The in vivo precision of broad-band ultrasonic attenuation (BUA) was measured in one normal male volunteer over periods of one day and six weeks and in a group of 10 volunteers over a period of two weeks using a Walker Sonix UBA 1001 bone mineral analyser in which the foot is immersed in a water tank. Coefficients of variation ranged from 1-3% and, in general, there was no difference in BUA variance between right and left feet. However, right foot BUA values were consistently greater than those of the left, but there was no difference between mediolateral and lateromedial transmission. The use of preboiled water improved precision and reduced measurement time. In addition, factors affecting BUA precision were investigated in the individual volunteer. Equipment stability and conditions of immersion (immersion time, water depth, water temperature and concentration of a detergent wetting agent) had comparatively little effect on BUA values. On the other hand, the effects of foot positioning were more pronounced: rotation about the long axis of the foot, translation across the water tank and foot movement in the dorsal-plantar direction each contributed a maximum of 1.5-2.0% variation in BUA. However, both rotation about the long axis of the leg and foot movement in the heel-toe direction could change BUA by as much as 9%. It is likely that foot positioning will limit the precision achievable with this technique.