What if it's lexicalization?.

Abstract

In this paper I show the possibility of, and in fact a need for applying lexicalization criteria to P doubling. Whenever doubling appears in non-extractions and over short distances, it indicates a cross-linguistic preference for adjacency of a verb and preposition which form a semantically connected sequence. The same effect has been shown to be operative in relative and interrogative clauses (Nykiel, forthcoming); in such extractions, however, other factors are also involved, for example, locality and wh-processing (Hawkins 2004, Arnon et al 2005, Staum and Sag 2008). While lexicalization effects have been noticed before (Wasow 2002, Hawkins 2004), they have not yet been appealed to with respect to P doubling. The data presented here strongly suggest that theories of grammar can account for P doubling if it is interpreted as a performance strategy that interacts with competence.