To determine the rates and factors affecting cytomegalovirus transmission from children infected in day care to their seronegative mothers, we prospectively monitored 96 seronegative mothers. Of 46 seronegative mothers without infected children, 2 seroconverted. Among 50 mothers with infected children, 19 seroconverted and of these 19, 9 shed cytomegalovirus and all 9 shed the same isolate as their child. The annual seroconversion rate for these women was 30%, significantly higher than the 3% rate for mothers without infected children (P less than 0.001; relative risk, 10.2; 95% confidence interval, 2.4, 43.8). Maternal infection was not associated with maternal age, race, duration of observation, duration of viral shedding by their children or the DNA pattern of each isolate but was associated with the age when a child's infection was identified. Only 3 of the 19 mothers who seroconverted had children older than 20 months of age (26, 28 and 28 months). Sixteen (57%) of 28 mothers with infected children 20 months of age or younger became infected compared with only 3 (13%) of 22 mothers with infected children more than 20 months (P less than 0.007), Fisher's exact test, two tailed; relative risk, 3.9; 95% confidence interval, 1.3, 11.8). For mothers with infected children younger than 20 months of age the interval between identification of her child's infection and maternal infection ranged from 1 to 26 months (8 +/- 6 (SD) months). Survival estimates revealed that mothers of infected children younger than 20 months of age acquired cytomegalovirus significantly more rapidly than mothers of older children (chi square, 9.34; P less than 0.0022).