Non-progression and duration of preclinical neoplastic lesions of the cervix uteri were studied using screening data from a previously unscreened population, Maribo County, Denmark (1966-82). To estimate regression rates, the incidence of clinical cancer before the screening programme was related to the prevalence and incidence of preclinical lesions estimated from the detection rates of first smear and third and subsequent smears respectively. Duration was estimated from the time lag between the cumulative incidence of preclinical lesions and the combined cumulative incidence of clinical cancer and the estimated 'incidence of regression'. Of all preclinical lesions in women aged 25-50, 24% progressed, 39% regressed and 38% remained. Even if we assume no onset of preclinical lesions above age 50, we estimated that 48% of the preclinical lesions would not progress to clinical cancer in the women's lifetime. The estimated mean duration of preclinical lesions was 16 years. In Maribo County during the 1970s, the positive rate (1.6%) was low compared with current rates in several countries. We conclude that the detection of non-progressive lesions was outweighed by the prevention of clinical cancer.