Longitudinal studies of subjects with autopsy-proven Alzheimer's disease in one skilled nursing home and of clinically diagnosed cases (NINCDS/ADRDA criteria) in three community cohorts are compared with regard to the annual rate of change in the error score of the Blessed information-memory-concentration test (IMC) in which the maximum number of errors possible is 33. The four cohorts differed significantly from each other in regard to age, education, sex, and the degree of dementia as measured by the initial IMC score. Subjects spanned the age range of 52 to 96 years and had 2 to 20 years of education. The rate of change in error score per year was similar whether the initial error score was 0 to 7, 8 to 15, or 16 to 23; however, the rate was reduced when the initial error score was 24 or above, due to a ceiling effect of the test. Among subjects with initial IMC scores less than 24, the annual rate of change varied considerably. However, the mean annual rate of change, 4.4 errors (SD +/- 3.6, SEM +/- 0.3) per year, was independent of residence in a nursing home, location of the study site, and of the patient's sex or education. Of particular importance was the finding that the rate of change in mental test score was independent of age. It can be concluded that the rate of cognitive deterioration in patients with Alzheimer's disease is quite variable among individuals and is independent of the patient's age and whether the patient resides in the community or in a nursing home.(ABSTRACT TRUNCATED AT 250 WORDS)