A phenomenon in probabilistic logic and finite model theory: for a fixed logical language and a probability distribution on finite structures (commonly uniform on n-element structures or G(n,p) graphs), every sentence has a limiting probability as the size grows, and the Zero-One Law asserts that this limit is always either 0 or 1.

Principle

Symmetry and indistinguishability of large finite structures under the chosen distribution force every definable property to be almost surely true or almost surely false; combinatorial concentration and finite-variable equivalences underlie the dichotomy.

Demonstration

In the Erdős–Rényi random graph model G(n,1/2) and first-order logic, any first-order sentence has a limit probability of either 0 or 1 as n → ∞; typical proofs use quantifier-rank locality and extension axioms to show that, with high probability, large graphs satisfy the same finitely many local constraints.

Misapplication

Assuming a Zero-One Law holds for stronger logics (for example full second-order logic) or for different probability regimes (very sparse graphs, e.g. p = c/n) without checking hypotheses; this leads to false predictions about almost-sure properties.

Consequence

When applicable, it collapses the study of asymptotic behaviour of definable properties to a binary classification (almost-sure vs almost-never), simplifying typical-case reasoning and enabling transfer of model-theoretic limit results to probabilistic combinatorics.

Reversal

A nontrivial limit law in which some sentences have limiting probabilities strictly between 0 and 1 (a convergence law), or a regime with no limiting probabilities at all; these exhibit richer asymptotic variability.

Boundary

Depends on the logical language (first-order often satisfies zero-one under classical distributions), the probability model (dense vs sparse), and the signature; many fragments or alternative measures fall outside the law's scope.

Semantic Tension

Tension between 'almost sure' (measure-theoretic determinacy) and meaningful structural variation: a property that is measure-one may still capture important typical structure, yet the zero-one collapse hides quantitative frequencies and rates of convergence.

Synthesis

The Zero-One Law captures how, under symmetry-preserving random models and limited expressive resources, definability and probabilistic concentration combine to make every sentence either almost always true or almost always false, while exceptions arise when either the logic or the distribution introduces greater distinguishing power.