Kent Bach, "Seemingly Semantic Intuitions"

Gibbs and Moise's tests for what is said are faulty; the best test for whether something is (semantically) said or something that is pragmatically contributed is whether it is cancellable or not--if not, it is what is said.