In a step-wise design of systems, models of components are being developed on several levels of abstractions. In such a design process model transformations are used to change or replace (parts of) models. Model transformations are required to be behaviour preserving: component models at lower levels should adhere to the descriptions given in higher levels thus achieving substitutability. Moreover, for complex components, models usually consist of descriptions of different views or aspects (e.g. data and protocols). Consequently, different kinds of transformations take place on different views, and together they should guarantee behaviour preservation. In this paper we discuss the applicability of formal methods concepts to model transformations. Formal methods come with build-in notions of transformations between models, or more precisely, with refinement and subtyping concepts which provide means for comparing component models on different levels with respect to their behaviour. Moreover, refinement and subtyping concepts for different views can be shown to neatly fit together. This is achieved by giving a common semantics to all views, which furthermore opens the possibility of checking consistency.