Recent research on the overdistribution principle implies that episodic memory is infected by conjunction illusions. These are instances in which an item that was presented in a single context (e.g., List 1) is falsely remembered as having been presented in multiple contexts (e.g., List 1 and List 2). Robust conjunction illusions were detected in source-monitoring designs in which conjunctive probes ("Was bagpipe presented on List 1 and List 2?") were added to the traditional nonconjunctive probes ("Was bagpipe presented on List 1?"). In Experiment 1, the levels of those illusions were comparable to what would be expected on the basis of results from prior overdistribution experiments. In Experiments 2 and 3, conjunction illusions were neither eliminated nor reduced by a manipulation that should have had such effects if the illusions are by-products of subjective differences in retrieved memory support. Also, conjunction illusions sometimes rose to the level of conjunction fallacies: In certain conditions, subjects thought that items were more likely to have occurred in all the presentation contexts than in any single context, which is impossible. Two general approaches to explaining overdistribution, representational accounts and retrieval accounts, are considered.