A discourse recognition theory derived from more general memory formulations would be broad in its psychological implications. This study compared discourse recognition with some established profiles of item recognition. Participants read 10 stories either once or twice each. They then rated their confidence in recognizing explicit, paraphrased, inference, and false (lure) probes. False alarms were higher for unrepeated than for repeated stories, signifying a recognition criterion shift: Readers can apparently adjust their criterion, item-by-item, according to probe strength. Story repetition "reduced" people's recognition of story paraphrases and inferences, an outcome labeled recollection-to-reject. Functions (receiver operating characteristics) relating correct (hits) and erroneous (false alarms) acceptances bore the curvilinear, asymmetric shape observed in item memory, an outcome suggestive if not definitively diagnostic of the contributions of recollection and familiarity to discourse retrieval. Such similarities between discourse memory and item memory are consistent with the memory-based text processing framework.