Natural ventilation has the potential to provide good indoor air quality, thermal comfort for occupants, and can also save energy and reduce CO2 emissions. Computational fluid dynamics (CFD) offers detailed information about indoor flow patterns, air movement, temperature and local draught distribution in buildings, so it has unique advantages as an efficient and cost-effective tool for optimum design in a complex built environment. This thesis shows the use of CFD to simulate the coupled external and internal flow field around a 6m cubic building with two small openings. To study both wind driven and combined wind and buoyancy driven cross ventilation through a full-scale cubic structure, un-structured grid CFD and a steady envelope flow model were applied to calculate mean ventilation rates. To validate the CFD results, full-scale experiments were undertaken under various weather conditions in England. For wind driven ventilation RANS model predictions were proved reliable when wind directions were near normal to the ventilation openings, i.e. 0o~30o. However, when the fluctuating ventilation played a more dominant role than the mean flow (90o) RANS models were incapable of predicting the total ventilation rate. Improved results may be expected by applying more sophisticated turbulence models, such as LES, weighted quasi-steady approximations, or unsteady envelope flow models. In the thesis experience on the modelling of combined wind and thermal effects is outlined and feedback is provided to CFD code developers to enable further improvements for building ventilation studies. The full-scale field testing data from this study is valuable for comparison with wind tunnel results and validation of CFD applications.