This paper explores the temporal course of discourse updating after reading counterfactual events. To test the accessibility to discourse information, readers were asked to identify probes related to initial events in the text, previous to the counterfactual, or probes related to the critical counterfactual events. Experiment 1 showed that 500 ms after reading events in counterfactual format, initial events were more accessible than after reading the same critical events in factual format. This suggests that discourse updating occurs in factuals, but not in counterfactuals. However, the critical events were equally accessible in both formats, indicating that the alternative "as if" scenario was also activated in counterfactuals. Experiment 2 demonstrated that the initial events continued accessible 1500 ms after reading counterfactuals, but the alternative "as if" scenario becomes less accessible. In sum, the experiments indicate that the realistic meaning of counterfactuals prevents discourse updating at both the early and the late stage, whereas the "as if" meaning was only activated at the initial stage. (Contains 1 table and 2 figures.)