The Linguistic Relevance of Quasi-Trees

We discuss two constructions (long scrambling and ECM verbs) which challenge most syntactic theories (including traditional TAG approaches) since they seem to require exceptional mechanisms and postulates. We argue that these constructions should in fact be analyzed in a similar manner, namely as involving a verb which selects for a " defective " complement. These complements are defective in that they lack certain Case-assigning abilities (represented as functional heads). The constructions differ in how many such abilities are lacking. Following the previous analysis of scrambling of Rambow (1994), we propose a TAG analysis based on quasi-trees.