Abstract: Policy-oriented learning is important for decision-makers who are confronted with complex societal problems. Learning can be supported by linking policies to implementation actions and expected outcomes, which then in turn can be monitored. Unfortunately, the use of monitoring information in decision-making is often diffuse and indirect. If one accepts the importance of, and the practical limitations to the use of monitoring information in relation to policy-oriented learning, what can reasonably be expected? To what extent do actors in a policy process adhere to these expectations regarding their collection and use of information to support decision-making over time? We have studied these questions for two policy processes in coastal management, one in the Netherlands and one in South Africa; both related to infrastructures and spanning a 25-year period. Results show that actors who were driving the original policy decisions on coastal infrastructures devised monitoring strategies that addressed the issues most salient to their core responsibilities. Other issues, raised by other actors in early policy decisions, were monitored less intensively, if at all. Although understandable, this omission affected the possibilities for policy-oriented learning on these other issues, and limited the information base in subsequent policy games when these issues became more pertinent. This raises questions regarding the responsibilities for monitoring design and knowledge development in multi-actor settings. [Copyright &y& Elsevier]