Previous comparisons between the Reference and Definitive Methods for measuring serum cholesterol have demonstrated a small but persistent positive bias in the Reference Method, averaging about +1.6%. Here we describe the results of further investigations designed to better characterize the nature of this bias. Analysis of a well-characterized model serum sample (SRM 909) suggests that more than half of the difference in cholesterol values determined by the two methods is the result of small contributions from cholesterol precursor sterols and phytosterols, which are also measured for the Reference Method. An additional significant contribution may be from cholesterol oxidation products, particularly 7-hydroxycholesterol isomers, which are active in the Liebermann-Burchard reaction. The 7-hydroxycholesterol in SRM 909, most of which appeared to be already present in the serum rather than formed during saponification, may account for as much as 20% of the observed difference between the methods. Contributions from other possible sources, including impurities in the cholesterol standard and incomplete saponification of cholesteryl esters, are very small. Because the observed bias is both quite small and consistent among samples, the cholesterol Reference Method continues to meet all of the requirements generally expected for a dependable and effective Reference Method.