The state of cytomegalovirus (CMV) after the resolution of acute infection is an unsolved problem in CMV research. While the term "latency" is in general use to indicate the maintenance of the viral genome, a formal exclusion of low-level persistent productive infection depends on the sensitivity of the assay for detecting infectious virus. We have improved the method for detecting infectivity by combining centrifugal infection of permissive indicator cells in culture, expansion to an infectious focus, and sensitive detection of immediate-early RNA in the infected cells by reverse transcriptase PCR. A limiting-dilution approach defined the sensitivity of this assay. Infectivity was thereby found to require as few as 2 to 9 virion DNA molecules of murine CMV, whereas the standard measure of infectivity, the PFU, is the equivalent of ca. 500 viral genomes. Since murine CMV forms multicapsid virions in most infected tissues, the genome-to-infectivity ratio is necessarily >1. This assay thus sets a new standard for investigating CMV latency. In mice in which acute infection was resolved, the viral DNA load in the lungs, a known organ site of CMV latency and recurrence, was found to be 1 genome per 40 lung cells, or a total of ca. 1 million genomes. Despite this high load of CMV DNA, infectious virus was not detected with the improved assay, but recurrence was inducible. These data provide evidence against a low-level persistent productive infection and also imply that intermittent spontaneous recurrence is not a frequent event in latently infected lungs.