Background. Early diagnosis of dementia is important for those who might benefit from treatment. We designed a brief comprehensive neuropsychological test battery to help differentiate control subjects from patients with mild cognitive impairment (MCI) and dementia.

Method. The battery included tests of memory, attention, executive function, speed, perception and visuospatial skills. It was administered to subjects from the OPTIMA cohort: 51 controls, 29 with MCI, 60 with ‘possible’ or ‘probable’ Alzheimer's disease (AD) (NINCDS/ADRDA) and 12 with cerebrovascular disease (CVD). Mann–Whitney U tests were used to compare performance of controls with other diagnostic groups. The sensitivity and specificity of the tests were determined using Receiver Operating Characteristic curve analyses. The effects of age, gender and years of education on test performance were determined with Spearman's rank correlations.

Results. The AD group performed worse than controls on all tests except an attention task. The Hopkins Verbal Learning Test and The Placing Test for episodic memory showed significant discriminative capacity between controls and other groups. Attention and processing speed tests discriminated CVD from controls. Category fluency, episodic memory tests and the CLOX test for executive function distinguished MCI from AD. Spearman's correlations showed negative associations between age and processing speed. Years of education affected performance on all tests, except The Placing Test.

Conclusions. Certain neuropsychological tests have been shown to be sensitive and specific in the differential diagnosis of various types of dementia and may prove to be useful for detection of MCI.