It was demonstrated previously that abrupt transfer of vigorously aerated cultures of Mycobacterium tuberculosis to anaerobic conditions resulted in their rapid death, but gradual depletion of available O2 permitted expression of increased tolerance to anaerobiosis. Those studies used a model based on adaptation of unagitated bacilli as they settled through a self-generated O2 gradient, but the model did not permit examination of homogeneous populations of bacilli during discrete stages in that adaptation. The present report describes a model based on culture of tubercle bacilli in deep liquid medium with very gentle stirring that keeps them in uniform dispersion while controlling the rate at which O2 is depleted. In this model, at least two stages of nonreplicating persistence were seen. The shift into first stage, designated NRP stage 1, occurred abruptly at a point when the declining dissolved O2 level approached 1% saturation. This microaerophilic stage was characterized by a slow rate of increase in turbidity without a corresponding increase in numbers of CFU or synthesis of DNA. However, a high rate of production of glycine dehydrogenase was initiated and sustained while the bacilli were in this state, and a steady ATP concentration was maintained. When the dissolved O2 content of the culture dropped below about 0.06% saturation, the bacilli shifted down abruptly to an anaerobic stage, designated NRP stage 2, in which no further increase in turbidity was seen and the concentration of glycine dehydrogenase declined markedly. The ability of bacilli in NRP stage 2 to survive anaerobically was dependent in part on having spent sufficient transit time in NRP stage 1. The effects of four antimicrobial agents on the bacilli depended on which of the different physiologic stages the bacilli occupied at a given time and reflected the recognized modes of action of these agents. It is suggested that the ability to shift down into one or both of the two nonreplicating stages, corresponding to microaerophilic and anaerobic persistence, is responsible for the ability of tubercle bacilli to lie dormant in the host for long periods of time, with the capacity to revive and activate disease at a later time. The model described here holds promise as a tool to help clarify events at the molecular level that permit the bacilli to persist under adverse conditions and to resume growth when conditions become favorable. The culture model presented here is also useful for screening drugs for the ability to kill tubercle bacilli in their different stages of nonreplicating persistence.