One of the most controversial explanations of class inequalities in health is the health selection hypothesis or drift hypothesis which suggests there is a casual link between the health status of individuals and their chances of social mobility, both inter- and intra-generational. This study tests this hypothesis, and tries to answer three related questions: (a) to what extent does health status influence the chances of intra-generational mobility of individuals? (b) what is the impact on health inequalities of the various kinds of social mobility (both mobility in the labour market and exit from employment)-do they increase or reduce inequalities? (c) to what extent does health-related intra-generational social mobility contribute to the production of health inequalities? The data analysed in this paper were drawn from the records of the Turin Longitudinal Study, which was set up to monitor health inequality of the Turin population by combining census data, population registry records and medical records. Occupational mobility was observed during the decade 1981-1991. To evaluate the impact of the various processes of social mobility on health inequalities, mortality was observed over the period 1991-1999. The study population consists of men and women aged 25-49 at the beginning of mortality follow-up (1991), and registered as resident in Turin at both the 1981 and the 1991 censuses (N = 127,384). Health status was determined by observing hospital admission. For the purpose of the study healthy individuals were those with no hospital admissions during the period 1984-1986, while those admitted were classed as unhealthy. Social mobility in the labour market was measured via an interval data index of upward and downward movements on a scale of social desirability of occupations, designed for the Italian labour force via an empirical study carried out by de Lillo and Schizzerotto (La valutazione sociale delle occupazioni. Una scala di stratificazione occupazionale per l'Italia contemporanea, Il Mulino, Bologna, 1985). Movement out of the labour market was described by a discrete variable with four conditions: employed, unemployed, early retired and women returning from work to the housewife status. The relationship between health status and occupational mobility was analysed via analysis of variance and multinomial logistic regression. Health inequalities were measured by the ratio of standardised mortality rates in the unskilled working class and the upper middle class. The study found a weak relationship between health status and occupational mobility chances. Decidedly stronger was the impact on occupational mobility of gender, education and "ethnicity" (being born in the South of Italy). The relationship between occupational mobility and health takes two different forms. Occupational mobility in the labour market decreases health inequalities; occupational mobility out of the labour market (early retirement, unemployment, housewife return) widens them. The maximum contribution health-related intra-generational social mobility can make towards health inequalities was estimated at about 13% for men.