In French, a quantifier can appear in various positions outside of the NP it quantifies over, whether this NP is the subject or the (direct or indirect) object of the sentence. This phenomenon, often referred to as ‘floating’, has been investigated since the early stages of the generative framework, and several analyses have been proposed to account for both the quantifier subject and the quantifier object in a unified way. However, to my knowledge, none of them has succeeded in providing such a unified account without recourse to non-explanatory restrictions. The main aim of this paper is to propose an analysis that does not require any such restrictions. The focus will be on anaphoric quantifiers (i.e. quantifiers that have to be linked to some other argument position in order to be interpretable), the analysis of which will be shown to extend straightforwardly to pronominal and adverbial quantifiers, according to the principles of Government and Binding theory.

The study of floating quantifiers raises the broader question of how to account for locality requirements in a satisfactory way. Basically, there are two possible ways to account for the restrictions on the distribution of floating quantifiers: either they flow from derivational restrictions, or they are subject to representational restrictions. I will argue in favour of the latter.

The analysis proposed here is essentially syntactic. However, reference will be made to the semantic interpretation of various structures: the position occupied by the floating quantifier at S-structure will be shown to constrain its interpretation. The semantics of floating quantifiers will however not be investigated beyond this.