A commercial polymerase chain reaction (PCR) assay (Amplicor, Roche) for Chlamydia trachomatis was compared with a direct fluorescent antibody (DFA) test using urethral and cervical samples, many of which on the basis of prior testing by DFA contained small rather than large numbers of elementary bodies. Urine samples were collected from patients in a sequential unselected manner. Of 244 clinical specimens (138 male urethral and cervical; 106 male and female urine), 66 were positive by both DFA and PCR and 141 were negative by both tests. Nine samples were DFA negative and PCR positive, and 28 samples were DFA positive and PCR negative. However, 24 (86%) of the latter samples contained fewer than ten elementary bodies. When serial dilutions of laboratory stock strains (serovars E and H) were tested, the DFA test detected Chlamydia trachomatis at a dilution tenfold greater than the PCR. Furthermore, of five DFA-positive clinical samples, three that were PCR negative when tested according to the manufacturer's instructions were positive when they were diluted less. A modification of the PCR assay along these lines might improve sensitivity.