The main goal of this paper is to present a new account of improper movement phenomena first discussed in Chomsky (1973). Improper movement phenomena provide us with an interesting puzzle regarding learnability: why is it that A-to-A, A'-to-A' and A-to-A' movement are allowed, yet A'-to-A movement is by hypothesis excluded? The question: "What is the nature of the mechanism or constraint responsible for the prohibition?" has been a central concern since Chomsky (1973). The authors claim that improper movement is excluded by virtue of Agree failure between a moving element and a finite T as a consequence of "feature-splitting", which the authors argue is the most natural implementation of Chomsky's phi-feature inheritance system and Richards' (2007) value-transfer simultaneity. In addition, this analysis is empirically supported by and seeks to explain, without stipulation, A'-opacity intervention effects discussed in Rezac (2003). Furthermore, the proposed account enables the authors to rule out improper movement without appeal to the Activity Condition.