Research ethics is an integral part of research, especially that involving human subjects. However, concerns have been expressed that research ethics has come to be seen as a procedural concern focused on a few well-established ethical issues that researchers need to address to obtain ethical approval to begin their research. While such prospective review of research is important, we argue that it is not sufficient to address all aspects of research ethics. We propose retrospective review as an important complement to prospective review. We offer two arguments to support our claim that prospective review is insufficient. First, as currently practiced, research ethics has become for some a ‘tick box’ exercise to get over the ‘hurdle’ of ethics approval. This fails to capture much of what is important in ethics and does not promote careful reflection on the ethical issues involved. Second, the current approach tends to be rules-based and we argue that research ethics should go beyond this to develop people’s capacity to be sensitive to the relevant moral features of their research, their ethical decision-making skills and their integrity. Retrospective review of a project’s ethical issues, and how they were addressed, could help to achieve those aims better. We believe that a broad range of stakeholders should be involved in such retrospective review, including representatives of ethics committees, participating communities and those involved in the research. All stakeholders could then learn from others’ perspectives and experiences. An open and transparent assessment of research could help to promote trust and understanding between stakeholders, as well as identifying areas of agreement and disagreement and how these can be built upon or addressed. Retrospective review also has the potential to promote critical reflection on ethics and help to develop ethical sensitivity and integrity within the research team. Demonstrating this would take empirical evidence and we suggest that any such initiatives should be accompanied by research into their effectiveness. Our article concludes with a discussion of some possible objections to our proposal, and an invitation to further debate and discussion.