Regular Category¶

A finitely complete category in which every morphism has a pullback-stable regular-epimorphism–monomorphism image factorization—equivalently, kernel-pair quotients exist and regular epimorphisms remain regular under pullback.

Version: v2 · 2026-08-30 · History
Domain-specific #: 2646
Origin domain: category theory
Subdomain: categorical exactness

Core Idea¶

A regular category is a category with enough finite limits and stable quotients to support a well-behaved notion of image. Concretely, every morphism factors as a regular epimorphism followed by a monomorphism, and this image factorization survives pullback. An equivalent axiom package requires finite limits, coequalizers of all kernel pairs, and pullback stability of regular epimorphisms.^[1]^[2]

This is categorical exactness without assuming additive structure. In the category of sets, every function \(f:X\to Y\) first surjects onto its image and then includes that image into \(Y\); pulling this factorization back along any map \(Y'\to Y\) gives the corresponding image factorization over \(Y'\). A regular category isolates precisely enough structure for that familiar behavior to make sense when objects have no elements and arrows need not be functions between sets.

The word “regular” therefore does not mean merely orderly, smooth, typical, or satisfying a repeated pattern. It names a specific categorical axiom class. Regularity provides canonical images, a stable calculus of internal relations, and semantics for regular logic—the fragment generated by equality, truth, finite conjunction, and existential quantification.^[3]^[4] It is weaker than Barr exactness, which additionally requires every internal equivalence relation to be effective, and independent of Quillen's different use of “exact category.”

Structural Signature¶

Category + all finite limits + kernel pairs for every morphism + coequalizers of those kernel pairs + regular epimorphisms stable under arbitrary pullback -> stable regular-epi/mono image factorizations, internal relations, and existential-image semantics.

The mandatory roles are:

underlying category \(\mathcal C\): objects, morphisms, identities, and associative composition;
finite-limit structure: in particular a terminal object and pullbacks, hence binary products and equalizers;
kernel pair: for each \(f:X\to Y\), the pullback \(R=X\times_YX\) with projections \(r_1,r_2:R\rightrightarrows X\);
kernel-pair quotient: a coequalizer \(e:X\to I\) of \(r_1,r_2\);
induced image map: a unique \(m:I\to Y\) with \(f=m\circ e\);
regular epimorphism: a morphism that is a coequalizer of some parallel pair, and in this setting of its own kernel pair;
pullback stability: every pullback of a regular epimorphism is again a regular epimorphism;
image factorization: every \(f\) factors as \(X\twoheadrightarrow I\hookrightarrow Y\), uniquely up to the expected isomorphism.

The two common definitions are equivalent:

\(\mathcal C\) has finite limits; each kernel pair has a coequalizer; and regular epimorphisms are pullback-stable.
\(\mathcal C\) has finite limits; each morphism has a regular-epi/mono factorization; and these factorizations are stable under pullback.

The stability clause is load-bearing. It says that forming an image commutes with change of base. Without it, quotients may exist globally but cease to behave like the same kind of quotient when restricted along another morphism.

Recognition test. For an arbitrary \(f:X\to Y\), construct its kernel pair, quotient that pair, prove the induced map into \(Y\) is monic, and then pull the construction back along an arbitrary \(g:Y'\to Y\). If the pulled-back quotient remains regular epic and gives the image of the pulled-back map, the characteristic mechanism is present.

What It Is Not¶

It is not merely a category. Every regular category has objects, arrows, identities, and associative composition, but most of the definition lies beyond those axioms: finite limits, kernel-pair quotients, image factorization, and base-change stability.

It is not the claim that every epimorphism is regular. The privileged left class consists of regular epimorphisms—coequalizers—not arbitrary epis. In a regular category the regular epis and monos form the stable image factorization system; an ordinary epi may require separate analysis.

It is not a category with all coequalizers or all finite colimits. Only coequalizers of kernel pairs are required. Nor does regularity say that pulling back an arbitrary coequalizer preserves it. The preservation statement is specifically about regular epimorphisms or, equivalently, kernel-pair quotients.

It is not Barr-exactness. A kernel pair is automatically an internal equivalence relation, and regularity makes its quotient behave well. Barr-exactness adds the converse effectiveness requirement: every internal equivalence relation must arise as the kernel pair of its quotient.^[5] A regular category can fail that stronger condition.

It is not a Quillen exact category, where an additive category is equipped with a distinguished class of short exact sequences. The shared word “exact” belongs to different axiom traditions.

It is not a coherent category or a topos. A coherent category adds stable finite unions of subobjects. An elementary topos has substantially more structure and is regular, but regularity alone does not supply power objects, exponentials, or a subobject classifier.

It is not regular logic itself. Regular logic is a syntactic fragment; a regular category is the semantic environment in which its conjunctions, equations, and existential images are interpreted.

Scope of Application¶

Regular categories sit at the meeting point of categorical algebra, categorical logic, universal algebra, topos theory, and relation calculus. They are useful whenever one needs images and quotients that remain meaningful after changing parameters or base objects.

In categorical algebra, regularity captures exactness properties common to sets, groups, rings, algebraic varieties, abelian categories, and topoi without imposing addition on every hom-set.^[2] It is therefore broad enough to compare nonadditive algebra with homological examples while retaining a common image-factorization language.

In categorical logic, subobjects of an object represent predicates in a context. Finite limits interpret equality, truth, and conjunction. For \(f:X\to Y\), the direct image of a subobject \(A\hookrightarrow X\) is the image of \(A\to X\to Y\), interpreting existential quantification. Pullback-stable images ensure that this existential operation respects substitution of parameters through the Beck–Chevalley behavior expected of regular logic.^[3]

In the calculus of internal relations, a relation from \(A\) to \(B\) is a subobject \(R\hookrightarrow A\times B\). To compose \(R\hookrightarrow A\times B\) with \(S\hookrightarrow B\times C\), form \(R\times_BS\) and then take its image in \(A\times C\). Stable image factorization makes this composition well defined and associative, generalizing ordinary relational composition in Set.^[6]^[4]

Regular completion and exact completion constructions show another scope. They freely add missing stable images or effective quotients under universal properties, allowing a weakly complete presentation to be embedded in a setting with the required exactness.^[7]

Clarity¶

Regular Category replaces the vague phrase “images behave well” with four checkable questions:

Do finite limits exist?
Does each map's kernel pair have a coequalizer?
Does that quotient induce a monomorphism into the codomain?
Do regular epimorphisms remain regular epic after every pullback?

It also clarifies three morphism notions that are easy to conflate. A monomorphism is left-cancellable. An epimorphism is right-cancellable. A regular epimorphism has constructive quotient evidence: it is a coequalizer. Regularity privileges the quotient-evidenced class because it interacts correctly with pullback and images.

Finally, it distinguishes a quotient one already knows to arise from a map's equality relation from a quotient of an arbitrary internal equivalence relation. The first is required by regularity; the second is the additional effectiveness demand of Barr exactness. This single distinction prevents a common overstatement that every regular category already has all well-behaved equivalence-relation quotients.

Manages Complexity¶

Stable image factorization compresses arbitrary maps into a two-stage normal form:

\[ X \xrightarrow{e} \operatorname{Im}(f) \xrightarrow{m} Y, \]

with \(e\) regular epic and \(m\) monic. Arguments can separate the quotient-like part from the inclusion-like part instead of reasoning about every map as an indivisible object. Because the factorization is unique up to isomorphism, the image is structural rather than presentation-dependent.

Pullback stability makes the normal form modular. A construction or proof carried out over \(Y\) can be transported to a parameterized fiber over \(Y'\) without re-proving that the quotient still behaves as a quotient. This is why regularity is often described as stable exactness under change of base.

The internal relation calculus is a second compression. A potentially element-heavy statement such as “there exists an intermediate \(b\) with \(aRb\) and \(bSc\)” becomes pullback followed by image. The same operation works in sets, groups, topoi, and other regular categories. Regular logic similarly replaces repeated element arguments with a small categorical kit: pullback for substitution, intersection for conjunction, diagonal for equality, and image for existence.

Abstract Reasoning¶

Start with \(f:X\to Y\). Its kernel pair \(R=X\times_YX\) records pairs of generalized elements that \(f\) identifies. Suppose

\[ R \underset{r_2}{\overset{r_1}{\rightrightarrows}} X \xrightarrow{e} I \]

is the coequalizer. Because \(f r_1=f r_2\), there is a unique \(m:I\to Y\) satisfying \(f=me\). Pullback stability of regular epis forces \(m\) to be monic: informally, any further equality created by \(m\) would pull back to an equality already collapsed by \(e\), contradicting the universality of the kernel-pair quotient. Thus \(I\hookrightarrow Y\) is the categorical image.

For sets, the construction is familiar. \(R\) consists of \((x,x')\) with \(f(x)=f(x')\). Quotienting \(X\) by this relation gives its fibers as equivalence classes, and the induced map \(X/R\to Y\) injects onto \(f[X]\). Pulling along \(Y'\to Y\) restricts the function and its image compatibly.

Several deductions follow from the axioms:

regular epimorphisms compose and are stable under pullback;
every regular epi is the coequalizer of its own kernel pair;
regular epis and monos form a factorization system;
images and inverse images of subobjects satisfy the stability needed for existential semantics;
internal relations compose by pullback-then-image;
a regular functor can be characterized as preserving finite limits and regular epimorphisms, equivalently the relevant image or kernel-pair quotient structure.

These conclusions are not separate axioms pasted together. They express one invariant: quotienting identified information and embedding the residual image commutes with contextual restriction.

Knowledge Transfer¶

The identity transfers literally across different categorical substrates. In Set, regular epis are surjections and monos are injections. In Grp, a homomorphism factors through its quotient by the kernel and then into its image subgroup. In an abelian category, the familiar coimage-to-image factorization supplies the same pattern. In a topos, the logical and sheaf-theoretic structure includes regularity as a foundational layer.

The same reasoning transfers across tasks: algebraic image arguments, semantic existential quantification, relational composition, descent and quotient questions, and universal completions all use stable image factorization. The internal objects and morphisms change, but the recognition test does not.

Outside category theory, phrases such as “stable quotient,” “image,” and “regular structure” appear widely, but they do not by themselves instantiate Regular Category. The portable high-level ideas—factorization, stability under transformation, equivalence relations, and compositionality—belong to existing primes. Literal recurrence requires an actual category, finite limits, categorical kernel pairs, regular epimorphisms, and pullback.

Examples¶

Sets. Every function factors as a surjection onto its ordinary image followed by inclusion. Surjections are coequalizers of their kernel pairs and remain surjective under pullback. Set is not only regular but Barr-exact, because every equivalence relation is the kernel pair of its quotient.

Groups and algebraic varieties. A group homomorphism \(f:G\to H\) factors as \(G\twoheadrightarrow G/\ker f\cong\operatorname{Im}(f)\hookrightarrow H\). More generally, categories of models of finitary algebraic theories are regular and, in standard set-based settings, exact. This shows that additivity is not required.^[2]

Abelian categories. Kernels, cokernels, and the coimage–image isomorphism give stable epi–mono factorization. Every abelian category is regular and Barr-exact, but the definition of Regular Category deliberately retains only the nonadditive exactness core.

Elementary topoi. A topos is regular; images and existential quantification are stable under pullback. Yet the topos also carries exponentials and a subobject classifier, none of which regularity alone implies.

Internal relation composition. Given \(R\hookrightarrow A\times B\) and \(S\hookrightarrow B\times C\), their composite is the image of \(R\times_BS\to A\times C\). In Set this says that \(a\) is related to \(c\) when some \(b\) connects them. The categorical construction uses no element notation.

Nonexample: topological spaces. The category Top has finite limits and quotient maps, but quotient maps are not generally preserved by pullback. Since its regular epimorphisms fail the stability test, Top is not a regular category.^[2]

Structural Tensions¶

T1 — Quotients versus base change. A quotient may exist but fail after contextual restriction. Diagnostic: pull it back along an arbitrary map and test whether the new arrow remains regular epic.

T2 — Regular versus arbitrary epimorphism. Cancellation alone need not provide a stable quotient presentation. Diagnostic: exhibit the arrow as a coequalizer and compare it with its kernel pair.

T3 — Image existence versus image stability. Every arrow may admit some epi–mono decomposition without those choices commuting with pullback. Diagnostic: require the pulled-back decomposition to be the image factorization, not merely another factorization.

T4 — Regular versus exact. Kernel-pair relations are effective by construction, while arbitrary internal equivalence relations may not be. Diagnostic: test whether each internal equivalence relation is the kernel pair of its quotient before claiming Barr exactness.

T5 — Expressive power versus controlled semantics. Regular logic supports equality, conjunction, and existence but not unrestricted disjunction, negation, or universal quantification. Diagnostic: if a semantic argument uses those operations, identify additional coherent, Heyting, or topos structure.

T6 — Generality versus element intuition. Set-based pictures explain the axioms but can smuggle in elements, choice, or concrete surjectivity. Diagnostic: restate the proof using universal properties, pullbacks, and regular-epi/mono factorization.

Structural–Framed Character¶

Regular Category is structural within a highly technical mathematical frame. Its content is invariant under categorical equivalence and defined entirely by universal properties, factorization classes, and preservation under pullback. It is not evaluative, institution-defined, or dependent on a preferred presentation.

It nevertheless remains domain-specific because its indispensable vocabulary—finite limits, kernel pairs, coequalizers, regular epimorphisms, monos, pullbacks, internal relations—presupposes category theory. Translating “quotient plus inclusion survives restriction” into another field may be insightful, but unless the recipient has literal categorical structure, the transfer is analogy.

Structural Core vs. Domain Accent¶

The structural core is stable two-stage factorization: collapse exactly the distinctions a map identifies, then embed the resulting image, with both stages preserved by contextual restriction. Factorization, Stability, Equivalence Relation, and Composition carry pieces of this structure across domains.

The domain accent makes the core exact: maps are morphisms, contextual restriction is pullback, identified pairs form a kernel pair, collapse is a coequalizer and regular epimorphism, and embedding is a monomorphism. The interaction of these roles—not the bare word “regular”—constitutes the node.

Stripping the categorical accent does not reveal a missing prime. It decomposes into existing general abstractions. Retaining the exact roles yields a recognized category-theoretic class with theorems, examples, counterexamples, functors, internal logic, and completion constructions.

Regular Category strictly specializes Category. It retains objects, morphisms, identities, and associative composition while imposing finite-limit and stable-image exactness. The proposal-only DAG therefore uses one subsumption/specializes/strict edge to live prime:category.

Equivalence Relation is related through kernel pairs and exactness. Factorization is related through the regular-epi/mono normal form, though the live prime's broad wording does not itself supply categorical factorization. Stability appears through pullback invariance. Kernel and Functor are catalog neighbors but do not define the class: a kernel pair is not the same as the zero-based Kernel node, and a regular functor is a structure-preserving map between already-regular categories.

Relationships to Other Abstractions¶

Current abstraction Regular Category Domain-specific

Parents (1) — more general patterns this builds on

Regular Category is a kind of Category Prime

Regular Category strictly specializes Category.

Hierarchy paths (3) — routes to 3 parentless roots

Regular Category → Category → Associativity → Invariance

Show alternative paths (2)

Neighborhood in Abstraction Space¶

Regular Category sits in a sparse region of the domain-specific corpus (72^nd percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Categorical Algebra & Model Systems (8 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-09-08

Not to Be Confused With¶

Category supplies the base objects–arrows–composition structure; Regular Category adds a specific exactness package. Functor maps categories while preserving identity and composition; a regular functor further preserves finite limits and regular epis, but is not the category class. Kernel in the live catalog is the preimage of the identity under an algebraic map; a categorical kernel pair is instead the pullback \(X\times_YX\rightrightarrows X\) and requires no zero object.

Barr-exact category means regular plus effectiveness of every internal equivalence relation. Quillen exact category means an additive category with chosen admissible short exact sequences. Abelian category is additive and exact with kernels and cokernels. Coherent category adds stable finite unions of subobjects. Topos adds still more logical and exponential structure.

Regularity elsewhere is unrelated without explicit categorical context. Castelnuovo–Mumford Regularity is a numerical invariant in algebraic geometry; Regularization is a soft complexity penalty in model fitting; regular languages, regular graphs, and regular spaces use different definitions. Theme and Variation, Classification, Functional Fixedness, Prototype Theory, Axial Coding, Isomorphism, and Universality are embedding neighbors but omit the axiom package.

References¶

[1] Grillet, Pierre Antoine. “Regular Categories.” In Michael Barr, Pierre A. Grillet, and Donovan H. Van Osdol, Exact Categories and Categories of Sheaves, Lecture Notes in Mathematics 236, 121–222. Springer, 1971. Foundational extended treatment of regular categories. registry ↩

[2] Borceux, Francis. “Regular Categories.” In Handbook of Categorical Algebra 2: Categories and Structures, 89–121. Cambridge University Press, 1994. Authoritative account of equivalent definitions, image factorization, exactness properties, examples, and counterexamples. registry ↩a ↩b ↩c ↩d

[3] Butz, Carsten. Regular Categories and Regular Logic. BRICS Lecture Series LS-98-2, Aarhus, 1998. Detailed treatment of regular categories, regular logic, soundness, internal language, and generic models. registry ↩a ↩b

[4] Fong, Brendan, and David I. Spivak. “String Diagrams for Regular Logic.” Electronic Proceedings in Theoretical Computer Science 323 (2020): 196–229. Relates regular categories, relation composition, and the equality–truth–conjunction–existential fragment. registry ↩a ↩b

[5] Barr, Michael. “Exact Categories.” In Exact Categories and Categories of Sheaves, Lecture Notes in Mathematics 236, 1–120. Springer, 1971. Source for the stronger effectiveness condition now called Barr exactness. registry ↩

[6] Carboni, Aurelio, and Robert F. C. Walters. “Cartesian Bicategories I.” Journal of Pure and Applied Algebra 49, nos. 1–2 (1987): 11–32. Develops the categorical calculus of relations arising from regular categories. registry ↩

[7] Carboni, Aurelio, and Enrico M. Vitale. “Regular and Exact Completions.” Journal of Pure and Applied Algebra 125, nos. 1–3 (1998): 79–116. Establishes universal regular and exact completions for categories with weak limits. registry ↩

[8] “Regular category,” Wikipedia, frozen revision 1363815749, 12 July 2026. Discovery provenance only; the axiom equivalences, boundaries, examples, and placement were independently checked against the sources above. registry