One of the primary public health functions of a tuberculosis (TB) program is to arrest the spread of infection. Traditionally, TB programs have relied on epidemiological information, gathered through contact tracing, to infer that transmission has occurred between people. The ability of drawing such inferences is extensively context dependent. Where epidemiological information has been strong, such as 2 cases of TB occurring sequentially within a single household, confidence in such inferences is high; conversely, public health authorities have been less certain about the significance of TB cases merely occurring in the same wider social group or geographic area. Many current laboratory tests for TB used globally may be sufficient to confirm a diagnosis and guide appropriate therapy but still be insufficiently precise for distinguishing two strains reliably. In short, drawing inferences regarding a chain of transmissions has always been as much art as science.