The nominalization of several French idioms constructed with the verb mettre contrasts with that of derived nominais such as la destruction de Rome par les barbares. This contrast is mainly due to the quasi mandatory shift of the internal arguments of NPs such as la mise en oeuvre de X ‘the implementation of X’ which corresponds to mettre X en oeuvre ‘to implement X’. A standard morphological treatment of the nominalization of mettre cannot account for the “parasitic gap” associated with this shift. A treatment in terms of “derivational syntax” appears to be more adequate, in that it attributes to the verbal suppletion the preservation of the categorial identity and the semantic unity of the lexical entry shared by the NP and the VP. A major theoretical consequence of this analysis is that it challenges the generally agreed upon claims that X-bar structures are necessarily endocentric and that lexical entries always constitute atomic categories.

This article proposes a solution to two unresolved issues regarding the V2 structure in the history of French: to which type of V2 languages — symmetrical or asymmetrical—does Old French belong, and why did V2 later disappear? It appears that Old French is not homogeneous with respect to V2: it is symmetric up to the 12th century and then goes through a period of grammatical competition between the two types of V2 structure. This indicates a syntactic change in progress, from a symmetrical to an asymmetrical structure. But French never completed this transition and lost V2 after the period of competition, contrary to Continental Scandinavian which underwent the same change. This contrast follows from the fact that Old French failed to meet the learnability conditions of the asymmetrical V2 structure. This analysis thus raises the issue of the relation between syntactic change and learnability.

This article investigates the nature of argumenthood and adjuncthood, through an examination of the behaviour of the internal arguments of two classes of motion verbs in English. A highly modular view is put forth, in which three separate distinctions influencing argument-like or adjunct- like behaviour must be recognized: aspectual versus thematic licensing, structural versus inherent case assignment, and referentiality versus non-referentiality. Of these three, only referentiality is a graded rather than a binary distinction. The distinction between aspectual and thematic licensing is developed and elucidated. A picture emerges in which aspectual structure may itself be thematically licensed by a verb, and this aspectual structure may have its own arguments, which are then indirectly licensed by the verb. Cognate objects and Romance measure phrases are also discussed in light of these theoretical conclusions.