There has been a growing interest in the role of theory within Software Engineering (SE) research. For several decades, researchers within the SE research community have argued that, to become a real engineering science, SE needs to develop stronger theoretical foundations. A few authors have proposed guidelines for constructing theories, building on insights from other disciplines. However, so far, much SE research is not guided by explicit theory, nor does it produce explicit theory. In this paper we argue that SE research does, in fact, show traces of theory, which we call theory fragments. We have adapted an analytical framework from the social sciences, named the Validity Network Schema (VNS), that we use to illustrate the role of theorizing in SE research. We illustrate the use of this framework by dissecting three well known research papers, each of which has had significant impact on their respective subdisciplines. We conclude this paper by outlining a number of implications for future SE research, and show how by increasing awareness and training, development of SE theories can be improved.