Despite the overwhelming success of immunization in reducing, and even eliminating, the global threats posed by a wide spectrum of infectious diseases, attempts to do the same for tuberculosis (TB) have failed to date. While most effective vaccines act by eliciting neutralizing antibodies, T cells are the primary mediators of adaptive immunity against TB. Unfortunately, the onset of the T cell response after aerosol infection with Mycobacterium tuberculosis (Mtb), the bacterium that causes TB, is exceedingly slow, and systemically administered vaccines only modestly accelerate the recruitment of effector T cells to the lungs. This delay seems to be orchestrated by Mtb itself to prolong the period of unrestricted bacterial replication in the lung that characterizes the innate phase of the response. When T cells finally arrive at the site of infection, multiple layers of regulation have been established that limit the ability of T cells to control or eradicate Mtb. From this understanding, emerges a strategy for achieving immunity. Lung resident memory T cells may recognize Mtb-infected cells shortly after infection and confer protection before regulatory networks are allowed to develop. Early studies using vaccines that elicit lung resident T cells by targeting the lung mucosa have been promising, but many questions remain. Due to the fundamental nature of these questions, and the need to understand and manipulate the early events in the lung after aerosol infection, only coordinated approaches that utilize tractable animal models to inform human TB vaccine trials will move the field towards its goal.