The analysis of mortality in urban settings for the Cities of Florence (1991-95) and Leghorn (1987-95), based on data from the Tuscany Longitudinal Study, is reported in the present paper. The data came from a census-based cohort study, all residents at the census day 1981 (Leghorn) or 1991 (Florence) being enrolled and followed-up by automated procedures of record-linkage. The cause of death certificate had been eventually collected by the Regional Mortality Register. For each city, internally standardized mortality ratios (SMR) had been calculated by sub-urban areas (city sectors or wards). The analysis was restricted to age groups > 15 years to have interpretable results on socio-economic variables derived from census questionnaires. Bayesian estimates (Besag, York e Mollié) of mortality relative risks had been calculated to overcome extra-variability of SMRs. In the city of Florence two wards showed about 10% excess risk for overall mortality. In the city of Leghorn one sector was at higher risk while one showed a significant lower mortality. For both cities such risk gradients were still present after adjustment for deprivation index at individual level.