Two enzyme-linked immunosorbent assays (ELISAs) for herpes simplex virus (HSV) detection were compared with culture in a prospective, blinded study with 153 patients with suspected recurrent oral or genital HSV. A subset of 15 of these subjects were studied daily until symptom resolution during a single episode of recurrent HSV. Direct-site specimens were collected and either placed in viral transport media (for Ortho ELISA and fresh inoculation into primary rabbit kidney cells) or frozen in ELISA collection media (DuPont). One hundred eighty-six culture-ELISA comparisons were analyzed. On the basis of culture positivity, the DuPont and Ortho ELISAs differed substantially with regard to sensitivity (93 versus 35%) but had similar specificities (95 versus 100%) and positive (85 versus 100%) and negative (98 versus 85%) predictive values. There were seven DuPont ELISA-positive, culture-negative samples which were confirmed positive for HSV by blocking antibody test (revised specificity, 100%; positive predictive value, 100%). Six of these discrepant samples were from previously culture-positive subjects. These results demonstrate that currently available ELISA kits vary substantially as to their sensitivities in detecting HSV antigen from direct-site specimens. In addition, antigen detection, by ELISA technology, is not always synonymous with state of viral infectivity as judged by tissue culture cytopathic effect.