The construction of reference ranges that accurately represent the population at large is essential for the correct identification of osteoporosis from bone mineral density (BMD) measurements. In this study, reference data supplied by the manufacturer of the Lunar DPX+ bone densitometer were compared with data obtained locally. Lumbar spine, proximal femur and total body BMD measurements were made in an age-stratified random sample of 702 Southampton women aged 20 to 89 years. Relevant demographic and medical data were recorded for each subject using a questionnaire. Reference curves of BMD (mean +/- SD) were plotted against age for each measurement site and were found to be higher than the manufacturer's reference values at all ages and sites. Exclusion of women with factors known to affect bone mass only served to increase this discrepancy. According to World Health Organisation definitions, osteoporosis may be identified from BMD values alone. Based upon neck of femur BMD values, 100 (14.8%) of the women in this study group were categorized as osteoporotic using local young normal reference data, compared with only 39 (5.8%) using the manufacturer's data. By normalizing for age distribution, these findings were extrapolated to the local population where it was predicted that 26.0% and 10.1% of females over 50 years of age would be classified as osteoporotic using the respective reference ranges. This study clearly illustrates how the numbers of women diagnosed as osteoporotic vary with the use of different reference populations.