Every form of learning draws upon a specific view of the world, on a particular construction of the environment as “reality”. Experiential learning is based on the placement of events within these patterns of interpretation. 1 A placement of this kind is plunged into crisis, when it is no longer possible to establish any convergence within a paradigm, about what has “really” happened. In the following, two large-scale accidents will be examined, which featured precisely such a loss of evidence. They represent a category of events that tend to lead to problems in the stabilization of consistent patterns of interpretation, due to their specific production of uncertainty. What we are referring to here, are contamination disasters that emanate from large-scale technological plants. This particular type serves well as an example of how starting points for learning processes are socially constructed by the affected groups of stakeholders. For this purpose, two historical events provide the basis of the analysis: the dioxin emissions of Seveso (Italy) in 1976, and the accident in a nuclear power station near Harrisburg (USA) in 1979. The focus is placed on the structures of interpretation used by each of the stakeholder groups, their subsequently derived action strategies, and the question of whether their handling of the respective disaster can be qualified in terms of learning processes. The argumentation applied here makes fruitful use of Michel Foucault’s concept of biopolitics (1977).