Three experiments were designed to clarify the perplexing ability of subjects to discriminate between to-be-remembered (TBR) and to-be-forgotten (TBF) items in memory. After the presentation of each pair of words in a list, subjects were first required to solve one to four arithmetic problems and were then cued whether to remember or forget that pair. When subjects were free to use a remember (R) or forget (F) cue in any way they saw fit, their subsequent ability to differentiate TBR and TBF items was impressive, but when subjects were required to retrieve both TBR and TBF word pairs in response to an R or F cue, respectively, their subsequent ability to differentiate TBR and TBF items deteriorated. The results implicate within-list retrieval of TBR items as a potent tagging or strengthening operation that provides a basis on which those items may later be discriminated from TBF items. The "potency" of such events, in terms of their influence on later recall, depends in a clear-cut way on the "depth" (i.e., delay) of retrieval involved, but there are no such comparable effects on later recognition. It is critical to the reasonably efficient functioning of human memory that we be able to differentiate to-be-remembered information from to-be-forgotten information. In a variety of experiments on directed forgetting, in which cues denote which items are to be remembered and which items are to be forgotten, subjects have shown a remarkable ability to accomplish such differentiation, often under circumstances where items are cued arbitrarily and there is apparently no free time available to use those cues. A particular case in point is the itemby-item cuing procedure introduced by Woodward and Bjork (1971). In this procedure, items (typically, individual words)