Vaught's Conjecture in Model Theory

Updated 8 July 2026

Vaught’s Conjecture asserts that a complete first-order theory has either countably many or continuum many non-isomorphic countable models, excluding an intermediate ℵ1 case.
The topic leverages Scott analysis and Lω1,ω logic to link model spectra with descriptive set theory, providing structured insights into countable models.
Recent work strengthens the conjecture by showing counterexamples must demonstrate cofinal Scott rank growth and uniform definability across models.

Vaught’s Conjecture is the assertion that if $T$ is a complete first-order theory in a countable language, then the number $I(\aleph_0,T)$ of non-isomorphic countable models of $T$ is either countable or $2^{\aleph_0}$ . Morley proved that for an $L_{\omega_1,\omega}$ -sentence the number of countable models is always either $\le \aleph_0$ , $\aleph_1$ , or $2^{\aleph_0}$ , so the conjecture is precisely the exclusion of the intermediate $\aleph_1$ -case for complete first-order theories; in modern usage it is also studied in its infinitary form for $L_{\omega_1,\omega}$ -sentences (Pillay et al., 9 Aug 2025, Ackerman et al., 2017).

1. Statement, notation, and basic model spectra

For a complete first-order theory $I(\aleph_0,T)$ 0 in a countable language with infinite models, $I(\aleph_0,T)$ 1 denotes the number of countable models of $I(\aleph_0,T)$ 2 up to isomorphism. Since every countable model is isomorphic to one with domain $I(\aleph_0,T)$ 3, there are at most $I(\aleph_0,T)$ 4 such isomorphism types, so $I(\aleph_0,T)$ 5 (Pillay et al., 9 Aug 2025).

The conjecture belongs to a broader classification problem for countable model spectra. Theories with $I(\aleph_0,T)$ 6 are the $I(\aleph_0,T)$ 7-categorical theories; the survey literature lists the theory of an infinite set, theories of infinite-dimensional vector spaces over a finite field, and the theory of dense linear orders as standard examples. Theories with $I(\aleph_0,T)$ 8 are said to have few countable models. Ehrenfeucht theories are those with more than one but only finitely many countable models, and Vaught himself proved that $I(\aleph_0,T)$ 9, so finite spectra begin at $T$ 0 rather than $T$ 1 (Pillay et al., 9 Aug 2025).

The conjecture is usually formulated in first-order logic, but much of the structural work passes through $T$ 2. In that setting one asks whether every sentence has either countably many or continuum many countable models. This infinitary form is stronger than the first-order one and is tightly connected with Scott analysis, descriptive set theory, and computability-theoretic reformulations (Pillay et al., 9 Aug 2025, Gonzalez et al., 2022).

2. Scott analysis and structural reformulations

A central bridge between model counting and infinitary logic is Scott analysis. A countable structure $T$ 3 has a Scott sentence, namely an $T$ 4-sentence characterizing $T$ 5 up to isomorphism among countable models. The Scott rank of $T$ 6 measures the least complexity needed for such a description; one robust formulation used in recent work defines $T$ 7 as the least ordinal $T$ 8 such that $T$ 9 has a $2^{\aleph_0}$ 0 Scott sentence (Harrison-Trainor et al., 6 Feb 2026, Larson, 2014).

The survey literature records two equivalent infinitary reformulations of the conjecture. One is Steel’s strong form, according to which an $2^{\aleph_0}$ 1-sentence should have either countably many countable models or a perfect set of pairwise non-isomorphic countable models. Another is a Scott-rank formulation stating that every $2^{\aleph_0}$ 2-sentence with few countable models has countable Scott rank. These formulations make clear that Vaught’s Conjecture is not only a statement about cardinal arithmetic; it is also a claim about the descriptive complexity of isomorphism classes (Pillay et al., 9 Aug 2025).

A more recent strengthening is the $2^{\aleph_0}$ 3-Vaught’s Conjecture. For an $2^{\aleph_0}$ 4-sentence $2^{\aleph_0}$ 5, the associated Vaught ordinal $2^{\aleph_0}$ 6 is defined as the least ordinal at which one already sees either continuum many $2^{\aleph_0}$ 7-types realized among models of $2^{\aleph_0}$ 8, or a bound on the Scott ranks of all countable models. The $2^{\aleph_0}$ 9-version requires uniform bounds of the form $L_{\omega_1,\omega}$ 0 for $L_{\omega_1,\omega}$ 1-sentences, and linear orders satisfy this stronger property (Gonzalez et al., 2022).

3. Consequences for hypothetical counterexamples

The conjecture remains open in general, but the known structure of any counterexample is highly constrained. Using Scott processes, Larson gave a proof of an unpublished theorem of Harrington showing that a counterexample to Vaught’s Conjecture has models of cofinally many Scott ranks below $L_{\omega_1,\omega}$ 2. More precisely, if $L_{\omega_1,\omega}$ 3 is a counterexample and $L_{\omega_1,\omega}$ 4 is the quantifier depth of $L_{\omega_1,\omega}$ 5, then for every limit ordinal $L_{\omega_1,\omega}$ 6 with $L_{\omega_1,\omega}$ 7, $L_{\omega_1,\omega}$ 8 has a model of Scott rank $L_{\omega_1,\omega}$ 9 (Larson, 2014).

The same Scott-process framework also recovers a theorem of Harnik and Makkai: if a counterexample exists, then there is one whose uncountable models all satisfy the same $\le \aleph_0$ 0-theory, and which has a model of Scott rank $\le \aleph_0$ 1. This shifts the search for counterexamples away from low-rank or sporadic behavior and toward theories with a very rich and highly organized Scott-rank spectrum (Larson, 2014).

Recent work strengthens this picture below the Vaught ordinal. One result shows that any counterexample to Vaught’s Conjecture must have at least two models of every parameterized Scott rank, in contrast with the unparameterized setting, where minimal counterexamples have only one model at many ranks. The same paper also proves that theories with fewer than continuum many models have trivial Scott spectra and classifies low Scott-rank models under countability assumptions on realized $\le \aleph_0$ 2-types (Gonzalez et al., 13 Jun 2026).

For trees, Scott-rank methods also yield consequences independent of a direct proof of Vaught’s Conjecture for the class. It has been shown that if a Borel class of trees is definable by a $\le \aleph_0$ 3 sentence, then there is a model of Scott rank at most $\le \aleph_0$ 4. This gives another proof that trees are not faithfully Borel complete, and it does so without first proving Vaught’s Conjecture for trees (Harrison-Trainor et al., 6 Feb 2026).

4. Verified cases and sharp variants

A large part of the subject consists of positive results for specific classes. The general survey literature lists proofs for colored orders, one unary operation, $\le \aleph_0$ 5-stable theories, stable theories with Skolem functions, superstable theories of finite $\le \aleph_0$ 6-rank, varieties, $\le \aleph_0$ 7-minimal theories, and trees (Pillay et al., 9 Aug 2025).

Several later papers sharpen these results to exact dichotomies.

Class	Conclusion for countable models	Source
Linear orders	$\le \aleph_0$ 8	(Kurilić, 2022)
Almost chainable theories	$\le \aleph_0$ 9	(Kurilić, 2019)
Theories with a definable infinite discrete linear order	$\aleph_1$ 0	(Tanović, 2022)
Unions of products of rooted trees	If $\aleph_1$ 1, then $\aleph_1$ 2 for $\aleph_1$ 3, and otherwise $\aleph_1$ 4	(Kurilić, 10 Jun 2026)
Partial orders from finite products and disjoint unions of linear orders or rooted-tree classes	Sharp VC: $\aleph_1$ 5 for the classes treated	(Kurilić, 2022)
FLD $\aleph_1$ 6-theories of partial orders	VC holds iff the theory is large or its atomic model has a VC-decomposition; VC $\aleph_1$ 7 holds when there is a VC $\aleph_1$ 8-decomposition	(Kurilić, 6 Jan 2026)

The almost chainable case is especially transparent. A complete theory is almost chainable when its models are definable, up to a finite kernel, inside a linear order with finitely many unary predicates. Kurilić proved that an almost chainable theory has either one or continuum many non-isomorphic countable models, confirming Vaught’s Conjecture for that class; in finite relational languages this condition is equivalent to bounded profile in the sense of Fraïssé (Kurilić, 2019).

A different route to the continuum case arises from definable discrete order. If a countable complete first-order theory admits an infinite, parameter-free definable discrete linear order, then it has continuum many countable models. The proof is purely first-order and proceeds by coding arbitrary countable discrete linear orders into the locus of a simple type (Tanović, 2022).

Partial orders assembled from rooted trees exhibit a more varied but still controlled spectrum. For $\aleph_1$ 9, where the $2^{\aleph_0}$ 0 are rooted trees and $2^{\aleph_0}$ 1, one has $2^{\aleph_0}$ 2 when $2^{\aleph_0}$ 3, and otherwise $2^{\aleph_0}$ 4. Thus Vaught’s Conjecture holds for this whole closure class, even though the sharp trichotomy need not (Kurilić, 10 Jun 2026).

5. Strengthenings, analogues, and neighboring problems

The sharp form of the conjecture, often written VC $2^{\aleph_0}$ 5, replaces the countable-or-continuum dichotomy by the stronger alternative $2^{\aleph_0}$ 6. Rubin’s theorem for linear orders is the prototype, and later work extends VC $2^{\aleph_0}$ 7 to broad classes of partial orders obtained from linear orders, Boolean algebras, and certain trees by finite products, disjoint unions, and finite lexicographic decompositions (Kurilić, 2022, Kurilić, 6 Jan 2026).

The $2^{\aleph_0}$ 8-Vaught’s Conjecture is another strengthening, formulated via Vaught ordinals and Scott ranks. It predicts that for $2^{\aleph_0}$ 9-sentences one should already see the dichotomy by stage $\aleph_1$ 0, and all infinitary sentences whose models are linear orders satisfy this stronger principle (Gonzalez et al., 2022).

There are also measure-theoretic analogues. For a properly ergodic $\aleph_1$ 1-invariant probability measure $\aleph_1$ 2 on the space of countable structures and any countable fragment $\aleph_1$ 3, the almost-sure $\aleph_1$ 4-theory $\aleph_1$ 5 has continuum many countable models, while the full almost-sure $\aleph_1$ 6-theory $\aleph_1$ 7 has no models at all. This is explicitly described as an analogue of Vaught’s Conjecture in the ergodic setting (Ackerman et al., 2017).

Another related line replaces isomorphism by weaker equivalence relations. If two countable models are called distinguishable when some formula has different numbers of realizations in them, then the resulting equivalence relation is Borel on the standard Polish space of countable structures. Consequently, if a theory has an uncountable set of pairwise distinguishable countable models, then it has continuum many such models. As a concrete corollary, Vaught’s Conjecture holds for the language with only one unary relation symbol (Assem, 2013). Algebraic-logic methods extend this kind of analysis to the case without equality and to certain infinitary settings via cylindric and quasi-polyadic representation theory (Assem et al., 2013).

The conjecture also has consequences outside spectrum problems. A sufficient condition for an $\aleph_1$ 8-theory to have an independent axiomatization is formulated in terms of the size of the part of the theory consisting of sentences with exactly $\aleph_1$ 9 countable models, and an immediate corollary is that, assuming Vaught’s Conjecture, every $L_{\omega_1,\omega}$ 0-theory in a countable language has an independent axiomatization (Hjorth et al., 2010).

6. Open directions and current status

Despite the breadth of positive results, the general conjecture remains open. Survey work highlights several unresolved classes, including arbitrary superstable theories, stable and superstable theories of modules, and broad regions of the weakly $L_{\omega_1,\omega}$ 1-minimal and weakly quasi- $L_{\omega_1,\omega}$ 2-minimal landscape (Pillay et al., 9 Aug 2025).

The finite-model-spectrum side also remains structurally mysterious. Ehrenfeucht theories supply examples with exactly $L_{\omega_1,\omega}$ 3 countable models, but the conjectural internal structure of such theories is unsettled. One open conjecture recorded in the survey is that every Ehrenfeucht theory has the strict order property; another is Martin’s conjecture for Ehrenfeucht theories, which predicts very low-complexity invariants for countable models with few-model spectra (Pillay et al., 9 Aug 2025).

Recent work suggests that progress may come from finer Scott-analysis invariants rather than from raw model counting alone. The classification of Scott spectra below the Vaught ordinal, the $L_{\omega_1,\omega}$ 4-Vaught’s Conjecture for linear orders, and the bounded Scott spectral gaps for trees all indicate that the obstruction to a counterexample is not merely combinatorial cardinality, but also the inability to sustain the required pattern of infinitary definability and rank growth across models (Gonzalez et al., 13 Jun 2026, Gonzalez et al., 2022, Harrison-Trainor et al., 6 Feb 2026).

Vaught’s Conjecture therefore occupies a singular position in model theory. It is simultaneously a problem about countable model spectra, infinitary definability, Scott rank, Borel structure, and the internal geometry of classification-theoretic tame classes. The known results show that many natural theories fall into sharply controlled trichotomies, but the full conjecture still marks the boundary between current structure theory and a general understanding of countable models.