Data from the screening programme in British Columbia are used to test hypotheses about the natural history of cervical cancer, especially about progression and regression of preclinical lesions (dysplasia and carcinoma in situ). Three models are considered. A model without regression does not give an adequate fit of the data (P less than 0.001), and results in an implausible estimate of 33 years for the mean duration of pre-invasive lesions. A model with an equal regression rate at all ages still does not result in a good reproduction of the data. A good fit is achieved for a model with different regression rates in lesions that develop under and over age 34. Under age 34, 84% of the new lesions will regress spontaneously, with a 95% confidence interval of 76-92% regression. Over age 34, we estimate that 40% of the new lesions will regress. The average duration of dysplasia + CIS is 11.8 years, and the sensitivity of the Pap-smear is 80%. It is concluded that a considerable proportion of pre-invasive lesions in young women do not progress. The findings about progression and duration of pre-invasive lesions do not support the still prevailing tendency of frequently making Pap smears in young women.