This register-based longitudinal study of 392 969 Norwegians examined associations between birth order, gender and educational attainment at age 25 years within families (fixed effects regression) and between families (ordinary OLS regression). Data were retrieved from national registers for births of mothers with single births only and a first birth 1967-1976. Mean education was 12.61 years (SD 2.09). In within-family analysis education decreased with increasing birth order, compared to first born, birth orders two and three were associated with mean decreases of 0.24 years (95% confidence interval [CI] 0.21, 0.26) and 0.32 years (95% CI 0.27, 0.37), respectively. Females had an educational advantage of 0.29 years (95% CI 0.28, 0.31). Birth order effects were moderately stronger for females. With few exceptions, the effects showed similar patterns for subsets of families with different characteristics. Between-family analysis showed that social rather than biologic rank was associated with educational attainment, and that only children and last second children had moderately lower education than their counterparts from larger families. A substantial part of the birth order effect on education was mediated by intellectual performance (General Ability score at conscription). The birth order effect was small, but could have some impact on educational attainment at the population level. (Contains 6 figures and 4 tables.)