In the medical and epidemiologic literature, a registry denotes a data base in which registrants share some common characteristic such as disease category. One criticism of registries is that they frequently collect subjects in a haphazard fashion and, hence, are "nonrepresentative of the population purportedly being represented." In this report, we compare two registries: an incident-based Alzheimer's Disease Patient Registry (ADPR) recruiting subjects for epidemiologic studies from a large health maintenance organization; and an Alzheimer's Disease Research Center (ADRC) registry recruiting subjects for phenomenologic, biologic, and pharmacologic studies. While these registries share personnel, overlap geographically, and use similar diagnostic procedures, they differ substantially in their missions and resulting recruitment strategies. We compared these registries with respect to demographic characteristics and cognitive features at subject entry. Subjects enrolled in the incident-based registry are older and report shorter time between symptom onset and recruitment. They are less demented and mirror the general population demographically more closely than do subjects in the other registry. The ADRC registry contains a much greater proportion of subjects with higher educational attainment.