The trajectories and stability of self-reported sleep duration recorded at ages 13, 15, and 23 years on reported sleep duration at age 30 years among 1105 students (55% male) who participated in the Norwegian Longitudinal Health and Behaviour Study were examined. Questionnaire data were used to obtain demographic and sleep variables. Dichotomised short sleep duration was based on normative values and set as ≤8.5 h (age 13 years), ≤8 h (age 15 years) and ≤7 h (ages 23 and 30 years). Results indicated a significant overall reduction in total sleep duration (h per night) across age groups. Sleep duration (continuous) at age 15 and 23 years (whole group) was moderately but positively correlated with sleep duration at age 30 years (P < 0.01). When split by sex, at age 15 years, this association was present among females only (P < 0.01); however, at age 23 years, this association was present in both male and females (both P < 0.001). Categorical short sleep at age 23 years (whole group) was associated with short sleep at age 30 years (unadjusted odds ratio = 3.67, 95% confidence interval 2.36–5.69). Following sex stratification, this effect was significant for both males (unadjusted odds ratio = 3.77, 95% confidence interval: 2.22–6.42) and females (unadjusted odds ratio = 2.71, 95% confidence interval: 1.46–5.04). No associations were noted for categorical short sleep at ages 13 or 15 years, and subsequent short sleep at 30 years. Habitual short sleep duration during middle adulthood is not sustained from the time of early adolescence. Rather, these trends appear to be formed during early adulthood.