Age (Model Theory)¶

The isomorphism-closed class of finitely generated structures that embed into a fixed model, recording exactly its finite or finitely generated local patterns.

Version: v1 · 2026-08-30 · History
Domain-specific #: 1250
Origin domain: mathematics
Aliases: Age of a structure

Core Idea¶

In model theory, the age of a structure \(M\), written \(\operatorname{Age}(M)\), is the class of isomorphism types of finitely generated structures that embed into \(M\). Equivalently, it is the isomorphism closure of the finitely generated substructures of \(M\). The language/signature is held fixed, and embeddings preserve the language's relations, functions, and constants. In a purely relational language with no function symbols, a substructure generated by finitely many elements is finite; with functions, finite generation need not imply a finite underlying set.^[1]^[2]

The age records local structural possibilities while discarding their locations, multiplicities, and global arrangement inside \(M\). It answers “which finitely generated patterns occur?” rather than “how many occur?”, “where do they occur?”, or “which first-order sentences are true?” Any age is nonempty (for a nonempty convention as applicable), closed under isomorphism and finitely generated substructures—the hereditary property (HP)—and has the joint embedding property (JEP), because two finitely generated substructures of \(M\) lie in a substructure generated by their union.^[1]^[3]

The recognition invariant is fixed language + finite generation + embedding into one ambient structure + identification up to isomorphism. This role system distinguishes a model-theoretic age from demographic age, historical period, a theory's finite models, or an arbitrary class of structures.

Structural Signature¶

Recognition roles:

A fixed language \(L\) — relation, function, and constant symbols whose interpretations define the relevant structure type.
An ambient \(L\)-structure \(M\) — the model whose local configurations are profiled.
Finitely generated \(L\)-structures \(A\) — candidate local pieces generated by finite tuples, not automatically finite in functional languages.
Embeddings \(A\hookrightarrow M\) — structure-preserving injective maps, not merely homomorphisms or arbitrary set injections.
Isomorphism closure — renamed copies count as the same structural possibility; representatives do not matter.
Hereditary closure — finitely generated substructures of members remain in the age.
Joint embeddability — any two age members embed into a third age member generated within the common ambient model.
Local-profile boundary — multiplicity, placement, global cardinality, and elementary theory are deliberately not encoded.

Recognition test: given \(A\), can one exhibit an \(L\)-embedding into \(M\), and is membership unchanged when \(A\) is replaced by an isomorphic copy? If only \(A\models\operatorname{Th}(M)\) is known, or only a homomorphism exists, membership has not been established.^[2]

What It Is Not¶

An age is not the chronological age of a model, nor a sequence indexed by time. It is also not a model's elementary theory. Structures may share an age while disagreeing on global first-order properties; age membership sees embeddable finitely generated patterns rather than truth of arbitrary sentences.

It is not the class of all finite models of \(\operatorname{Th}(M)\). A finite structure can embed into \(M\) without satisfying the same complete theory, and a finite model of a theory need not appear as a substructure of a selected model. It is not an elementary class because closure under ultraproducts or elementary equivalence is not its defining relation.

It is not a Fraïssé class automatically. Every age has HP and JEP, but the amalgamation property (AP) is additional. Under countability hypotheses, a class with HP, JEP, AP, isomorphism closure, and countably many isomorphism types is a Fraïssé class and determines a countable ultrahomogeneous limit. AP need not hold for an arbitrary age.^[1]^[3]

Finally, “finite” and “finitely generated” are not universal synonyms. In groups, rings, or other functional languages, a finitely generated substructure may be infinite. Any statement that silently replaces one by the other must restrict to a suitable relational or locally finite setting.

Scope of Application¶

Ages organize Fraïssé theory, homogeneous structures, permutation-group/model-theory connections, structural Ramsey theory, and classification by finite configurations.^[4] They are used for graphs, orders, metric or enriched structures under adapted frameworks, groups and algebras, and relational expansions. The exact class changes with the language: adding a distinguished relation or constant can alter embeddings and therefore the age even when the underlying set is unchanged.

In classical relational examples, ages convert infinite structures into manageable catalogs of finite objects. The countable random graph has the class of all finite simple graphs as its age. The ordered rationals \((\mathbb Q,<)\) have all finite linear orders as their age. These classes possess HP, JEP, and AP and yield the familiar ultrahomogeneous limits.^[1]

Use outside first-order model-theoretic structure must be qualified. Category-theoretic subobject profiles or database schemas can resemble ages, but without fixed-language embeddings and finite generation they are analogies, not literal instances.

Clarity¶

The abstraction separates local possibility from global assembly. If a finite graph embeds into the random graph, the age certifies that the pattern occurs; it does not tell how often or in which larger neighborhoods. If two structures have different ages, they necessarily differ in some finitely generated embeddable configuration. If they have the same age, however, they can still differ globally.

It also clarifies the roles of HP, JEP, and AP. HP and JEP characterize the basic shape any age must have under standard countability conditions for realization as an age. AP is the extension-consistency condition that supports a homogeneous Fraïssé limit. Treating all three as automatic obscures why some local-profile classes lack canonical homogeneous assembly.^[3]^[1]

Manages Complexity¶

An infinite structure may contain unboundedly many elements and relations, while each age member is controlled by finite generators. Passing to isomorphism types removes irrelevant naming. This compression lets proofs work with finite diagrams, embeddings, amalgams, and extension properties rather than the entire ambient model at once.

The discarded information is substantial. Age does not record how copies intersect, their multiplicities, definability, measures, or the full automorphism group. Additional expansions, Ramsey properties, forbidden configurations, or extension axioms may recover useful organization. The age is powerful precisely as a selected invariant, not as a complete code for every structure.

Abstract Reasoning¶

From \(A\in\operatorname{Age}(M)\) and a finitely generated substructure \(B\le A\), composition of embeddings gives \(B\in\operatorname{Age}(M)\); this is HP. For \(A,B\in\operatorname{Age}(M)\), choose embedded copies inside \(M\), take the substructure generated by their finite generating tuples, and obtain \(C\in\operatorname{Age}(M)\) into which both embed; this is JEP.

Conversely, in a countable language, a nonempty isomorphism-closed class of finitely generated structures with countably many isomorphism types, HP, and JEP can be realized as the age of a countable structure under the usual formulation. Adding AP supports construction of a unique countable ultrahomogeneous Fraïssé limit with that age.^[3]^[1] The uniqueness statement is restricted to the ultrahomogeneous limit: an arbitrary age need not identify an arbitrary ambient structure up to isomorphism.

Knowledge Transfer¶

Literal transfer occurs from graphs to orders, relational systems, algebraic structures, and suitable enriched settings because the same roles—language, finite generation, embedding, isomorphism class—remain intact. Proof strategies based on HP/JEP/AP also transfer, although the meaning of generated substructure and amalgam changes with the language.

The broader transferable parent is Representation: an age represents selected local structure of \(M\) by a class of smaller structures with an explicit faithfulness boundary. Classification and Finite-to-Infinite Bridge are related primes. A generic parts list is not an age; the embedding and isomorphism conventions must travel.

Examples¶

The ordered rationals. Every finite substructure of \((\mathbb Q,<)\) is a finite linear order. Conversely, every finite linear order embeds into \(\mathbb Q\) by assigning its ordered elements to increasing rationals. Hence \(\operatorname{Age}(\mathbb Q,<)\) is exactly the isomorphism class of all finite linear orders. The language is \(\{<\}\), generation is finite, embeddings preserve order, and labels are quotiented by isomorphism.

The countable random graph. Its extension property ensures every finite simple graph embeds, so its age is all finite simple graphs. This does not mean every countable graph is isomorphic to the random graph; age records finite occurrence, while ultrahomogeneity and the extension property determine the Fraïssé limit.^[1]

Functional-language warning. Let \((\mathbb Z,+,-,0)\) be a group-language structure, with unary inverse shown explicitly. The substructure generated by \(1\) is all of \(\mathbb Z\), infinite despite one generator. Its isomorphism type belongs to the age. This example falsifies the unqualified claim that ages contain only finite structures.

Same-age boundary. Different infinite linear orders can have all finite linear orders as their age. Their global order types differ even though every finite pattern agrees. Age is therefore a local profile, not complete global identity.

Structural Tensions¶

Compression versus completeness. Isomorphism types of finitely generated pieces make infinite structure tractable but erase multiplicity and placement. Diagnostic: does the question concern which patterns occur, or how they are globally arranged?

Finite generation versus finiteness. Relational intuition simplifies examples but misstates functional structures. Diagnostic: inspect the language's closure operations before replacing “finitely generated” with “finite.”

Autonomy versus reduction. Representation and Classification describe the broad act, yet neither entails embeddings of finitely generated \(L\)-structures or HP/JEP consequences. Diagnostic: if fixed-language embeddings and isomorphism closure disappear, the model-theoretic age residual disappears too; current generic nodes do not close it.

Structural–Framed Character¶

Age is almost entirely structural. Membership is mathematical once language, ambient model, substructure, and embedding conventions are fixed. It carries no evaluative weight and does not depend on institutional authority.

Its frame dependence is technical: changing the language changes generated substructures and embeddings. A reduct and an expansion of the same underlying object can have different ages. That domain-semantic dependence is enough to keep the node within model theory rather than promoting “age” as a universal prime.

Structural Core vs. Domain Accent¶

The portable core is representation of a large object by the isomorphism types of its finitely generated local pieces. Similar local-to-global profiles appear throughout mathematics and computation.

The indispensable accent is first-order model-theoretic structure: fixed language, generated substructure, embedding, HP/JEP, and Fraïssé-theoretic extension. Removing those commitments leaves a loose finite-profile pattern, not \(\operatorname{Age}(M)\). The abstraction is autonomous but domain-specific.

Age directly instantiates Representation because it maps an ambient structure to a selected, invariant local profile and explicitly states what is preserved and omitted. It relates to Classification through isomorphism types and to Finite-to-Infinite Bridge through Fraïssé construction. Representation is the proposed direct parent; the other two describe operations and consequences rather than the complete identity.

Universal Property is a nearby mathematical abstraction but not a genus: an age is defined by embeddability, not by initial/terminal mapping uniqueness.

Relationships to Other Abstractions¶

Current abstraction Age (Model Theory) Domain-specific

Parents (1) — more general patterns this builds on

Age (Model Theory) presupposes Representation Prime

Age directly instantiates Representation because it maps an ambient structure to a selected, invariant local profile and explicitly states what is preserved and omitted.

Hierarchy path (1) — routes to 1 parentless root

Age (Model Theory) → Representation → Abstraction

Neighborhood in Abstraction Space¶

Age (Model Theory) sits in a sparse region of the domain-specific corpus (70^th 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¶

Elementary theory \(\operatorname{Th}(M)\): class of sentences true in \(M\), not embeddable local structures.
Finite models of a theory: satisfaction-based class, not necessarily substructures of a chosen model.
Fraïssé class: an age-like class satisfying AP and countability conditions; narrower than age generally.
Fraïssé limit: the countable ultrahomogeneous structure constructed from a Fraïssé class, not the class itself.
Profile/enumeration function: counts substructures of given sizes; age records which isomorphism types occur without multiplicity.
Hereditary class: HP alone does not guarantee JEP or specify one ambient realization.
Chronological age: a linguistic homonym with no model-theoretic relation.

References¶

[1] Wilfrid Hodges, A Shorter Model Theory, Cambridge University Press, 1997, chapter on Fraïssé construction, ISBN 978-0-521-58713-6. Publisher-indexed book record. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g

[2] Mathlib contributors, “The Age of a Structure and Fraïssé Classes,” formalized documentation for Mathlib.ModelTheory.Fraisse. Documentation. registry ↩a ↩b

[3] Roland Fraïssé, “Sur l'extension aux relations de quelques propriétés des ordres,” Annales scientifiques de l'École Normale Supérieure 71(4), 363–388 (1954). doi:10.24033/asens.1027. registry ↩a ↩b ↩c ↩d

[4] Alexander S. Kechris, Vladimir G. Pestov, and Stevo Todorčević, “Fraïssé Limits, Ramsey Theory, and Topological Dynamics of Automorphism Groups,” GAFA 15, 106–189 (2005). doi:10.1007/s00039-005-0503-1. registry ↩