Objective: Fever is one of the most common causes of hospital admissions. In presence of fever, a number of laboratory tests, radiological imaging and treatment are considered. The aim of this study is to investigate the accuracy of subjective fever perception and estimation of fever by touching by the health personnel. Material and Methods: Sixteen years and above patients with or without fever who were admitted to Hacettepe University Medical Faculty Emergency Department whose fever was not measured with a termometer before coming to hospital were included in the study between February 1, 2010 and April 30, 2010. The study was double blind and prospective. Results: A total of 390 patients, 213 (55%) with fever and 177 (45%) without fever participated in the study. Of subjective fever perception of patients, sensitivity was 88%, selectivity was 51%, positive predictive value (PPV) was 68%, negative predictive value (NPV) was 78%, accuracy was 71%. For accurate estimation of fever with 0.5 degrees C of difference by touching, sensitivity was 85%, selectivity was 91%, PPV was 92%, NPV was 84% for specialists; sensitivity was 87%, selectivity was 92%, PPV was 93%, NPV was 86% for nurses; sensitivity was 84%, selectivity was 90%, PPV was 91%, NPV was 82% for resident doctors; sensitivity was 80%, selectivity was 91%, PPV was 91%, NPV was 79% for interns. Ratio of accurate estimation of fever was 89% for nurses, 88% for specialists, 87% for resident doctors, 85% for interns, and there was not a significant difference between them (p=0.356). Conclusion: Because of high PPV and NPV values, both subjective perception of the patient and estimation of fever by touching by health personnel may be taken into consideration for decision of diagnosis and treatment initiation when a termometer cannot be used, for example in case of infections and rheumatic diseases involving skin and joints.