Witnesses are likely to describe an event to a police investigator or 911 operator soon after the event and prior to any exposure to misinformation. Recent studies have found that recalling an event can increase people's suggestibility (e.g., Chan, Thomas, & Bulevich, 2009) while other studies have reported that retrieval can reduce subsequent eyewitness suggestibility (e.g., Pansky & Tenenboim, 2011). In this dissertation, I examined whether differences in the way misinformation is presented can modulate the effects of testing on suggestibility. Participants watched a video of a robbery and some were questioned about the event immediately afterwards. Later, participants were exposed to misinformation in a narrative (Experiment 1a) or in questions (Experiment 1b). Consistent with previous studies, testing increased suggestibility when misinformation was presented via a narrative; however, when misinformation was presented in questions, testing decreased suggestibility. In Experiments 2 and 3, I sought to uncover why the retrieval-enhanced suggestibility (RES) effect was eliminated when misinformation was presented in questions. Experiment 2 was designed to address whether differences in the presentation duration of misinformation can account for the opposite patterns of results in Experiment 1 - they cannot. In Experiment 3, I manipulated whether a) misinformation was presented in questions or a narrative and b) the amount of contextual detail presented with the misinformation. I found that an RES effect was present only when misinformation was embedded in rich contextual details, regardless of whether misinformation was presented in questions or a narrative. Consistent with these data, previous studies that have found an RES effect have used misinformation narratives that included many additional contextual details (e.g., Chan & LaPaglia, 2011), whereas work showing a testing benefit has consistently presented misinformation in relative isolation (e.g., LaPaglia & Chan, 2012). These results suggest that the way in which misinformation is presented determines whether initial testing enhances or reduces suggestibility. The main findings are discussed within the context of discourse comprehension and narrative persuasion.