Diagnosis and treatment of infection is a common procedure in the clinical management of patients in the ICU. Infection in the ICU is an important area for study, but requires well-defined and proven diagnostic criteria. The diagnosis of infection, like any diagnosis, is based on probability, and diagnostic criteria are therefore selected according to the physician's objectives and the acceptable margin of error. It is easier to diagnose correctly a full-blown, severe bacterial infection than one that is just beginning, and the same criteria cannot be used to identify accurately both conditions. We should diagnose an infectious complication at the time it needs treatment, but there is often a lack of clear objectives in the diagnostic process, and up to now, few reliable criteria have been available. Before considering the sensitivity and specificity of single diagnostic procedures it is important to trace the evolution of the infection. The problem may be approached in two steps, by describing or defining (i) the minimum level of severity of a probable infection which requires/justifies specific treatment as the first end-point of the diagnosis, and (ii) the ways the diagnosis may be confirmed using the best available procedure (which might not be always available or applicable in all cases in the short term).