Animal models for tuberculosis vaccine assays have evolved through a series of sequential experiments aimed at optimizing the activity of a particular vaccine product. As a result, studies that use different animal models do not agree on the potency ranking of antituberculosis vaccines, and each major test-system variable contributes to the disagreement. Disagreements among laboratories about the efficacy of vaccines are in part due to differences in test systems. A survey of potency assays of tuberculosis vaccines suggests that, based on the choice of a specific combination of variables in the test model, an investigator can show that any given vaccine product has superior potency. In view of these problems with animal models for research on tuberculosis vaccines, we recommend that attention should be focused on those animal models that replicate the key aspects of the natural history of human tuberculosis. The development of vaccines by means of new technologies requires animal models to assay the protective potency of vaccines that, ideally, inhibit tubercle bacilli at the point of infection. The development of new tuberculosis vaccines may be handled most efficiently in a joint venture that includes both laboratories involved in vaccine production and laboratories concerned with animal models.