Focusing on how people select topics to mention when reporting on a personal event, a study investigated the hypotheses that people would be more likely to mention topics (1) of high prior concern than topics of low concern and (2) with an unusually good or bad outcome than topics with an ordinary outcome. Subjects, 12 couples expecting babies, responded to prenatal and postpartum questionnaires and tape-recorded the phone conversations in which they announced the birth of their baby. Results supported both hypotheses, as well as the notion that topic selection depends on memory search. Findings showed that topics chosen for spoken and written discourse are generated by accessing information that is salient in long-term memory. Alternately, these results suggested that discourse topics are chosen for their relevance and information value. Findings further implied that writers, like speakers, depend on memory processes to generate ideas. It is important, in the revision process, to compensate for memory bias by allowing the writer to focus on the concerns of the reader. Overall, results showed that, given knowledge of prior concerns and of what actually happened, it was possible to predict which speakers would address which general topics in naturally occurring conversations. However, there was considerable variety in how subjects handled the topics, i.e., with humor, suspense, or philosophically. (JD)