Current economic conditions and the increasing competition for government funding are leading to an increased focus on the impact of research. Measuring the impact of research is difficult because not all impacts are direct and some can be negative or result from the identification of problems that require a non-research response. The time between the performance of research and when its benefits become apparent can be significant, unpredictable and differ for different kinds of research. Attempts to measure impact, especially if these are ongoing, can distort behaviours such that they might diminish the probability of research reaching its maximum impact. A rigorous assessment of research impact has also to develop appropriate counterfactuals and consider opportunity costs. There is a wide range of methods that it is possible to use to evaluate research impact. They can operate at the level of individual projects or programs, institutions and nations. Each method has its own characteristics and advantages. While different methods can appeal to different target groups, none is complete in itself and none offers unambiguous or certain results. Measuring the impact of research is necessary and can be useful but it is important to use the results of such evaluations with care, recognising their fragility. In particular, it is important to assess impact in terms of the impacts the research aimed to achieve, not across all impacts which are possible. (Contains 10 footnotes.)