This paper tries to contribute towards the solution of an important question raised in the SE literature: What is a Software Engineering (SE) specific theory? Which are the main features of a theory that is endemic to SE? In this paper we will use 'theory' as the term is used in traditional sciences. Other uses of the term 'theory' are discussed. Finally, we propose to focus on the reference class and on the structuring of SE theories as a basis for further progress.