During text comprehension, readers maintain a mental representation of the described events in a situation model. When new information is encountered, this model must be updated or a new model created. The additional processing that this requires is usually reflected in increased reading times for sentences that convey a shift along one of several situation model dimensions. However, these increases have not been consistently found for all variables. This thesis addresses these inconsistencies by examining the role of causal connections in the creation of event boundaries. That is, to what degree do the processing differences observed with event boundaries reflect model updating and to what degree do they reflect an unexpected causal break in the flow of narrative action? In three experiments, participants read stories in which an event shift was foreshadowed or not. The results of Experiment 1 revealed that readers do not expect event shifts, but providing prior knowledge of a shift eliminates the difference in expectedness. Experiment 2 showed that providing foreshadow information does not affect the realization that a shift has taken place. Finally, Experiment 3 found that reading times were slower when a shift was not foreshadowed but were not different in the other conditions. Overall, the results support the idea that causal connections play a role in the amount of online processing required to mentally process an event shift.