Summary. Data from the Office for National Statistics Longitudinal Study are used to investigate the effect of mobility between occupationally defined social classes between 1991 and 2001 on health inequality in men and women. Logistic regression models related movement into more or less advantaged employment conditions to limiting long-term illness in 2001, controlling for social class in 1991 and 2001. When class in 1991 was controlled (‘class of origin’) those who moved into more advantaged social classes were least likely and those moving into less advantaged classes most likely to report a limiting illness. However, when social class in 2001 (‘class of destination’) was controlled, those moving from less to more advantaged positions were most likely to report limiting illness. The same patterns were seen in women. This means that social mobility did not increase the extent of health inequality over the time period that was observed, but rather served to constrain or dilute it. The results are interpreted in terms of an accumulation model of health inequality, and the policy implications are discussed.