Hip axis length (HAL), a measure of femoral geometry, has been shown to predict hip fracture in white women over the age of 67 years, independently of bone mineral density at the femoral neck. A cross-sectional study of 304 pairs of female twins [176 monozygous (MZ) and 128 dizygous (DZ)], aged between 10 and 89 years, was performed to examine the influence of age, constitutional, lifestyle, and genetic factors on HAL, HAL was calculated from dual energy X-ray absorptiometry scans of the proximal femur using an automated technique with an Hologic QDR-1000W. Lean mass, fat mass, height, and weight were also measured. Maximum mean HAL was achieved by the age of 15 years. After this age there was no discernible dependency of mean HAL on age. Using within-pair differences, after adjusting for height there were no other independent constitutional or lifestyle predictors. Cross-sectionally, after adjustment for height, MZ and DZ correlations were 0.79 (95% CI: 0.73 - 0.84) and 0.54 (95% CI: 0.39 - 0.68), respectively, and independent of age. The MZ correlation exceeded the DZ correlation (p < 0.001). The best-fitting model apportioned 79% (SE 7%) of variation in height-adjusted HAL to additive genetic factors. There was marginal evidence that an environmental influence shared by twins explained 31% (SE 16%) of height-adjusted variance (p = 0.07), in which case the genetic variance was reduced to 51% (SE 15%). Adjustment for height had reduced the magnitude of total variance by 26%, and 95% of this reduction was in the additive genetic component. Applying a previously described theoretical model, approximately 10% of the increased risk of hip fracture associated with a maternal history of hip fracture could be attributed to the genetic factors determining HAL. We conclude that, in women, adult HAL is achieved by midadolescence. After adjustment for height, which is itself largely under genetic influence, other genetic factors appear to play the predominant role in explaining variation in HAL.