A method commonly used for localization of brain regions on positron emission tomographic (PET) images is direct visualization and designation of structure on the image itself. This technique, however, is limited to portions of the brain having sufficient differential radionuclide uptake to permit the recognition of structure by observers familiar with brain anatomy. Two other methods commonly used instead of direct visualization are magnetic resonance imaging (MRI) and stereotactic methods of localization. This report compares the accuracy of a MRI method versus a stereotactic method of brain structure localization of PET. The three localization methods were analyzed for sixteen different brain regions in 5 different subjects by two independent observers. The results were converted to PET pixel size (1 pixel = .127 cm) for comparison. MRI localization differed from direct visualization by a mean and standard deviation of approximately 2 +/- 1 pixels in both the X and Y dimensions. Stereotactic localization differed from direct visualization by approximately 2 +/- 1 pixels in the X dimension and by approximately 6 +/- 2 pixels in the Y dimension. This larger variation seen with the stereotactic method may be attributed to the questionable assumption of linearity of structure location with respect to size of the inner table of the calvarium.