Bone mineral density (BMD) measurements are frequently performed repeatedly for each patient. Subsequent BMD measurements allow reproducibility to be assessed. Previous studies have suggested that reproducibility may be influenced by age and clinical status. The purpose of the study was to examine the reproducibility of BMD by dual energy X-ray absorptiometry (DXA) and to investigate the practical value of different measures of reproducibility in three distinct groups of subjects: healthy young volunteers, postmenopausal women and patients with chronic rheumatic diseases. Two hundred twenty-two subjects underwent two subsequent BMD measurements of the spine and hip. There were 60 young healthy subjects, 102 postmenopausal women and 60 patients with chronic rheumatic diseases (33 rheumatoid arthritis, 10 ankylosing spondylitis and 10 other systemic diseases). Forty-five patients (75%) among the third group were receiving corticosteroids. Reproducibility was expressed as the smallest detectable difference (SDD), coefficient of variation (CV), least significant change (LSC) and intraclass correlation coefficient (ICC). Sources of variation were investigated by linear regression analysis. The median interval between measurements was 0 days (range 0-7). The mean difference (SD) between the measurements (g/cm2) was -0.0001 (+/-0.003) and -0.0004 (+/-0.002) at L1-L4 and the total hip, respectively. At L1-L4 and the total hip, SDD (g/cm2) was +/-0.04 and +/-0.02, CV (%) was 2.02 and 1.29, and LSC (%) 5.60 and 3.56, respectively. The ICC at the spine and hip was 0.99 and 0.99, respectively. Only a minimal difference existed between the groups. Reproducibility in the three groups studied was good. In a repeated DXA scan, a BMD change, the least significant change (LSC) or the SDD should be regarded as significant. Use of the SDD is preferable to use of the CV and LSC because of its independence from BMD and its expression in absolute units. Expressed as SDD, a BMD change of at least +/-0.04 g/cm2 at L1-L4 and +/-0.02 g/cm2 at the total hip should be considered significant. This reproducibility seems independent from age and clinical status and improved in the hips by measuring the dual femur.