Multiplex PCR analyses for both bacterial and viral pathogens were conducted in a blinded manner on 33 archival specimens, of known culture status, procured from chinchilla models of both single- and mixed-pathogen-induced otitis media and from a pediatric patient. These specimens had been maintained at -70 degrees C for up to 6 years. Experimental specimens evaluated included middle-ear effusions, nasopharyngeal lavage fluids and middle-ear lavage fluids from animals which were immunologically naive, sham-immunized or actively immunized with nontypeable Haemophilus influenzae antigens. Sampling times used ranged from the day of bacterial or viral challenge to 42 days after challenge. Initial PCR analyses of the 33 specimens matched the traditional culture data in 24 instances (73%), correctly identifying nontypeable H. influenzae, Moraxella catarrhalis, Streptococcus pneumoniae, or adenovirus as the causative agent. A PCR-positive signal for the microbe(s) inoculated was also obtained in four animal model specimens (12%) which were culture negative. One of two culture-negative human effusions was also PCR positive. Thus, overall, results obtained by blinded PCR were 85% concordant with traditional culture methods or correctly indicated the specific pathogen introduced in four specimens that were sterile. In no instance was a false-positive signal obtained for any of the five etiologic agents being evaluated. We conclude that the multiplex PCR analyses are rapid and accurate methodologies when they are used to retrospectively evaluate diverse archival specimens of limited volume from experimental models of otitis media.