Rabbit antisera to certain strains of Gram-negative bacteria are reported to be cytotoxic for the lymphocytes of about 80% of HLA-B27 positive patients with ankylosing spondylitis (AS) but not for the lymphocytes of healthy HLA-B27 positive individuals. The lymphocytes of normal individuals can, however, be made susceptible to lysis by antibacterial sera by incubation in spent supernatant from appropriate bacterial cultures. In an attempt to explain the failure of certain laboratories to reproduce these results we have tested a number of variables in the 51Cr-release complement-dependent cytotoxicity assay. We conclude that AS patients in London and Sydney carry the same antigen, that several different incubation media can be used for both the cytotoxicity assay of HLA-B27 positive AS cells and modification of normal B27 positive cells and that the cells used may be collected either in heparin containing media or after defibrination. Major requirements for success include a healthy sample of lymphocytes (preferably at a fairly high concentration), suitable antiserum from a rabbit repeatedly immunised with large numbers of bacteria, appropriate complement and efficient technique. However, as we have failed to repeat this test consistently in London, even using materials tested in this study, it seems that another, probably non-biological, factor is also important.