Assessment instruments used to measure psychological outcome in clinical trials in dementia include rating scales of activities of daily living and scales of cognitive function. Each has specific advantages and limitations. The popularity of certain dementia scales results from their demonstrated correlation with measures of pathological change, implying that they accurately reflect dementia severity. It is argued that the inference of test validity on the basis of such clinicopathological correlation may be unfounded. Alzheimer's disease is a heterogeneous disorder, encompassing varying profiles of cognitive disability. Dementia scales, which are weighted in favour of specific areas of dysfunction, may be differentially sensitive to the deficits of certain patients relative to others, thus yielding a spurious measure of dementia severity.