Institutional transmission of tuberculosis is an important public health problem, especially in low resource settings where protective measures such as negativepressure isolation rooms are difficult to implement. Natural ventilation by simply opening windows and doors is recommended by the World Health Organisation for tuberculosis control in resource limited settings, but efficacy is poorly defined. In the first part of this thesis, the rates and determinants of natural ventilation were investigated in a wide variety of clinical settings in 8 hospitals in Lima, Peru. Natural ventilation was measured in 70 clinical rooms using a carbon dioxide tracer-gas technique, compared with 12 mechanically ventilated negative-pressure rooms. Results were used to model tuberculosis transmission risk. Natural ventilation by opening windows and doors resulted in high rates of air exchange, especially in oldfashioned wards with high ceilings and large windows. Modelling predicted reduced tuberculosis transmission with natural ventilation compared with mechanical ventilation. However, natural ventilation is not appropriate in cold climates and other environmental controls are needed. Evaluation of such controls is complicated by the difficulty of measuring M. tuberculosis in air. Therefore the second part of this thesis deals with the creation and optimisation of a guinea pig air sampling facility to detect airborne tuberculosis transmission. Situated above a tuberculosis ward, an average 92 guinea pigs over 16 months breathed exhaust air from the ward which was occupied by 97 patients co-infected with HIV and pulmonary tuberculosis. DNA fingerprinting of animal and patient tuberculosis strains demonstrated great heterogeneity of patient infectiousness, tuberculosis transmission from a minority of patients, and highly infectious individual multidrug-resistant TB-HIV cases. This guinea pig air sampling model is now in use evaluating upper room ultraviolet light and negative air ionisation for reducing institutional tuberculosis transmission.