On the Nature of Edge Feature Movement

On the Nature of Edge Feature Movement

&nbsp;&nbsp;Edge feature movement (EF-movement) is supposed to be pure internal merge (IM), which is solely induced by edge feature (EF) just like pure external merge (EM). Both pure IM and pure EM, which constitute "free Merge", are supposed to come free in natural language (Chomsky 2005). In this paper, in fact, it is shown that EF-movement is solely constrained as well as interpreted by interface conditions, which means that EF-movement is not subject to last resort or crash at narrow syntax but to the extended interface conditions.<BR>&nbsp;&nbsp;It is not necessary or even possible to assume that accounting for various syntactic deviancies entirely depends on syntactic operations like feature checking in narrow syntax; for example, syntactically deviant or "crashed" sentences may receive appropriate interpretations at the interface. In this paper, it is argued that at least EF-movements are licensed in terms of the D-effect interpretation at the interface. In other words, the notion of Agree in narrow syntax should be replaced by the notion of D-effect in terms of the D-effect Condition at the interface for EF-movement. In place of the various minimality conditions for Agree-movements at narrow syntax, the Interpretive Economy Condition is proposed at the interface to constrain the interpretation of derivations by EF-movement. For A-type EF-movements the Case Licensing Condition is proposed at the interface for the generalizations on case that are independent of Agree for the theory of EF-movement. Given the three conditions at the interface, D-effect Condition, Interpretive Economy Condition, and Case Licensing Condition, it is argued that EF-movements are "free-ride movements" in the sense that they are simply induced by free Merge in narrow syntax only to be licensed as well as constrained by the three interface conditions. It is speculated that ultimately not only EF-movements but all movements need to be checked or licensed through Agree or D-effect at the interface, for proper characterization of their interactions with other factors at the interface such as possible interpretations of even "crashed" sentences in terms of linguistic and/or non-linguistic contexts at the interface.