In the presentation and discussion of epidemiologic study results, investigators often argue that their results have to be viewed as conservative due to the occurrence of presumedly nondifferential exposure misclassification. Sometimes attempts are made to correct for this supposed bias toward the null. Particularly, various dual exposure measurement strategies that may eliminate or at least reduce the bias if the exposure misclassification is truly nondifferential have gained much popularity in recent years. However, even in cohort studies, it is often questionable whether the exposure misclassification rates are completely independent of measured disease risk. In this article, I assess the effects of violating the assumption of nondifferential misclassification on the direction and magnitude of bias and on the misclassification correction procedures. Even slight violation of the nondifferentiality assumption can lead to large bias away from the null or to crossover bias, particularly if dual measurement procedures are used to correct for exposure misclassification. Consequently, inferences on the potential effects of presumedly nondifferential exposure misclassification in any given study should only be made after careful sensitivity analyses that take into account the full range of plausible misclassification rates as well as of plausible deviations from nondifferentiality.