The effects of focus on syntax differ across languages: some languages encode focus in situ, while in other languages focus induces an array of constructions that deviate from the canonical configuration, such as non-canonical orders or clefts. This article presents semi-spontaneously produced data from American English, Québec French, Hungarian, and Georgian which shows that speakers of these languages select different structures under identical discourse conditions. The observed cross-linguistic differences are accounted for by means of grammatical properties of the object languages that hold independently of information structure. This account leads to the conclusion that a non-compositional mapping between information structural concepts and structural configurations is an unnecessary complication of the grammatical model.