The Structured Interview for the Diagnosis of Dementia of the Alzheimer Type, Multi-Infarct Dementia, and Dementias of Other Aetiology (SIDAM) includes the Mini-Mental State Examination extended by a number of additional items and allows the diagnosis of dementia according to ICD-10 and DSM-III-R criteria. The authors proposed to summarize selected items to form syndrome scores. These syndrome scores are supposed to measure different aspects of cognition. However, these syndrome scores have not been empirically confirmed. The present article presents a principal component analysis performed on the SIDAM test performances of 456 elderly subjects. The subjects were recruited in the course of a family study on dementia of the Alzheimer's type and depression in the elderly. One hundred four of these subjects met the criteria of dementia according to the ICD-10 criteria. Thus, the sample represnets the population in which the SIDAM is frequently used in routine clinical and epidemiological studies. We found a six-factor structure accounting for 57.1% of the variance with some similarities to the predefined structure of different syndrome scores proposed by the authors of the SIDAM. The first factor found in principal component analysis represented different higher cortical functions that all depend on language and comprehension. Three factors covered three different aspects of memory, i.e., orientation for time, orientation for place, and short-term memory. One factor represented visuoconstructive skills and, finally, there was a factor representing intellectual abilities and education. This empirically found factor structure characterizes the dimensions of cognitive deficits in demented subjects measured by the SIDAM. Syndrome scores should reflect these dimensions. Consequently, we propose to consider the empirically found factor structure in a new version of the SIDAM.