It has been proposed that bone area on two scans should be very close (within 2%) in order to ensure a valid comparison, but this recommendation has not been critically evaluated. We assessed the importance of bone area in test precision within a regional testing program that maintains a large quality assurance database (208 analyzable scan-pairs). Differences in bone area exceeding the "2% rule" were common (22-46% depending upon site). Scan-pairs that exceeded the 2% rule were associated with significantly greater bone density measurement error for the lumbar spine (p<0.005) and femoral neck (p=0.001), but not the total hip. A simpler "1-cm(2) rule" for the lumbar spine also predicted significantly worse spine precision (p<0.005). In conclusion, differences in bone area are sources of short-term bone density measurement error.