The study involved a sample of eight moderately to severely demented elderly people who used psychogeriatric day services. Counselling skills were used by the interviewer to investigate informants' recall of emotional memories. Interviews were normally carried out individually each week over a number of months. Between thirteen and twenty-five interviews with individual informants were recorded and transcribed. The data form a series of longitudinal case-studies, analysed using quasi-judicial methods (Bromley 1986), and with a grounded theory approach (Glaser and Strauss 1967). Over time, it became apparent that each case-study revealed fragmented pieces of an informant's personal narrative. The emotions associated with their past experiences appeared to provide a strong cue to recall and formed a significant feature of their accounts as well as providing all informants with narrative identity. For some informants, this sense of narrative identity began to dissolve as their illness progressed and their stories faded from memory. For other informants, whose memories were not so devastated by their illness, it remained with them. Although outcomes varied for all informants, all experienced varying levels of increased well-being. The data may have important therapeutic implications for the care of dementia patients through the development of reminiscence work. Attention is drawn to some theoretical implications for understanding the relationship between emotion, memory and dementia.