The flows that may result from the release of a buoyant fluid from a small source in a confined space are varied and complex, depending on the source characteristics, the confining geometry and container ventilation. Previous work has generally been based on the 'filling-box' model (Baines & Turner 1969). This model, however, may only be applied when the source has little or no initial momentum, is in a container with a height/width aspect ratio less than unity and does not interact with the side boundaries. In this thesis some situations in which the 'filling-box' model may not be applied are investigated. In chapter 1 the 'filling-box' model and the work based on it are reviewed and its limitations discussed. Sources are usually modelled as a turbulent jet, plume or buoyant jet; thus in chapter 2 the established properties of jets and plumes are summarised using established theoretical arguments and experimental results. In order to improve some of the theoretical predictions, the effects of previously neglected second order terms and intermittency factor variation are investigated. In most practical situations the source is a buoyant jet. In chapter 3 the flow of an initially horizontal buoyant jet is examined, concentrating on obtaining simple analytical results from the conservation equations and investigating the effect of the nature of the entrainment assumption. The effect of source momentum is examined in chapter 4, in which the flow of a vertical jet with high initial momentum flux in a long, ventilated tunnel is studied. The bulk flow variables are deduced and compared with experimental measurements. A method for calculating the total number of vents required to vent the source fluid is given - this result is particularly important in applications to hazardous releases. In chapter 5 the effect of the aspect ratio, a, on the 'filling-box' model is investigated. It is found that when the aspect ratio is very large (a > 6) the flow is quite different to that observed in the 'filling-box' case. Theoretical models are derived both when a > 6 and when a < 6, and the results are compared with experimental measurements. In chapter 6 the study of chapter 5 is extended, investigating the effect of placing the tank at an angle to the vertical. The flow in the large aspect ratio case is significantly different to that of chapter 5 and is modelled theoretically, comparing the predictions with experimental measurements. Finally, in chapter 7 the general results are reviewed. The current knowledge of flows resulting from buoyant releases in a confined space is then summarised in a simple tabular form, which also indicates areas for future research.