A new method is described for employing family data to test for significant haplotype effects on continuously distributed variables, using likelihood-ratio tests of linear models in which haplotype effects are parameterized and familial correlations taken into account. The method is applied to the apolipoprotein B (Apo B) gene, using 5 polymorphisms (Insertion/deletion, Bsp1286I, XbaI, MspI, EcoRI) to define haplotypes in 121 French nuclear families. Eleven haplotypes were found, five of which, combined, account for over 95% of the sample. A haplotype phylogeny is proposed, and is used to define a nested set of models for testing the effects of Apo B variation on total-, low-density-lipoprotein (LDL)-, and high-density-lipoprotein (HDL)-cholesterol, triglyceride, and Apo B levels. Apo B haplotype effects account for about 10% of the genetic variance and 5% of the total variance in HDL-cholesterol and triglyceride levels. Clusters of evolutionarily-related haplotypes with similar phenotypic effects are identified for HDL-cholesterol and triglycerides. Single haplotypes with statistically significant effects are identified for cholesterol, LDL-cholesterol, and Apo B levels.