Analytic expressions were derived for estimating the error in PET RCBF measurements associated with the time lag between brain and blood radioactivity following bolus H2(15)O injection and during non-steady-state CO15O inhalation. This lag time reflects the physiological difference in arrival times of 15O activity at brain and radial arterial sampling site as well as the experimentally introduced resistance to flow offered by the arterial catheter/stopcock assembly. Multiple measurements of this time lag ranged between 1 and 10 s. For non-steady-state CO15O PET measurements, estimated errors in RCBF ranged from 0.02 to 30% for delays of 2-8 s and scan lengths of 30-180 s. In the range 20-100 ml min-1 per 100 g, variations in RCBF only marginally affected these errors. Errors increased with longer delays but decreased sharply with scan durations greater than 60 s. For 30-180 s scans, even larger errors are associated with the H2(15)O injection technique (peak blood activity at 10 s): 1-60% for delays of 2-8 s. A 'slow' bolus peaking at 20 s decreased the error by 40%. For the H2(15)O method it is essential to estimate the time shift to within 2 s if accurate flow measurements (error less than 5%) are to be obtained from 40-60 s scans.