The memory advantage of bimodal encoding in the retrieval of isolated stimulus have been extensively studied, but researchers have not investigated this advantage for naturalistic events. This study reports both behavioral and functional magnetic resonance imaging (fMRI) data on whether memory advantage of bimodal encoding exists for retrieval of naturalistic events. In Experiment 1, participants took memory tests after learning naturalistic events via three different encoding modes (1) text reading, (2) story listening, and (3) video watching. The results showed that, at immediate recall, participants made few errors in the text reading and video watching conditions than the story listening condition; at delayed recall, these differences disappeared. In Experiment 2, participants similarly read texts, listened to stories, watched videos, and underwent fMRI scanning during a recall task. Our fMRI data showed stronger activation in the right angular gyrus for retrieving bimodal naturalistic events (i.e., video watching) than unimodal ones (i.e., text reading and story listening). These results suggest a memory advantage of the bimodal encoding for retrieving complex episodic memories, given the rich, multisensory events across encoding modes over time.