Semantic dementia (SD) is a rare clinical syndrome, assigned to the group of frontotemporal lobar degenerations (FTLD). Histopathological analysis has not revealed the deposition of amyloid plaques in the majority of SD cases, in contrast to dementia of the Alzheimer type (AD). However, based on clinical examination alone a reliable differentiation of the underlying pathology cannot be guaranteed, i.e. AD and SD may be confused in some cases. Our aim was to determine, whether AD and SD can be differentiated in vivo by means of amyloid plaque PET imaging. In groups of AD and SD patients, matched for gender, age and overall degree of cognitive impairment, cerebral glucose metabolism was examined with [(18)F]Fluorodeoxyglucose (FDG)-PET and cerebral amyloid plaque density was assessed using [(11)C]6-OH-BTA-1 (PIB)-PET. A volume-of-interest analysis (VOI), using the cerebellum as a reference region, and voxel-based statistical group comparisons (SPM2) were carried out between the patient groups and matched groups of healthy controls. To control for a potential influence of atrophy on the PET findings, a correction of partial volume effects was performed. Characteristic patterns of hypometabolism could be demonstrated in both clinically defined AD and SD with some regional overlap and subtle differences (AD: bilateral posterior cingulate, temporoparietal and frontal cortex; SD: left>right polar temporal, frontal mesial cortex). Compared with healthy controls, significant [(11)C]PIB amyloid plaque tracer binding was observed only in patients with AD (in bilateral temporoparietal, frontal and posterior cingulate cortex and the precuneus) but not in SD. This difference in amyloid plaque deposition could be reproduced in direct statistical comparison of AD and SD and clearly extended the metabolic differences between the patient groups. These findings support the notion that SD can be diagnosed in vivo as a separate entity from AD using amyloid plaque imaging. In general, amyloid plaque PET may complement neuropsychological assessment regarding reliable differential diagnosis of AD and FTLD dementias based on characterization of underlying pathology and may improve the definition of individual prognosis and the selection of patients for scientific trials.