This paper examines several mathematical models that have been developed to investigate cervical cancer screening. Each model is discussed from the perspective of the inferences that can be drawn about the natural history of cervical cancer. The modelling analyses of Barron and Richart, Coppleson and Brown, Albert, Knox and Habbema and colleagues are examined in an attempt to gain insight about the transition and duration properties of the preclinical stages of cervical cancer. The picture that emerges is one of a complex preclinical natural history. There is a clear indication of regression of carcinoma in situ, as well as an age-dependence of transition probabilities and duration of disease status.