On the notion of a patterning property in model theory

Abstract

Different kinds of definable patterns in the models of a first-order theory, such as the order property, the tree property, or the ($n$-)strict order property, allow us to distinguish theories according to their logical complexity. The complexity distinctions given by these definable patterns play a central role in model theory. However, a rigorous definition of the notion of a model-theoretic patterning property has yet to be established. We start by discussing different proposals from the literature for making the notion of a model-theoretic patterning property rigorous. Some examples will include the straight definability of Shelah, which will describe properties definable by a pattern of consistency and inconsistency in a formula and its negation, and the poset definability of Garcia and Mennuni, covering properties definable by interpreting a partial order embedding a given poset. We will also introduce a higher-arity version of straight definability. In our first main result, we will answer open questions of Bailetti and Garcia-Mennuni, showing that the $n$-strict order property $\mathrm{SOP}_{n}$ is straightly definable and poset definable even for integers $n \geq 4$. This will complete the categorization of all of the classical classification-theoretic properties as straightly definable. Our other main result will concern properties that are straightly definable without negation: the positively straightly definable properties defined by Bailetti. We will show using Saracino's theorem and results of Bodirsky, Bodor and Marimon that, in any countably categorical theory, implications between positively straightly definable properties must be exhibited at the level of $\exists\forall$-formulas. This will have special consequences under the assumption that $\mathrm{SOP}_{2}$ is equal to $\mathrm{SOP}_{3}$.