Forking in finite models (Q2354623)

Let \( \mathcal L=L^k, k\geq2\) be a first order language with exactly \(k\) distinct variables. Further, let \(T\) be a complete \(\mathcal L\) theory and \(\mathbf K=\{\mathcal A\mid \mathcal A\models T\text{ and }\mathcal A\text{ finite}\}\). It is assumed that there are infinitely many isomorphism types among the elements of \(\mathbf K\), \(\mathbf K\) has amalgamation over subsets of \(\mathbf K\)-models, and \(\mathbf K\) is stable. The author assumes familiarity with his earlier article [Arch. Math. Logic 39, No. 2, 89--102 (2000; Zbl 0972.03034)]. In this article, the stability of \(\mathbf K\) means that the ranks \(R^n(a,\mathcal A,\omega)\) exist. Hence, forking can be defined. The first part of the article under review defines forking and establishes its usual properties. The second section deals with the construction of \(\mathbf K\)-models under the assumption that forking is trivial; e.g., if forking is trivial, for every \(m\in \mathbb N\) there is an \(m\)-saturated \(\mathbf K\)-model \(\mathcal A\).