Corrections are not always effective at reducing belief in misinformation. Negated corrections, which state a piece of information is not true, may only be effective at inhibiting information an observer has already encountered. We compared the effectiveness of negated corrections and replacements while manipulating initial exposure to a target concept. Subjects read one (Experiment 1) or six (Experiment 2) passages presenting a target concept (e.g., blue car) or not, followed by a negated correction (e.g., not blue), replacement (e.g., red), or no correction, then answered open-ended questions which were scored for mentions of the target concept. When subjects were exposed to the target concept, negated corrections reduced mentions of the misinformation relative to no correction; however, when not exposed to the concept, negated corrections increased mentions relative to no correction. These results demonstrate that negated corrections can increase belief in misinformation when observers have not been exposed to the misinformation.