Although a number of inviscid numerical methods have been developed for the calculation of three-dimensional transonic flow in turbomachinery blade passages it is often difficult to compare the results of these calculations directly with experiment. For instance in real machines the overall flow is very complex and it is usually impossible to model the full geometry of the machine, thus it is difficult to trace the sources of the discrepancies which inevitably arise between the measured and calculated results. Even in simpler test cases the experimental results are often strongly affected by viscous effects and flow separations. This paper presents the results of an experimental investigation which generates three-dimensional transonic flow with shock waves using relatively simple geometries and where the influence of viscous effects has been reduced as far as possible. Comparisons with the output of various calculation methods show that even with these simple geometries it is essential to use fully three-dimensional methods to predict the flow development.