Two previously demonstrated sensitive and specific enzyme immunoassays (EIAs) for rotavirus, one using polyclonal and monoclonal antisera (TestPack Rotavirus [TPK]; Abbott Laboratories) and the other using only monoclonal anti-rotavirus antibodies (Rotaclone [RTC]; Cambridge BioScience Corporation), were evaluated as potential reference assays for rotavirus testing in comparison with direct negative-staining electron microscopy (EM), the current laboratory standard. Two hundred and seven stool samples collected consecutively during the winter of 1989 from children with acute diarrhea admitted to a ward for infants from 0 to 2 years of age were tested by the EIAs and by EM. TPK specimens were read visually; RTC results were read spectrophotometrically. Specimens with discordant EIA and EM results were further evaluated by a fluorescent focus assay. Specimens positive by EM and those negative by EM but positive by fluorescent focus assay were considered to be positive for rotavirus. Of the 207 stools tested, 35 (17%) were positive for rotavirus by these criteria. EM had a sensitivity of only 80%. Specificities were 100% for RTC and EM and 89% for TPK. These findings indicate that EM, although very specific, is relatively insensitive compared with a highly sensitive monoclonal antibody-based EIA. An EIA with high sensitivity and specificity, such as RTC, is a more appropriate reference standard for rotavirus testing.