Several polymerase chain reaction (PCR)-based methods have recently been developed for diagnosing malarial infections in both birds and reptiles, but a critical evaluation of their sensitivity in experimentally-infected hosts has not been done. This study compares the sensitivity of several PCR-based methods for diagnosing avian malaria (Plasmodium relictum) in captive Hawaiian honeycreepers using microscopy and a recently developed immunoblotting technique. Sequential blood samples were collected over periods of up to 4.4 yr after experimental infection and rechallenge to determine both the duration and detectability of chronic infections. Two new nested PCR approaches for detecting circulating parasites based on P. relictum 18S rRNA genes and the thrombospondin-related anonymous protein (TRAP) gene are described. The blood smear and the PCR tests were less sensitive than serological methods for detecting chronic malarial infections. Individually, none of the diagnostic methods was 100% accurate in detecting subpatent infections, although serological methods were significantly more sensitive (97%) than either nested PCR (61-84%) or microscopy (27%). Circulating parasites in chronically infected birds either disappear completely from circulation or to drop to intensities below detectability by nested PCR. Thus, the use of PCR as a sole means of detection of circulating parasites may significantly underestimate true prevalence.