The major aim of this study was to describe the conversational abilities of a group of severely injured TBI speakers, at a minimum of 2 years post-injury. The association between conversational impairment and (a) selected measures of executive function and (b) psychosocial handicap was also examined. Twenty-four members of the group of 26 severely injured TBI speakers who had initially been assessed between 3 and 6 months post-injury were reviewed at a minimum of 2 years post-injury (mean = 2 years, 10 months). At initial assessment, TBI speakers were compared with non-TBI orthopaedic patients and with a group of university students. At follow-up, however, they were compared only with the orthopaedic patients. Conversational assessment was carried out using a modified form of Damico's Clinical Discourse Analysis. As a group, the TBI speakers' conversational abilities did not improve over time. There was, however, a subgroup (n = 8) of speakers who did improve, and these could be distinguished by greater initial severity of injury and a significantly longer period of speech-language pathology intervention than the speakers who either remained the same or worsened over time. Modest associations between conversational discourse skills and measures of executive function and a measure of psychosocial handicap were identified. These findings indicate that disruptions in conversation persist into the longer term. More assiduous efforts may need to be made to (a) identify subtle discourse changes in the early months after injury and (b) engage TBI speakers in speech-language pathology services. Such services are also required over a longer time frame, in community-based models of service provision.