Towards an interface definition of root phenomena*

Root phenomena are those that typically occur in matrix clauses but are also allowed in a restricted set of embedded (“root-like”) clauses. This paper explores root phenomena with an interpretive import, and identifies three kinds of data that a purely syntactic approach cannot account for: the gradience in acceptability within clause types, the variable behaviour of peripheral adverbial clauses, and the existence of root phenomena in “fragments”. I argue that far from being noise in otherwise harmonious behaviour, these data are essential to consider for a full understanding of root phenomena, and that a strictly syntactic approach cannot capture them. An interface account is called for, where (most of) the burden of licensing befalls on the interpretive component.