User attention is a scarce resource, and users are susceptibleto interruption overload. Systems do not reason about theeffects of interrupting a user during a task sequence. In thisstudy, we measure effects of interrupting a user at differentmoments within task execution in terms of taskperformance, emotional state, and social attribution. Taskmodels were developed using event perception techniques,and the resulting models were used to identify interruptiontimings based on a user’s predicted cognitive load. Ourresults show that different interruption moments havedifferent impacts on user emotional state and positive socialattribution, and suggest that a system could enable a user tomaintain a high level of awareness while mitigating thedisruptive effects of interruption. We discuss implicationsof these results for the design of an attention manager.