In this talk, I present results from eye-tracking studies on the use of depicted events in human incremental syntactic disambiguation. Confirming the results of prior studies that used highly non- stereotypical stimuli, we found an even earlier effect for plausible event scenes. For scenes that were not inspected before the utterance, the effect was delayed. Implications of the time course of disambiguation for the accessability of past event information are discussed. In the end, I give an outlook on future experiments investigating the role of visual working memory in an account of situated language comprehension.