Summary  This paper describes an attempt to explore how far a categorisation of citations could be used as part of an assessment of the outcomes from health research. A large-scale project to assess the outcomes from basic, or early clinical, research is being planned, but before proceeding with such a project it was thought important to test and refine the developing methods in a preliminary study. Here we describe the development, and initial application, of one element of the planned methods: an approach to categorising citations with the aim of tracing the impact made by a body of research through several generations of papers. The results from this study contribute to methodological development for the large-scale project by indicating that: only for a small minority of citing papers is the cited paper of considerable importance; the number of times a paper is cited can not be used to indicate the importance of that paper to the articles that cite it; and self-citations could play an important role in facilitating the eventual outcomes achieved from a body of research.