SUMMARYSerological tests are very frequently used in epidemiological surveys of leishmaniasis and other parasitoses. Their sensitivity and specificity are generally defined with respect to parasitism and disease, rather than infection. The reason is that known positives are those individuals most likely to yield parasites, or who have distinctive clinical signs, and concomitantly high antibody titres. This paper investigates the performance of one serological method, the indirect fluorescent antibody test (IFAT), in detecting Leishmania infantum infection during an intensive 2-year cohort study of dogs in southern France. The results show that sensitivity and specificity with respect to infection can be simultaneously high, but maximum sensitivity is probably < 80%, and lasts for a relatively short period of 2-3 months after a lengthy incubation period. The IFAT gave the incidence of infection as 18-65% in the first year, whereas the best estimate of incidence based on parasite isolation and clinical observation was 72 %. But data from the second year suggest that the 72 % was itself an underestimate. We argue that, during epidemiological surveys, the IFAT in particular, and serological tests for leishmania in general, will underestimate prevalence, incidence and hence the scale of the control problem. However, there is evidence that tests for canine leishmaniasis employing high threshold titres will identify the most infectious animals, allowing selective treatment or culling of those which contribute disproportionately to transmission.