Purpose: We address how the national prevalence of cognitive impairment can be estimated from two nationally representative surveys. Design and Methods: Data are from the 1999-2001 National Health Interview Survey (NHIS) and the 1999 National Nursing Home Survey (NNHS). The NHIS represents all community-dwelling people living in the United States, and the NNHS is representative of all nursing home residents. Results: NHIS data show that there are approximately 800,000 community-based elders aged 65 and older with reported confusion or memory loss, and 2.3 million elders with reported limitation of activity caused by senility or dementia. There are an estimated 632,000 nursing home residents aged 65 and older with a reported diagnosis of dementia. Implications: Estimates of the prevalence of cognitive impairment that are based on nationally representative data are rare, because comprehensively evaluating a national sample by using standard, validated cognitive-impairment assessment methods is difficult and expensive, and because most national surveys are broad based and designed to cover a wide variety of topics. Crude measures of cognitive impairment, such as the presence of confusion or memory loss or limitations caused by senility or dementia, that are included in these multipurpose surveys may be only rough proxies for clinically evaluated cognitive impairment, but they do appear to produce prevalence estimates that are similar to estimates found with the use of more precise case-ascertainment methods. These nationally representative data sets may be used to generate hypotheses related to the prevalence, epidemiology, and health care utilization patterns of people with cognitive impairment that can be tested in studies using more specific case-ascertainment criteria.