Our research goal was to explore limits to the impact of misinformation on eyewitness memory. In two studies using the standard eyewitness misinformation paradigm, we showed participants a series of slides accompanied by narratives including a total of 15 misleading details. After an initial cued recall test (allowing us to capture the size of the misinformation effects), we asked participants to retrospectively indicate for which details they had noticed a discrepancy between original and post-event detail information. We found (in Study 1; N = 81) that the misinformation effect (operationalised through both misinformation endorsement and accuracy for original details) was significant and stable up to about seven detected discrepancies and clearly declined with higher numbers of discrepancy detection, suggesting that there are indeed limits to how much misinformation people are willing to take. In Study 2 (planned N = 221), we extended the generality of this finding using a different set of materials and a slightly different procedure, replicating the essential findings of Study 1.