It is not known if murine cytomegalovirus (MCMV) establishes a state of molecular latency independent of low-level persistent infection. The presence of low levels of infectious MCMV distinguishes persistence from molecular latency. Thus, the distinction between persistence and latency has depended on the sensitivity of plaque assays for detecting low levels of infectious virus in tissue of previously infected mice. To determine whether MCMV establishes molecular latency or remains persistent, we developed two assays for detecting low levels of MCMV in tissue. Using prolonged in vitro culture of virus with either mouse embryonic fibroblasts or the murine 3T12 fibroblast cell line, we reproducibly detected a single PFU of MCMV. Inclusion of undiluted sonicated tissue in this assay decreased sensitivity by up to 100-fold. However, sensitivity was improved to 1 PFU of MCMV when sonicated tissue was appropriately diluted. Severe combined immunodeficient (SCID) mice were also used to detect MCMV in sonicated tissue. Infection of SCID mice with a single PFU of MCMV killed two of eight SCID mice, and the 50% lethal dose of MCMV in SCID mice was 2 to 3 PFU. Applying these two methods, we detected infectious virus in 0 of 34 spleens, 1 of 34 kidneys, and 0 of 37 salivary glands from latently infected mice. Spleens and kidneys assessed for persistent virus contained MCMV DNA by PCR and reactivated after 10 to 50 days in explant cultures. Latently infected kidney cells reactivated after adoptive transfer to SCID mice. Quantitation of the MCMV genome by PCR showed that latently infected spleens without detectable infectious MCMV contained about 3,000,000 copies of the MCMV genome. These results demonstrate that MCMV latency in spleen and kidney exists in the absence of low-level persistent infection. Use of assays with defined sensitivity for detection of MCMV in tissue provides a basis for evaluation of cytomegalovirus gene expression in the spleen and kidney during molecular latency.