The tracer appearance time relative to the radial artery-sampling site has been evaluated in six brain locations in five human subjects using dynamic positron emission tomography (PET) following the bolus injection of H2(15)O. There was a maximum difference of +/- 2 s from the average in each location. To globally adjust the timing difference between the measured arterial curve and the PET scan, a correction method was developed based on a nonlinear least-squares fitting procedure. This new technique determined the global time delay with an accuracy of +/- 0.5 s. On the other hand, the linear backward extrapolation method resulted in a systematic error of 4 s.