A revision of the criteria of fever of unknown origin (FUO), established in 1961, is desirable because of important evolutions in medical practice and the emergence of new patient populations. The development of rapid laboratory tests and powerful diagnostic tools, such as ultrasonography, computed tomography and magnetic resonance imaging often makes hospitalization unnecessary and new categories of patients such as those with HIV infection, neutropenia, immunosuppression and nosocomial illness require an approach different from classical FUO. The more then 200 reported causes of FUO can be classified into four diagnostic categories; infections, tumours, noninfectious inflammatory diseases (NIID) and miscellaneous. A uniform classification system is highly wanted to allow comparison between different series. The reports of the 1990s show slight changes in the distribution of causes, namely less infections, less tumours, more NIID and more undiagnosed cases. A uniform diagnostic strategy cannot be determined. The initial investigation should be directed by potentially diagnostic clues revealed by extensive history, meticulous physical examination and a standard set of laboratory tests. 18Fluoro-deoxy-glucose-positron-emitted-tomography is a new valuable total body scintigraphy in the search for the site of origin of the fever. In view of the rather good long-term prognosis, a wait-and-see strategy may be more appropriate than a systematic staged approach. Elderly patients and patients with episodic fever represent two specific groups of classical FUO that require a distinct approach. HIV-associated, nosocomial and neutropenic FUO should be considered as separate clinical entities.