Pervasive computing introduces complex systems consisting of many users, sensors, and applications that should react to context data, provide services, and manipulate devices in a predictable and reliable manner. Context data sensed from the environment is largely uncertain due to lack of precision and
imperfect or faulty sensors. Uncertainty is generally dealt with at the level of individual context data. Due to the difficulties associated with catering for data of such fine granularity in applications, the environment can be divided into larger chunks of context called situations. Situations, comprising of finer-grained
events in the form of context data, warrant a different approach to dealing with uncertainty. Furthermore, we demonstrate that the uncertainty threshold of an event that triggers a behaviour is determined by the severity of the behaviour, making this task non-trivial. In this paper we detail an approach to dealing with
uncertainty at the level of situations that takes into account the severity of the behaviour that it is triggering.