Cornell University maintains two genetic lines of specific-pathogen-free chickens in a filtered-air, positive-pressure house as a closed colony. Offspring from each generation are maintained in the same house as the parents without clean-out between successive generations. The two lines have been persistently infected with chicken infectious anemia virus (CIAV) since the mid-1990s. All flocks were monitored from 1999 to 2008 for the presence of CIAV antibodies two to four times over the 65-wk life span of each flock, starting at approximately 15 wk of age. The serologic data were modeled using the logistic mixed model for seroprevalence and the Poisson generalized linear mixed model for seroconversion. We defined seroprevalence as the percentage of seropositive birds on a sampling date; seroconversion was defined as the difference in the percentage of seropositive birds between two subsequent bleeding dates. Seroprevalence varied between flocks from 1% to 95% but was never zero. Strain and gender in general did not influence seroprevalence or seroconversion rates, but sires of the P2a line had a significantly higher seroprevalence than all other groups. There are at least two different explanations possible for the extreme variation in seroconversion. The first one is that a low level of continuous horizontal infection from seropositive to seronegative birds occurs in the facility. The second explanation is based on the concept of latency of infection, with reactivation occurring during and after sexual maturity. Latency may occur in both seropositive and seronegative chickens. Our data are compatible with reactivation from latency, perhaps followed by limited horizontal spread as well as with a low level of continuous horizontal transmission. Although the fitted Poisson model supports both options, we propose that the reactivation from latency is the likely explanation for the observed data.