Serological surveys using the schizont indirect fluorescence antibody test (IFAt) are routinely carried out to monitor the Theileria parva infection prevalence. The present study evaluates the diagnostic accuracy of the IFAt in eastern Zambia, where the transmission of T. parva is highly seasonal. The data set resulted from a sentinel herd (n = 105 animals) study carried out between 1995 and 2000 and was split into an epidemic period, during which the majority of the cattle became infected, and an endemic period with seasonal disease incidence in calves. In the epidemic period the T. parva seroprevalence followed closely the build up of the herd immunity. In the endemic period the seroprevalence fluctuates considerably although most of the animals had been infected. Overall, the diagnostic sensitivity of the IFA test was 55% at cut-off titre 1:40 and 28% at cut-off 1:160. The specificity of the test was 86 and 95%, respectively. A logistic regression model demonstrates that the sensitivity is significantly lower when the T. parva transmission is low (p < 0.01). The analysis of receiver operator characteristic curves classifies the test as moderately accurate (area under the curve, AUC = 0.79) during the epidemic period and less accurate in the endemic period (AUC = 0.63). Neonatal serology surveys yield a better estimate of the infection prevalence. The sensitivity of the neonatal test was 73% at cut-off titre 1:40 and 24% at cut-off 1:160.