This article examines the feasibility of using molecular genetic information for diagnostic and predictive testing in Alzheimer's disease (AD). The scope is limited largely to early onset familial cases, but a brief update on genetic research in late onset disease is included. The usefulness and accuracy of such testing is determined by the results and interpretation of current research data which are presented herein. The ethical and legal issues implicated in such procedures are not covered but are examined elsewhere in this issue. During the course of the genetic analysis of AD, it has become clear that it is an etiologically heterogeneous disorder. In understanding the genetic predisposition to AD there should be continual attention to this message. Age of onset and familiality have emerged as the most useful clinical features demarcating subgroups with common origins. At least two genes predispose (independently) to early onset familial AD: the beta-amyloid precursor protein gene (beta APP) and an unidentified gene on chromosome 14. Germ line mutations in these genes act dominantly. The accurate use of genetic data in AD, for diagnostic, screening and predictive purposes relies on the most up-to-date knowledge of the transmission of the disorder in relation to mutations in these genes. The interpretation of genetic data is examined for each of the known early onset genes. In addition, we review the data pertaining to late onset disease and risk conferred by the APOE locus to both familial and nonfamilial cases. The use of these data in genetic risk analysis must be reserved until a full explanation of the association and its meaning is forthcoming. Isolated cases of AD occur at all ages of onset, but no AD mutations in the beta APP gene have yet been identified in isolated cases.