In the current experiments, we attempted to elicit nonbelieved memories (NBMs) using the Deese/Roediger-McDermott (DRM) false memory paradigm. Furthermore, by using this approach, we explored the consequences of nonbelieved true and false memories. In Experiments 1 and 2, participants received several DRM wordlists and were presented with a recognition task. After the recognition task, participants' statements were contradicted by giving them feedback about true and false items. In this way, we succeeded in eliciting nonbelieved true and false memories. In Experiment 2, participants were also involved in a modified perceptual closure task after receiving belief-relevant feedback. In this task, participants received degraded visual representations of words (e.g., false and true) that became clearer over time. Participants had to identify them as fast as possible. We also measured dissociation, compliance, and social desirability. We found that undermining belief had contrasting consequences for true and false memories. That is, nonbelieved true memories were identified more slowly whereas nonbelieved false memories were identified more quickly. We did not find any relation between our individual differences measures and the formation of NBMs.