A new method of measuring cerebral atrophy using a ratio of brain parenchyma to ventricular and subarachnoid space is described. It uses digitized brain CT. This ratio was measured prospectively on 117 consecutive elderly patients referred for evaluation of cognitive dysfunction. Diagnosis was determined by preestablished criteria and confirmed by follow-up. Despite the improved accuracy and reproducibility of this method, its ability to differentiate persons with senile dementia of the Alzheimer's type (SDAT) from those suffering from pseudodementia was confounded by age, and was hence of limited utility. We conclude that even with sophisticated measures of cerebral atrophy, CT is unable to discriminate among common causes of cognitive dysfunction in the elderly.