A statistical method has been developed which can be used to distinguish between various possible explanations for the associations that have been observed between HLA and certain diseases. By analyzing the numbers of B27 positives with and without a second detectable B locus antigen in patients with ankylosing spondylitis and in a control series, three possible hypotheses have been tested: 1. the B27 antigen itself makes an individual more susceptible; 2. a recessive susceptibility gene is closely linked with HLA; 3. a dominant susceptibility gene is closely linked with HLA. In four groups of patients suffering from another disease the procedure has also been applied and it turns out that at least in some of these the second hypothesis can be rejected while the third seems to be the most likely.