The property of a reduction system (rewriting or computational calculus) that every possible reduction sequence starting from any term is finite; equivalently, every term has no infinite descending chain of reductions and therefore reaches a normal form.

Principle

Well-foundedness of the reduction relation: there is no infinite chain t0 → t1 → t2 → …, so every reduction sequence must terminate at a form with no further reductions.

Demonstration

In the simply typed lambda calculus, one proves strong normalization by assigning types and using a logical relations or reducibility candidates argument to show every typed term admits no infinite beta-reduction sequence and thus evaluates to a normal form.

Misapplication

Assuming untyped lambda calculus or a system with unrestricted recursion is strongly normalizing; or confusing strong normalization with normalization relative to a particular reduction strategy (weak normalization or strategy-dependent termination).

Consequence

Programs corresponding to terms in a strongly normalizing system always terminate; strong normalization together with confluence yields decidability of convertibility and supports consistency proofs for type theories.

Reversal

Negation of strong normalization is the existence of at least one term admitting infinite reduction sequences (divergence); in such systems there exist computations that never reach a normal form.

Boundary

Applies to abstract reduction relations and calculi without side effects; it does not directly cover systems with non-terminating primitives (general recursion, I/O loops), nor does it coincide with weak normalization, which only guarantees existence of some terminating strategy.

Semantic Tension

Tension with weak normalization (which requires only one terminating reduction path) and with confluence (which concerns uniqueness of normal forms); a system can be confluent but not strongly normalizing, or strongly normalizing but not confluent in pathological definitions.

Synthesis

Strong normalization is the global termination guarantee for a rewriting or computational system: it asserts that every term admits no infinite descent under the reduction rules and thus universally reaches a normal form, enabling reliable extraction of canonical results and supporting consistency and decidability reasoning.