Typing for histocompatibility antigen (HLA)-B27 has been suggested as a clinically valuable diagnostic test for ankylosing spondylitis and Reiter's syndrome, although some decry its use for this purpose. Diagnoses can be made in most patients with these diseases on the basis of the history, physical examination, and roentgenographic findings. The B27 test cannot be used to screen an asymptomatic population to detect these diseases and should not be thought of as a routine diagnostic test. We present probability graphs derived from Bayes' theorem, which show that for certain patients the B27 test, when used properly, is of clinical value as an aid to diagnosis. Proper application of the B27 test in clinical medicine is discussed. The test result does not absolutely confirm or exclude the presence of these diseases; it merely provides a probability statement on their existence in the patient. The test is therefore most useful to physicians who understand the use of probability reasoning in clinical decision making.