Genetic factors affect many aspects of sleep, such as sleep length. We investigated the contribution of genetic factors to stability and change of sleep length among adults over a 15-year period. In this representative follow-up study we used the Finnish Twin Cohort as the study population. Questionnaire surveys were performed in 1975 (response rate 89%, 11 041 twin pairs; age ≥18 years), 1981 (84%, 9323; ≥24 years) and 1990 (77%, 4507; 33–60 years). Sleep was categorized as short (<7 h), average or long (>8 h). Pairwise similarity in monozygotic and dizygotic pairs was examined at each survey by age group and sex. Quantitative genetic modelling was used to estimate cross-sectional and longitudinal genetic effects. The proportion of variance in sleep length at one point in time that was accounted for by genetic effects was very stable over the study period, being 0.31 in 1975, 0.32 in 1981 and 0.30 in 1990. Longitudinal genetic modelling indicated that the correlations of genetic effects between the three measurement points were high: 0.85 between 1975 and 1981; 0.93 between 1981 and 1990; and 0.76 between 1975 and 1990. Despite a high contribution of environmental effects, their correlations over time were modest: 0.31 between 1975 and 1981; 0.33 between 1981 and 1990; and 0.18 between 1975 and 1990. In conclusion, genetic factors have a modest but stable effect on the evolution of sleep length over a long time span in adults. Multiple measures are a more robust basis for genetic analyses than a single cross-sectional measure.