The Radford Memorial LectureIn the last 20 years a large body of literature on ways of synthesising research in education has developed (Dunkin 1994). Approaches to gleaning the accumulated findings of that research have varied from the narrative, through vote-counting or box-scores, to meta-analysis. Some of these approaches make more demands upon the conceptual and interpretative skills of the reviewer than others and, therefore, contain more scope for error and bias than others, although all approaches are subject to the fallibility of the synthesisers. It is important that the reliability of all syntheses be tested, for they are the main ways in which assessments can be made about the accumulation ar, d development of researchbased knowledge.