The aim of this project was to compare three phantoms used for cross-calibration of dual-energy X-ray absorptiometers with an in vivo cross-calibration. The phantoms used were the Bona Fide Phantom (BFP), the European Spine Phantom (ESP) and the GE Lunar Aluminum Spine Phantom (ASP). The cross calibration was for L2-L4 lumbar spine bone mineral density (BMD) on a GE Lunar DPX-L and Hologic QDR 2000. The in vivo cross-calibration was obtained using 72 subjects (61 female, 11 male; mean age 49 years, range 14-84 years). The phantoms were measured 10 times without repositioning on both instruments. A further, long-term cross-calibration was obtained with the BFP over a 9 month period. The true linear relationship between the two instruments was calculated used a standardized principal components method. The mean residuals were calculated between each phantom cross-calibration line and the in vivo data to obtain a measure of the goodness of fit between the phantom cross-calibration and the in vivo data. There was no significant difference between the in vitro and in vivo cross-calibrations. The long-term BFP cross-calibration gave an in vitro cross-calibration that is closest to the in vivo cross-calibration in this group of subjects. When calculating Hologic QDR BMD from results on the GE Lunar DPX-L, the ASP underestimates Hologic QDR 2000 BMD by 4% at high BMD and overestimates by 4% at low BMD. The ESP cross-calibration overestimates Hologic QDR2000 BMD by 1% at high BMD and 4% at low BMD. The BFP performs best, overestimating Hologic QDR2000 BMD by between 1.2% and 1.8%, whilst the difference between the long-term BFP cross-calibration and the in vivo data is less than 1% over the range of BMD covered.