One third of the world's population is estimated to be infected with Mycobacterium tuberculosis, representing a huge reservoir of potential tuberculosis (TB) disease. Risk of progression to active TB is highest in those with HIV coinfection. However, the nature of the host-pathogen relationship in those with "latent TB infection" and how this is affected by HIV coinfection are poorly understood. The traditional paradigm that distinguishes latent infection from active TB as distinct compartmentalised states is overly simplistic. Instead the host-pathogen relationship in "latent TB infection" is likely to represent a spectrum of immune responses, mycobacterial metabolic activity, and bacillary numbers. We propose that the impact of HIV infection might better be conceptualised as a shift of the spectrum towards poor immune control, higher mycobacterial metabolic activity, and greater organism load, with subsequent increased risk of progression to active disease. Here we discuss the evidence for such a model and the implications for interventions to control the HIV-associated TB epidemic.