Two methods for the detection of cytomegalovirus (CMV) in 457 clinical specimens were compared: (1) centrifugal inoculation of MRC-5 cells seeded on coverslips in 24-well plates and staining with a monoclonal antibody to CMV early nuclear antigen after incubation for both 16-18 hours (EA-1) and four days (EA-4); and (2) conventional tube cell culture. CMV was identified in 50 (11%) specimens from 34 different patients. EA-1 and EA-4 had positive results for CMV in 32 (64%) and 36 (73%) of the specimens, respectively. Positive inclusions were present on only one coverslip in 31% of the cases by EA-1 and in 10% by EA-4. The number of inclusions was not necessarily predictive of tissue culture results. CMV was recovered by conventional tissue culture from 27 specimens (54%) after an average of 17 days (range, 6-26 days). One specimen, positive for CMV by EA-4, yielded herpes simplex virus (HSV), and from 9 of the 407 CMV-negative specimens, another virus was recovered: HSV from 6 specimens and varicella zoster virus, adenovirus, and enterovirus from one specimen each. CMV was detected in significantly more specimens by EA-4 than by tissue culture (P = 0.037). However, there was no significant difference in the detection of CMV between EA-1 and EA-4 or between EA-1 and conventional culture. The authors' data suggest that for maximum recovery of CMV from clinical specimens, both an early antigen assay and conventional tissue culture should be performed. For urine specimens it appears that inoculation of two coverslips followed by staining after overnight incubation is adequate. To optimize the yield of the early antigen assay when testing specimens other than urine, the authors recommend inoculating three coverslips, two of which should be stained after overnight incubation, and, if necessary, the third coverslip could be stained after a more prolonged incubation period.