In pervasive systems, exceptions are likely occurrences, and not rare events as is usually the case in less dynamic and more predictable systems. The uncertainty and dynamism surrounding pervasive systems require new and sophisticated approaches to defining, detecting and handling complex exceptions. In this paper, we propose exception definition strategies based on "situations" - a new and extended notion of context. Our approach enables programmers to express situation-based temporal properties of exceptions effectively, which significantly improves the overall safety of the pervasive system. In this paper, we present the notion of situation and present a programming interface for the definition of situation-based exceptions. We then present two implementation algorithms and analyze and evaluate their performance and effectiveness. © 2011 IEEE.