The variation of serum cholesterol level was studied in a cohort of 16,281 individuals, with repeated measurements of cholesterol. The mean correlation coefficient between the two cholesterol values taken with a six-week interval on two occasions was 0.74. This correlation coefficient indicates a short-term variation of serum cholesterol, and reflects measurement errors and intra-individual fluctuations in cholesterol level, eg due to variations in dietary habits. The correlation coefficient for serum cholesterol values taken with a two-year interval was 0.66 and reflects a long-term variation. The ratio between these coefficients may be interpreted as a rough estimate of a correlation between the true cholesterol levels at a two-year interval, free from measurement errors and short-term intra-individual variations in dietary habits. The validity was assessed with the correlation between a single serum cholesterol value and mortality from myocardial infarction, in a cohort of 92,839 individuals followed-up for 14-16 years. The relative risks (RR) for death in myocardial infarction increased with increasing cholesterol levels. The RRs were in accordance with the well-established correlation between serum cholesterol and death from myocardial infarction. The results indicated a fairly high stability of the cholesterol level in blood and that the cholesterol values in the studied cohort were not less reliable than in comparable studies.