Higher recall of bizarre images relative to common images (the bizarreness effect) is consistently found when bizarreness is varied as a within-subject (mixed-list) variable. In Experiment 1, mixed lists, rather than the smaller number of bizarre sentences typically used in such lists, determined the occurrence of the bizarreness effect. Contrary to predictions from expectation-violation theory, Experiments 2 and 3 showed that manipulations designed to augment or attenuate surprise reactions to bizarre sentences had little impact on the bizarreness effect. Experiments 4 and 5 indicated that mixing affected the degree to which participants differentially encoded order information for bizarre and common items. A new account of the bizarreness effect is presented that combines considerations of distinctiveness with the differential use of order information across bizarre and common items.