Studies of personal relationships have often been conducted in the laboratory, on the self-report questionnaire or in the interview. In contrast, this paper studies relationships via a corpus of thirty British 'speed-dating' encounters between 30-45 year old heterosexual couples, in which talk about previous relationships was pervasively relevant. The analysis examines how talk about prior relationships, and current relationship status, was occasioned and accounted for. The first section shows that, in the overall structure of the date, talk about relationship histories was located after talk about other matters (e.g., occupation, place of residence). Second, relationship history questions contained design features for managing the delicacy of answering them (e.g., trail-off 'or' turn endings), as well as paired categorial items (e.g., a question about 'children' was answered in terms of one's 'marital' status and vice versa). In the final section, the analysis shows that and how participants treat some relationship histories as more accountable than others (e.g., being 'never married'). The analysis revealed a more general set of accountabilities: of being 'single', of being previously unsuccessful in relationships, of being unable to meet people in 'natural' settings, and, therefore, for attending speed-dating events. Overall, the paper demonstrates the importance of examining the richly detailed brief encounters of social life, in order to better understand people's understandings of, troubles with, and goals for their personal relationships.