Experimental models of arthritis and their human counterparts fall into three distinct classes: (a) responses of T cells to disseminated microbial antigens (Ags) as such; (b) responses of T cells to cartilage autoAgs; and (c) responses of T cells to major histocompatibility complex (HLA-B27, DRB1) or other membrane components (LFA-1) expressed on bone marrow-derived cells. The primary immune response is driven, in naturally occurring disease, by microbial infection, e.g. with streptococci, enteric gram-negative rods or spirochetes, or is experimentally induced with mycobacterial and other adjuvants. The response to cartilage components, such as collagen type-II and various proteoglycans, may be driven by cross-reactive microbial Ags, heat shock proteins (HSPs) in particular, or the adjuvant effect of intense primary joint inflammation, as in rheumatoid arthritis and the spondyloarthropathies. Adjuvant disease appears to be purely T-cell-mediated, whereas both T cells and antibody play a role in collagen and many other forms of arthritis. Experimental evidence suggests a pathogenetic role for T-cell receptor gammadelta T cells in some lesions. Arthritis may be regulated by microbial and tissue HSPs, when these are administered by a nonimmunizing route or as altered peptide ligands, by anti-idiotypic responses that block the action of effector T cells, and by competing Ags. Immune regulation involving natural killer (NK), NK T and certain subsets of gammadelta and alphabeta T cells, which may affect the occurrence, localization and character of this group of diseases, presents a challenge for further investigation.