Each language has a strategy for the formation of constituent ('wh'-) questions. Cross-linguistically, while the interpretation of these questions can be invariant, the strategies employed in the world's languages differ. I propose a typology of constituent question formation strategies (CQFSs) based on two empirically observable variables: (i) the structural level at which interrogative scope is marked by one (focused) question word, and (ii) the structural level at which any question word not involved in scope marking (i.e. in a multiple question) is focused. One striking consequence of this approach is that English, instead of exemplifying the theoretically expected but in reality arbitrary norm of a single movement, is seen rather as a hybrid of two other CQFSs, exemplified by Hungarian and Japanese. By refining and fully exploiting its parallel architecture, the proposed LFG analysis captures the relevant generalizations, as well as the notion that CQFSs represent different means to the same end. This development of a typologically informed, formally explicit, non-derivational framework represents a new approach to constituent question formation in natural language, and reinforces the importance of LFG as a framework for investigating linguistic structure.