In order to test the detection feasibility of human immunodeficiency virus (HIV) in saliva, a three-method blind screening analysis was conducted. Sixty-eight individuals were studied, comprising 34 HIV carriers and 34 noncarriers (controls) of matched gender and age. An oral examination preceded saliva and blood sampling of studied individuals. All samples were tested blind for HIV by using two immunological methods [Oraquick-compatible enzyme-linked immunosorbent assay (ELISA) and a fluorescent immunoenzymatic method (ELFA)], confirmed by western blotting, and a simple molecular method (polymerase chain reaction amplification of a relatively constant viral DNA region), confirmed by DNA hydridization. Compared with the controls, about twice as many HIV carriers had oral health problems, including periodontal disease. ELFA resulted in 33/34 positives and 34/34 negatives in saliva, while it detected 34/34 positives and 34/34 negatives in blood. ELISA performed even better, with correct assignment of all positives and negatives in both saliva and blood. The PCR method, at three annealing temperatures, surprisingly detected all positive samples, while it gave no false-positive result. In conclusion, the detection of anti-HIV in saliva may achieve accuracy of 97.1-100%, comparable with that in blood. Furthermore, this study suggests that a highly accurate molecular method of HIV detection may be feasible, although the studied carriers had rather homogeneous characteristics.