BACKGROUND: Individuals with severe expressive aphasia often have difficulty sharing adequate amounts of specific information to sustain topical conversations. AIMS: This single subject experiment investigated whether graphic representations of topics increased conversational duration, number of information exchanges, proportion of participant initiations, and percentage of successful communication exchanges during dyadic conversations involving a communicator with severe, nonfluent aphasia (SD). METHODS AND PROCEDURES: SD conversed with two partners about personal and current events in ''no treatment'' and ''graphic topic-setter'' conditions. He also participated in an additional ''instruction'' condition with Partner 2. Conversations were videotaped, transcribed, and coded for the dependent variables. Results were averaged by condition, graphed, and analysed for statistical significance using randomisation testing. OUTCOMES AND RESULTS: Despite variability in conversational parameters across sessions, mean data from each condition revealed that graphic topic setters increased the average duration of interactions and mean number of communication exchanges per topic across both partners. Proportion of initiations increased significantly with graphic context for Partner 1 but not Partner 2; this effect was more pronounced for current event topics than personal events. Information transmission was significantly more successful when graphic context was present with both partners, but differences were more noticeable with Partner 1. CONCLUSIONS: Participants appeared better able to co-construct conversations when graphic topic setters were available to supplement the natural communication signals of a communicator with severe aphasia. Clinical implementation issues are discussed.