The proportion of people with mobility limitations (difficulties with running, walking and stairs) decreased between 1968 and 1991 in the Swedish population aged 18-75. The distribution of predictors of late life morbidity, e.g., social class, health behaviour and childhood conditions, also changed during this period. This study explored whether the changes in these predictors over time were related to the decrease in the proportion of the population with mobility limitations. In two nationally representative interview samples (n=4468) from 1974 and 1991 of persons aged 45-71 the odds for limitations were cross-sectionally compared in ordered logistic regression models. In addition, predictors for the mobility outcome in 1974 were collected from an earlier 1968 survey and predictors for the 1991 outcome were collected from 1981. In 1974 the odds for limitations in the population was 50% higher than in 1991. Had the population composition regarding social class and housewives in 1991 been identical to 1974, the odds for limitations would have been similar in 1974 and 1991. Period improvement in social classes with poorer mobility also contributed to the overall period improvement. Health behaviours were examined as possible mediating factors. The increase of physically active people between 1968 and 1974 was related to the period improvement in mobility between 1974 and 1991. Smoking showed an increased association with mobility limitations during the period, indicating that mobility improvement would have been greater if everyone had been a non-smoker. Results indicate how sensitive disability rates may be for cohort or period effects.