Morse–Kelley set theory (MK), also called Kelley–Morse set theory (KM), is a powerful axiomatic theory of sets and classes that extends the foundational framework of ordinary Zermelo–Fraenkel set theory. Whereas ZFC speaks formally only about sets, Morse–Kelley set theory allows both sets and proper classes to participate directly in its foundational language. This makes it possible to reason explicitly about collections too large to constitute sets, such as the class of all sets, the class of all ordinals, and the class of all cardinals. MK therefore provides a richer foundational language for describing the global architecture of set theory while retaining ordinary set mathematics within its set-sized portion.

The distinction between a set and a proper class is fundamental. A set is a collection that may itself occur as an element of another set. A proper class is a collection so extensive that it cannot itself be a set without generating contradiction. The collection of every ordinal, traditionally written Ord, is a standard example. If Ord were itself an ordinal, it would necessarily exceed every ordinal while simultaneously being one of them, producing the Burali–Forti paradox. Set theory therefore treats Ord not as a set but as a proper class. Likewise, the total collection V of all sets is a proper class rather than a set.

Within ZFC, proper classes are normally treated indirectly. One may informally write expressions such as “the class of all ordinals,” but classes are not independent objects quantified over by the formal first-order language of ZFC. They function as shorthand for definable conditions. Morse–Kelley set theory changes this situation by incorporating classes explicitly into the theory. Statements may quantify over classes themselves, allowing one to reason directly about global collections and operations extending throughout the entire cumulative hierarchy.

The fundamental relationship between sets and classes can be expressed by saying that every set determines a class, but not every class is a set. Sets are the small collections of the theory, while proper classes are collections that cannot occur as members. Thus the collection of natural numbers ω is a set, the real numbers constitute a set, and even enormously large cumulative ranks such as Vκ are sets whenever κ is an ordinal. By contrast, V itself, Ord, and the collection of all cardinals are proper classes because no single set can contain their complete extent.

One of the defining features of Morse–Kelley set theory is its powerful Class Comprehension principle. Roughly stated, whenever a property can be expressed in the language of MK—including quantification over classes—there exists a class consisting precisely of the sets satisfying that property. Schematically,

∃A ∀x (x ∈ A ↔ φ(x)),

where φ may contain quantification over both sets and classes, subject to the appropriate restrictions on free variables.

This is significantly stronger than the class-comprehension principle available in von Neumann–Bernays–Gödel set theory (NBG). In NBG, class comprehension is restricted so that the defining formula quantifies only over sets. MK permits class quantifiers within comprehension formulas. This seemingly technical difference dramatically increases the strength of the theory because classes may now be constructed according to properties involving the totality of other classes.

The comparison between ZFC, NBG, and MK is therefore especially important. ZFC is formulated entirely in terms of sets. NBG adds explicit classes but is a conservative extension of ZFC for statements purely about sets: if NBG proves a sentence expressed entirely in the language of sets, then ZFC already proves it. Consequently, NBG provides a convenient language for discussing classes without increasing the ordinary set-theoretic consequences beyond ZFC.

Morse–Kelley set theory is different. MK is not conservative over ZFC. Its stronger class-comprehension principles allow it to establish set-theoretic consequences that cannot in general be proved within ZFC alone. In particular, MK proves the consistency of ZFC, assuming MK itself is consistent. By Gödel's Second Incompleteness Theorem, a consistent ZFC cannot prove its own consistency. This immediately demonstrates that MK possesses genuinely greater proof-theoretic strength than ordinary ZFC rather than merely supplying a more convenient vocabulary.

This additional strength arises because quantification over arbitrary classes enables MK to describe global constructions unavailable to first-order ZFC. Instead of examining only individual sets within the cumulative hierarchy, one can discuss class-sized functions, class relations, class recursions, global well-orderings, satisfaction predicates for set-sized structures, and constructions extending through all ordinals. The theory can therefore express certain aspects of the entire cumulative hierarchy as a unified object of mathematical investigation.

Class recursion provides a particularly important example. Ordinary transfinite recursion constructs sequences indexed by set-sized ordinals. In a class theory such as MK, one may formulate recursion extending throughout the proper class Ord. This permits global constructions whose stages continue through every ordinal without requiring the total resulting collection to constitute a set. Such methods are extremely useful in higher set theory because many naturally occurring constructions—inner models, cumulative hierarchies, global functions, and canonical class sequences—are inherently proper-class-sized.

Morse–Kelley set theory also provides a natural framework for discussing elementary embeddings involving proper classes. Large-cardinal theory frequently considers expressions such as

j : V → M

or even hypothetical embeddings such as

j : V → V.

Within ordinary ZFC, these expressions require careful metatheoretical interpretation because V is not a set. A sufficiently expressive class theory allows such embeddings to be treated directly as class functions. This does not automatically guarantee that the corresponding large cardinals exist, but it provides a formal environment in which global embeddings and class-sized structures can be discussed more naturally.

The relationship between MK and the cumulative hierarchy is consequently deeper than merely adding another type of collection. ZFC primarily describes objects occurring inside V. MK can additionally reason about classes extending across V as a whole. A class may intersect every cumulative rank without ever becoming an element of any rank. The distinction therefore introduces a fundamental separation between objects internal to the set hierarchy and global collections defined throughout that hierarchy.

Morse–Kelley set theory should nevertheless not be interpreted as creating a “set of all sets.” The very purpose of the set/class distinction is to prevent precisely this collapse. The collection V exists as a proper class, but V is not permitted to become a member of another ordinary set. Likewise, Ord can be discussed as a complete class without becoming an ordinal or a set. MK thereby permits global mathematical totalities to be discussed while preserving the structural restrictions required to avoid the classical paradoxes of unrestricted set formation.

There are several formulations of Morse–Kelley set theory, and details concerning primitive notions, Global Choice, and particular axiom systems may vary between presentations. Some formulations use a one-sorted language in which sets are distinguished among classes by whether they belong to another class, while others employ separate variables for sets and classes. These differences affect presentation but not the central conceptual feature: MK provides impredicative class comprehension strong enough to quantify over classes while defining new classes.

The word impredicative is especially significant here. A definition is impredicative when an object may be defined through quantification over a totality that includes objects of the same general kind as the object being defined. MK's class comprehension permits this kind of reasoning for classes. NBG's elementary comprehension does not permit unrestricted class quantification in the same way. The impredicative character of MK is therefore one of the principal sources of its increased foundational strength.

From a foundational perspective, Morse–Kelley set theory demonstrates that the distinction between sets and proper classes creates another major level of mathematical organization beyond ordinary set construction. ZFC develops an enormous cumulative hierarchy of sets, but the hierarchy itself cannot become one of its own set-sized objects. Class theory provides a language for discussing structures whose scope extends throughout that hierarchy without falsely reducing them to sets. MK then strengthens this framework by allowing the totality of classes to participate in powerful comprehension principles.

This distinction becomes increasingly important as mathematical investigations move from local structures to global set-theoretic architecture. Questions concerning all ordinals, global choice functions, class forcing, elementary embeddings of V, inner models extending throughout Ord, and global reflection principles naturally invoke proper classes. Morse–Kelley set theory supplies a rigorous framework in which these objects can be handled directly rather than merely treated as informal abbreviations external to the theory.

Ultimately, Morse–Kelley set theory is a powerful axiomatic theory of sets and proper classes distinguished by its impredicative class-comprehension principle. It extends the ordinary set-theoretic environment of ZFC by permitting direct quantification over proper classes and exceeds conservative class theories such as NBG in consistency and proof-theoretic strength. Sets continue to constitute the ordinary mathematical objects of the cumulative hierarchy, while proper classes describe global collections such as V and Ord that extend beyond every set-sized stage. Through this distinction, MK provides a foundational framework capable of reasoning not merely about objects within the set-theoretic hierarchy, but about global structures extending throughout the hierarchy as a whole.