The reproducibility of single photon absorptiometry (SPA) results for detection of changes in bone mineral content (BMC) was evaluated in a clinical setting. During a period of 18 months with 4 different sources, the calibration scans of an aluminium standard had a variation of less than 1% unless the activity of the 125I source was low. The calibration procedure was performed weekly and this was sufficient to correct for drift of the system. The short term reproducibility in patients was assessed with 119 duplicate measurements made in direct succession. The best reproducibility (CV = 1.35%) was found for fat corrected BMC results expressed in g/cm, obtained at the site proximal to the 8 mm space between the radius and ulna. Analysis of all SPA scans made during 1 year (487 scans) showed a failure of the automatic procedure to detect the space of 8 mm between the forearm bones in 19 scans (3.9%). A space adjacent to the ulnar styloid was taken as the site for the first scan in these examinations. This problem may be recognized and corrected relatively easy. A significant correlation was found between BMC of the lower arm and BMC of the lumbar spine assessed with dual photon absorptiometry. However, the error of estimation of proximal BMC (SEE = 20.0%) and distal BMC (SEE = 19.4%) made these measurements of little value to predict BMC at the lumbar spine in individuals. The short term reproducibility in patients combined with the long term stability of the equipment in our clinical setting showed that SPA is a reliable technique to assess changes in bone mass at the lower arm of 4% between 2 measurements with a confidence level of 95%.