This thesis presents a detailed investigation of the capability of a modern three-dimensional Navier-Stokes solver to predict the secondary flows and losses in a linear cascade of high turning turbine rotor blades. Three codes were initially tested, to permit selection of the best of the available numerical solvers for this case. This program was then tested in more detail. Results showed that although very accurate prediction of the effects of inviscid fluid mechanics is now possible, the Reynolds stress modelling can have profound effects upon the quality of the solutions obtained. Solutions using two different calculation meshes, have shown that the results are not significantly grid dependent. The flowfield of the cascade was traversed with hot-wires to obtain measurements of the turbulent Reynolds stresses. A turbulence generating grid was placed upstream of the cascade, to produce a more realistic inlet turbulence intensity. Results showed that regions of high turbulent kinetic energy are associated with regions of high total pressure loss. Calculation of eddy viscosities from the Reynolds stresses showed that downstream of the -cascade the eddy viscosity is fairly isotropic. Evaluation of terms in the kinetic energy equation, also indicated that both the normal and shear Reynolds stresses are important as loss producing mechanisms in the downstream flow. The experimental Reynolds stresses have been compared with those calculated from the eddy viscosity and velocity fields of Navier-Stokes predictions using a mixing length turbulence model, a one equation model, and K - ϵ model. It was found that in the separated, shear flows, agreement was poor, although the K - ϵ model performed best. Further experimental work is suggested to obtain data with which to determine the accuracy of the models within the blade and endwall boundary layers.