# 6.5 Classification Programs Classification theory studies first order theories by organizing them according to the complexity of their definable sets and types, with the goal of distinguishing theories that admit a systematic structural description from those whose models exhibit uncontrolled combinatorial behavior. The central principle is that the behavior of formulas, when evaluated across different parameter configurations, determines the global structure of models, and therefore one can classify theories by analyzing the combinatorial patterns that formulas can encode. A key insight is that not all theories are equally complex. Some theories behave in a highly regular and predictable way, where types are controlled and definable sets resemble geometric objects, while other theories allow arbitrary configurations that lead to a proliferation of types and prevent meaningful classification. The role of classification theory is to isolate precise dividing lines that separate these regimes. ### Dividing Lines and Combinatorial Properties A dividing line is a property of a theory defined in terms of the behavior of formulas, and it separates theories into classes with fundamentally different structural characteristics. These properties are usually formulated in terms of the existence or absence of certain configurations that encode combinatorial complexity. The most basic such property is the order property. A formula $\varphi(x,y)$ has the order property if it can encode an infinite linear order, meaning that there exist sequences $(a_i)$ and $(b_j)$ such that: $$ \mathcal M \models \varphi(a_i,b_j) \quad \text{if and only if} \quad i < j. $$ This configuration allows the formula to distinguish infinitely many different patterns, and it leads to a large number of distinct types over parameter sets. The presence of the order property is therefore a source of instability. A second important property is the independence property. A formula $\varphi(x,y)$ has the independence property if it can encode arbitrary subsets of an infinite set. More precisely, there exist sequences $(a_i)$ and $(b_I)$ indexed by subsets $I \subseteq \mathbb N$ such that: $$ \mathcal M \models \varphi(a_i,b_I) \quad \text{if and only if} \quad i \in I. $$ This means that the formula can represent any pattern of membership, and therefore it introduces a high degree of combinatorial freedom. The absence of this property defines the class of NIP theories. These combinatorial properties are not merely technical conditions, but they reflect deep structural differences between theories. ### Stability A theory is stable if the number of types over a parameter set is controlled by the size of that set. Formally, for every infinite set $A$: $$ |S_1(A)| \leq |A|. $$ This condition ensures that the space of types does not grow too quickly, and it leads to strong structural consequences. In stable theories, types behave in a regular way, and one can define a notion of independence that shares many properties with linear independence in vector spaces. Stability excludes the order property. If a formula had the order property, then one could construct many distinct types by varying the position of an element relative to an ordered sequence, and this would violate the bound on the number of types. Stable theories include important algebraic examples, such as algebraically closed fields, where definable sets correspond to algebraic varieties and types correspond to generic points of these varieties. ### Simplicity The class of simple theories generalizes stable theories by weakening the constraints while preserving a well behaved notion of independence. A theory is simple if it admits a notion of independence, called forking independence, that satisfies key properties such as symmetry, transitivity, and extension. In simple theories, types may be more numerous than in stable theories, but they still obey a coherent structure. Simple theories include stable theories as a special case, but also include additional examples such as the random graph, where the combinatorial behavior is richer but still controlled in a precise sense. The development of simplicity theory shows that many techniques from stability can be extended beyond the stable context, although often with additional complexity. ### NIP Theories A theory has NIP if no formula has the independence property. This condition ensures that formulas cannot encode arbitrary subsets of infinite sets, which places a strong restriction on definable sets. NIP theories form a broad class that includes stable and simple theories, but also includes structures that are not simple, such as ordered fields. In these theories, definable sets exhibit combinatorial regularity, and one can develop tools analogous to measure theory and probability. The study of NIP theories connects model theory with other areas of mathematics, including combinatorics and analysis. ### Hierarchy of Theories These dividing lines organize theories into a hierarchy: stable theories, simple theories, NIP theories, general theories. Each level allows more complexity than the previous one, and the techniques required to study theories become more sophisticated as one moves down the hierarchy. ### Types and Their Structure Types are the central objects in classification theory. A type describes all possible properties of an element relative to a parameter set, and the collection of all types over a set forms a space that encodes the possible behaviors of elements. In stable theories, types over models are well behaved. They are often definable, and they can be organized using independence relations. This allows one to treat types as geometric objects, where definable sets correspond to regions and types correspond to points or directions. In unstable theories, types can be much more numerous and less structured. The presence of combinatorial configurations such as the order property or the independence property leads to a rapid growth in the number of types, which makes classification difficult. ### Definability of Types In stable theories, types over models are definable. This means that for each formula $\varphi(x,y)$, there is a definable condition that determines whether $\varphi(x,b)$ belongs to the type. This property allows one to encode types using formulas, and it plays a crucial role in the development of geometric stability theory. ### Independence and Geometry One of the main achievements of classification theory is the introduction of independence relations that generalize familiar notions from algebra. In stable theories, independence satisfies properties similar to those of linear independence in vector spaces. For example, it is symmetric, transitive, and has a notion of dimension. This leads to a geometric interpretation of models, where elements behave like points in a space, and definable sets behave like geometric objects. ### Strongly Minimal Sets A strongly minimal set is a definable set in which every definable subset is either finite or cofinite. These sets are the simplest infinite definable sets, and they serve as building blocks for more complex structures. The study of strongly minimal sets reveals deep connections between model theory and algebraic geometry, where such sets correspond to irreducible varieties of dimension one. ### Morley Rank and Dimension In stable theories, one can define a notion of dimension for definable sets, such as Morley rank. This assigns an ordinal valued measure of complexity to definable sets, and it allows one to classify them according to their structure. Morley rank behaves in many ways like dimension in geometry, and it provides a powerful tool for analyzing stable theories. ### Lemma 6.73 (Stability Excludes Order Property) If $T$ is stable, then no formula in $T$ has the order property. Proof Assume that some formula $\varphi(x,y)$ has the order property. Then there exist sequences $(a_i)$ and $(b_j)$ such that: $$ \mathcal M \models \varphi(a_i,b_j) \quad \text{if and only if} \quad i < j. $$ As shown earlier, this configuration can be used to construct many distinct types over the parameter set $\{b_j : j \in \mathbb N\}$, yielding: $$ |S_1(A)| \geq 2^{\aleph_0}. $$ Since $A$ is countable, this implies: $$ |S_1(A)| > |A|, $$ which contradicts stability. Therefore no such formula exists. ### Towards Classification The classification program seeks to describe models of a theory up to isomorphism using invariants derived from types and definable sets. In stable theories, this program has been highly successful. Models can often be decomposed into simpler components, and their structure can be described using geometric and algebraic methods. In less tame theories, classification becomes more difficult, and the focus shifts to identifying partial structure and understanding the sources of complexity. ### Perspective Classification theory reveals that first order logic contains a rich internal structure, where the behavior of formulas determines the geometry of models. By identifying and studying dividing lines, one can navigate this landscape and understand which theories admit a coherent and tractable description. This perspective continues to guide modern research in model theory, connecting logic with algebra, geometry, and combinatorics.