In a screening programme for cervical cancer, coverage of the target population is a major determinant of effectiveness and cost-effectiveness and is one of the parameters for programme monitoring recommended by the "European Guidelines for Quality Assurance". An organised screening programme was started in Turin, Italy, in 1992. Spontaneous screening was already largely present, but coverage (proportion of women who had at least a test within 3 years) was low (< 50%) and distribution of smears uneven. No comprehensive registration of spontaneous smears was available. All women were invited for the first round, independently of their previous test history. Coverage was estimated by integrating routine data from the organised programme with data on spontaneous screening obtained by interviews of a random sample of 268 non-compliers to invitation and 167 compliers. Overall (spontaneous + organised) coverage was estimated to be 74% (95% CI, 71-78%). The proportion of the target population covered as an effect of invitation was estimated to be 17% (95% CI, 15-20%). Invitations were successful in increasing coverage in previously poorly screened groups. Although 20-25% of compliers was estimated to have had further tests before the end of the round, we estimated that switching to a 3-year interval saved approximately 0.26 tests per complier. This suggests that invitations to an organised programme even to previously covered women, can be a cost-effective policy. Our method of estimating overall coverage can be useful in many other European areas where a comprehensive registration of smears is not available.