This project consists of a two-pronged computational and experimental approach to the study of flow in closed, thin rectangular ducts with a partial cubic blockage. Results are presented at three different bulk Reynolds numbers, ReD = 5600, 10400 and 15600, based on the channel height, which is also the blockage dimension. The new experimental data produced consists of fluctuating pressure measurements at the cube surface, with 2D-2C PIV snapshots captured simultaneously in the wake region. In addition to this, DNS data is produced at the lowest Reynolds number of ReD = 5600, allowing more detailed comparisons where PIV laser access was not possible. Comparisons are drawn between the data and URANS CFD simulations. A literature review and preliminary testing process narrowed down the considered URANS models to the two-layer k−ε model and the Elliptic Blending Reynolds Stress Model, or EBRSM. In the light of the new data, these two URANS models are compared in order to better understand their strengths and weaknesses. Particular regard is given to the prediction of large-scale unsteady behaviour, with a focus on vortex shedding. This unsteady phenomenon was found to be present and to have a significant effect on the flow in the near-cube and wake regions. Results show that certain aspects of this behaviour are captured with only limited accuracy by the URANS models tested. As a result, inaccuracies are also found in the mean simulated velocity fields. The shortcomings appear more pronounced at higher flow rates. At a given flow rate, they are more severe in regions of the flow where organised unsteadiness is large relative to the mean values. It is suggested that inaccuracies in mean URANS predictions are a result of limitations in model capability for unsteady flows, and that validation cases may be pertinent to address this.