Abstract. Instructional design is not a linear process: designers have to weigh the advantages and disadvantages of alternative solutions, taking into account different kinds of conflicting and changing constraints. To make sure that they eventually choose the most optimal one, they have to keep on collecting information, reconsidering continuously whether their own decisions are still justified in the light of the latest insights. We have studied the role of iteration during instructional design. For our research, we have used an ISD-based method for the specification of training simulators. During our empirical evaluation study, we introduced five events that are likely to cause iteration. The results show that the quality of the designs is not directly related to the amount of iteration. We conclude that there are different kinds of iteration, triggered by different kinds of actions and events. We propose a list of triggers for iteration some of which originate from outside (new information, new opinions/arguments and acquisition procedures); others are caused by, or evolve from interaction with the design process itself (discovery of missing input, need to repair errors, new insights based on work later on in the design process, and new ideas of the designers).