A method for identification and quantitative evaluation of task-specific changes of the regional cerebral blood flow (rCBF) measured with PET in activation studies of individual subjects is presented. The method is based on the statistical distributions of the quantitative and spatial information of regions of interest in rCBF subtraction images. For validation, a cylindrical phantom of 20 cm diameter containing six spheres of 10-30 mm in diameter was used. The spheres representing the specific signals were filled with 18F, while one-tenth of this activity concentration was filled into the background compartment of the phantom representing "noise." Of a sequence of dynamically recorded frames, subtraction images with different signal-to-noise ratios were calculated. In these subtraction images, our method allowed us to identify the larger spheres accurately and to quantify the signals. Comparison with t map analysis in averaged subtraction images revealed a high correspondence with the results obtained by our method in individual subtraction images. Based on this phantom validation, the method was applied for mapping of rCBF changes in humans. The rCBF was measured with [15O]butanol in four subjects during unilateral somatosensory discrimination and during rest. The method proved to be capable of identifying task-specific rCBF changes in the contralateral motor, premotor, and sensory cortex accurately and with high quantitative and anatomical precision in each subject.