RESEARCH DOSSIER
Sets and Classes
Abstract. Sets and classes organize mathematical discourse about plurality, totality, infinity, and the domain over which a theory quantifies. Their distinction is mathematically exact within particular axiomatic frameworks and philosophically consequential whenever mathematical language appears to name a totality too extensive to be a set. This monograph reconstructs the historical transition from Cantorian collection to axiomatic set formation, explains the role of Russell's antinomy, analyzes the Zermelo–Fraenkel universe and its proper classes, compares definitional treatment in ZFC with class-quantifying systems such as Gödel–Bernays and Kelley–Morse, and investigates ontological commitment in relation to Quinean quantification, Boolos's plural interpretation, structuralism, category theory, and set-theoretic pluralism. Its principal analytical conclusion is that ontological commitment to classes must be read through the formal language, comprehension principles, semantics, and intended explanatory use of a theory. The mathematical assertion that a collection is a proper class identifies a precise limit on set formation inside an adopted foundation; its philosophical import requires an additional argument concerning what its class variables, definitions, or plural expressions designate. The work develops this conclusion through proofs, comparative interpretation, and explicitly constructed examples. It proposes no newly named theory and claims no new mathematical discovery.
Keywords: set; class; proper class; set theory; ZFC; Gödel–Bernays set theory; Kelley–Morse set theory; ontology; quantification; comprehension; cumulative hierarchy; Russell's paradox; transfinite infinity; mathematical structuralism; category theory; philosophical foundations of mathematics.
The Problem of Mathematical Totality
A set is a mathematical object governed by a membership relation. A class is, depending on the foundational language, a collection specified by a property or an entity over which class variables range. The distinction matters because mathematics routinely refers to collections whose members themselves exhaust an indefinitely extensible field: all sets, all ordinals, all groups, or all sets of a given algebraic kind. Such expressions are indispensable in ordinary research even when the collection cannot belong to the same domain of set-sized objects as its members.
The pivotal philosophical question is therefore one of scope. A sentence can purport to describe all entities of a certain kind while the accepted foundational axioms prevent those entities from being assembled into an additional set. As Bagaria (2023) explains in his account of the set-theoretic universe, the cumulative hierarchy encompasses every set without itself constituting a set. In a related survey, Holmes (2021) documents how alternative axiomatic theories of classes turn this practical manner of speaking into formal resources of differing strength.
This monograph asks how these resources should be interpreted ontologically. Formal mathematics supplies exact distinctions between membership, sethood, classhood, and quantification. Philosophical interpretation then asks whether a formal class commits us to a further abstract object, a predicate applicable to many sets, a plural reference to the sets themselves, or a useful level of mathematical description. These possibilities must be assessed at the level of the particular theory rather than from the mere grammar of the word "collection."
Research Method and Source Boundaries
The investigation combines a historical reconstruction, a formal comparison, and a philosophical assessment. The historical part reads foundational works and their documented bibliographic contexts: Cantor's 1895 and 1897 papers in Jourdain's English edition, Zermelo's 1908 axiomatization, Russell's antinomy as reconstructed from correspondence, and Gödel's 1938 consistency announcement. Later primary works include Quine (1948), Boolos (1984), Lawvere (1964), Cohen (1963), and Hamkins (2012). These primary texts are supplemented by research encyclopedias and contemporary technical work in second-order set theory.
A source's scope determines the claim for which it is cited. For example, Zermelo's 1908 original publication record confirms the work's identity and date; the exposition of its philosophical reception is assessed through Holmes's historical and formal comparison of Zermelo systems. Gödel's 1938 paper directly attests to a relative consistency argument, while broader claims about later independence results require Cohen's separate work. Treating a book listing, abstract, or encyclopedia entry as proof of an uninspected theorem would invalidate the source-to-claim relation.
The proofs reproduced below are elementary derivations within familiar axioms, not new results. Historical attribution is kept distinct from modern textbook formulations, and philosophical conclusions are expressly presented as interpretation. This method is suited to the research question because the ontological debate often arises from crossing a boundary between the formal object language and the informal descriptions through which mathematicians explain it.
Membership, Inclusion, Collection, and Class
Membership, written (x\in y), says that an object is an element of another object. Inclusion, written (A\subseteq B), abbreviates the statement that every member of A is a member of B. The relation (\in) is primitive in standard pure set theory; (\subseteq) is definable from it. An unordered pair, union, power set, and ordered pair are constructed using set-theoretic axioms and conventions, as in the standard ZFC presentation by Bagaria (2023). Their existence is governed by principles and cannot be inferred solely from an English phrase that refers to several things at once.
The word "collection" is deliberately neutral in this study. It may denote a set, a class, the extension of a predicate in a background model, a plurality of objects, or an informal range of discourse. The term "proper class" has a sharper technical meaning: a class that is not a set, in a framework recognizing that distinction. In a usual NBG formulation, sets are precisely the classes that belong to some class, while proper classes cannot occur as members. In a ZFC discussion the same words are metalinguistic shorthand for a condition represented by a formula with no corresponding set.
These distinctions already produce a diagnostic question. When an author writes "the class of all groups," is this a set whose elements are groups, a class-level object in a two-sorted theory, or merely a useful abbreviation for a predicate over group-structures? The answer depends on the underlying language and the operations proposed on this collection. The reader should demand that information before drawing existential conclusions.
Cantor's Mathematical Enlargement of Infinity
Cantor transformed infinity from a loose description of unbounded succession into a structured subject of mathematical comparison. His contributions to transfinite number theory, first published in 1895 and 1897 and translated by Jourdain in 1915, distinguish the cardinality of collections from the order types associated with well-orderings. This distinction enables a collection to be infinite in a precise mathematical sense while differing from another infinite collection in number or in order.
Cantor's achievement made a new demand on foundations. Once mathematics treats infinite totalities as determinate objects and defines operations upon them, it must say which totalities may be formed. That demand is especially urgent when transfinite procedures generate larger and larger collections. The historical set-theory overview by Cunningham places this development in the emergence of a powerful general foundation for mathematics. Its mathematical success created the conditions for sharper reflection on paradox and totality.
One philosophical temptation is to read every consistent description of a many as a completed mathematical object. Cantor's research does not authorize that unrestricted inference. Modern axiomatizations preserve the mathematical force of comparing infinite sets while specifying existence through axioms. The resulting discipline separates the study of infinitely many objects from the claim that absolutely every plurality of objects must itself form an object of the same kind.
Diagonalization and the Power-Set Boundary
Cantor's theorem establishes that for every set A there is no surjection from A onto its power set (\mathcal P(A)). The usual diagonal argument starts with a proposed function (f:A\to\mathcal P(A)) and defines (D={a\in A:a\notin f(a)}). Since D is a subset of A, it is in the power set. If some (d\in A) satisfied (f(d)=D), then (d\in D) if and only if (d\notin D), a contradiction. Thus (|A|<|\mathcal P(A)|) where the cardinal comparison is interpreted in the usual way.
The diagonal set exists because the defining predicate is applied within a previously given set A. This point will become central in the reconstruction of Russell's paradox. The ZFC axioms and power-set principle explicitly authorize the object (\mathcal P(A)), not an unrestricted power set of everything. The proof's strength comes from the interaction of a defined domain, Separation, and Power Set.
If a universal set U contained all sets in a system satisfying the relevant operations, (\mathcal P(U)) would also be a set and, since U contained every set, its members would lie in U. This would conflict with the strictly greater cardinality of (\mathcal P(U)). The contradiction points to a boundary in set formation. It does not mean that mathematical discourse cannot quantify over all sets using a domain-wide universal quantifier; it means that the domain itself cannot be packaged as another set within these axioms.
Russell's Paradox and Unrestricted Abstraction
Unrestricted comprehension promises a collection ({x:\varphi(x)}) for every condition (\varphi). Apply it to (x\notin x) and call the resulting collection R. If R belongs to itself, its defining property says it does not. If R does not belong to itself, its defining property says that it does. Therefore the intended object cannot exist under the assumed unrestricted comprehension principles.
Russell communicated the difficulty to Frege in June 1902; the historical reconstruction of the paradox by Deutsch, Marshall, and Irvine (2024) discusses the correspondence and Frege's difficulty with the unrestricted extension principle. Frege's response and later alternatives show that the problem extends beyond a clever contradiction: a foundational language must regulate how predicates determine object-like extensions.
Russell's derivation is often summarized as proving that the "set of all sets that are not members of themselves" is impossible. This sentence should be parsed carefully. In ZFC the displayed condition is a legitimate predicate, and one may use it in Separation relative to a specific existing set. The impossible step is treating its unrestricted extension as a set available for self-membership. In a class theory the same predicate may determine a class; the theory blocks the paradox by placing a proper class outside the membership relation's admissible left-hand domain. The lesson concerns the permissible passage from meaningful condition to object.
Zermelo's Axiomatic Restriction
Zermelo's 1908 investigation, Untersuchungen über die Grundlagen der Mengenlehre I, offers an explicit axiom-based treatment of set formation. The historical move was to identify operations and existence principles sufficient for substantial mathematics while restricting the power to construct sets from arbitrary predicates. In the later first-order development, the Separation schema forms ({x\in A:\varphi(x)}) from a given set A and a formula (\varphi). The previously available set supplies the range, and the formula selects its members.
This approach retains a rich theory of finite and infinite mathematics. It permits the empty set, pair formation, unions, power sets, and the infinite sequences supported by the axioms; it does not let a definition simply declare the existence of a containing universe. Holmes (2021) explains the transition from Zermelo to ZFC and why later formalization of Separation and addition of Replacement matter to the development of transfinite recursion.
The ontological interest lies in the role of prior existence. A mathematical specification can determine a set only when the accepted axioms justify the required object. That is an instructive alternative to treating grammar as ontology. The relation between definition and existence is mediated by formal rules, allowing unrestricted mathematical discussion of a property while withholding unrestricted object formation.
The Architecture of ZF and ZFC
The contemporary standard Zermelo–Fraenkel axioms with Choice—ZFC—use first-order logic with equality and the single nonlogical binary predicate (\in). Extensionality fixes equality of sets by common members. Empty Set, Pairing, Union, and Power Set support basic operations. Infinity secures an inductive infinite set. Separation extracts definable subsets of an already available set. Replacement collects the images of set domains under functional formulas. Foundation rules out infinitely descending membership chains in the ordinary well-founded formulation. Choice asserts a selection principle equivalent over ZF to several important well-ordering statements.
This list is explanatory rather than a claim that every named item is independent of the others. Bagaria's systematic account of ZFC axioms notes that some existence axioms follow from stronger components and that Separation and Replacement are schemas indexed by formulas. The formal detail matters: a schema is a collection of first-order axioms, whereas a single quantifier over arbitrary properties can change the logical strength of a theory.
Nothing in the ordinary language of ZFC requires a separate sort of proper-class objects. Nevertheless, the mathematics is saturated with constructions commonly described using classes. A foundation can therefore be ontologically economical at the level of its primitive domain while remaining expressively expansive in metatheoretical practice. That economy is an initial reason to distinguish formal ontology from ordinary research vocabulary.
Separation, Replacement, and the Meaning of Closure
Separation and Replacement perform different functions. Separation chooses elements from a set already in hand: (B={x\in A:\varphi(x)}). Replacement begins with a set A and a formula that defines one value y for each x in A and concludes that the range of those values forms a set. Replacement is thus crucial for transfinite sequences of constructions, including the sets found at limit stages of the cumulative hierarchy.
The axiom-by-axiom explanation by Bagaria (2023) clarifies why this matters for arbitrary ordinal stages. At a limit ordinal (\lambda), a set-indexed sequence of earlier ranks can be collected by Replacement and united. Without adequate range formation, a familiar construction could be described but its claimed stage might lack justification within a weaker theory.
Both axioms illustrate a form of constrained closure. An admissible operation is performed on a set-sized input, and its output is a set. Neither says that every formula or operation with a class-wide domain has a set as its value. Consequently the formula "take all ordinals" and the formula "take the ordinals below this ordinal" have radically different existence consequences. The difference is often invisible in casual notation and decisive in formal proof.
Definable Classes in the Language of ZFC
Within a ZFC discussion, a class may be introduced as a convenient expression (C={x:\varphi(x,p_1,\ldots,p_n)}), where the parameters are sets and (\varphi) is a formula in the set-theoretic language. This is a metalinguistic abbreviation: the notation records which sets satisfy the formula. If a set y contains exactly those objects, C is a set; if no such y exists, it is termed a proper class. In ZFC there is no new class variable automatically introduced by this abbreviation.
For example, the class of all ordinals is definable by a first-order formula. The class of all sets is definable by (x=x). The class of groups is definable using a coding convention for ordered tuples, operations, and algebraic laws. Holmes's discussion of virtual proper classes emphasizes the explanatory role of such terminology even in theories where proper classes receive no independent objectual existence.
There is a subtle qualification. Not every externally distinguished subcollection of a model of ZFC is definable inside that model by a formula with set parameters. "Definable class" therefore describes an available linguistic resource; it does not enumerate every subset of the model's domain that an external mathematician may consider. This observation blocks an illicit argument that the set-theoretic language encompasses every conceivable way of grouping sets merely because it can talk about all sets in a universal quantification.
The Proper Class of All Sets
Let V denote all sets in a ZFC discussion. Suppose V were itself a set. By Separation, the expression (R={x\in V:x\notin x}) would then define a set. As V contains every set, R would belong to V. By its definition, (R\in R) if and only if (R\notin R), a contradiction. Hence V is not a set: it is a proper class in the usual metatheoretical vocabulary.
The common account of the universe V explains its relation to the cumulative hierarchy. It does not identify V with a set occupying a final level of that hierarchy. Instead, V expresses the unrestricted range of the set variables or the union across the ordinal-indexed ranks as class-level shorthand.
The word "universe" sometimes encourages a picture of a vast container. The proof shows why this metaphor is limited. A ZFC universe is the domain of all set-objects for the intended interpretation of the theory. If it were also an object in that same unrestricted domain, the theory would lose the stratification that makes its comprehension principles coherent. One may describe all the sets without assigning them an additional encompassing set. The legitimacy of the description and the existence of its object are distinct questions.
The Ordinals and Burali-Forti Reasoning
An ordinal is a transitive set well-ordered by membership in the standard von Neumann account. If the collection On of all ordinals were a set, its union would be an ordinal at least as large as each ordinal in the collection; taking a successor would produce a further ordinal beyond that bound, contradicting the claim that the set already contained every ordinal. The exact textbook proof can also use the fact that any set of ordinals has a supremum ordinal. The critical premise is that the supposed totality is a set to which the ordinary set operations apply.
The classic Burali-Forti difficulty thus parallels the universal-set argument while involving order type rather than only self-membership. The history of set-theoretic antinomies links this family of arguments to the crisis of unrestricted totality. Bagaria's treatment of transfinite ordinals situates ordinals as the indexing structure of the cumulative hierarchy itself.
Calling On a proper class allows mathematical statements such as "for every ordinal there is a larger ordinal" to remain meaningful. It does not introduce a last ordinal or a completed ordinal-shaped set. The logical content of the quantification and the restriction on its conversion into a set can be simultaneously preserved. This is an exemplary case of how class terminology stabilizes a rigorous generality without requiring set-sized closure.
The Cumulative Hierarchy and Its Limits
The cumulative hierarchy is recursively defined by (V_0=\varnothing), (V_{\alpha+1}=\mathcal P(V_\alpha)), and (V_\lambda=\bigcup_{\beta<\lambda}V_\beta) for limit ordinals (\lambda). Each rank (V_\alpha) is a set, with the needed construction justified by the ZF axioms. Under Foundation every set occurs at some stage of this hierarchy, so V may be described as the proper-class union of its ranks.
This picture supports a precise reading of mathematical accumulation. No final ordinal stage exhausts all stages. The description of the entire hierarchy is itself a class-level description, and its correctness cannot be reduced to the existence of one rank containing every set. As Bagaria states in the cumulative-hierarchy discussion, Power Set, Replacement, and Union each play identifiable roles in the recursion.
The hierarchy also carries philosophical weight. A set's members are obtained at earlier stages, making set formation appear iterative rather than unrestricted. Yet the iterative conception is a mathematical motivation for the axioms and is open to distinct metaphysical interpretations. One can accept the formal rank theorem without deciding whether ranks constitute independently existing abstract strata or an efficient conception of all set-theoretic consequences. The mathematical construction itself settles the sethood of each rank and the non-sethood of the total V.
Foundation, Well-Foundedness, and Rank
The Axiom of Foundation says that every nonempty set A has an element disjoint from A. In the ordinary ZF background, this excludes finite self-membership loops, forbids infinitely descending membership chains in the relevant set-theoretic sense, and supports the assignment of ordinal ranks to sets. The rank of x can be characterized as the least stage above the ranks of its members, equivalently (\mathrm{rank}(x)=\sup{\mathrm{rank}(y)+1:y\in x}).
Foundation's role is conceptual as well as technical. It organizes sets by membership depth and makes the hierarchy a systematic representation of the universe. However, the paradox of unrestricted comprehension can be derived even without using Foundation. The survey of alternative set theories discusses theories allowing non-well-founded or differently structured membership, demonstrating that anti-paradox restrictions and well-foundedness are distinct choices.
An ontological conclusion based solely on the visual metaphor of stages would therefore overreach. Foundation contributes a particular structural discipline; it does not independently establish that every possible mathematical object must be a well-founded set. The present dossier evaluates the set/class distinction within named axiomatic systems and does not treat a single foundation as a decree concerning every coherent mathematical ontology.
Class Terms, Predicates, and Metalanguage
A language may include a predicate (\varphi(x)) without including a term denoting its extension. Conversely, a language may quantify over class objects directly. To move from the first fact to the second requires a comprehension or abstraction principle. Russell's result shows why the transition cannot be universally unrestricted if membership and extension retain their naïve meaning.
The distinction between object language and metalanguage is especially important. Inside first-order ZFC, quantifiers range over sets. In a metatheoretical explanation, an author may speak of the collection of all elements satisfying a formula. The alternative-set-theory analysis explains how a theory with class variables brings some such discourse inside a richer formal language; the extension's exact status depends on the added axioms.
This produces a sound rule of interpretation: begin with the quantified variables and the grammar of allowable membership relations, then inspect which comprehension instances are provable. The phrase "all x such that" is a schema of description, not an automatic existence theorem. Proper-class talk becomes mathematically responsible when its ontological reading is matched to the language that licenses it.
Gödel–Bernays and von Neumann–Bernays–Gödel Systems
Class theories associated with von Neumann, Bernays, and Gödel formalize both sets and classes. In a typical one-sorted presentation, every object is a class, but an object is a set precisely when it can be a member of some class. Only sets occur as members. Proper classes are consequently classes that are not sets and can never belong to another class. A two-sorted notation distinguishes set variables and class variables for convenience. This is the structural reason that the Russell class can exist without generating Russell's contradictory self-membership step.
The formal account of class theories by Holmes (2021) emphasizes predicative class comprehension: an eligible defining formula may quantify over sets and use class parameters but does not quantify over arbitrary classes in the defining condition. The NBG discussion in Cunningham's survey similarly explains why the restriction matters. Variant naming conventions and the treatment of Choice should be stated explicitly; the names GB, NBG, and GBC are used somewhat differently across literature.
The system's philosophical advantage is grammatical clarity: mathematicians can write class functions, class relations, and class operations inside a formal language. But the appearance of bound class variables does not alone resolve what kind of abstract entity a class must be. One can read class theory as an extension of ontology, as a disciplined apparatus of definable predicates, or as a formal surrogate for genuinely plural discourse. The mathematical theory tells us its consequences; the ontology of its semantic domain remains open to philosophical argument.
Global Choice and Class Comprehension
Ordinary Choice says, in one familiar form, that every set of nonempty sets possesses an appropriate choice function. Global Choice asks for a class function selecting an element from every nonempty set across the entire universe. It is therefore a class-wide assertion, naturally expressed in class theory. The difference is structural: a set-indexed family and the proper class of all nonempty sets occupy different quantificational scopes.
GB with Global Choice is often called GBC. Class-comprehension principles still control which classes are guaranteed to exist; Global Choice does not entail that every conceivable external subcollection is a class in the model. The class-theoretic overview distinguishes the class ontology and limitations of size from first-order set principles. In this environment, talk of class well-orderings and class functions gains an exact object-language role.
A common mistake is to infer that since ordinary Choice and Global Choice involve selecting elements, their formal status is identical. They are related but assert different ranges of choice. Another mistake is to treat Global Choice as the existence of a set that lists all choices. Such a list would have a proper-class domain. Keeping its function class-valued in the formal sense is what allows the principle to be stated without pretending the universal domain is a set.
Kelley–Morse and Impredicative Classes
Kelley–Morse class theory, usually abbreviated KM or MK, extends the possibilities of class comprehension. Its comprehension schema may use class quantifiers within the formula defining a class. This is called impredicative comprehension because the defining condition can range over the very sort of totalities to which the defined class belongs. In the familiar Gödel–Bernays treatment, the comprehension formulas are restricted to set quantifiers. These differences are precise differences in axiom strength, not merely notational preferences.
Holmes (2021) traces the stronger Kelley–Morse comprehension and notes that KM proves statements about sets beyond what ZFC itself can prove, including the formalized consistency of ZFC under the appropriate background encoding. The argument should be formulated with attention to consistency assumptions: a sufficiently strong theory can prove consistency of a weaker theory without proving its own consistency. Gödel's incompleteness theorem remains fully relevant.
Impredicative class formation is an especially valuable philosophical comparison because it isolates the dependence of consequences on a quantification rule. Two theories can both use the word "class" and recognize all familiar definable collections while differing in whether class-quantified descriptions generate new class objects. If one simply counts class variables to measure ontology, one will overlook this difference in expressive and deductive resources. The relevant commitment concerns what kinds of entities and comprehension operations the formalism allows.
Relative Strength, Conservativity, and Interpretation
Conservativity is a relation between theories and a specified class of sentences. In the usual formulation, a class theory such as GBC is conservative over ZFC for first-order sentences about sets: it yields no new theorems in that language beyond the underlying set theory. The claim is about provability of set-language sentences, not about identity of all models, the irrelevance of classes, or an absolute proof that class objects do not exist.
The distinction is made explicit in the survey of set/class axiomatizations and in technical studies of intermediate second-order theories. Gitman, Hamkins, Holy, Schlicht, and Williams (2017) show that the class forcing theorem has a precise strength over GBC, namely equivalence to elementary transfinite recursion of length Ord. Their analysis demonstrates that class-theoretic statements can separate theories even where their ordinary set-theoretic arithmetic is aligned.
Conservativity is philosophically informative but not a complete criterion of existence. It demonstrates that class quantification may be integrated without altering a given stock of first-order set consequences. It does not specify whether class variables should be interpreted realistically or instrumentally. Stronger comprehension in KM yields further set-language consequences, showing that changing the expressive basis can change mathematical reach. The transition from formal strength to ontology demands an explicit interpretation of quantification and satisfaction.
Models, Internal Universes, and External Descriptions
A model of first-order set theory, when treated in ordinary model theory, is represented externally by a domain M and a membership-like relation E. The mathematical structure (M,E) may satisfy ZFC even though M is a set from the standpoint of an ambient metatheory. Inside the model, however, the statement "there exists a set containing all sets" is false. The external description of M as a set is therefore compatible with the internal failure of a universal set.
This distinction is a standard feature of model-theoretic reasoning about set theory and follows from the account of set-theoretic models given by Bagaria (2023). It shows that the status "proper class" is evaluated relative to an interpreted membership structure and its available objects. A collection externally coded by a set need not be internally set-sized. Indeed, if a countable external model of a suitable theory exists, it may internally satisfy the assertion that some of its sets are uncountable, because the externally available bijections need not belong to the model.
The ontological lesson is subtler than model relativism. Semantic interpretation supplies a context in which quantifiers have a domain; it does not erase the possibility of a preferred intended interpretation. But arguments about class size must identify their standpoint. Saying "M is a set" and "the collection of all sets is a proper class" is coherent when the first judgment is external and the second internal. A contradiction appears only if the two uses of "set" and the two scopes of quantification are conflated.
Ontological Commitment and Quine
In On What There Is, Quine (1948) connects ontology with the values that bound variables must take for an accepted theory to be true. A theory which requires numbers or sets as objects in its domain incurs a corresponding ontological commitment under the Quinean criterion. This move shifts attention from ordinary names and descriptions to regimented quantification.
Applied to ZFC, the criterion gives an immediate answer at the first-order level: its objectual variables range over sets. Applied to an explicitly class-quantifying theory, an analogous criterion can count classes among the domain's values. But class theory can also be interpreted through predicative definitions or semantic devices that do not reify every class in the same way. Bricker's (2014) analysis of ontological commitment argues that different accounts of theory, truth conditions, and quantification complicate any mechanical inference from syntax to existence.
The significance is methodological. Quine offers a disciplined question—what must the variables range over?—rather than a blanket answer that every use of the word "class" entails a new sort of being. One first selects an interpreted theory and a formal translation of its practice. Only then can one identify its existential commitments. If two translations preserve relevant mathematical inferences while differing in ontology, their comparative explanatory costs become a genuine philosophical issue.
Plural Quantification and Boolos
George Boolos (1984) argued in To Be Is to Be a Value of a Variable—with its parenthetical extension to "some values of some variables"—that plural quantification can express significant logical structure without always introducing a set or class object corresponding to the many. An English sentence about some things can concern those things plurally. It need not assert that there is a further object whose members are exactly those things.
The philosophy of plural quantification explains both the promise and the controversy of this program. In certain settings plural variables interpret second-order expressions, but the claim of complete ontological innocence is contested, particularly regarding strong comprehension principles and semantics. It would therefore be premature to declare that plural discourse automatically eliminates every commitment associated with formal class theory.
Plural quantification is nevertheless an important alternative for the philosophy of classes. It supplies a coherent conceptual contrast between speaking of several sets at once and speaking of another entity, a class, to which those sets belong. A class-theoretic assertion may admit a plural paraphrase in some contexts and fail to admit one preserving all desired higher-order operations in others. The work needed to establish equivalence is formal and semantic, not a matter of replacing "class" with "some things" in English.
Predication, Extensionality, and Identity
In extensional set theory, sets with the same elements are identical. In common class theories, classes likewise satisfy extensionality: if every set belongs to A exactly when it belongs to B, the classes are identical. This does not mean that two defining formulas with distinct meanings must be identical as formulas. Coextensive predicates can characterize the same class while retaining different definitions and explanatory roles.
The basic extensionality principle clarifies why identities of sets can be established by member-by-member comparison. The formal analysis of class theories extends the same concern to class-objects in the accepted language. In both cases a distinct piece of syntax need not correspond to a distinct entity. Thus the predicates (x=x) and (x\notin\varnothing) define the same universal extension of sets in ordinary ZFC discourse, although one expression derives from equality and the other from membership.
Extensionality settles an internal identity question only after the theory has specified which extensions exist. It cannot by itself validate an abstraction from every property. This makes three issues separable: whether a formula is meaningful, whether its extension exists in the designated ontology, and when two existing extensions are identical. Philosophical disputes become clearer when these three questions are addressed in that order.
Sets as Mathematical Representations
Foundational set theory can represent numbers, functions, relations, and structures using sets. Natural numbers may be coded by von Neumann ordinals, a function by its graph, an ordered pair by a standard set-theoretic construction, and a topological space by an ordered pair consisting of a set and a topology. Such encodings are central to the set-theoretic foundation of mathematics described by Bagaria (2023).
Representation, however, has more than one philosophical reading. On a strong reductionist interpretation, the relevant mathematical objects simply are the set-theoretic constructions. On a structural or methodological interpretation, set-theoretic representatives realize properties and relations needed by a mathematical practice without settling their unique intrinsic nature. A number-theoretic theorem remains valid under any isomorphic coding of the natural number structure, even if the concrete set identities of its representatives differ.
This observation bears on classes as well. The assertion "all groups" commonly refers to the structures satisfying the group axioms under a chosen coding. Group presentations can be shifted among equivalent set encodings. The corresponding class description has a stable structural use, while its exact extension depends on conventions of representation. It is therefore methodologically important to state whether a class is intended to collect presentations, isomorphism types, or another mathematically defined kind.
Structuralism and the Benacerraf Problem
Benacerraf (1965) used alternative set-theoretic constructions of the natural numbers to challenge the identification of a number with one unique particular set. The original argument in What Numbers Could Not Be distinguishes mathematical structural role from an arbitrary coding of that role. Later structuralist positions vary widely; Reck and Schiemer (2025) distinguish realist, eliminativist, and methodological versions in their survey of mathematical structuralism.
The relevance to sets and classes concerns identity and invariance. A foundational theory can choose representatives for an arithmetic progression without thereby identifying the concept of successor with the incidental membership details of its code. Similarly, a class of all structures satisfying an axiom set may be understood through invariance under isomorphism rather than through a privileged inventory of presentations. Such use remains compatible with rigorous set-theoretic definitions.
Structuralism does not automatically dismiss the ontological reality of sets or classes. Some structuralists are realists about structures; others seek eliminations or reconstructions. The point is that foundational adequacy and ontological identification are distinct achievements. A class can be technically apt as an extension of a property while the philosophical interpretation of the structures it ranges over remains plural.
Large Categories and the Size of Mathematical Practice
Category theory studies objects and morphisms, their composition, and identity morphisms. Mathematicians frequently consider the category Set whose objects are all sets and whose morphisms are functions between them. The objects of this category form a proper class in ordinary ZFC metatheory. Yet for individual sets A and B the collection (\mathrm{Hom}(A,B)) of functions from A to B is a set, making Set a familiar example of a locally small large category.
The category-theory reference entry explains how categories provide both a mathematical language and a possible alternative foundational perspective. Mac Lane's (1971) Categories for the Working Mathematician established a systematic treatment of functors, natural transformations, and universal constructions in the discipline. Size distinctions arise because mathematical practice moves between particular structures and categories surveying many structures at once.
The expression "the category of all categories" requires further care. The category of all small categories can be handled relative to a specified background level, while a category collecting every category of every size may lead to a new size problem. Another option uses Grothendieck universes, often with large-cardinal assumptions for the desired closure properties. These are concrete foundational choices, not proof that there is one final universal collection containing every mathematical object without restriction.
Categorical Reconstructions of Sets
Lawvere's 1964 Elementary Theory of the Category of Sets helped develop a foundation that characterizes set-like objects and their mappings in categorical terms. The relevant conceptual move is to give axioms for the organization of objects and arrows rather than to begin solely from the primitive membership relation. Such approaches can recover substantial ordinary mathematics while changing which operations and notions are primitive.
Category theory's philosophical account of structure considers this shift as part of a wider investigation into invariance and mathematical objects. It is important not to claim that a change of primitive vocabulary automatically removes every need for size control. The distinction between small and large categories appears as soon as one forms totalities of structures, functors, or categories beyond a fixed level.
The coexistence of categorical and set-theoretic foundations is philosophically productive. Each can organize broad regions of mathematics, and some of their technical results can be translated across frameworks. The fact that the same structural relationships admit different foundational presentations limits any argument that ontology must be uniquely read from a single notational primitive. The challenge is to identify which existence assumptions each presentation actually needs.
Independence, Constructibility, and Axiomatic Plurality
Gödel's 1938 announcement of relative consistency using constructible sets established that, assuming a suitable base set theory is consistent, adding Choice and the generalized continuum hypothesis does not produce a contradiction. His later Consistency of the Continuum Hypothesis (1940) gave the extended treatment. Cohen's (1963) Independence of the Continuum Hypothesis supplied the complementary forcing development, under suitable relative-consistency assumptions.
The mathematical upshot is that ZFC cannot decide every meaningful set-theoretic proposition if it is consistent. The continuum hypothesis is the celebrated example. This concerns the strength and limitations of particular axioms, not a license for arbitrary contradictory statements to be true in one and the same interpreted model.
For classes, the lesson is indirect but significant. A class-wide framework lives alongside choices about what sets exist, which higher-order principles hold, and which models are being described. It would be implausible to infer one final metaphysical interpretation of classhood solely from the fact that a widely used axiom system has become standard. The exact mathematical results are relative to formal systems; their philosophical meaning must be argued rather than asserted by association with independence.
The Set-Theoretic Multiverse and Ontological Variation
Hamkins (2012) advances a set-theoretic multiverse perspective, according to which a plurality of legitimate set-theoretic universes can illuminate mathematical experience with independence and forcing. A universe-oriented realist may instead posit one intended universe V whose statements have determinate truth values even when particular axioms do not decide them. Holmes's analysis of competing set conceptions records this disagreement within the wider landscape of alternative foundations.
For the present question, the contrast concerns how to interpret claims about all sets. Relative to one model, the totality is its internal V, described as a proper class. A different model or forcing extension may contain sets that the first does not. Multiverse interpretation gives a philosophical rationale for taking many such universes seriously; universe interpretation retains the possibility of a privileged absolute totality.
Both perspectives can use rigorous class notation. Their disagreement shows that technical use of V and class quantification leaves open substantive ontological commitments. It also demonstrates why the phrase "the class of all sets" should be accompanied by an account of the intended universe and the scope of the quantifier when the context is philosophically demanding.
Three Constructed Tests of Class Discourse
Consider first a researcher who defines a set of samples (S={a,b,c}) and its subcollection (T={x\in S:P(x)}). Under ordinary ZFC assumptions, T is a set by Separation. Calling it a "class" would add no necessary ontological entity; an existing set already serves as the relevant extension. The result depends on the set-sized domain S, not on how many predicates are used to describe its elements.
Second, suppose a researcher writes (G={x:x\text{ encodes a group}}). This generally ranges over group structures of unbounded set-theoretic size, under the chosen coding, and is a proper class in the standard universe. The researcher can still assert (\forall x(G(x)\rightarrow Q(x))) as shorthand for a first-order claim about all set-coded groups. If the argument manipulates G as a class object, a class-theoretic language can make that manipulation formal. The foundational role of set encodings explains why the specification is mathematically meaningful.
Third, a category theorist considers all small categories relative to a fixed universe U. These can be treated as a collection at one level while their aggregate may fall outside that level. The category-theoretic account of size makes the relevant dependency explicit. This example is constructed for analysis, not drawn from an empirical study: moving the universe boundary may change which objects count as small, and a new external viewpoint can survey the old one. The three tests jointly demonstrate that classhood is sensitive to axioms, coding, and scope, rather than simply the number of objects informally mentioned.
Realism, Nominalism, and Formalist Readings
A realist may interpret mathematical sets and classes as abstract entities, with existence independent of individual inscriptions or calculations. A nominalist may seek paraphrases that avoid commitment to the corresponding abstract objects. A formalist may emphasize structures of derivation and their relation to interpretations. These familiar positions supply distinct answers to the same formal evidence, as the philosophy of mathematical Platonism and Bricker's ontology-of-quantification discussion illustrate.
Neither the utility of set theory nor the paradoxes of naïve comprehension settle this philosophical competition. A realist can accept restrictions on set formation because genuine mathematical totalities obey principled laws. A nominalist can accept the same derivations while interpreting mathematical language as a powerful regimentation of inferential practice. A formalist can explain the rules that block inconsistency while acknowledging that model theory invokes mathematical structures of its own.
The strongest interpretive approach is to state the cost and benefit of each reading. Realism offers direct objectual semantics but owes an account of mathematical knowledge. Nominalism seeks ontological economy but must preserve mathematical inferential power. Formalism offers precision in derivation but must specify how formal results apply to mathematical understanding. Class discourse sharpens these traditional debates because the apparent domain of quantification is broader than any set permitted by the foundational axioms.
The Explanatory Work Performed by Proper Classes
Proper classes appear whenever a theory wishes to express systematic generality across all of its set-sized objects. They organize transfinite constructions, global functions, definable families, and categories of mathematical structures. Their usefulness is not an incidental convenience: it follows from the fact that a successful foundation must be able to speak about its entire domain while forbidding that domain from becoming one more set subject to the same unrestricted operations.
The results of contemporary class-forcing research illustrate an additional purpose. Proper-class operations can give rise to substantive theorems and independence phenomena inside second-order set theory; class forcing is not merely informal set forcing repeated with a larger input. The associated demands on recursion and satisfaction become mathematically significant.
But explanatory work should be distinguished from ontological proof. If a class term abbreviates a formula, its explanatory role can be genuine without adding an object to the first-order domain. If a theory quantifies over class objects, those objects play a more direct formal role. If plural language provides an adequate interpretation of a certain class use, one may preserve that use without objectifying the plurality. A disciplined philosophy of classes states which of these operations each argument actually performs.
Critical Objections and Replies
One objection holds that class quantification is clearly objectual, and therefore class realism follows immediately. The reply is that this conclusion is valid conditional on selecting that objectual semantics as ontologically authoritative. It is less decisive when the rival interpretation supplies definitional or plural translations preserving the relevant claims. Formal semantics is an indispensable part of inquiry, yet selecting one semantics as a metaphysical account requires reasons concerning explanation, truth, and mathematical practice. The debate over plural innocence exemplifies this unresolved issue.
A second objection treats all classes as merely definitional and therefore assumes that no class theory has consequences beyond ZFC. This misreads the role of impredicative comprehension. Kelley–Morse and intermediate class theories show that stronger higher-order principles can have substantive proof-theoretic consequences. The hierarchy of class-theoretic principles makes this distinction explicit; classes cannot be dismissed as a uniform shorthand.
A third objection says that a class larger than every set cannot be described consistently, because every description must denote a set. Russell's derivation shows the reverse: descriptions and predicates are available without unrestricted object-formation. A fourth objection says that if an external model treats an internal universe as a set, the distinction is meaningless. The internal and external domains differ; the distinction remains a mathematically informative assertion relative to their structures. Finally, a fifth objection invokes alternative foundations to suggest that no rigorous ontological answer is possible. The present study does not accept that consequence. Rigorous answers are possible once formal and semantic commitments are declared; philosophical dispute concerns the authority of those commitments.
Original Analytical Synthesis
The research yields four connected analytical judgments. First, sethood is a status within an axiomatic account of membership and permissible formation. The declaration that a class is proper records the impossibility, within that account, of giving the specified collection a set as its extension. Second, class discourse is underdetermined by ordinary grammar. A formula, a class variable, an externally given subset of a model, and a plural expression may perform overlapping communicative work while carrying different formal commitments. Third, an interpretation of classhood needs both an account of its quantificational domain and an account of what comprehension and other operations are licensed. Fourth, differences in conservative and stronger class theories demonstrate that philosophical assessment must attend to deductive strength as well as to the names of the entities being quantified over.
These judgments follow jointly from the distinction between ZFC and explicit class theories, the Russell and ordinal arguments, the separation of internal from external models, and philosophical debates over quantification. They constitute a synthesis of established mathematical and philosophical materials rather than a newly established theoretical system. In particular, the claim that proper-class talk can be exact without always involving an additional set is a direct consequence of the formal machinery; the claim that no single metaphysical reading follows from that fact is a comparative philosophical inference.
The positive conclusion is therefore precise: to understand the ontology of sets and classes, identify the accepted mathematical objects, identify the scope of the quantifiers, state the existence principles allowing classes or sets to form, and then give reasons for the preferred semantic interpretation. The technical picture of alternative class theories, the Quinean criterion of commitment, and the structuralist challenge to coding converge on the need for this disciplined reading even as they support different philosophical options.
Research Limitations and Further Questions
The study is conceptual and documentary. It establishes no new set-theoretic independence theorem, consistency result, completeness proof, or formal equivalence theorem. Its proof sketches reproduce standard mathematical results and depend on ordinary classical foundations. It does not supply a full formal interpretation theorem between every variation of GB, GBC, NBG, and KM, nor does it attempt an exhaustive survey of class forcing, large-cardinal theory, type-theoretic universes, topos foundations, or the philosophical literature on indefinite extensibility.
Several referenced primary publications were consulted as accessible full text or substantial excerpts; some older texts were checked at publisher or archival record level and then compared with specialist reference literature. For that reason, exact historical textual claims have been confined to documentary points supported by accessible material. Detailed source-verification scope and claim association are preserved in EVIDENCE.json. The external scholarly review of the manuscript is outside the present internal research and editorial cycle.
Further inquiry can focus on three questions: when a proposed translation from class quantification to plural discourse preserves the mathematical operations of a target field; how the availability of truth predicates and transfinite recursion alters the apparent ontology of classes; and which size foundations most economically support actual mathematical categories. These are questions for additional work. They are not represented here as newly named doctrines or discoveries.
Conclusion
Sets and classes are indispensable to the mathematical expression of totality and infinity. A set has the membership-governed status granted by an axiomatic foundation. A proper class is a collection recognized in a wider description or language that, under the operative set principles, cannot be a set. Russell's paradox and the ordinal arguments show why the collection of all sets and the collection of all ordinals cannot be added as ordinary set objects. Separation and Replacement make this limit compatible with a rich and rigorous mathematics. The cumulative hierarchy exhibits the total domain of sets while withholding a final set containing them all.
Class theories reveal more than one disciplined way to handle that domain. Predicative class comprehension in Gödel–Bernays systems, impredicative comprehension in Kelley–Morse, and intermediate principles studied through class forcing provide distinct formal resources. Model theory exposes the dependence of sethood and classhood judgments on internal and external standpoints. Quine's criterion of ontological commitment, Boolos's plural interpretation, structuralist analysis, and categorical foundations demonstrate that the mathematical classification alone does not settle the metaphysics of its discourse.
The definitive analytical result is that the ontology of sets and classes must be argued from explicit rules of formation, quantificational ranges, semantics, and explanatory practices. A proper class marks a principled limit within a foundation, while class discourse can remain mathematically expressive across that limit. The historical development from Cantorian infinity to modern class theories establishes a durable philosophical insight: exact mathematical generality depends upon specifying what can be collected, what can be quantified over, and what kind of existence those acts of description claim.
References
- Bagaria, Joan. 2023. "Set Theory." The Stanford Encyclopedia of Philosophy, substantive revision January 31, 2023. Set Theory.
- Benacerraf, Paul. 1965. "What Numbers Could Not Be." The Philosophical Review 74 (1): 47–73. DOI: 10.2307/2183530. Full article facsimile.
- Boolos, George. 1984. "To Be Is to Be a Value of a Variable (or to Be Some Values of Some Variables)." The Journal of Philosophy 81 (8): 430–449. Full journal article.
- Bricker, Phillip. 2014. "Ontological Commitment." The Stanford Encyclopedia of Philosophy, first published November 3, 2014. Ontological Commitment.
- Cantor, Georg. 1895 and 1897. "Beiträge zur Begründung der transfiniten Mengenlehre." Mathematische Annalen 46: 481–512; 49: 207–246. English translation by Philip E. B. Jourdain, Contributions to the Founding of the Theory of Transfinite Numbers, Open Court, 1915; Dover reprint. English translation and notes.
- Cohen, Paul J. 1963. "The Independence of the Continuum Hypothesis." Proceedings of the National Academy of Sciences 50 (6): 1143–1148. DOI: 10.1073/pnas.50.6.1143. Original paper bibliographic record.
- Cunningham, Daniel W. "Set Theory." Internet Encyclopedia of Philosophy. Accessed November 7, 2026. Set Theory encyclopedia article.
- Deutsch, Harry, Oliver Marshall, and Andrew David Irvine. 2024. "Russell's Paradox." The Stanford Encyclopedia of Philosophy, substantive revision December 18, 2024. Russell's Paradox.
- Gitman, Victoria, Joel David Hamkins, Peter Holy, Philipp Schlicht, and Kameryn Williams. 2017. "The Exact Strength of the Class Forcing Theorem." arXiv:1707.03700. Research preprint.
- Gödel, Kurt. 1938. "The Consistency of the Axiom of Choice and of the Generalized Continuum-Hypothesis." Proceedings of the National Academy of Sciences 24 (12): 556–557. Original announcement.
- Gödel, Kurt. 1940. The Consistency of the Axiom of Choice and of the Generalized Continuum-Hypothesis with the Axioms of Set Theory. Annals of Mathematics Studies 3. Princeton: Princeton University Press. Publisher record.
- Hamkins, Joel David. 2012. "The Set-Theoretic Multiverse." The Review of Symbolic Logic 5 (3): 416–449. DOI: 10.1017/S1755020311000359. Original journal article.
- Holmes, M. Randall. 2021. "Alternative Axiomatic Set Theories." The Stanford Encyclopedia of Philosophy, substantive revision September 21, 2021. Alternative Axiomatic Set Theories.
- Lawvere, F. William. 1964. "An Elementary Theory of the Category of Sets." Proceedings of the National Academy of Sciences 52 (6): 1506–1511. DOI: 10.1073/pnas.52.6.1506. Original journal record.
- Mac Lane, Saunders. 1971. Categories for the Working Mathematician. Graduate Texts in Mathematics 5. New York: Springer-Verlag. Publisher record, first edition.
- Linnebo, Øystein. 2022. "Plural Quantification." The Stanford Encyclopedia of Philosophy, substantive revision February 16, 2022. Plural Quantification.
- Quine, W. V. O. 1948. "On What There Is." The Review of Metaphysics 2 (5): 21–38. Original journal article.
- Reck, Erich H., and Georg Schiemer. 2025. "Structuralism in the Philosophy of Mathematics." The Stanford Encyclopedia of Philosophy, substantive revision September 17, 2025. Structuralism in the Philosophy of Mathematics.
- Marquis, Jean-Pierre. 2019. "Category Theory." The Stanford Encyclopedia of Philosophy, substantive revision August 29, 2019. Category Theory.
- Zermelo, Ernst. 1908. "Untersuchungen über die Grundlagen der Mengenlehre. I." Mathematische Annalen 65: 261–281. DOI: 10.1007/BF01449999. Digitized original and bibliographic record.
- Linnebo, Øystein. 2023. "Platonism in the Philosophy of Mathematics." The Stanford Encyclopedia of Philosophy. Platonism in the Philosophy of Mathematics.