Current theorizing suggests that critical lures in the Deese/Roediger-McDermott (DRM) procedure are often falsely remembered because they have received considerable relational processing (e.g., spreading activation or encoding of gist information). We used a repeated-testing paradigm to assess the amount of item-specific and relational processing given to the list items and the critical lures. Research has shown that items receiving itemspecific processing are more likely to be recovered across successive tests. They are also output more slowly but more steadily throughout the recall period. In two experiments, we manipulated the processing performed on list items and then used item gains and cumulative recall curves to assess the amount of item-specific and relational information encoded for both list items and lures. The results suggest that increasing the relational processing of list items increased item-specific processing of lures, whereas increasing item-specific processing of list items decreased item-specific processing of lures. We conclude that critical lures are typically rich in item-specific information, relative to list items.