Forking and Independence in Stability Theory

Core Idea

In stable theories, forking is a notion of dependence: a type p forks over a set A if it extends to two contradictory types over a larger set. Non-forking extension provides a notion of algebraic independence in arbitrary structures, generalizing the concept from field theory. Forking satisfies symmetry and transitivity, making it a fundamental concept in stability theory.

How It's Best Learned

Study forking in algebraically closed fields (ACF), where non-forking corresponds to algebraic independence. Verify the forking axioms: symmetry, transitivity, and finite character.

Explainer

From stability theory, you know that stable theories have well-controlled type spaces — the number of types over any set is bounded. From type spaces and Stone topology, you have a precise picture of what a type is: a maximal consistent set of formulas. But stability alone doesn't give you a canonical way to extend types from one parameter set to a larger one while preserving "independence." Forking is the device that solves this problem — it defines a notion of when a type extension is "generic" (non-forking) versus "dependent" (forking).

The definition is technical but the intuition is algebraic: in a field, an element a is algebraically independent over a set A if a does not satisfy any polynomial equation with coefficients from A. Forking generalizes this to arbitrary stable structures. A type p(x) over a set B forks over A ⊆ B if it implies a disjunction of formulas φ₁(x, b₁) ∨ ... ∨ φₙ(x, bₙ) where each φᵢ is "algebraically constraining" in a precise sense (each has finitely many realizations in any model). Informally, p forks over A when B contains information that dramatically constrains where x can land — information that was not already present in A. A non-forking extension of a type over A to a larger set B is an extension that adds no such new constraints.

The key properties that forking satisfies in stable theories are what make it a genuine independence relation. Symmetry: a is independent from b over A iff b is independent from a over A. Transitivity: if a is independent from bc over A and independent from b over Ac, then a is independent from b over A. Finite character: a forks over A iff some finite subset of its type forks. Extension: any type over A has a non-forking extension over any larger set B. These four axioms are axioms of an independence relation in the abstract sense, and in stable theories they have a unique solution — forking is the *only* relation satisfying them.

In algebraically closed fields (ACF), non-forking independence coincides exactly with algebraic independence: the transcendence degree over A of a tuple a is preserved in non-forking extensions. This gives you concrete grounding. In general stable theories, forking plays the same structural role that linear independence plays in vector spaces or algebraic independence plays in fields — it defines a dimension theory. The Morley rank and the forking geometry of a strongly minimal set determine whether the structure "looks like" a vector space (modular geometry) or a projective space or something more complex. This geometric classification of stable theories — one of the deepest results of model theory — rests entirely on forking as its foundation.