42.02.06 · mathematical-logic / model-theory

Categoricity: Ryll-Nardzewski, Morley, and Baldwin-Lachlan

shipped3 tiersLean: none

Anchor (Master): Marker 2002 *Model Theory: An Introduction* (Springer GTM 217) Ch. 4-6 (the Ryll-Nardzewski theorem and oligomorphic automorphism groups; strongly minimal sets, the exchange property, acl-pregeometry and dimension; ω-stable and totally transcendental theories, Morley rank and Morley degree; prime and saturated models over sets; the absence of Vaughtian pairs and the two-cardinal theorems of Vaught and Morley; the full proof of Morley's categoricity theorem via the equivalence of ω-stability with κ-categoricity for some uncountable κ, and the upward/downward transfer; the Baldwin-Lachlan theorem through strongly minimal sets and dimension, refining Vaught's spectrum); Shelah 1990 *Classification Theory and the Number of Non-Isomorphic Models* 2e (North-Holland Studies in Logic 92), the main gap; Baldwin and Lachlan 1971 *On strongly minimal sets* (J. Symbolic Logic 36); Pillay 1996 *Geometric Stability Theory* (Oxford); Tent and Ziegler 2012 *A Course in Model Theory* Ch. 3-6

Intuition Beginner

Here is a strange question you can ask about a rulebook. Fix a size — say, count the points and demand there are exactly so many. Among all the worlds that obey the rulebook and have that size, is there really only one shape, or are there many different shapes hiding behind the same count? When the answer is "only one shape," we say the rulebook is categorical at that size: the size alone pins the world down.

Most rulebooks are not like this. The whole numbers with a "next" arrow can be arranged in many different countable shapes — one long line, or a line plus extra loose chains floating beside it — all the same size, all obeying the rules, none the same shape. The size tells you almost nothing.

But the dense rational line is different. Any two dense orders with no first or last point, both countable, are secretly the same order relabelled. Count the points and the shape is forced. The rulebook for dense orders is categorical at the countable size.

The real surprise is what happens past the countable. For the algebraically closed numbers, once the world is uncountable, its size fixes everything — because such a world is built from a stack of independent free ingredients, and counting the world counts the ingredients. Morley's great theorem says that for any rulebook, this good behaviour above the countable is all-or-nothing: pin the world down at one uncountable size and you have pinned it down at every uncountable size. This unit is about why.

Visual Beginner

The picture contrasts two rulebooks at a fixed size. One splinters into many shapes; the other is forced into a single shape. The forcing comes from a dimension that counts independent ingredients, so equal size makes equal shape.

   A RULEBOOK THAT SPLINTERS (not categorical at this size)
   "next-arrow on the whole numbers", same count, different shapes:

      shape 1:   . -> . -> . -> . -> ...          (one line)
      shape 2:   . -> . -> ...   +   o -> o -> ... (a line plus a loose chain)
      same size, different shape  ->  size does NOT pin it down

   A RULEBOOK THAT IS FORCED (categorical at this size)
   "dense order, no ends", any two countable ones match up:

      a < b < c   densely filled, no first, no last
      relabel  ->  every countable such order is the SAME shape
      size DOES pin it down

   WHY SIZE CAN PIN IT DOWN:  a DIMENSION of free ingredients

      world = [ free ingredient #1 ][ #2 ][ #3 ] ... (independent)
      count the world  =  count the ingredients  =  the dimension
      two worlds, same size  ->  same dimension  ->  same shape

rulebook	one size, how many shapes?	size pins it down?
next-arrow on whole numbers	many	no
dense order, no ends (countable)	one	yes
algebraically closed numbers (uncountable)	one	yes
whole numbers with plus and times	many	no

Read top to bottom: a dimension counting independent ingredients is what lets size alone fix the shape.

Worked example Beginner

We test two rulebooks at the countable size by hand and see one pin its world down while the other splinters. No symbols beyond comparisons and counting.

Step 1. Take the dense rational order: points with an order, no first point, no last point, and a point between any two. Ask whether two such countable worlds must be the same shape. Line them up one pair at a time: match a first point on the left to a first point on the right, then for each new point on either side find a matching point in the same relative position. Density always supplies room, and no-ends means there is never a wall. The matching never gets stuck, so the two worlds are the same shape.

Step 2. Conclude: at the countable size, the dense-order rulebook has exactly one shape. Size pins it down.

Step 3. Now take the "next-arrow" rulebook: every point has exactly one next point and one previous point, going on forever. One countable world is a single line stretching both ways. Another countable world is that line plus a second, separate two-way line floating beside it. Both obey the rules. Both are countable.

Step 4. Are they the same shape? No. In the single line you can walk from any point to any other by stepping along arrows. In the two-line world you can never step from the first line to the second. A matching would have to send a reachable point to a reachable point, and the second world has points reachable from nowhere on the first line. The match is impossible.

Step 5. Conclude: at the countable size, the next-arrow rulebook has at least two shapes. Size does not pin it down.

What this tells us: categoricity is not automatic. Some rulebooks force a single shape per size; others allow many. The dense order forces; the next-arrow splinters. The whole project of this unit is to find the exact dividing line, and to discover that above the countable the dividing line is governed by a hidden count of independent ingredients.

Check your understanding Beginner

Formal definition Intermediate+

Fix a countable first-order language $L$ , a complete $L$ -theory $T$ with infinite models, the satisfaction and elementary-embedding hierarchy of 42.02.01 pending, the compactness/diagram machinery of 42.02.02 pending, the type spaces $S_{n} (A)$ of 42.02.03 pending, and the quantifier-elimination and strong-minimality apparatus of 42.02.05 pending. For an infinite cardinal $κ$ , $T$ is $κ$ -categorical when any two models of $T$ of cardinality $κ$ are isomorphic ^{[Marker Ch. 4]}.

$ℵ_{0}$ -categoricity. The Ryll-Nardzewski theorem gives, for countable complete $T$ with infinite models, the equivalence of: $T$ is $ℵ_{0}$ -categorical; $S_{n} (T)$ is finite for every $n$ ; every type in every $S_{n} (T)$ is isolated; for each $n$ there are finitely many $L$ -formulas in $n$ free variables up to $T$ -equivalence; and (the Engeler-Svenonius form) the automorphism group $Aut (M)$ of the countable model is oligomorphic — it has finitely many orbits on $M^{n}$ for every $n$ ^{[Marker Ch. 4]}. This restates the criterion of 42.02.03 pending. Standard $ℵ_{0}$ -categorical theories: $DLO$ (an $n$ -type is fixed by the order pattern of the variables), the random graph (the Fraïssé limit of finite graphs, where the extension axioms make every $n$ -type isolated by a quantifier-free formula), the pure infinite set $(Z; =)$ (an $n$ -type is fixed by the equality pattern), and infinite-dimensional vector spaces over a fixed finite field.

Strong minimality and dimension. A definable set $D$ in $M ⊨ T$ is strongly minimal when it is infinite and, in every elementary extension, every definable subset of $D$ (with parameters) is finite or cofinite (42.02.05 pending). On a strongly minimal $D$ the algebraic closure operator $acl$ — where $a \in acl (B)$ means $a$ lies in a finite $B$ -definable set — satisfies the exchange property: if $a \in acl (B \cup {c}) ∖ acl (B)$ then $c \in acl (B \cup {a})$ . Exchange makes $(D, acl)$ a pregeometry (matroid), so $D$ has a well-defined dimension $dim (D)$ , the cardinality of any maximal $acl$ -independent subset; two strongly minimal sets over the same parameters are isomorphic over them iff they have equal dimension.

$ω$ -stability and Morley rank. $T$ is $ω$ -stable (equivalently totally transcendental for countable $T$ ) when $∣ S_{n} (A) ∣ \leq ℵ_{0}$ for every $n$ and every countable $A \subseteq M ⊨ T$ . Morley rank $RM (φ)$ is the ordinal-valued Cantor-Bendixson rank of the definable set $φ (M)$ in the type space, defined by $RM (φ) \geq α + 1$ iff $φ (M)$ contains infinitely many disjoint definable subsets each of rank $\geq α$ ; Morley degree is the finite number of rank-maximal disjoint pieces. $T$ is $ω$ -stable iff every formula has ordinal Morley rank, and the strongly minimal sets are exactly those of rank $1$ and degree $1$ ^{[Morley 1965]}.

Key theorem with proof Intermediate+

The signature result is Morley's categoricity theorem, the resolution of Łoś's conjecture. It says cardinality alone, above the countable, has a single all-or-nothing verdict on whether it pins a model down.

Theorem (Morley). Let $T$ be a countable complete theory with infinite models. If $T$ is $κ$ -categorical for some uncountable cardinal $κ$ , then $T$ is $λ$ -categorical for every uncountable cardinal $λ$ ^{[Morley 1965]}.

Proof. The argument runs through $ω$ -stability and saturation. First, uncountable categoricity forces $ω$ -stability. If $T$ were not $ω$ -stable, some countable $A_{0} \subseteq M_{0} ⊨ T$ would carry $∣ S_{n} (A_{0}) ∣ = 2^{ℵ_{0}}$ types for some $n$ ; a counting/coding argument then builds two models of any uncountable cardinal $λ \leq 2^{ℵ_{0}}$ that realize different numbers of types over countable subsets — one saturated, realizing $λ$ types, and one realizing only $ℵ_{1}$ — and these are non-isomorphic. Choosing $λ = κ$ contradicts $κ$ -categoricity. So $T$ is $ω$ -stable.

Second, an $ω$ -stable theory has a saturated model in every uncountable cardinal $λ$ , and a prime model over every set (42.02.04 pending); the saturated model of size $λ$ is unique. The crux is that for an $ω$ -stable $T$ , $κ$ -categoricity is equivalent to every model of size $κ$ being saturated. The forward direction: if all $κ$ -models are isomorphic, the saturated one (which exists) is the only shape, so every $κ$ -model is saturated. The transfer of this property between uncountable cardinals is controlled by Vaughtian pairs. A Vaughtian pair is a proper elementary extension $M ≺ N$ , $M \neq = N$ , together with a formula $φ$ with $φ (M) = φ (N)$ infinite — a definable set that fails to grow. If $T$ has a Vaughtian pair, the two-cardinal theorem of Vaught and Morley produces, in some uncountable cardinal, a non-saturated model alongside the saturated one, breaking categoricity. Conversely the absence of all Vaughtian pairs lets the saturation of $κ$ -models transfer up and down: every uncountable model is then saturated, hence determined up to isomorphism by its cardinality. Thus $ω$ -stability plus no-Vaughtian-pairs is a cardinal-independent property, and it holds at some uncountable $κ$ iff it holds at all uncountable $λ$ , giving categoricity at every uncountable $λ$ . $□$

Corollary (the categoricity spectrum). A countable complete theory is categorical in no uncountable cardinal, in all of them, or — separately — in $ℵ_{0}$ ; the four spectrum classes are exactly ${$ totally categorical (all infinite $κ$ ), uncountably categorical only, $ℵ_{0}$ -categorical only, neither $}$ , with $ACF_{p}$ and infinite vector spaces realizing the first two cases.

Bridge. Morley's theorem is the foundational reason cardinality is a meaningful classifying invariant above the countable: the buried mechanism is that an uncountably categorical theory is $ω$ -stable, and $ω$ -stability supplies a Morley rank whose rank- $1$ strongly minimal sets carry a dimension, so two models of the same uncountable size carry the same dimension and this is exactly why they are isomorphic. It builds toward the Baldwin-Lachlan analysis below, where the single strongly minimal set's dimension is shown to determine the whole model, and it appears again in the stability theory of 42.02.07 pending, where Morley rank generalises to forking and the categoricity case becomes the simplest, one-dimensional node of Shelah's classification. The argument generalises the Ryll-Nardzewski finiteness-of-types criterion of 42.02.03 pending from $ℵ_{0}$ to the uncountable, replacing "finitely many types" with "ordinal-ranked types," and putting these together, the saturated-model existence of 42.02.04 pending is the bridge that turns rank control into the uniqueness of the $κ$ -model. The central insight is that uncountable categoricity, total transcendentality, and the absence of Vaughtian pairs are three faces of one cardinal-independent fact.

Exercises Intermediate+

Exercise 3 (medium, symbolic).

Show that $ACF_{p}$ is uncountably categorical, identifying the strongly minimal set, the closure operator, and the dimension.

Hint

By the QE of 42.02.05 pending, the home sort is strongly minimal. What is $acl$ on a field, and what does its dimension count?

Answer

By quantifier elimination (42.02.05 pending) every one-variable definable subset of a model of $ACF_{p}$ is finite or cofinite, so the home sort (the formula $x = x$ ) is strongly minimal. The model-theoretic algebraic closure $acl (B)$ coincides with the field-theoretic algebraic closure of the subfield generated by $B$ , which satisfies the Steinitz exchange property; thus $(\cdot, acl)$ is a pregeometry and the dimension of a model is the cardinality of a transcendence basis over the prime field — its transcendence degree. Two models of the same uncountable cardinality $κ$ have transcendence degree $κ$ , hence equal dimension, hence are isomorphic over the prime field. So $ACF_{p}$ is $κ$ -categorical for every uncountable $κ$ ; by Morley's theorem (or directly) it is uncountably categorical. Rubric: full credit for strong minimality from QE, $acl =$ field-algebraic-closure with exchange, and dimension $=$ transcendence degree controlling isomorphism.

Exercise 4 (medium, symbolic).

Prove the exchange property for the algebraic closure operator on a strongly minimal set: if $a \in acl (B \cup {c}) ∖ acl (B)$ then $c \in acl (B \cup {a})$ .

Hint

$a \in acl (B \cup {c})$ is witnessed by a formula $φ (x, c)$ with finitely many solutions. Consider, over $B$ , the set of $y$ for which $φ (x, y)$ has the right finite number of solutions, and use that definable subsets of the strongly minimal set are finite or cofinite.

Answer

Let $φ (x, c)$ over $B \cup {c}$ have exactly $k$ solutions, including $a$ , with $k$ minimal among formulas in $tp (a / B c)$ . Suppose for contradiction $c \in / acl (B \cup {a})$ , so the set $ψ (M) = {y : φ (a, y) \land ∣ φ (x, y) ∣ = k}$ , definable over $B \cup {a}$ and containing $c$ , is infinite (else $c$ is algebraic over $B a$ ); by strong minimality it is cofinite. For all but finitely many $y$ , then, $φ (a, y)$ holds and $φ (x, y)$ has exactly $k$ solutions. Now vary: the set of $x$ such that $φ (x, y)$ holds for cofinitely many $y$ is definable over $B$ ; it contains $a$ and, being a definable subset of the strongly minimal set, is finite or cofinite. If finite, $a \in acl (B)$ , contrary to hypothesis; so it is cofinite, hence the cofinitely many such $x$ each lie in $k$ -element solution sets shared across cofinitely many $y$ , forcing the same $k$ elements to recur for distinct $y$ — impossible once more than $k$ values $y$ are used, since the solution sets are pairwise distinct in a strongly minimal set. The contradiction gives $c \in acl (B \cup {a})$ . Rubric: full credit for the minimal-solution-count witness, the strong-minimality finite/cofinite dichotomy applied to both the $y$ -set and the $x$ -set, and the counting contradiction.

Exercise 5 (medium, symbolic).

Explain why $(R; +, \cdot, <)$ and $(N; +, \cdot)$ are not $ω$ -stable, hence not uncountably categorical.

Hint

$ω$ -stability bounds the number of types over a countable set by $ℵ_{0}$ . A definable linear order, or the coding of all subsets of a definable set, produces $2^{ℵ_{0}}$ types.

Answer

In $(R; +, \cdot, <)$ the order is definable, and over a countable dense set $A$ the cuts of $A$ give $2^{ℵ_{0}}$ distinct $1$ -types (each real determines a distinct cut), so $∣ S_{1} (A) ∣ = 2^{ℵ_{0}} > ℵ_{0}$ and the theory is not $ω$ -stable — the order property obstructs stability outright. In $(N; +, \cdot)$ first-order arithmetic codes every arithmetical set, so over a countable parameter set one realizes $2^{ℵ_{0}}$ types (formulas pin down membership in arbitrary coded sets), again violating the $ℵ_{0}$ bound. By Morley's theorem uncountable categoricity requires $ω$ -stability, so neither theory is uncountably categorical. Rubric: full credit for the cut-counting (order property) in $R$ , the arithmetical coding in $N$ , and the contrapositive via Morley.

Exercise 7 (hard, symbolic).

Prove that if a strongly minimal set $D$ has the property that $acl$ over the empty set is infinite-dimensional-free (no nontrivial algebraic dependencies beyond $acl (\emptyset)$ ), then the models of the theory built as the prime model over a $D$ -basis are classified by $dim (D) \in {0, 1, 2, \dots, ω, \dots}$ , and deduce the Baldwin-Lachlan dichotomy for the countable models.

Hint

A model of an uncountably categorical theory is prime and minimal over a strongly minimal set together with a basis. Isomorphism of such models reduces to equal dimension; count the possible countable dimensions.

Answer

By the Baldwin-Lachlan structure theorem an uncountably categorical $T$ is $ω$ -stable, has (after naming finitely many parameters) a strongly minimal formula $D$ , and every model $M$ is prime and minimal over $D (M)$ together with an $acl$ -basis of $D (M)$ ; the isomorphism type of $M$ is then determined by $dim D (M)$ , the basis cardinality. For uncountable $M$ of size $κ$ one has $dim D (M) = κ$ (the model has the cardinality of $acl$ of a $κ$ -basis), so all $κ$ -models share dimension $κ$ and are isomorphic — Morley's theorem. For countable $M$ , the dimension is a countable cardinal, i.e. a value in ${0, 1, 2, \dots, ω}$ . If every countable dimension yields the same model (which happens exactly when $T$ is also $ℵ_{0}$ -categorical, the dimension being forced), $I (T, ℵ_{0}) = 1$ . Otherwise the dimensions $0, 1, 2, \dots, ω$ give pairwise non-isomorphic countable models — distinct dimension forbids isomorphism since the dimension is an isomorphism invariant of the strongly minimal set — and this is an $ω$ -indexed ladder, so $I (T, ℵ_{0}) = ℵ_{0}$ . No intermediate value occurs because the dimension is the only invariant and it ranges over an initial segment of ${0, 1, \dots, ω}$ that is either a single point or all of ${0, 1, \dots, ω}$ . Hence $I (T, ℵ_{0}) \in {1, ℵ_{0}}$ . Rubric: full credit for the prime-over-basis structure, dimension as the complete invariant, the uncountable case giving Morley, and the countable dimension ladder giving the ${1, ℵ_{0}}$ dichotomy.

Exercise 8 (hard, symbolic).

Prove that uncountable categoricity implies $ω$ -stability: if a countable complete $T$ is not $ω$ -stable, it has two non-isomorphic models of size $ℵ_{1}$ (hence is not $ℵ_{1}$ -categorical).

Hint

Non- $ω$ -stability gives a countable set with uncountably many types. Build a saturated model of size $ℵ_{1}$ realizing $ℵ_{1}$ types over a countable set, and a model omitting all but $ℵ_{0}$ of them, by an elementary chain controlling realization.

Answer

Suppose $T$ is not $ω$ -stable: some countable $A_{0} \subseteq M_{0} ⊨ T$ has $∣ S_{n} (A_{0}) ∣ > ℵ_{0}$ , so $∣ S_{n} (A_{0}) ∣ = 2^{ℵ_{0}} \geq ℵ_{1}$ . Build a saturated model $N_{sat}$ of size $ℵ_{1}$ by an elementary chain of length $ω_{1}$ realizing, at each stage, all types over the countably many parameters accumulated so far; $N_{sat}$ then realizes $ℵ_{1}$ distinct types over (an isomorphic copy of) $A_{0}$ . Build a second model $N_{small}$ of size $ℵ_{1}$ by a Skolem-hull/omitting construction (downward Löwenheim-Skolem over a carefully chosen subset, 42.02.01 pending) realizing only $ℵ_{0}$ of the types over $A_{0}$ while still having cardinality $ℵ_{1}$ — possible because the types not realized are non-isolated and the omitting-types machinery of 42.02.03 pending, iterated, suppresses them. The cardinal invariant "number of types over the (definable, $ℵ_{0}$ -sized) copy of $A_{0}$ realized in the model" is $ℵ_{1}$ for $N_{sat}$ and $ℵ_{0}$ for $N_{small}$ , and it is preserved by isomorphism; so $N_{sat} \neq ≅ N_{small}$ . Two non-isomorphic models of size $ℵ_{1}$ contradict $ℵ_{1}$ -categoricity, so an $ℵ_{1}$ -categorical (hence by Morley any uncountably categorical) theory is $ω$ -stable. Rubric: full credit for extracting $2^{ℵ_{0}}$ types from non- $ω$ -stability, the saturated-versus-type-poor construction at $ℵ_{1}$ , and the isomorphism-invariant type count distinguishing them.

Advanced results Master

Categoricity organizes four developments: the Ryll-Nardzewski analysis of the countable case through finiteness of types and oligomorphy, the $ω$ -stability/Morley-rank machinery that converts uncountable categoricity into a dimension theory, Morley's transfer theorem with its Vaughtian-pair obstruction, and the Baldwin-Lachlan determination of the countable spectrum that launches classification theory.

Theorem 1 (Ryll-Nardzewski; the countable case). For countable complete $T$ with infinite models, $ℵ_{0}$ -categoricity is equivalent to finiteness of every $S_{n} (T)$ , to isolation of every type, to finitely many $n$ -formulas modulo $T$ for each $n$ , and to oligomorphy of $Aut (M)$ on the countable model ^{[Marker Ch. 4]}. The oligomorphic form locates $ℵ_{0}$ -categorical structures as exactly the countable structures whose automorphism group, as a permutation group, has finitely many orbits on tuples of each length — the permutation-group reading developed by Cameron. $DLO$ , the random graph, the pure infinite set, and infinite vector spaces over a finite field are the canonical examples; the random graph's homogeneity makes its automorphism group oligomorphic and even highly transitive on the level of the extension axioms, and Fraïssé's construction produces $ℵ_{0}$ -categorical structures as the limits of amalgamation classes with finitely many isomorphism types of $n$ -element configurations.

Theorem 2 ( $ω$ -stability, Morley rank, strong minimality). For countable $T$ , $ω$ -stability equals total transcendentality — every formula has ordinal Morley rank $RM$ , with Morley degree the count of rank-maximal pieces ^{[Morley 1965]}. The strongly minimal sets are the rank- $1$ , degree- $1$ definable sets; on each, $acl$ satisfies exchange, so $(D, acl)$ is a pregeometry and $dim$ is well-defined. Morley rank is a genuine dimension: $RM (φ \lor ψ) = max (RM φ, RM ψ)$ , rank drops under fibers (the additivity $RM (\exists y φ) \leq RM (φ)$ over definable families), and a type's rank measures how generic it is. $ω$ -stable theories have prime models over every set and saturated models in every uncountable cardinal (42.02.04 pending); the prime model over $A$ realizes only the isolated types over $A$ , the saturated model realizes all.

Theorem 3 (Morley's categoricity theorem and Vaughtian pairs). A countable complete theory categorical in one uncountable cardinal is categorical in all ^{[Morley 1965]}. The proof: uncountable categoricity $\Rightarrow$ $ω$ -stability (a non- $ω$ -stable theory has $2^{ℵ_{0}}$ types over a countable set, hence type-rich and type-poor models of the same uncountable size); $ω$ -stability $\Rightarrow$ saturated and prime models exist in each uncountable cardinal; and $κ$ -categoricity $\Leftrightarrow$ every $κ$ -model is saturated $\Leftrightarrow$ $T$ has no Vaughtian pair (a proper $M ≺ N$ with an infinite definable set unchanged). The two-cardinal theorems of Vaught and Morley convert a Vaughtian pair into models of distinct saturation in a common uncountable cardinal, so the no-Vaughtian-pair condition is cardinal-independent and transfers categoricity across all uncountable $λ$ . The Vaughtian-pair obstruction is the technical heart: categoricity is exactly the statement that every infinite definable set grows with the model, so no dimension is left frozen below the cardinality.

Theorem 4 (Baldwin-Lachlan; the countable spectrum and classification). For uncountably categorical $T$ , every model is prime and minimal over a strongly minimal set together with an $acl$ -basis, so its isomorphism type is fixed by the dimension of that strongly minimal set ^{[Baldwin-Lachlan 1971]}. The uncountable models have dimension equal to their cardinality (re-deriving Morley); the countable models have dimensions in ${0, 1, 2, \dots, ω}$ , giving $I (T, ℵ_{0}) = 1$ when the dimension is forced (the totally categorical case) and $I (T, ℵ_{0}) = ℵ_{0}$ otherwise — the Baldwin-Lachlan dichotomy $I (T, ℵ_{0}) \in {1, ℵ_{0}}$ , refining Vaught's $I (T, ℵ_{0}) \neq = 2$ (42.02.03 pending). This is the first node of Shelah's classification theory ^{[Shelah 1990]}: the stability hierarchy stratifies theories as stable, superstable, $ω$ -stable, and the main gap theorem on $I (T, κ)$ shows a sharp dichotomy — either $I (T, ℵ_{α})$ attains the maximum $2^{ℵ_{α}}$ for all uncountable $α$ (unclassifiable), or it is bounded by a slow function of $α$ and the models are classified by a tree of dimension-like invariants (classifiable, the descendants of the categoricity case).

Synthesis. Categoricity is the foundational reason cardinality became a structural invariant of theories, and putting these together it threads all four developments through one dimension-theoretic mechanism: a model is reconstructible from a count of independent ingredients, so equal cardinality forces equal count and equal count forces isomorphism. This is exactly what Theorem 2 makes precise — Morley rank supplies the ordinal dimension and its rank- $1$ strongly minimal sets carry the matroid $dim$ on which the whole reconstruction rests — and Theorem 1's Ryll-Nardzewski analysis is the countable shadow of the same idea, where "finitely many types" is the degenerate dimension that makes a single countable model. Morley's transfer of Theorem 3 generalises the finiteness criterion to the uncountable, replacing finite type-counts with ordinal-ranked ones, and the absence of Vaughtian pairs is dual to the demand that every definable dimension grow with the model — categoricity and no-frozen-dimension are the same cardinal-independent fact. The Baldwin-Lachlan analysis of Theorem 4 is the central insight that one strongly minimal set's dimension determines the entire model, so the countable spectrum collapses to ${1, ℵ_{0}}$ and the uncountable spectrum collapses to a single point.

The bridge from categoricity to classification theory is exactly this: Morley rank is the first instance of a dimension on definable sets, $ω$ -stability is the first stability class, and the absence of Vaughtian pairs is the first instance of the no-two-cardinal obstructions that Shelah's main gap turns into the structural dichotomy on $I (T, κ)$ (42.02.07 pending). The uncountably categorical theories — $ACF_{p}$ , infinite vector spaces — are the one-dimensional bottom of the hierarchy, where the dimension is transcendence degree or linear dimension, and the failure of $ω$ -stability for $(R; +, \cdot)$ and $(N; +, \cdot)$ marks the boundary past which the order property and the coding of arithmetic make cardinality powerless to classify. Categoricity is where dimension, stability, and the classification of models all begin.

Full proof set Master

Proposition 1 (Ryll-Nardzewski, finiteness $\Rightarrow$ categoricity). If $T$ is countable complete with infinite models and $S_{n} (T)$ is finite for every $n$ , then $T$ is $ℵ_{0}$ -categorical.

Proof. Finiteness of $S_{n} (T)$ makes every type isolated (a finite Hausdorff space is discrete, so each point is open, i.e. isolated). Hence the countable model is atomic, and finiteness of $S_{n} (A)$ for finite $A$ follows (each named parameter multiplies the type space by a finite factor), so the countable model is also saturated over finite sets. Two countable atomic-and-saturated models are isomorphic by back-and-forth (42.02.03 pending, Exercise 7 there): maintain a finite partial elementary map; to extend by a new element, the type of the enlarged tuple is isolated, so isolation supplies a matching witness on the other side, and saturation over finite sets supplies witnesses for the back step. The union is an isomorphism, so $T$ is $ℵ_{0}$ -categorical. $□$

Proposition 2 (exchange on a strongly minimal set). If $D$ is strongly minimal then $acl ↾ D$ satisfies the exchange property, and $(D, acl)$ is a pregeometry.

Proof. Suppose $a \in acl (B \cup {c}) ∖ acl (B)$ , witnessed by $φ (x, c)$ over $B$ with $a \in φ (M)$ and $∣ φ (M, c) ∣ = k$ minimal. Consider $θ (y) := φ (a, y) \land \exists^{= k} x φ (x, y)$ , definable over $B \cup {a}$ and holding of $c$ . If $θ (M)$ is finite, $c \in acl (B a)$ and exchange holds. If $θ (M)$ is infinite, strong minimality makes it cofinite, so for cofinitely many $y$ the formula $φ (x, y)$ has exactly $k$ solutions including $a$ . The set ${x : φ (x, y) for cofinitely many y}$ is definable over $B$ , contains $a$ , and is finite or cofinite. Finite gives $a \in acl (B)$ , excluded. Cofinite forces a fixed $k$ -element set to lie in $φ (x, y)$ for cofinitely many distinct $y$ ; but distinct $y$ give distinct fibers $φ (x, y)$ in a strongly minimal set once more than finitely many are used (two infinite families of disjoint $k$ -sets cannot share a common $k$ -set cofinitely), a contradiction. So $θ (M)$ is finite and $c \in acl (B a)$ . Exchange plus the finite-character and monotonicity of $acl$ give a pregeometry. $□$

Proposition 3 (dimension is well-defined and an isomorphism invariant). In a pregeometry $(D, acl)$ any two maximal independent sets have the same cardinality $dim D$ , and an isomorphism of strongly minimal sets over the base preserves $dim$ .

Proof. This is the matroid basis-exchange theorem from exchange: given independent $I, J$ maximal, the Steinitz exchange argument swaps elements of $I$ for elements of $J$ one at a time, preserving independence and spanning, so $∣ I ∣ = ∣ J ∣$ (the finite case by induction; the infinite case by a counting argument on the finite-character closure). An isomorphism $f : D \to D^{'}$ over the base maps $acl$ -independent sets to $acl$ -independent sets (algebraicity is a first-order, hence isomorphism-invariant, condition) and bases to bases, so $dim D = dim D^{'}$ . Thus two strongly minimal sets over the same base are isomorphic over it iff they have equal dimension: equal dimension lets a base-bijection extend, by exchange, to an isomorphism. $□$

Proposition 4 (non- $ω$ -stable $\Rightarrow$ not uncountably categorical). A countable complete $T$ that is not $ω$ -stable is not $ℵ_{1}$ -categorical.

Proof. Non- $ω$ -stability gives a countable $A_{0}$ with $∣ S_{n} (A_{0}) ∣ \geq ℵ_{1}$ for some $n$ . Construct a saturated $N_{1} ⊨ T$ of size $ℵ_{1}$ realizing $ℵ_{1}$ types over a copy of $A_{0}$ , by an elementary chain of length $ω_{1}$ realizing all finitely-based types at each stage. Construct $N_{2} ⊨ T$ of size $ℵ_{1}$ realizing only $ℵ_{0}$ types over the copy of $A_{0}$ , by iterating the omitting-types construction (42.02.03 pending) over a transfinite Skolem hull to suppress all but countably many non-isolated types while keeping cardinality $ℵ_{1}$ . The number of types over the definable copy of $A_{0}$ realized in a model is an isomorphism invariant, equal to $ℵ_{1}$ for $N_{1}$ and $ℵ_{0}$ for $N_{2}$ , so $N_{1} \neq ≅ N_{2}$ and $T$ is not $ℵ_{1}$ -categorical. $□$

Proposition 5 (categoricity $\Leftrightarrow$ no Vaughtian pair, for $ω$ -stable $T$ ). An $ω$ -stable countable $T$ is $κ$ -categorical for some/all uncountable $κ$ iff $T$ has no Vaughtian pair.

Proof. If $T$ has a Vaughtian pair $M ≺ N$ with $φ (M) = φ (N)$ infinite, the two-cardinal theorem of Vaught-Morley builds, in a common uncountable cardinal $λ$ , two models: a saturated one (where $φ$ has $λ$ solutions) and one where $φ$ has $< λ$ solutions (the frozen definable set transferred upward). These are non-isomorphic, so $T$ is not $λ$ -categorical. Conversely, with no Vaughtian pair, every infinite definable set grows with the model; an $ω$ -stable theory with this growth has every uncountable model saturated — a non-saturated $λ$ -model would omit a type over a small set, producing a frozen definable set and hence a Vaughtian pair. All saturated models of size $λ$ are isomorphic (42.02.04 pending), so $T$ is $λ$ -categorical, and the no-Vaughtian-pair condition is cardinal-independent, giving categoricity at every uncountable $λ$ . $□$

Proposition 6 (Baldwin-Lachlan dichotomy). For uncountably categorical $T$ , $I (T, ℵ_{0}) \in {1, ℵ_{0}}$ .

Proof. By Proposition 4 and Theorem 3, $T$ is $ω$ -stable with no Vaughtian pair, and (Baldwin-Lachlan structure theorem) after naming finitely many parameters there is a strongly minimal formula $D$ such that every model is prime and minimal over $D (M)$ with an $acl$ -basis; by Proposition 3 the isomorphism type is determined by $dim D (M)$ . For countable $M$ , $dim D (M) \in {0, 1, 2, \dots, ω}$ . If the dimension is forced to a single value across all countable models — the totally categorical case, where $T$ is also $ℵ_{0}$ -categorical by Proposition 1 — then $I (T, ℵ_{0}) = 1$ . Otherwise every value $0, 1, 2, \dots, ω$ is attained by a countable model, and distinct dimensions give non-isomorphic models (Proposition 3), an $ω$ -indexed family, so $I (T, ℵ_{0}) = ℵ_{0}$ . The dimension being the sole invariant ranging over an initial segment of ${0, 1, \dots, ω}$ , no value strictly between $1$ and $ℵ_{0}$ occurs. $□$

Connections Master

Types and the omitting types theorem 42.02.03 pending supplies the countable-case engine this unit ascends. The Ryll-Nardzewski theorem restated here — $ℵ_{0}$ -categoricity iff every $S_{n} (T)$ is finite iff every type is isolated iff $Aut (M)$ is oligomorphic — is the finite-type degenerate limit of the ordinal Morley-rank dimension this unit builds for the uncountable case, and Vaught's $I (T, ℵ_{0}) \neq = 2$ of that unit is exactly what Baldwin-Lachlan refines to $I (T, ℵ_{0}) \in {1, ℵ_{0}}$ for uncountably categorical theories. The prime and atomic models constructed there by omitting types are the prime-over-a-basis models the categoricity reconstruction here depends on.
Quantifier elimination and strong minimality 42.02.05 pending proves the strong minimality of $ACF$ that is this unit's central uncountably-categorical example. The finite/cofinite dichotomy of one-variable definable sets established there is precisely the strong-minimality hypothesis under which $acl$ satisfies exchange and the transcendence-degree dimension controls isomorphism; the QE that makes $ACF_{p}$ strongly minimal is the precondition that this unit's $κ$ -categoricity of $ACF_{p}$ presupposes, and Morley rank generalises the strong-minimality rank- $1$ case to all ordinals.
Saturation and homogeneous models 42.02.04 pending, co-produced this wave, supplies the existence and uniqueness of saturated and prime models in each uncountable cardinal that Morley's proof runs on. The equivalence " $κ$ -categorical iff every $κ$ -model is saturated" and the transfer of saturation across cardinals are statements about the saturated-model theory of that unit; Morley's own proof of the categoricity theorem is, in modern form, the statement that an $ω$ -stable theory's saturated models exist and absorb every model of the same uncountable size, exactly the saturation apparatus that unit develops.
Stability and the main gap 42.02.07 pending, co-produced this wave, is where the categoricity theory of this unit launches classification theory. Morley rank is the first dimension on definable sets, $ω$ -stability the first stability class, the absence of Vaughtian pairs the first no-two-cardinal obstruction; that unit generalises rank to forking and the U-rank, stability to the full stable/superstable/ $ω$ -stable hierarchy, and the categoricity dichotomy to Shelah's main gap on $I (T, κ)$ , with the uncountably categorical theories sitting as the one-dimensional bottom node.
Cardinal arithmetic and aleph exponentiation 42.03.03 supplies the cardinal-counting framework in which the categoricity spectrum and the function $I (T, κ)$ are stated. The two-cardinal theorems and the $2^{ℵ_{0}}$ -many-types arguments of this unit are cardinal-arithmetic facts; the main gap dichotomy $I (T, ℵ_{α}) = 2^{ℵ_{α}}$ versus a bounded slow function is a statement about aleph exponentiation, and the size-equals-dimension reconstruction of an uncountably categorical model uses the absorption $max (ℵ_{0}, κ) = κ$ for uncountable $κ$ from that unit.

Historical & philosophical context Master

The categoricity question in its modern form was posed by Jerzy Łoś in 1954, who conjectured that a countable theory categorical in one uncountable cardinal is categorical in all, having observed that the known examples — algebraically closed fields, divisible torsion-free abelian groups, vector spaces — were categorical in every uncountable power but split into many countable models. The countable case had been settled around 1959 by Czesław Ryll-Nardzewski, Erwin Engeler, and Lars Svenonius independently, characterizing $ℵ_{0}$ -categoricity by the finiteness of the type spaces and, in the Engeler-Svenonius form, by the oligomorphy of the automorphism group ^{[Marker Ch. 4]}. Michael Morley proved Łoś's conjecture in his 1962 Chicago thesis, published as "Categoricity in power" in 1965, introducing the transfinite rank now bearing his name and the notion of a totally transcendental theory; the proof's identification of uncountable categoricity with $ω$ -stability and the absence of Vaughtian pairs created stability theory as a subject ^{[Morley 1965]}.

John Baldwin and Alistair Lachlan, in their 1971 "On strongly minimal sets," recast Morley's theorem around strongly minimal sets and the pregeometry of algebraic closure, proving that every model of an uncountably categorical theory is controlled by the dimension of a single strongly minimal set and deducing that the number of countable models is exactly $1$ or $ℵ_{0}$ ^{[Baldwin-Lachlan 1971]}. The dimension-theoretic core they isolated — exchange, pregeometry, the basis invariant — became the geometric stability theory of Boris Zilber, Gregory Cherlin, Ehud Hrushovski, and Anand Pillay. Saharon Shelah's classification theory, developed through the 1970s and codified in Classification Theory and the Number of Non-Isomorphic Models (1978; 2nd edition 1990), generalized forking and the stability hierarchy and proved the main gap theorem on the function $I (T, κ)$ , with Morley's uncountably categorical theories as the simplest classifiable case ^{[Shelah 1990]}.

Bibliography Master

@book{marker2002modeltheory,
  author    = {Marker, David},
  title     = {Model Theory: An Introduction},
  series    = {Graduate Texts in Mathematics},
  volume    = {217},
  publisher = {Springer},
  year      = {2002}
}

@article{morley1965categoricity,
  author  = {Morley, Michael},
  title   = {Categoricity in power},
  journal = {Transactions of the American Mathematical Society},
  volume  = {114},
  year    = {1965},
  pages   = {514--538}
}

@article{baldwinlachlan1971,
  author  = {Baldwin, John T. and Lachlan, Alistair H.},
  title   = {On strongly minimal sets},
  journal = {Journal of Symbolic Logic},
  volume  = {36},
  year    = {1971},
  pages   = {79--96}
}

@book{shelah1990classification,
  author    = {Shelah, Saharon},
  title     = {Classification Theory and the Number of Non-Isomorphic Models},
  edition   = {2},
  series    = {Studies in Logic and the Foundations of Mathematics},
  volume    = {92},
  publisher = {North-Holland},
  year      = {1990}
}

@article{ryllnardzewski1959,
  author  = {Ryll-Nardzewski, Czes{\l}aw},
  title   = {On the categoricity in power $\leq \aleph_0$},
  journal = {Bulletin de l'Acad\'{e}mie Polonaise des Sciences},
  volume  = {7},
  year    = {1959},
  pages   = {545--548}
}

@incollection{los1954categoricity,
  author    = {{\L}o\'{s}, Jerzy},
  title     = {On the categoricity in power of elementary deductive systems and some related problems},
  journal   = {Colloquium Mathematicum},
  volume    = {3},
  year      = {1954},
  pages     = {58--62}
}

@book{tentziegler2012,
  author    = {Tent, Katrin and Ziegler, Martin},
  title     = {A Course in Model Theory},
  series    = {Lecture Notes in Logic},
  volume    = {40},
  publisher = {Association for Symbolic Logic / Cambridge University Press},
  year      = {2012}
}

@book{pillay1996geometric,
  author    = {Pillay, Anand},
  title     = {Geometric Stability Theory},
  series    = {Oxford Logic Guides},
  volume    = {32},
  publisher = {Oxford University Press},
  year      = {1996}
}

Prerequisites

42.02.05

Tier anchors

beginner: Marker 2002 *Model Theory: An Introduction* (Springer GTM 217) §4.3, Ch. 6 (categoricity read informally as 'cardinality alone pins down the structure' — for a fixed size, any two models obeying the rules are secretly the same object relabelled; the dense rational line and the dense real line being the same shape once you only count points, so the dense-order rulebook has one countable model; the surprise that for the algebraically closed numbers the size of the world fixes everything past the countable, because the world is determined by how many independent free coordinates it carries; the picture of a 'dimension' counting independent ingredients, two worlds of the same size carrying the same dimension and so matching up); checking by hand that the dense rational order has exactly one countable shape while the successor-line rulebook splinters into many countable shapes
intermediate: Marker 2002 *Model Theory: An Introduction* §4.3 and Ch. 4-6 (κ-categoricity; the Ryll-Nardzewski theorem characterizing ℵ_0-categoricity by finiteness of S_n(T), isolation of every type, and oligomorphy of the automorphism group, with DLO, the random graph, and (ℤ as a pure set) as worked examples; strongly minimal sets and the pregeometry of algebraic closure giving a well-defined dimension; ω-stability and counting types; Morley rank and degree; the statement of Morley's categoricity theorem that a countable theory categorical in one uncountable cardinal is categorical in all, and of the Baldwin-Lachlan theorem that an uncountably categorical theory has 1 or ℵ_0 countable models); Chang and Keisler 1990 *Model Theory* 3e (North-Holland) Ch. 7; Tent and Ziegler 2012 *A Course in Model Theory* (ASL/Cambridge) Ch. 3-6
master: Marker 2002 *Model Theory: An Introduction* (Springer GTM 217) Ch. 4-6 (the Ryll-Nardzewski theorem and oligomorphic automorphism groups; strongly minimal sets, the exchange property, acl-pregeometry and dimension; ω-stable and totally transcendental theories, Morley rank and Morley degree; prime and saturated models over sets; the absence of Vaughtian pairs and the two-cardinal theorems of Vaught and Morley; the full proof of Morley's categoricity theorem via the equivalence of ω-stability with κ-categoricity for some uncountable κ, and the upward/downward transfer; the Baldwin-Lachlan theorem through strongly minimal sets and dimension, refining Vaught's spectrum); Shelah 1990 *Classification Theory and the Number of Non-Isomorphic Models* 2e (North-Holland Studies in Logic 92), the main gap; Baldwin and Lachlan 1971 *On strongly minimal sets* (J. Symbolic Logic 36); Pillay 1996 *Geometric Stability Theory* (Oxford); Tent and Ziegler 2012 *A Course in Model Theory* Ch. 3-6

References

Marker, D. — Model Theory: An Introduction · Graduate Texts in Mathematics 217, Springer (2002), Chapters 4-6. A complete L-theory T with infinite models is κ-categorical (categorical in power κ) if any two models of T of cardinality κ are isomorphic. ℵ_0-categoricity is governed by the RYLL-NARDZEWSKI theorem: for countable complete T with infinite models the following are equivalent — T is ℵ_0-categorical; for every n the type space S_n(T) is finite; every type in every S_n(T) is isolated; for each n there are finitely many L-formulas in n free variables up to T-equivalence; and (Engeler-Svenonius) the automorphism group Aut(M) of the countable model M is oligomorphic, acting with finitely many orbits on M^n for every n. Worked ℵ_0-categorical examples: DLO (dense linear orders without endpoints; a type in n variables is fixed by the order pattern of the variables, S_n finite), the random (Rado) graph (the Fraïssé limit of finite graphs, ℵ_0-categorical and homogeneous via the extension axioms), the pure infinite set (ℤ or ℕ with equality only, S_n fixed by the equality pattern), infinite-dimensional vector spaces over a fixed finite field, and atomless boolean algebras. Uncountable categoricity is governed by MORLEY's categoricity theorem (Łoś's conjecture, 1954; proved Morley 1965): a countable complete theory categorical in one uncountable cardinal is categorical in every uncountable cardinal. The proof architecture: ω-STABILITY (for all countable A ⊆ M ⊨ T, |S_n(A)| ≤ ℵ_0) equals total transcendentality and is equivalent, for countable T, to κ-categoricity for some uncountable κ; MORLEY RANK RM(φ) and MORLEY DEGREE giving an ordinal-valued dimension on definable sets, with strongly minimal sets exactly the rank-1 degree-1 definable sets; STRONGLY MINIMAL sets, where the algebraic-closure operator acl satisfies the EXCHANGE property and so defines a PREGEOMETRY (matroid) with a well-defined DIMENSION equal to the cardinality of an acl-basis; PRIME and SATURATED models over sets (existence and uniqueness over ω-stable theories, cross-ref 42.02.04); the absence of VAUGHTIAN PAIRS (a Vaughtian pair is M ≺ N with M ≠ N and an infinite definable set the same in both) and the two-cardinal/upward-downward transfer theorems (Vaught, Morley) used to move categoricity between cardinals. The BALDWIN-LACHLAN theorem (Baldwin-Lachlan 1971): an uncountably categorical theory has exactly 1 or exactly ℵ_0 countable models (I(T, ℵ_0) ∈ {1, ℵ_0}), refining Vaught's no-two-countable-models theorem; the countable models are stratified by the DIMENSION of a strongly minimal set, ranging over a finite or ω-indexed ladder (cross-ref strong minimality 42.02.05, 42.02.07). The MORLEY categoricity SPECTRUM and the launch of CLASSIFICATION THEORY / STABILITY: Shelah's stability hierarchy (stable, superstable, ω-stable) and the MAIN GAP theorem on the function I(T, κ) counting models of size κ — either it grows as fast as 2^κ (the unclassifiable case) or it is bounded below 2^κ by a tame invariant (the classifiable case, controlled by dimension-like data). Examples: ACF_p (algebraically closed fields of characteristic p) is uncountably categorical, the unique model of each uncountable cardinal κ being determined by transcendence degree κ over the prime field, strongly minimal via the QE of 42.02.05 with acl = field-theoretic algebraic closure and dimension = transcendence degree; infinite vector spaces over a fixed field, strongly minimal with acl = linear span and dimension = linear dimension; the FAILURE for (ℝ, +, ·) and for (ℕ, +, ·), which are not ω-stable (the order, resp. the coding of arithmetic, produces 2^{ℵ_0} types) and so not uncountably categorical. The categoricity question 'when does cardinality alone pin down a structure?' is answered by Morley — uncountable categoricity is cardinal-independent — and the proof machinery (rank, stability, dimension, the avoidance of Vaughtian pairs) is the foundation of modern model theory. Cross-ref QE and strong minimality 42.02.05, types and Ryll-Nardzewski 42.02.03, saturation 42.02.04, stability and the main gap 42.02.07, cardinal arithmetic 42.03.03.
Morley, M. — Categoricity in power · Transactions of the American Mathematical Society 114 (1965), pp. 514-538. The original proof of Łoś's conjecture: a countable first-order theory categorical in some uncountable cardinal is categorical in every uncountable cardinal. Morley introduces the rank that now bears his name — an ordinal-valued dimension RM on the definable sets of a model, defined by a Cantor-Bendixson-style derivative on the type spaces — and the notion of a totally transcendental theory (every definable set has ordinal Morley rank), equivalent for a countable theory to ω-stability (|S_n(A)| ≤ ℵ_0 for every countable parameter set A). The proof shows: (i) categoricity in one uncountable κ forces ω-stability, since a non-ω-stable countable theory has, in some model, 2^{ℵ_0} types over a countable set and hence non-isomorphic models of size ℵ_1 (one realizing few types, one realizing many); (ii) an ω-stable theory has saturated and prime models in every uncountable cardinal, built by a transfinite construction realizing or omitting types under rank control; (iii) categoricity is equivalent to every model of size κ being saturated, which transfers up and down via the two-cardinal theorems and the absence of Vaughtian pairs. The technical heart is the Morley rank and degree calculus and the two-cardinal transfer: a Vaughtian pair would give two non-isomorphic models in a larger cardinal, so categoricity is exactly the omission of Vaughtian pairs over the relevant definable sets. Morley's paper inaugurated stability theory and classification theory; the program of computing the function I(T, κ) (the number of isomorphism types of models of T of cardinality κ) is the direct descendant, culminating in Shelah's main gap.
Baldwin, J. T. and Lachlan, A. H. — On strongly minimal sets · Journal of Symbolic Logic 36 (1971), pp. 79-96. A self-contained proof of Morley's categoricity theorem organized around strongly minimal sets, together with the determination of the number of countable models of an uncountably categorical theory. A definable set D in a model of T is strongly minimal if it is infinite and every definable subset of D (with parameters, in every elementary extension) is finite or cofinite; on a strongly minimal set the model-theoretic algebraic closure acl satisfies the Steinitz EXCHANGE property, making (D, acl) a pregeometry/matroid, so D carries a well-defined DIMENSION — the cardinality of any maximal acl-independent subset — and any two strongly minimal sets of the same dimension are isomorphic over the parameters. Baldwin and Lachlan prove that an uncountably categorical theory T (i) is ω-stable, (ii) has, after naming finitely many parameters, a strongly minimal formula, and (iii) every model is prime and minimal over a strongly minimal set together with a basis for it, so the isomorphism type of a model is determined by the dimension of the strongly minimal set. Consequently the models of T are linearly ordered by this dimension: the uncountable models all have dimension κ = |M| (forcing κ-categoricity for each uncountable κ, Morley's theorem), while the countable models have dimensions running through a possibly-finite-or-ω ladder. The BALDWIN-LACHLAN theorem reads off the countable spectrum: I(T, ℵ_0) is either 1 (when the strongly minimal set's dimension is determined, the ℵ_0-categorical case) or exactly ℵ_0 (the dimensions 0, 1, 2, ..., ω of a non-locally-modular or non-ℵ_0-categorical strongly minimal set give an ω-chain of pairwise non-isomorphic countable models), never any other value — refining Vaught's I(T, ℵ_0) ≠ 2 to I(T, ℵ_0) ∈ {1, ℵ_0} for uncountably categorical T. The strongly-minimal-set technology and the dimension invariant are the geometric core of the later geometric stability theory (cross-ref 42.02.05, 42.02.07).
Shelah, S. — Classification Theory and the Number of Non-Isomorphic Models · Studies in Logic and the Foundations of Mathematics 92, 2nd edition, North-Holland (1990). The systematic development of the stability hierarchy and classification theory growing out of Morley's categoricity theorem. A complete theory T is STABLE in κ if |S_n(A)| ≤ κ for all |A| ≤ κ; T is stable if stable in some κ, superstable if stable in all κ ≥ 2^{|T|}, and ω-stable (= totally transcendental for countable T) if stable in ℵ_0. The stability spectrum (stable in exactly the κ with κ^{|T|} = κ, or κ ≥ 2^{|T|}, or for all κ ≥ |T|) is determined by combinatorial properties of formulas (the order property, the independence property, the strict order property). The central object is FORKING — an abstract notion of independence generalizing acl-independence on strongly minimal sets and algebraic independence in ACF — giving a dimension theory (the U-rank, RM rank, weight) on stable theories. The culminating MAIN GAP theorem on the function I(T, κ) (the number of isomorphism classes of models of cardinality κ): for countable complete T, either T is classifiable — superstable, with no two-cardinal obstructions (NDOP, no DOP, shallow) — and then I(T, ℵ_α) is bounded by a slow function of α (the models are classified by a tree of cardinal invariants/dimensions), or T is unclassifiable and I(T, ℵ_α) = 2^{ℵ_α} attains the maximum for all uncountable α. Morley's uncountably categorical theories sit at the bottom of this hierarchy (ω-stable, one-dimensional, with a single strongly minimal set controlling the model), the simplest non-trivial case of the classification program; the main gap is the structural dichotomy the categoricity theorem first hinted at. Cross-ref stability and the main gap 42.02.07.

Estimated time

beginner: 22m
intermediate: 56m
master: 94m