The flows in a model of a ventricular assist device (VAD) were investigated numerically and experimentally for two different Newtonian test fluids. These were a blood analogue fluid and a much higher viscosity fluid. A finite volume method was employed to solve the governing equations for a three-dimensional unsteady laminar flow on a transient grid. The numerical solutions were compared with experimental results from an identical physical model. The experimental flows were investigated by flow visualization and by laser Doppler velocity measurements at selected points in the flow field. The validation was based on comparisons of flow patterns and of non-component velocity-time histories. The maximum Reynolds numbers in the inflow tube of the model VAD were approximately 460 and 3300 using the high- and low-viscosity fluids respectively. The investigation showed that the flow patterns were better predicted for the high-viscosity fluid. However, the agreement between the velocity-time histories was found to be slightly better for the low-viscosity fluid. The discrepancies in the flow patterns may be due to intermittent turbulence with a further contribution from numerical diffusion.