To explore the possible mechanisms of phantom limb discomfort after amputation, three amputees with phantom limb pain were studied. This study examined the change of regional cerebral blood flow using technetium-99m hexamethylpropyleneamine oxime-single photon emission computerized tomography, which was arranged at the time of severe phantom limb discomfort and after the discomfort subsided or was completely relieved. Nine representative transverse slices parallel to the orbitomeatal line were selected for quantification. The cortical ribbon (2-cm thickness) was equally subdivided into 12 symmetrical pairs of sector regions of interest in each slice. The irregularly shaped regions of interest were drawn manually around the right thalamus and basal ganglion and then mirrored to the left thalamus and basal ganglion. The contralateral to ipsilateral ratio of regional cerebral blood flow for each area was calculated. The intensity of phantom limb pain was evaluated on a 0 to 10 visual analog scale. In Cases 1 and 2, the contralateral to ipsilateral regional cerebral blood flow ratios of multiple areas of the frontal, temporal, or parietal lobes were increased at the time of more severe phantom limb pain, and the ratios were normalized or even decreased when the phantom limb pain subsided. In Case 3, increased contralateral to ipsilateral regional cerebral blood flow ratios were also found over the frontal, temporal, and parietal lobe. However, most of the increased regional cerebral blood flow ratios of regions of interest in the first study persisted in the follow-up study. Also, the regional cerebral blood flow ratios of greater number of regions of interest of the same gyrus and new gyrus were increased. There was no significant right-left difference of regional cerebral blood flow over bilateral thalami and basal ganglia in all three cases. The results suggested that phantom limb pain might be associated with cortical activation involving the frontal, temporal, or parietal cortex, and it may imply the possibility of the existence of an ascending polysynaptic pathway that conveys the uncomfortable phantom limb sensation to the cerebral cortex. These findings may also indicate that reorganization of the cortical blood flow occurs in amputees. However, it is still difficult to conclude that the changes in regional cerebral blood flow were attributable directly to pain. With no comparison group of amputees and because of the small number of cases, it is hard to generalize about cerebrocortical involvement in phantom pain, and it is possible that the findings represent a normal phenomenon seen after amputation. Another possibility is that the findings represent increased arousal caused by pain rather than an intrinsic pain pathway. Further study is worthwhile.