Significant progress has been made in recent years to improve the quality of the evaluation of international aid. Increasingly, this includes an interest in improving the way evaluations are used to improve policies and programmes. However, the prevalence of symbolic use--a phenomenon that is often mentioned but rarely studied--reflects an uncomfortable gap that continues to emerge between evaluation practice and rhetoric. Drawing on literature from several contexts, this paper argues that efforts to improve evaluation use in this complex sector must be, post-Busan, more firmly rooted in evidence describing how its principals and agents actually behave.