Methodological concerns: longitudinal studies of dementia.

Common methodological problems in longitudinal research on dementia are discussed, applying terms and constructs from the literature on normal aging. Critique of prevailing methods suggests that longitudinal studies of dementia could be improved by more extensive pilot examination of dependent measures and by more stringent control of instrumentation bias. The impact of attrition also needs to be more thoroughly evaluated before generalizations are stated about rates and patterns of decline in dementing illness.