Extensive studies have been performed on the association between the estrogen receptor alpha (ER-alpha) gene and bone mineral density (BMD). Despite considerable efforts, the studies using limited markers and relatively small sample size have yielded largely inconsistent results. In this study, 1873 Caucasian subjects from 405 nuclear families containing 1512 sib pairs were recruited. BMD at the lumbar spine (LS) and femoral neck (FN) was measured by dual-energy X-ray absorptiometry (DXA). Seven single-nucleotide polymorphisms (SNPs) spanning from exon 1 to 8 in the ER-alpha gene were genotyped. The program QTDT (quantitative transmission disequilibrium test) was applied to test linkage and/or association of the ER-alpha gene and BMD variation using individual SNP markers and reconstructed haplotypes. Linkage disequilibrium (LD) was generally detected for SNPs in the ER-a gene (P < 0.05). Associations were observed between SNP rs932477 and FN BMD (P = 0.028), and between the most predominant three-marker haplotype (GCG) containing SNP rs932477 and FN BMD (P = 0.010). Within-family association (present only with both linkage and association) between SNP rs2228480 (G2014A) and FN BMD (P = 0.015) was observed. The most predominant seven-SNP haplotype (TCGCGGG) was associated with higher LS BMD (P = 0.015). However, after correction for multiple testing, these associations did not reach statistical significance. Denser markers may be necessary to better define the relationship between the ER-alpha gene and BMD variation in our sample.