We administered a battery of neuropsychological tests to 62 female patients with systemic lupus erythematosus (SLE), 12 female patients with rheumatoid arthritis (RA), and 35 normal control subjects. By applying objective decision rules to individual test protocols, an overall prevalence of cognitive impairment of 66% was obtained in the SLE patient sample. Independent clinical, radiological, and laboratory data were used to determine neuropsychiatric (NP) symptomatology and to group SLE patients as 1) "active" (N = 21), 2) "inactive" (N = 15), and 3) "never" (N = 26) NP-SLE. More than 80% of the patients in groups 1 and 2 and 42% in group 3 showed significant cognitive impairment as compared with 17% of the RA patients and 14% of the normal control subjects. Neither steroid medication nor psychological distress could account for these findings. The unexpectedly high prevalence of cognitive impairment in SLE patients with either inactive or absent neuropsychiatric symptomatology provides evidence for subclinical nervous system involvement in SLE.