To account for the wide variations in the prevalence of cytomegalovirus infections among day-care centers we serially tested 309 children at three day-care centers for 3 years. Based on the DNA restriction endonuclease pattern of each isolate, the rate of infection for children differed significantly (P less than 0.001) among centers: at Center 1, 50% (46 of 93) of children acquired cytomegalovirus in day care; at Center 2, 62% (64 of 104); and at Center 3, 25% (21 of 84). Infection rates were associated with the number of infants enrolled, and half or more of infected children were younger than 24 months of age. Six of 7 new isolates were introduced by children 18 months of age. Based on DNA patterns the prevalent isolates at Centers 1 and 2, although different, were shed for an average of 22 and 23 months, respectively, compared with an average of 15 months for other isolates (P less than 0.001). Reinfections with the prevalent isolates were observed for 2 of 34 children tested. The most important factors affecting day-care center transmission are the number of infants enrolled and prolonged viral shedding, possibly enhanced by reinfection.