Many popular models conceptualize working memory as consisting of three or four discrete slots or bins. This conceptualization, however, has been seemingly refuted by Bays and Husain (2009), who reported perfect performance on a working memory task with a large number of very simple items. We show, however, that this perfect-performance result likely reflects a design flaw rather than mnemonic structure. The flaw is that the test array itself in Bays and Husain's study provides information about the correct answer without recourse to working memory. We show perfect performance on eight items for 18 participants when this information is present. We show that performance is poorer, however, when this information is removed. Hence, the Bays and Husain result does not threaten models that stipulate that working memory is composed of limited slots.