We evaluate the persistence of social inequalities in overall mortality or mortality by causes (lung cancer, liver cirrhosis; AIDS and overdose) in the Tuscany Longitudinal Study (SLTo), a record linkage-study on the census population identified at the 1981 and 1991 censuses in Leghorn (1981: 175,741 subjects; 1991: 167,512 subjects), and 1991 in Florence (403,294 subjects), Central Italy. The census data allow an evaluation of socio-economic status of each subject, using variables such as education or occupation, or constructing indexes inclusive of deprivation indexes. Follow-up is from the census up to 1995 and for specific causes of death from 1987 (Leghorn) or 1991 (Florence). Estimates of risk are computed comparing rates of mortality among socio-economic groups, by means of poisson regression models, or by means of Standardized Mortality Ratios using the indirect method. The causes of deaths have been selected mainly because explained by known and strong determinants. As far as overall mortality, the results suggest the persistence of gradients by social class, more often negative among males. Mortality from lung cancer has a strong negative social class gradient among males, and a divergent gradient among women in the two towns, which is interpreted as the effect of a different prevalence of smoking by period, social class and sex. High RRs among lower socio-economic groups have been detected for liver cirrhosis mortality. Excess mortality of AIDS and overdose, an expression of the recent drug crisis, concentrates on the lower social strata and in young adults. Whereas mortality from AIDS has been detected among both sexes, deaths from overdose are occurring among males only. Being AIDS and overdose recent diseases, they stress the persistence of social inequalities over time.