Questions: Joint Embedding Property and Universality

Question 1

Two finite graphs G₁ and G₂ belong to a class 𝒦. A student argues: 'JEP requires G₁ and G₂ to share a common induced subgraph before they can be jointly embedded into a third graph.' What is wrong with this claim?

• A. Nothing — JEP and the amalgamation property make exactly the same demand
• B. JEP makes no such shared-substructure requirement — it only asserts there exists some C ∈ 𝒦 into which both G₁ and G₂ embed, regardless of whether they share anything
• C. JEP requires the graphs to be isomorphic, not just to share a subgraph
• D. JEP applies only to linear orders, not to graphs

Answer: (B)

The student is confusing JEP with the amalgamation property (AP). AP does require a shared substructure: given A and B that both extend a common C, amalgamation finds D extending both while respecting the shared C. JEP is strictly simpler — it only asks that any two structures A, B ∈ 𝒦 can be embedded into some common C ∈ 𝒦, with no condition whatsoever on what A and B share. JEP rules out incompatible pairs that cannot coexist in any common structure, but it imposes no requirement about shared substructures.

Question 2

What is the semantic consequence of a class of models satisfying the joint embedding property?

• A. Every model in the class is isomorphic to every other model
• B. The theory of the class is complete — no sentence can be true in one model and false in another
• C. The class has a unique countable model up to isomorphism
• D. Every model in the class is a substructure of a single universal model

Answer: (B)

JEP is equivalent to completeness of the theory (when the class is axiomatizable). If any two models must fit into a common larger model, they cannot disagree on any sentence: if φ were true in one model and false in another, both cannot embed into a common model satisfying some consistent theory. This is why JEP is called a coherence condition — it ensures all structures in the class are compatible at the logical level, pointing toward the same complete theory. Without JEP, two models could be logically incompatible, living in different 'branches' of the theory.

Question 3

A class 𝒦 that satisfies the joint embedding property automatically satisfies the amalgamation property, since JEP is the stronger condition.

• T. True
• F. False

Answer: False

AP is strictly stronger than JEP, not the other way around. Amalgamation requires embedding over a specified shared substructure — given A ← C → B, find D with embeddings from A and B that agree on C. JEP makes no such demand: it only asks for a common host for any two structures, ignoring shared substructures. Every class satisfying AP satisfies JEP (take C to be the empty structure), but the converse fails. A class can have JEP without AP if two structures sharing a common substructure cannot be extended compatibly over that substructure, even though they can always be jointly embedded when you don't need to respect the sharing.

Question 4

The joint embedding property alone (without the amalgamation property) is sufficient to guarantee the existence of a Fraïssé limit for a countable class of finitely generated structures.

• T. True
• F. False

Answer: False

Fraïssé's theorem requires both JEP and AP (plus countability and closure under substructures). JEP ensures the limit is universal — every structure in the class embeds into it, so no structure is 'left out.' But universality alone does not give homogeneity: the ability to extend any partial isomorphism between finite substructures to a full automorphism. Homogeneity requires AP, which ensures that whenever two copies of a structure are embedded into a common host over a shared substructure, they can be merged consistently. JEP and AP play complementary roles: JEP gives 'no orphans,' AP gives 'no conflicts.'

Question 5

Explain why the joint embedding property ensures universality while the amalgamation property ensures homogeneity in the Fraïssé limit.

Model answer: JEP ensures universality because it guarantees every structure in the class can be embedded into the limit. If any two structures can always be jointly embedded into a third, then no structure is incompatible with the growing union used to construct the limit — every structure can be 'absorbed.' AP ensures homogeneity because it guarantees that any partial isomorphism between finite substructures of the limit extends to a full automorphism. Whenever two finite substructures are isomorphic, AP ensures they can be amalgamated over their shared part in a consistent way, and iterating this construction produces a structure where all partial isomorphisms are realized globally.

The division of labor is clean: JEP is about coverage (nothing is left out), AP is about coherence (things fit together consistently over shared parts). A universal but non-homogeneous structure can exist — it contains every finite structure but has no automorphism extending a given partial isomorphism. AP rules this out by ensuring consistent amalgamation over shared substructures, which is exactly what homogeneity requires: any bijection between finite substructures that preserves structure must extend to a global automorphism.