Loading field map
Set theory began with a simple, explosive question: what is the size of infinity? In the 1870s, Georg Cantor discovered that infinite sets come in different sizes—the natural numbers are countable, the real numbers are uncountable, and there is an endless hierarchy of ever-larger infinities. This discovery opened a new mathematical universe, but it also revealed a deep tension: the intuitive principle that every property defines a set (the unrestricted comprehension principle) leads directly to paradoxes, most famously Russell's paradox of the set of all sets that do not contain themselves. The history of set theory is the story of how mathematicians responded to that tension, building frameworks that could capture Cantor's insights while avoiding contradiction, and then discovering that those frameworks left fundamental questions about the nature of the infinite permanently unsettled.
Cantor's work between 1874 and the 1890s created what is now called Naive Set Theory. He defined sets as any collection of definite, distinct objects, and he assumed that for any property, there is a set of all objects satisfying that property. On this foundation, he proved that the real numbers are uncountable, developed the arithmetic of infinite cardinal numbers, and formulated the continuum hypothesis (CH): the conjecture that every infinite subset of the real numbers is either countable or the same size as the whole set. Naive set theory was enormously productive, but it was also unstable. By 1901, Bertrand Russell had shown that the unrestricted comprehension principle produces a contradiction. The paradox did not merely embarrass Cantor's framework; it forced a complete rethinking of what a set is and how set theory should be axiomatized.
Two competing strategies emerged to replace naive set theory. The first, Typed Set Theories, was pioneered by Russell himself in his theory of types (1908). Russell's solution banned self-membership by arranging sets into a hierarchy of types: elements of type 0 are individuals, sets of type 1 contain only type-0 objects, sets of type 2 contain only type-1 sets, and so on. This blocks Russell's paradox because no set can contain itself, but it also makes ordinary mathematics cumbersome—real numbers, for instance, must be reconstructed at multiple types. The second strategy, Zermelo-Fraenkel Set Theory (ZFC, 1908–1930), took a different approach. Instead of typing, ZFC restricts the comprehension principle to a separation axiom: given any existing set, you can form the subset of its elements that satisfy a property. This prevents the formation of a universal set while preserving the flexibility mathematicians need. ZFC also introduced the axiom of choice, which allows selecting an element from each set in an infinite family—a principle that turned out to be independent of the other axioms and deeply controversial. Over the following decades, ZFC became the dominant foundation for mathematics, largely because its iterative conception of sets (the cumulative hierarchy) felt natural and its axioms were strong enough to develop all of classical mathematics. Typed set theories survived as active alternatives—New Foundations, for instance, remains a living research program—but ZFC's simplicity and power made it the default.
Even ZFC had a limitation: it could not talk directly about collections that are too large to be sets, such as the class of all sets or the class of all ordinals. Class-Theoretic Set Theory, developed by von Neumann, Bernays, and Gödel (NBG) in the 1920s–1930s, extended ZFC by adding proper classes—collections that are not members of any other collection. NBG is a conservative extension of ZFC: every theorem about sets in NBG is already a theorem of ZFC, but NBG provides a convenient language for discussing large-scale structure. At the same time, the Cumulative Hierarchy (V) gave ZFC its intuitive picture. The universe V is built in stages: start with the empty set, then at each stage form all subsets of the previous stage, and continue through all ordinal numbers. Every set appears at some stage, and the hierarchy grows without bound. This iterative conception became the standard way to think about what ZFC describes, and it remains the background picture for most set-theoretic research.
Once ZFC was established, set theorists began asking whether the universe V could be extended. Large Cardinal Axioms assert the existence of cardinals so large that their existence cannot be proved in ZFC alone. The first such axiom, the existence of an inaccessible cardinal, was proposed by Sierpiński and Tarski in 1930. Later axioms grew stronger: Mahlo cardinals, measurable cardinals, Woodin cardinals, and supercompact cardinals, each implying the consistency of the previous ones. Large cardinals are not idle curiosities; they serve as a calibration tool for consistency strength. Many natural statements—such as the determinacy of projective sets or the existence of certain combinatorial objects—turn out to be equiconsistent with a large cardinal axiom. This has made large cardinals a unifying thread across set theory: they provide a yardstick for measuring how much power is needed to settle a given question.
In 1938, Kurt Gödel introduced The Constructible Universe (L), an inner model of ZFC that is built by a more restrictive process than V. Instead of taking all subsets at each stage, L only takes subsets that are definable using first-order logic. The result is a universe that satisfies ZFC plus the axiom of choice and the generalized continuum hypothesis (GCH). Gödel's construction showed that if ZFC is consistent, then so is ZFC + CH—the continuum hypothesis cannot be disproved. But L also has a restrictive character: it satisfies combinatorial principles like ◇ (diamond) that imply CH and many other statements that are independent of ZFC. For decades, set theorists wondered whether L might be the whole universe. The answer turned out to be no, but L became the prototype for a whole family of inner models.
Running parallel to the development of large cardinals and inner models, Combinatorial Set Theory emerged as a distinct framework focused on the combinatorial structure of infinite sets. Its central concerns include partition relations (Ramsey theory for infinite cardinals), cardinal arithmetic (especially the behavior of exponentiation under the axiom of choice), and combinatorial principles such as ◇ and □ (square) that hold in L but can fail in forcing extensions. Combinatorial set theory provides the technical tools for many independence proofs and remains an active area, with results that often interact with large cardinals and forcing.
The landscape of set theory was transformed in 1963 when Paul Cohen invented Forcing. Forcing is a technique for extending a given model of ZFC to a larger model that satisfies new statements. Using forcing, Cohen proved that the continuum hypothesis is independent of ZFC: ZFC cannot prove CH, and it cannot prove its negation. Forcing also showed that the axiom of choice is independent of ZF. The method works by adding new sets—typically new subsets of ω—to a model while carefully controlling which statements remain true. Forcing revealed that ZFC is radically incomplete: many natural questions, including CH and a host of combinatorial statements, are not settled by the axioms. This discovery reshaped set theory, turning it from a search for the one true universe into an investigation of the space of possible universes.
If forcing shows how to build wide universes, Inner Model Theory explores how to build narrow ones. Starting from Gödel's L, set theorists asked: can we construct inner models that contain large cardinals? The answer is yes, but the construction becomes increasingly intricate. For measurable cardinals, the inner model L[U] adds a measure U to the construction; for stronger cardinals, the models become more complex, culminating in the Steel core model for superstrong cardinals. Inner model theory derives directly from the Constructible Universe: it generalizes L's definability hierarchy to accommodate large cardinals, producing models that are as close to L as possible while still containing the large cardinal. These models have deep connections to determinacy axioms: the minimal model containing all Woodin cardinals, for instance, satisfies projective determinacy.
Determinacy Axioms assert that certain infinite games are determined—one of the two players has a winning strategy. The axiom of determinacy (AD) states that every game on the natural numbers is determined, but AD contradicts the axiom of choice. The fruitful version is definable determinacy: restrict attention to games whose payoff set is definable in some complexity class, such as projective sets. Projective determinacy (PD) is consistent with ZFC and, remarkably, is equivalent to the existence of certain large cardinals. The connection runs through inner model theory: the inner model L(R) (the constructible closure of the reals) satisfies AD if and only if there are infinitely many Woodin cardinals in V. This result, due to Martin, Steel, and Woodin in the 1980s, showed that large cardinals are not just consistency-strength markers—they directly imply structural properties of the real line. Descriptive Set Theory, which studies the definable subsets of the real numbers, provides the natural setting for determinacy. Originally developed in the early 1900s by Borel, Lebesgue, and Lusin, descriptive set theory classifies sets of reals by their topological complexity. Determinacy axioms resolve many classical questions about these sets (e.g., every projective set is Lebesgue measurable) that are undecidable in ZFC alone.
Forcing Axioms are principles that assert that the universe is saturated with respect to certain forcing extensions. The simplest is Martin's Axiom (MA), which says that no cardinal less than the continuum can be collapsed by a certain class of forcings. Stronger axioms, such as the Proper Forcing Axiom (PFA) and Martin's Maximum (MM), assert that the universe is closed under a much wider class of forcing constructions. Forcing axioms typically imply that the continuum is ℵ₂ and that many combinatorial statements that hold in L (like ◇) fail. This creates a direct rivalry with the Constructible Universe: L represents a narrow, well-ordered universe where CH holds and combinatorial principles abound; forcing axioms represent a wide, saturated universe where CH fails and the continuum is relatively small. The rivalry is not merely philosophical—it is a living disagreement about which principles are true of V. Large cardinal axioms influence forcing axioms: PFA, for instance, is consistent relative to a supercompact cardinal, and its consequences often mirror those of large cardinals.
The independence results produced by forcing, combined with the proliferation of mutually incompatible extensions of ZFC, led some set theorists to question whether there is a single intended universe. Multiverse Set Theory, articulated by Joel David Hamkins in 2012, proposes that set theory studies a plurality of universes, each equally legitimate. On this view, the continuum hypothesis does not have a determinate truth value; instead, different universes satisfy different versions of CH, and the set theorist's job is to understand the structure of the multiverse of all models of ZFC. This framework contrasts sharply with the traditional view that V is a unique, maximal universe. The multiverse position remains controversial: many set theorists continue to search for new axioms (such as large cardinals or forcing axioms) that would settle CH, while others embrace pluralism as the correct interpretation of the independence phenomenon.
Today, several frameworks remain active and in productive tension. ZFC is the undisputed common foundation, but research extends far beyond it. Large cardinal axioms provide a hierarchy of consistency strength that anchors most modern work. Inner model theory and forcing axioms represent two competing visions of the universe: one building narrow, canonical models that accommodate large cardinals, the other building wide, saturated universes that violate CH. Determinacy axioms, through their connection to large cardinals and inner models, have become a central tool for understanding the real line. Descriptive set theory continues to develop, now deeply intertwined with large cardinals and determinacy. The multiverse view offers a philosophical interpretation of this diversity, but it has not displaced the search for a single true theory. What the leading frameworks agree on is that ZFC is only the starting point: the interesting questions lie in the realm of statements independent of ZFC, and the methods of forcing, inner models, and large cardinals are the tools for exploring that realm. What they disagree on is whether those tools will eventually converge on a unique picture of V or whether the plurality of universes is the final word.