Polymerase chain reaction (PCR) represents a major breakthrough for the diagnosis of infectious diseases. However, the absence of standardized kits for commercially unattractive targets, such as most of the parasites and the fungi, has led to the development of numerous in-house PCR assays. The performances reported, both for the sensitivity and the specificity of these assays are very divergent. For instance, for the antenatal diagnosis of toxoplasmosis, the sensitivity is either 97.4%, or 64%. For the diagnosis of toxoplasmosis in HIV-positive patients, the PCR on blood is either of limited value with a sensitivity of 13% or of excellent yield with a sensitivity of 87.5%. Similar results are reported for the diagnosis of invasive aspergillosis in bone-marrow-transplant recipients. The patients and the clinical specimens tested are often different. This can explain some of the discrepancies. However, when performed, the quality controls on identical specimens show different results depending on the laboratories. An analysis of the PCR techniques used shows that the control of false positive results as a result of carry-over and false negative results owing to PCR inhibitors is far from being systematic. These shortcomings of 'classical' PCR should be solved when real-time PCR assays are developed, leading to some standardization. Automated DNA extraction should also be useful to achieve this goal. Comparison between laboratories should then be possible and regular quality controls will be necessary to ensure the reliability of real-time PCR assays.