Advanced Type Theory: Omission and Realization

Core Idea

Advanced results characterize when models can simultaneously realize and omit specific families of types. The interaction between which types are realized and which are omitted determines the model class structure and classification theory. Techniques include Löwenheim-Skolem, compactness, and omitting types theorem.

Explainer

You already know the Omitting Types Theorem for countable theories: a type p(x) can be omitted in a countable model if and only if p is non-isolated — that is, no single formula in the theory isolates p by entailing it. And you know that realized types (types actually witnessed by elements of the model) and omitted types (consistent but unwitnessed) together shape what a model looks like from the inside. Advanced type theory asks a harder question: can you simultaneously control which families of types are realized and which are omitted across an entire model, and what does the answer tell you about the theory?

The key tool is the interplay between isolation and density. A type is isolated if there is a formula φ such that every model satisfying φ realizes the type. Isolated types *must* be realized in any model built via a Henkin-style construction — you cannot avoid them. Non-isolated types can be omitted, but requiring their omission constrains the construction. When you try to simultaneously omit a countable family of non-isolated types, you are essentially running a Baire category argument: each type you omit is a comeager condition on the space of models, and countably many comeager conditions can be simultaneously satisfied. The Baire category theorem underlies why countably many non-isolated types can be jointly omitted, but uncountably many cannot be handled the same way.

Realization tells the opposite story. A saturated model realizes every type consistent with a finite set of parameters from the model — it is maximally type-rich. A homogeneous model realizes all types consistent with smaller cardinal-sized sets of parameters. These constructions are dual to omitting: instead of building a sparse model that avoids types, you build a rich model that includes every possible type. The structure theorem says that any two saturated models of the same cardinality are isomorphic, which means the collection of realized types completely determines the isomorphism type at that cardinality.

The classification-theoretic payoff appears when you ask which theories have "few" models. A theory is ω-categorical if it has exactly one countable model up to isomorphism. By the Ryll-Nardzewski theorem, this happens precisely when the theory has only finitely many types over any finite parameter set — a strong constraint on type space that forces all countable models to look alike. More generally, Shelah's stability theory classifies theories by how their type spaces grow: stable theories have well-behaved type spaces with no order-like structure, while unstable theories have type spaces too complex for classification. The advanced study of type omission and realization is thus the microscope through which logicians examine the classification of all first-order theories.