In this study we analyzed the effect of variations in bone area size, baseline soft tissue composition represented by the R-value, and bone region of interest positioning on the precision in vivo of bone mineral density (BMD) and content (BMC) as measured by dual X-ray absorptiometry (DXA). The posterior-anterior (PA) spine, decubitus lateral, and femur modes were evaluated. Eleven (PA-spine), 9 (dec-lat), and 14 (femur) postmenopausal women were scanned twice on a Norland XR-26 with repositioning to determine short-term precision of BMD, BMC, AREA, and the R-value. Phantom precisions (CV[%] of 10 consecutive scans) for BMD (BMC) were PA spine: 0.66% (0.57%), neck: 1.1% (1.2%), and trochanter: 0.55% (1.0%). Precisions in vivo (CV[%]; two consecutive scans averaged over all patients) were PA spine: 0.9% (1.0%), dec-lat: 7.1% (18%), neck: 1.3% (1.9%), and trochanter: 2.5% (4.9%). BMD precision could be fully explained by BMC and AREA variations. However, BMC alone was a particularly poor predictor of BMD in the dec-lat (r2 = 0.05) and in the neck (r2 = 0.13) modes. AREA was a strong predictor for BMC precision explaining between 41% and 88% of the BMC changes. Changes in soft tissue composition contributed significantly in explaining the BMC changes in the dec-lat projection. A higher dependence of BMC changes on AREA changes resulted in a larger difference between BMC and BMD precision.(ABSTRACT TRUNCATED AT 250 WORDS)