Adjuvant-induced arthritis in the rat was considered as an agent-host disease and was studied from 3 major aspects: properties of the agent, distribution of agent in host, and the manifestations of host response. To induce arthritis in the rat, the adjuvant must consist of oily suspension of mycobacteria, nocardia, or other bacteria with lipid or hydrophobic surfaces. The intradermal route is the most effective. The intravenous route is relatively ineffective and blocks the induction of arthritis by a subsequent intradermal injection. The intracutaneous depot and the granulomatous response in the host is unnecessary and may inhibit the arthritis. C14 labelled mycobacteria rapidly leave the intradermal injection site attaining the highest concentration in the lymph nodes. After the ineffectual intravenous injection, the lymph nodes contain a similar concentration of C14 but other tissues, and especially the lung, liver and spleen, achieve even higher concentrations. There are only 3 possible mechanisms for adjuvant-induced arthritis-sensitivity to mycobacteria, auto-immunity and a microorganismal infection. The disease resembles an activated infection in its course, chronicity and morphology, but a culturable or transmissible agent is not demonstrable. Limited transmission of the arthritis with splenic or lymph node cells from adjuvant-injected donors was achieved, but such cellular suspensions also contain mycobacteria or their products. Delayed hypersensitivity to old tuberculin of M. butyricum, as determined by cutaneous or corneal reaction, does not appear to have an important role since it can be dissociated from the occurrence of the arthritis. The cutaneous reaction may be present without arthritis and absent in the presence of arthritis induced by adjuvant. While the pathogenesis of adjuvant-induced arthritis remains uncertain, the disease appears to be a peculiar response to myco-bacterial products which are injected intradermally in oil and which attain relatively high concentration only in the lymph nodes. Because of the morphological dissimilarity, adjuvant-induced arthritis is not the ideal experimental model for the study of the non-infectious arthritides in man.