The objective of this study was to suggest a new formulation of the core research diagnostic consensus criterion ''loss of insight'' in frontotemporal dementia (FTD). Eight patients with FTD (diagnoses made by interviews, medical and neuropsychological examination, CT scan, and regional cerebral glucose metabolism measured by positron emission tomography (PET) participated in the study). The results indicated that insight was present in three out of eight patients, and that insight appears to be a heterogeneous concept. Two types of insight emerged: Emotional insight associated with frontotemporal functions, and cognitive insight, related to posterior cognitive functions. These results suggest that loss of insight should not serve as a core criterion on FTD, but serves well as a supportive criterion of the disease.