Background Physical examination is a traditional outcome measure in epidemiological research. Its value as a reliable measure depends, in part, on the prevalence of positive ยฎndings. The purpose of this paper is to determine the empirical reliability of physical examination and anthropometry in a ยฎeld study of upper extremity disorders among keyboard operators. Methods Two experienced examiners independently performed common provocative tests and procedures in physical examinations of the neck and upper extremity among 160 keyboard operators. Two additional examiners conducted anthropometric surveys among 137 workers. Inter-examiner reliability was assessed with observed agreement, kappa statistics, and intra-class correlations (ICC). Results Observed agreement was between 96% and 100% for neck and upper extremity signs, muscle stretch reยฏexes, and muscle strength, however, with the exception of provocative tests, reliability statistics were unstable. Among the provocative tests, Phalen and Tinel tests had modest agreement after adjusting for chance ( range: 0.20ยฑ0.43). The carpal compression test had the best reliability ( 0.60 and 0.67, left and right side, respectively). The ICCs for anthropometry ranged from 0.36ยฑ0.91. Conclusions Results from the study showed that statistically, except for the carpal compression test, physical examination contributed minimal reliable information. This was attributed mainly to the low prevalence of positive ยฎndings, and generally mild nature of upper extremity disorders in this population. The results are the best estimate of what would be found in a ยฎeld study with experienced examiners. While it may reduce bias, separating physical examination from medical history may contribute to the poor reliability of ยฎndings. With a shift toward reliable measures, resources can be allocated to more effective tools, like questionnaires, in epidemiological research of upper extremity disorders among keyboard operators.