Nasal aspirates from 238 infants hospitalized with acute respiratory infections during the winter of 1994 and 1995 were tested for respiratory syncytial virus (RSV) by immunofluorescence assay (IFA) and the viral isolation technique (VIT) and by two PCR and hybridization methods: reverse transcription PCR 1 (RT-PCR1), which amplifies the RNAs of all RSV strains, and RT-PCR-2, which allows subgroup classification of RSV. RT-PCR-1 and RT-PCR-2 detected viral sequences in 56.7% (135 of 238) and 48.3% (115 of 238) of the samples, respectively, while only 80 (33.6%) samples were found to be positive by IFA and VIT. Of the PCR-positive specimens, 57 were missed by these routine techniques in RT-PCR-1 and 45 were missed in RT-PCR-2. Although the RSV-PCR-1 and RSV-PCR-2 techniques amplified two different sequences of the RSV genome, they gave similar results for 218 (91.6%) nasal aspirates. Compared with conventional methods, the sensitivity, specificity, and agreement were 97.5, 63.9, and 75.2%, respectively, for RT-PCR-1 and 89.7, 71.9, and 77.7%, respectively, for RT-PCR-2, and for these two RT-PCR assays, the positive predictive value (PPV) and the index of agreement (kappa) were comparable and moderate, respectively: PPV was 57.8% and kappa was 0.52 in RT-PCR-1, and PPV was 60.9% and kappa was 0.54 in RT-PCR-2. However, there was a perfect correlation between the two RT-PCRs, with a PPV of 100% and an excellent index of agreement (kappa = 0.88). Therefore, most RT-PCR results were really true positive, and VIT and IFA, which missed some of them, appeared to be less sensitive.