Previous genome-wide association (GWA) studies have identified SNPs associated with areal bone mineral density (aBMD). However, this measure is influenced by several different skeletal parameters, such as periosteal expansion, cortical bone mineral density (BMD(C)) cortical thickness, trabecular number, and trabecular thickness, which may be under distinct biological and genetic control. We have carried out a GWA and replication study of BMD(C), as measured by peripheral quantitative computed tomography (pQCT), a more homogenous and valid measure of actual volumetric bone density. After initial GWA meta-analysis of two cohorts (ALSPAC n = 999, aged ∼15 years and GOOD n = 935, aged ∼19 years), we attempted to replicate the BMD(C) associations that had p<1×10(-5) in an independent sample of ALSPAC children (n = 2803) and in a cohort of elderly men (MrOS Sweden, n = 1052). The rs1021188 SNP (near RANKL) was associated with BMD(C) in all cohorts (overall p = 2×10(-14), n = 5739). Each minor allele was associated with a decrease in BMD(C) of ∼0.14SD. There was also evidence for an interaction between this variant and sex (p = 0.01), with a stronger effect in males than females (at age 15, males -6.77mg/cm(3) per C allele, p = 2×10(-6); females -2.79 mg/cm(3) per C allele, p = 0.004). Furthermore, in a preliminary analysis, the rs1021188 minor C allele was associated with higher circulating levels of sRANKL (p<0.005). We show this variant to be independent from the previously aBMD associated SNP (rs9594738) and possibly from a third variant in the same RANKL region, which demonstrates important allelic heterogeneity at this locus. Associations with skeletal parameters reflecting bone dimensions were either not found or were much less pronounced. This finding implicates RANKL as a locus containing variation associated with volumetric bone density and provides further insight into the mechanism by which the RANK/RANKL/OPG pathway may be involved in skeletal development.