Typing for histocompatibility antigen HLA-B27 has been suggested as a useful diagnostic test for ankylosing spondylitis (AS) in certain clinical situations. The appropriate use of any diagnostic test requires the clinician to estimate the likelihood of disease before the test is performed. One clinical situation in which B27 testing has been suggested to be useful is in the investigation of a patient with low back pain suggestive of AS but with normal sacroiliac radiographs. We analyze here the sequence of steps taken by the clinican in estimating the likelihood of AS. The assumptions that must be made to render B27 typing useful are calculated.