Aim: To determine if subjects' cholesterol levels tracked relative to their peers from early childhood to adult mid-life. Methods: Longitudinal study using subjects from the Busselton Population Study. Data were available from 1967 on a triennial basis until 1983, and a re-survey held in 1994. The study included 1764 subjects aged 5-18 y at first measurement. Pearson's correlation coefficient, adjusted for age and survey year, was used to examine cholesterol tracking. The proportion of children who persisted with cholesterol measurements in the extreme high quartile was assessed over time before and after adjusting data for regression to the mean. The variability of the children's cholesterol level was examined by track width using the method of Porkka. Results: The correlation coefficients for tracking were from 0.35 to 0.55. Stronger correlations coincided with shorter time periods between measurements. Before adjustment for regression to the mean, 55-60% of children from the highest quartile at first measurement remained in the same quartile 27 y later. After adjustment for regression to the mean, the number of children with levels in the extreme high quartile decreased, but 80% of these persisted in that quartile.Conclusion: Intra-individual variations in cholesterol levels have an important influence on population tracking levels and need to be considered when interpreting tracking results from early childhood to adult mid-life. However, removing the effect of regression to the mean by taking multiple measurements of cholesterol will probably improve prediction for individuals.