A need exists to measure the absolute sensitivity of a positron emission tomographic (PET) scanner in units of counts.s-1.MBq-1. At present sensitivity is generally determined by measurement of a radionuclide of known concentration distributed in a water-filled cylindrical phantom, usually 20 cm in diameter. The measurement is confounded by self-attenuation of the source and scatter within the cylinder and does not give a true absolute sensitivity measurement. Due to variations in the magnitude and treatment of these factors, meaningful comparison between different manufacturer's scanners is difficult, as are comparisons between different acquisition geometries (e.g. with and without interplane septa present). A method has been developed in our laboratory that provides measurements of absolute sensitivity in air for a scanner independent of attenuation and scatter within the source. The method involves measurements of a thin-line source of fluorine 18 contained within an aluminium housing to which successive aluminium sleeves are added. The extrapolation of these measurements allows an effective counts.s-1.MBq-1 measurement to be made for zero thickness of aluminium. Measurements have yielded absolute sensitivities of 3926 +/- 61 counts.s-1.MBq-1 (0.39% efficiency), 5079 +/- 26 counts.s-1.MBq-1 (0.51%), and 32312 +/- 544 counts.s-1.MBq-1 (3.2%) for a whole-body PET scanner with interplane septa and for a NeuroPET operating with and without interplane septa, respectively.