Participants took part in a two-stage investigation examining the changes in memory for conversation over time. The impact of memory expectancy and mode of recall were also examined. Participants could recall only about 10% of their conversations immediately after the conversations. One month later this figure had dropped to 4%. Examination of recall protocols revealed that after a one month delay, participants recalled less content and reported more descriptive statements, made more inferences, and were less accurate than when they had recalled immediately. In addition, expectancies about memory and mode used to report recollections affected the amount and type of information reported. Finally, participants recalled more of their partner's contributions to the conversations than their own contributions.