In earlier generative grammar, sentences associated with Easy-type adjectives such as easy, hard, simple, difficult, tough, and so on had been derived by transformations, and assumed to be derived from a single underlying structure. And the difference of sentential types resulted from transformations. But this earlier analysis raises constituency problem in that each of `easy to please` and `easy to please John` forms a different constituent regardless of having the same underlying structure. And this NP-Movement results in the violation of the Specified Subject Condition in Chomsky(1977). To solve these problems Chomsky(1977) proposed that in `tough` constructions, the matrix subject NP is related indirectly to the gap in the complement clause by means of two distinct rules.; Wh-Movement and Rule of Predication. But in Chomsky(1977)`s approach, the interpretation of the constructions is not determined by only the pure Wh-Movement and is supplemented by the Rule of Predication and other as hoc conditions. And Chomsky(1981) introduces the Rule of Reanalysis to `tough` constraction to solve Chomsky(1977)`s problems, but the Reanalysis approach damages the fundamental principles, such as the projection principle, the Thematic theory, and the control theory. Recently, in the sentence like `John is easy to please.`, the matrix subject position which is base-generated has known to be non-θ-position. So, D-structure cannot be a pure representation of θ-structure. To solve these two problems, I assume that Easy-type adjectives have two separate lexical structures and the unit of `Easy-type adjective+to+ Verb` forms a single constituent, a complex lexical item which is lexically base-generated in the lexicon. But this approach has problems that produce an uninterpretable trace in D-structure, and that need a new device to provide a suitable interpretation for the trace. To solve these problems, I propose a new model of `tough` constructions based on Chomsky(1992)`s Minimalist approach.