This paper presents an approach for investigating in a predictive way potential disruptive effects of interruptions on task performance in a multitasking environment. The approach combines previous work in the field of interruption analysis, formal description techniques for interactive systems and stochastic processes to support performance analysis of user activities constrained by the occurrence of interruptions. The approach uses formal description techniques to provide a precise description of user tasks, and both system and interruptions behavior. The detailed mechanism by which systems and interruptions behave is first described using a Petri nets-based formal description technique called Interactive Cooperative Objects (ICO). The use of a formal modeling technique for the description of these three components makes it possible to compare and analyze different interaction techniques. In particular, it allows us to determine which of the system states are most affected by the occurrence of interruptions. Once composed together, models describing the system, user tasks and interruptions behavior are transformed into PEPA models (i.e. Performance Evaluation Process Algebra) that are amenable to performance analysis using the PRISM model checker. The approach is exemplified by a simple example that models two interaction techniques for manipulating icons in a desktop environment.