AbstractThe CAMCOG, ADAS-cog, and MMSE, designed to grade global cognitive ability in dementia have inadequate precision and accuracy in distinguishing mild dementia from normal ageing. Adding neuropsychological tests to their scale might improve precision and accuracy in mild dementia. We, therefore, pooled neuropsychological test-batteries from two memory clinics (ns = 135 and 186) with CAMCOG data from a population study and 2 memory clinics (n = 829) and ADAS-cog data from 3 randomized controlled trials (n = 713) to estimate a common dimension of global cognitive ability using Rasch analysis. Item difficulties and individuals’ global cognitive ability levels were estimated. Difficulties of 57 items (of 64) could be validly estimated. Neuropsychological tests were more difficult than the CAMCOG, ADAS-cog, and MMSE items. Most neuropsychological tests had difficulties in the ability range of normal ageing to mild dementia. Higher than average ability levels were more precisely measured when neuropsychological tests were added to the MMSE than when these were measured with the MMSE alone. Diagnostic accuracy in mild dementia was consistently better after adding neuropsychological tests to the MMSE. We conclude that extending dementia specific instruments with neuropsychological tests improves measurement precision and accuracy of cognitive impairment in mild dementia. (JINS, 2012, 18, 314–322)