Discusses conditions that must be satisfied for a model in a wind-tunnel to give the same air-flow as a full-sized building. Reports two series of tests on interior and exterior air flow patterns, made on a full-sized building and a scale model of the building. Air flow patterns were observed using titanium tetrachloride smoke. Tests were also made to determine the limits by which the product of the height of the model by the air speed may vary without serious error. Concludes that it is possible to use convenient size models and convenient air speeds to predetermine the natural ventilation characteristics of buildings. Finds that two conditions must be observed; changes in density and viscosity of the air must be negligible and the effects of thermal convection must be negligible.