The respective contribution of total, daytime and nighttime sleep duration in childhood obesity remains unclear. To assess the longitudinal association between developmental trajectories of sleep duration and BMI z-score in early childhood. Data were from the Melbourne INFANT program, a prospective cohort with 4-month-old infants being followed-up until age 60 months (n = 528). Sleep duration (total, daytime, nighttime) and BMI z-score were measured using questionnaire at ages 4, 9, 18, 43 and 60 months. Group-based trajectory modelling was used to describe longitudinal trajectories from ages 4 to 60 months. Multivariable logistic regression was conducted to assess the association between sleep duration and BMI z-score trajectories. Three nighttime sleep duration trajectory groups were identified: "Long stable" (10.5 to 11.0 hours, 61%), "catchup long" (8.0 to 11.5 hours, 23%) and "short stable" (8.7 to 9.8 hours, 16%) nighttime sleepers. BMI z-score trajectory groups were classified as "low-BMIz" (-1.5 to -0.5 unit, 21%), "mid-BMIz" (-0.5 to 0.5 unit, 58%) and "high-BMIz" (0.8 to 1.4 unit, 21%). With adjustment for child and maternal covariates, both "catchup long" (OR 3.69 95%CI 1.74, 7.92) and "long stable" nighttime sleepers (OR 4.27 95%CI 2.21, 8.25) revealed higher odds of being in the "mid-BMIz" than the "high-BMIz" group. By contrast, total or daytime sleep duration trajectories were not associated with BMI z-score trajectories. Longer nighttime, but not total or daytime, sleep duration was associated with lower BMI z-score trajectories in early childhood. Our findings reinforce the importance of nighttime sleep for healthy body-weight development in early childhood.