We performed ultrasound measurements in the calcaneus of 512 healthy women. Broadband ultrasonic attenuation (BUA) and speed of sound (SOS) were obtained with a Lunar Achilles ultrasonic instrument. Subjects studied were one group of 67 women working in our hospital (group A) and two groups which are part of two large prospective cohort studies (groups B and C). Group B consisted of 244 women aged 31-79 years randomly selected from a large insurance company, and group C consisted of 201 women aged 74-91 years randomly selected from the electoral rolls. Dual-energy X-ray absorptiometry (DXA) measurements of femoral neck and total body were performed with a Hologic QDR 2000 for group B and with a Lunar DPX Plus for group C. The in vitro precision of the Achilles, estimated by measuring a phantom daily for 45 days, was 0.84% for BUA and 0.12% for SOS. We assessed the in vivo short-term precision in 20 healthy volunteers working at the hospital, measured three times each. The coefficients of variation were 0.93% (+/-0.21) for BUA and 0.15% (+/-0.03) for SOS. The precision error was compared with the true variation, to obtain a standardized coefficient of variation. We analysed the three groups pooled together (n=512) and found for BUA an average 20% decrease and for SOS a 5% decrease between the ages of 20 and 90 years. We also performed separate analyses of subjects younger than 50 and older than 50 years, and within each 10-year age group we found that BUA was stable or slightly increased from 20 to 50 years and then decreased after 50. In contrast, SOS did not increase but decreased from the age of 20. We compared DXA measurements of the femoral neck and the total body with ultrasound measurements in groups B and C. In both groups the correlations were better with total body DXA than with femoral neck and spine DXA.