Initial misinformation in news reports can continue to influence readers’ judgments and inferences even after a correction has been provided (Lewandowsky et al., 2012). Successful correction is cognitively demanding; readers must extract the implications of the correction and inhibit reliance on the misinformation. Hence, we assessed whether the effectiveness of a correction at reducing reliance on misinformation is associated with working memory. Participants recruited online read a news story containing misinformation that was later either corrected or not depending upon condition. Reliance on the misinformation was assessed by examining how often participants answered open-ended questions with inferences based on the misinformation. Consistent with past findings, we expected that No Correction participants would provide more inferences consistent with the misinformation than Correction participants. To measure working memory, participants completed a 2-back task where they had to determine whether the current letter in a continuous stream of letters matched the letter two previous. The total inferences generated were modeled using a regression with Correction condition, 2-back performance, and the interaction as predictors. As expected, performance on the 2-back task was a significant moderator of the relationship between Correction and total inferences generated such that better 2-back performance was associated with a greater reduction in inferences following Correction. The present study provides evidence that ability to correct misinformation encountered in a news story is moderated by working memory.