OBJECTIVE: To assess distress in a sample of carers who were selected from a community survey rather than recruited via community-service agencies. METHODS: A community survey was carried out on 630 people aged 75 or over living in Sydney, Australia. Informants nominated by these elderly people were divided into full carer (n = 21), partial carer (n = 187) and non-carer groups (n = 344). Informants completed the General Health Questionnaire (a continuous measure of psychiatric symptoms), the life satisfaction index (a measure of well-being) and the interpersonal bonding measure (a measure of quality of the relationship with the elderly person). Elderly participants had a medical examination, were assessed for disability and were questioned about use of services. RESULTS: Elderly people who had a full carer were more disabled and had more medical diagnoses. Full, but not partial, carers reported more psychiatric symptoms and lower life satisfaction. In multivariate analysis, the main determinant of carer distress was a relationship in which the carer felt controlled by the elderly person. CONCLUSION: When carers are selected from a population-based sample, only those who are full carers are more distressed. However, relationship factors are the most important determinant of distress.