The accurate measurement of the concentration of a radioisotope in small structures with positron emission tomography (PET) requires a correction for quantitation loss due to the partial volume effect and the effect of scattered radiation. To evaluate errors associated with measures in the human basal ganglia (BG), a unilateral model of the BG has been inserted in a 20-cm cylinder. The recovery coefficient (RC=measured activity/true activity) for the BG phantom has been measured on a CTI tomograph with different background concentrations (contrast) and at different axial locations in the gantry. The BG was visualized on four or five slices, depending on its position in the gantry and on the contrast used. The RC was 0.75 with no background (contrast equal to 1.0). Increasing the relative radioactivity concentration in the background increased the RC from 0.75 to 2.00 when the contrast was -0.7 (BG<background). The RC was also affected by the size and the shape of the region of interest (ROI) used (RC from 0.75 to 0.67 with ROI size from 0.12 to 1.41 cm(2)). These results show that accurate RC correction depends not only on the volume of the structure but also on its contrast with its surroundings, as well as on the selection of the ROI.