Three experiments were conducted to investigate the process by which location information in narratives is represented in memory and the nature of the resulting memory representation. In Experiments 1 and 2, the results of a recognition task demonstrated that location shifts led to an immediate decrease in the accessibility in memory of protagonists' former locations. In Experiment 3, regardless of the amount of backgrounding after the last mention of the critical location ("the forest"), reference to an implied, location-typical entity ("the trees") was read equally fast as long as the protagonist remained in that location. In contrast to previous findings, we conclude that when location information is salient in a narrative it is included in readers' situation models, being updated immediately and remaining highly accessible even several sentences after it was last mentioned.