The concepts presented in this paper are motivated by the assumption that process knowledge is distributed knowledge and not completely known just by one person. Driven by this assumption we deal in this paper with the following questions: How can partial process knowledge be represented? How can this partial knowledge be used to define something more complete? To use higher level artefacts as building blocks to new applications has a long tradition in software engineering to increase flexibility and reduce modeling costs. In this paper we take a first step in applying this concept to processes, by defining process building blocks and operations which compose process building blocks. The building blocks will be referred to as process fragments in the following. The process fragment composition may take place either at design or runtime of the process. The design time approach reduces design costs by reusing artefacts. However the runtime fragment composition approach realizes high flexibility due to the possibility in the dynamic selection of the fragments to be composed. The contribution of this work lies in a fragment definition that enables the fragment modeler to represent his ’local’ and fragmentary knowledge in a formal way and which allows fragment models to be composed.