One of the important uses of bone absorptiometry is to examine the rate of bone mineral change in order to evaluate therapy and to identify individuals who need therapy. Generally, this involves comparing the difference between two scans obtained months to years apart. This study investigates the precision of dual photon absorptiometry using a human torso phantom, normal subjects, and abnormal patients. These studies showed that bone mineral calculated as g/cm2 was more precise than g/cm. Reanalysis of the same scan by the same individual produced an average error equivalent to that produced by scanning and analyzing the same subject on multiple occasions. Interobserver analysis error was essentially equal to the intraobserver error. In order to obtain maximum precision, care must be taken that the integrated area of a repeat scan is identical to the previous scan. Our findings indicate that to be confident (95%) of a real change between two scans a difference of at least 5.6% must be measured.