enough to explain productive understanding, while actual conceptual use can be flexible, goal-directed, and influenced by background knowledge. Determining which aspects of context alter content and which merely alter processing or typicality remains a central task.

The fourth concerns normative authority and psychological realization. Why does a representational transition count as correct? How can a public standard be realized through diverse cognitive mechanisms? A satisfactory answer should account for error, learning, and disagreement without identifying correctness with a majority response or a subject's strongest association.

The fifth concerns the unification of empirical research. Prototype, exemplar, theory-based, and embodied models may predict different aspects of the same performance. Whether they constitute a single research kind depends on causal and explanatory integration rather than on conventional terminology alone. The philosophical theory of concepts has a continuing role in specifying the common explanatory targets by which that integration can be assessed.

Conclusion

Concepts occupy a central position in the explanation of thought because they make possible stable general reference, structured judgment, productive recombination, and the revisability of understanding. The history of the subject reveals a continuous movement from the problem of universal predication through the origin of general ideas, the rules of judgment, the distinction between sense and reference, and the empirical study of categorization and cognitive development. Each historical development identifies a genuine problem that a theory of concepts must address.

The decisive analytical result is that conceptual content, conceptual structure, concept possession, and cognitive deployment are connected questions requiring explicitly distinguished answers. Definitions may fix some contents; prototypes organize some applications; exemplars preserve learning histories; causal theories structure some reflective classifications; atomism explains one route to referential stability; inferentialism illuminates normative mastery; and pluralist approaches attempt to coordinate these dimensions. Their explanatory strengths become intelligible once their constitutive claims are separated from their accounts of processing.

A comprehensive philosophy of concepts consequently requires two forms of accountability. It must explain how determinate content contributes to indefinitely variable thoughts, and it must explain how finite, developing, socially situated thinkers acquire and correct their use of that content. This dual requirement preserves the achievements of semantic philosophy and cognitive science while establishing the conditions under which their theories can be meaningfully compared. Concepts remain a live philosophical problem precisely because the unity of a thought, the organization of a cognitive system, and the public standards of meaning must be explained together.