The CAMCOG, which forms part of the CAMDEX interview (Roth et al., 1986, 1988), is a brief neuropsychological battery designed to assess the range of cognitive functions required for a diagnosis of dementia, and to detect mild degrees of cognitive impairment. It was administered to a population sample of 418 elderly people (aged 77 and above) in their place of residence. The data show that in contrast to the Mini-Mental State Examination, total CAMCOG scores are well distributed and there is no ceiling effect. Examination of the association between CAMCOG scores and socio-demographic variables (age, sex, education and social class) shows that each exerts a significant, and independent, effect upon performance. CAMCOG also includes a number of subscale which assess individual areas of cognitive function. Of the eight major subscales (orientation, language, memory, attention, praxis, calculation, abstract thinking, perception), age was significantly related to all but attention; sex with attention, praxis, calculation and perception; education with language and abstract thinking; and social class with language and perception. In all these analyses, the results were adjusted for the effects of the other socio-demographic variables using analysis of variance. However, education and social class are highly correlated variables and when the association with education is examined without adjusting for social class, attention and praxis are also found to be significantly related to education. Caution must therefore be taken when using the CAMCOG (or any other cognitive test) as a screening test for dementia, using a single, predetermined cutpoint. In general, the combination of brevity and breadth of the CAMCOG, along with its distributional properties, makes it an attractive neuropsychological test for use in the community or the clinic.