Tuberculosis (TB) has re-emerged over the past two decades: in industrialized countries in association with immigration, and in Africa owing to the human immunodeficiency virus epidemic. Drug-resistant TB is a major threat worldwide. The variable and uncertain impact of TB control necessitates not only better tools (diagnostics, drugs, and vaccines), but also better insights into the natural history and epidemiology of TB. Molecular epidemiological studies over the last two decades have contributed to such insights by answering long-standing questions, such as the proportion of cases attributable to recent transmission, risk factors for recent transmission, the occurrence of multiple Mycobacterium tuberculosis infection, and the proportion of recurrent TB cases attributable to re-infection. M. tuberculosis lineages have been identified and shown to be associated with geographical origin. The Beijing genotype is strongly associated with multidrug resistance, and may have escaped from bacille Calmette-Guérin-induced immunity. DNA fingerprinting has quantified the importance of institutional transmission and laboratory cross-contamination, and has helped to focus contact investigations. Questions to be answered in the near future with whole genome sequencing include identification of chains of transmission within clusters of patients, more precise quantification of mixed infection, and transmission probabilities and rates of progression from infection to disease of various M. tuberculosis lineages, as well as possible variations in vaccine efficacy by lineage. Perhaps most importantly, dynamics in the population structure of M. tuberculosis in response to control measures in high-prevalence areas should be better understood.