Variation in serum high density lipoprotein cholesterol (HDL-C) concentrations is believed to be largely explained by triglycerides, but this has been mainly explored in cross-sectional analyses. Eight hundred and eighty one white male participants in a health screening program attended on a total of 2158 occasions for measurements that included fasting HDL-C, low density lipoprotein cholesterol (LDL-C) and homeostasis model assessment insulin resistance (HOMA-R). Baseline, change between-visit and repeated-measures regression models were used to analyse predictors of between- and within-individual variation in HDL-C. Independent predictors of between-individual variation included serum triglycerides (20.3 or 19.6% of the variance explained, depending on the model used), body mass index (BMI: 4.7 and 4.3%), cigarette smoking (3.3 and 1.5%) and alcohol consumption (0.4 and 1.1%). Within-individual variation in HDL-C was explained by changes in serum triglycerides (4.7 and 7.5%) and BMI (5.3 and 2.9%). In multivariate models, 24.3 and 24.9% of between-individual variation in HDL-C, and 7.9 and 8.8% of within-individual variation could be explained, depending on the model used. Sixty percent of the variation in HDL-C was due to unobserved factors. The majority of variation in HDL-C remains to be explained by influences other than the conventional variables: triglyceride and LDL cholesterol concentrations, insulin resistance, smoking and alcohol.