Evaluations of practice have often adopted experimental or quasi-experimental approaches and have focused heavily on outcomes of innovative programmes. This creates problems in replicating studies and effectively creates obstacles to the sharing of good practice. In this paper, ideas from Pawson and Tilley's (1997) approach to Realistic Evaluation are used. The approach explores contexts, mechanisms and outcomes of programmes, so that the particular characteristics of a setting are accounted for in drawing conclusions about programme effectiveness. As examples, two earlier studies reported in "Support for Learning" are reexamined from the perspectives of Realistic Evaluation. It is suggested that Realistic Evaluation might be an appropriate framework to assess innovation in education and may also provide opportunities for researchers and practitioners to work together.