Witnesses are likely to describe a crime many times before testifying or encountering misinformation about that crime. Research examining the effect of retrieval on later suggestibility has yielded mixed results. LaPaglia and Chan manipulated whether misinformation was presented in a narrative or misleading questions, and they found that retrieval increased suggestibility when misinformation was presented in a narrative, but reduced suggestibility when the same misinformation was presented in questions. In the current study, we aimed to address why these differences occurred. Specifically, we examined whether contextual detail and narrative coherence during misinformation exposure influenced the relation between retrieval and suggestibility. Participants watched a robbery video and some were questioned about the event afterwards. They were then exposed to misinformation presented in a narrative (Experiment 1) or questions (Experiment 2) before taking a final memory test. Testing enhanced suggestibility when the misinformation phase reinstated contextual information of the event, but not when the misinformation phase included few contextual details–regardless of whether the misinformation was in a narrative or questions. In Experiment 3, disrupting narrative coherence by randomizing the order of contextual information eliminated retrieval-enhanced suggestibility. Therefore, context processing during the post-event information phase influences whether retrieval enhances or reduces eyewitness suggestibility.