The sensitivity and specificity of PCR, serology (ELISA) and lymphoproliferative response to Leishmania antigen for the detection of Leishmania infantum infection were evaluated in a cohort of 126 dogs exposed to natural infection in Brazil. For PCR, Leishmania DNA from bone-marrow was amplified with both minicircle and ribosomal primers. The infection status and time of infection of each dog were estimated from longitudinal data. The sensitivity of PCR in parasite-positive samples was 98%. However, the overall sensitivity of PCR in post-infection samples, from dogs with confirmed infection, was only 68%. The sensitivity of PCR varied during the course of infection, being highest (78-88%) 0-135 days post-infection and declining to around 50% after 300 days. The sensitivity of PCR also varied between dogs, and was highest in sick dogs. The sensitivity of serology was similar in parasite-positive (84%), PCR-positive (86%) and post-infection (88%) samples. The sensitivity of serology varied during the course of infection, being lowest at the time of infection and high (93-100%) thereafter. Problems in determining the specificity of serology are discussed. The sensitivity and specificity of cellular responsiveness were low. These data suggest that PCR is most useful in detecting active or symptomatic infection, and that serology can be a more sensitive technique for the detection of all infected dogs.