SummaryFever of unknown origin strictly defined by Petersdorf in 1961 may be caused by a whole array of conditions. A rational approach should be based upon the relative frequencies of the different causes and their importance for the health and life of the patient. Risks, discomfort and, to a limited degree, costs must be taken into consideration.The investigation protocol we propose takes into account all these factors and is mainly based upon an extensive literature survey and our personal experience with a series of 199 patients studied in the 1980 s.