Insomnia is widely reported and widely treated in general practice, yet relatively little research has focused on the natural history of the condition in primary care settings. As a result, there is at present little information to enable clinicians to assess insomnia risk, or anticipate outcomes in older general practice populations. To estimate, using 8-year longitudinal data, the risk of insomnia onset associated with selected health and lifestyle factors. Survivors from a nationally representative sample (n = 1042) of elderly people originally interviewed in 1985 were reassessed in 1989 (n = 690) and 1993 (n = 410). At the first follow up in 1989, 84 new cases of insomnia were identified (a weighted incidence rate per person per year at a risk of 3.1%; 95% CI = 2.7-3.5). In logistic regression analyses controlling for age and sex, the risk of insomnia onset was then assessed in relation to the selected factors. Three factors assessed in 1985 were significantly and independently related to incident insomnia: psychometric ratings consistent with depressed mood odds ratio (OR) = 4.41; 95% CI = 3.32-5.43); health index scores indicating lower physical health status (OR = 1.19; 95% CI = 1.06-1.31 per unit change in scale score); and moderate and low levels of physical activity (OR = 1.91 and 2.14; 95% CI = 1.91-3.62 and 2.14-3.64 respectively). However, although depressed mood represented a major risk factor, the most likely source of risk was physical rather than mental ill-health. Psychiatric, somatic and lifestyle factors significantly and independently increase the risk of insomnia in older general practice patients. In predicting incident sleep disturbance, these factors exceed in importance the age and sex of patients.