Hypertension is a major public health problem in both developed and developing countries. Ample evidence supports the concept that the roots of essential hypertension extend back to childhood. However, the long-term natural history of blood pressure in children is not well understood. Data are particularly sparse on the transitional period from late adolescence to early adulthood. This community-based study in China examined the sex-specific pattern of blood pressure and its major determinants between age 3 and 24 years. This is one of the first studies on blood pressure in rural Chinese population that accounts for 85% of the total Chinese population. Blood pressure, height, weight (via direct measurement); and related epidemiological data (via an interview) were obtained from all eligible subjects. A generalized additive model was used to explore independent relations between blood pressure and relevant covariates. The sample for analyses included a total of 14,277 subjects (7,244 males, 7,033 females). Systolic blood pressure increased linearly with age to an apparent plateau around the age of 15 years in girls and 19 years in boys. Diastolic blood pressure essentially paralleled systolic blood pressure. Until the age of 10 years both systolic and diastolic blood pressure were similar in boys and girls; thereafter, values for boys surpassed those for girls with an average difference of 9 mmHg and 4 mmHg for systolic and diastolic pressure, respectively, at 20 years of age. This sex difference appeared to persist into early adulthood. The independent relation between blood pressure and each of the three major determinants (age, height, and weight) is nonlinear over the age range studied. Sex-specific regression models with linear terms for age, height and weight were developed for each of four age groups (3-9, 10-14, 15-19, 20-24 years). The regression equations can be expanded to incorporate and assess other risk factors and interactions. The 5th, 10th, 90th and 95th percentiles based on the percentage of predicted values can serve as the cutoffs for low, normal-low, normal-high, and high blood pressure in this population.