As part of a multicenter study, we examined the intersite reproducibility of bone mineral content (BMC) and areal density (BMD) among three fan-beam dual-energy X-ray absorptiometry (DXA) instruments from one manufacturer, all using the same software version. Spine, femur, and body-composition phantoms were each scanned nine times at each center. Over a 3-wk period, the same 10 adults were scanned once at each of the three centers. For the spine and femur phantoms, the precision errors were 0.3-0.7%. For the body-composition phantom, the precision errors were 0.8-2.8%. The intersite coefficients of variation for the human measurements varied from 1.1 to 6.8%, depending on the bone site. We conclude that even when using the same fan-beam DXA model and software, an intersite cross-comparison using only phantoms may be inadequate. Comparisons based solely on the use of a spine phantom are insufficient to ensure compatibility of human bone mineral data at other bone sites or for the whole body.