For quantitative traits associated with risk to complex diseases, such as heart disease, single major locus models are likely to be too simplistic. Currently, researchers have begun to use oligogenic models of inheritance, but the resolving power of these models remains to be determined. As the major apoprotein of high density lipoprotein (HDL), apolipoprotein A1 (apo-A1) is generally accepted as a protective factor for coronary artery disease. Although familial aggregation of apo-A1 levels has been reported, the mode of inheritance of apo-A1 remains ill defined. In the present study, we conducted a segregation analysis comparing a series of one-locus and two-locus univariate models for apo-A1 levels in a sample of 137 families ascertained through probands undergoing elective, diagnostic coronary angiography. A two-locus Mendelian model fit these data significantly better than any one-locus model. The incorporation of the second major locus into the model of inheritance leads to a significant improvement in the fit, and a significant decrease of the residual heritability. The best-fitting model included two loci with a reciprocal pattern of epistasis generating 4 distinct genotypic distributions. Taken together, these two major loci account for 58% of the variance of adjusted apo-A1 levels. This demonstration of a second major locus controlling apo-A1 levels may explain the equivocal results obtained from previous studies. This two-locus model may be more powerful for linkage analysis to map one or both of these quantitative trait loci.