A rule of grammar is a generalization over possible and impossible constructions. During the process of analysis, it is a hypothesis and therefore subject to falsification.

A first launch of the rule may be too strict. For instance, the Cabecar enclitic lé presumptive is first only seen at the end of the clause and in Wackernagel's position; so these constitute the contexts for the rule of its distribution. One falsifies this hypothesis by putting the clitic in other positions, e.g. immediately following the predicate. It turns out that this is grammatical. So the initial rule must be widened: The enclitic is mobile.

Contrariwise, the first launch of a rule may be too generous; a formative or construction may seem to be allowable in all kinds of contexts. One then seeks context factors that might reasonably exclude it. For instance, a particle may not be admissible in an interrogative or in a negative clause.