To measure regional skeletal kinetics using fluorine-18 fluoride positron emission tomography (PET) it is necessary to know the concentration of radioactive tracer being delivered to bone by arterial plasma with relation to time, the arterial input function (IFa). Methods by which IFa can be derived without arterial sampling are attractive because of their relative technical simplicity and the reduction in possible morbidity to the subject. We have compared the use of a scaled population input function (IFp) and a corrected image-derived input function from the aorta (IFi) with an IFa directly measured from a radial artery line in ten normal postmenopausal women. Both of the aforementioned methods rely only on a small number of discrete venous samples. Each subject had a dynamic PET acquisition of the lumbar spine performed after the intravenous injection of 180 MBq (18)F-fluoride. Both the IFp and the IFi were compared with the IFa in terms of the accuracy of determination of six parameters. These were: plasma clearance of fluoride to bone mineral (K(i)), unidirectional plasma clearance to total bone tissue (K(1)) and individual rate constants k(2), k(3) and k(4), calculated using non-linear regression with a three-compartment model, and the plasma clearance to bone mineral calculated using the Patlak method (K(pat)). For both the IFp and the IFi method the root mean square errors for K(pat) and K(i) were similar and small (<8.2%). The errors in determining K(1) and the rate constants k(2) to k(4) are larger by either method, but with a small advantage using the IFp method. It is concluded that the use of either non-invasive method for determining the arterial plasma input function is suitable for the measurement of the most important parameters, K(i) and K(pat), in these subjects.