Elementary Theory of the Category of Sets (ETCS)

Origin. F. William Lawvere, "An elementary theory of the category of sets" (PNAS, 1964), written two years after his thesis and rejected by referees who could not see that it was a set theory. Lawvere and Rosebrugh's Sets for Mathematics (2003) is the textbook; Tom Leinster's "Rethinking set theory" (2014) is the readable case for it. The structural alternative to the cumulative hierarchy.

Models. ZFC says what a set is made of: membership is primitive, and every set's identity is fixed by its elements, which are themselves sets, all the way down. ETCS says what a set does: the primitives are functions and composition, an element is a map from a one-element set, and nothing is asked about what a set contains beyond what its maps to and from other sets reveal. The question "is 3 ∈ 7?" has an answer in ZFC and is not well formed here.

Formalism.

Language: A category: objects, arrows, composition, identities. No membership relation. x ∈ A is not a primitive.

Elements as maps: An element of A is an arrow 1 → A from a terminal object. Two arrows f, g : A → B are equal iff they agree on all elements — which is the axiom that 1 is a generator.

The axioms (a well-pointed topos with NNO and choice):

1. Finite limits exist (terminal object, products, equalizers)
2. Exponentials exist: B^A, so the category is cartesian closed
3. A subobject classifier Ω exists: subsets are maps A → Ω
4. A natural numbers object exists
5. Every epi splits (the axiom of choice)
6. 1 is a generator (well-pointedness)
7. Ω has exactly two elements (Boolean, so classical logic)

Strength: ETCS is bi-interpretable with BZC — bounded Zermelo with choice: Zermelo's axioms with separation restricted to bounded formulas, plus choice. Weaker than ZFC: no replacement. ETCS + replacement (as a categorical axiom scheme) is bi-interpretable with ZFC.

What changes in practice: No global membership: elements of different sets are not comparable. No cumulative hierarchy: sets are not built in stages. Isomorphic sets are interchangeable — structural invariance is automatic rather than a convention. Functions are primitive; ZFC's coding of a function as a set of ordered pairs is unnecessary.

Symbols.

Metatheory. Bi-interpretability with BZC is the result that makes ETCS a set theory rather than a manifesto: the two prove the same theorems about sets, so the choice between membership and maps is a choice of presentation and not of content — and the ZFC-strength version is one axiom scheme away. The absence of global membership is the point rather than a limitation: in ZFC, "3 ∈ 7" has a truth value that depends on the von Neumann coding and that no mathematician uses, and structural invariance has to be imposed by convention. ETCS makes it a theorem. Well-pointedness and the two-element Ω are what keep the theory classical; drop them and the axioms describe a topos, which is where the categorical logic in this collection lives.

Applies to. Foundations, as the structural alternative to the iterative conception. Category theory's relation to set theory. Type-theoretic foundations, where ETCS is the closest set-theoretic relative. Any argument about whether membership is fundamental or an artifact of a coding.

Limitations. Weaker than ZFC without a replacement scheme, and the categorical formulation of replacement is technical enough that most presentations either omit it or defer it. The axioms presuppose category theory, which makes ETCS a poor foundation for anyone who does not already have one — the circularity charge is standard and the standard answer, that the axioms are elementary and first-order, is correct and unpersuasive to the objectors. Large-cardinal strength has no natural ETCS formulation, so the whole upper reach of set theory is out of view.

© 2026 Lingenic LLC