The rheumatoid factor (RF) is a frequently ordered diagnostic test, yet it possesses significant limitations in sensitivity, specificity, and predictive value. Recognition of these limitations could improve the test's utility by encouraging more selective test ordering and more circumspect interpretation of test results. An analysis of 563 requests for RF from a teaching hospital revealed a positive predictive value of only 24% to 34%. The RF performs best under conditions of moderate pretest likelihood of rheumatoid arthritis, and otherwise has rather limited clinical utility.