In Model Driven Engineering, bidirectional transformations are considered a core ingredient for managing both the consistency and synchronization of two or more related models. However, current languages still lack of a common understanding of their semantic implications hampering their applicability in practice. This paper illustrates a set of relevant properties pertaining to bidirectional model transformations. It is a first step towards a taxonomy that can help developers to decide which bidirectional language or tool is best suited to their task at hand. This study is based on the existing literature and characteristics of existing approaches.