Detailed flow measurements made in a 1 metre diameter, shrouded, centrifugal (Ghost) impeller running at 500 rpm are presented. Relative velocities and rotary stagnation pressures $(p\ast =p-12\backslash \rho \omega 2r2+12\backslash \rho W2)$ were measured on five cross-sectional planes between the impeller inlet and the outlet, using probes which were traversed within the rotating impeller passage. The reduced static pressures $(pr=p-12\backslash \rho \omega 2r2)$, calculated from the flow measurements, are also presented. Measurements were made in a 'design' flow (approximately zero incidence at the blade leading edge), a 'below design' flow and an 'above design' flow. A wake flow was observed in all three flows and there were two major sources of the wake fluid. Firstly, from the separation of the shroud boundary layer and secondly, from the accumulation of low p* fluid from the other boundary layers by secondary flows. The results showed that the wake's position at the outlet moved from the suction side in the 'below design' flow, to the suction-side/shroud corner region in the 'design' flow and to the shroud in the 'above design' flow, because of the change in the relative strengths of the secondary flows generated by rotation and curvature. The modifications to turbulent mixing, by curvature and rotation, probably influenced the wake size. In order to predict the wake's location at the impeller discharge, a simple secondary flow model, which represented the impeller as a pipe bend, was devised. This model was successfully tested on two analytically soluble flows, in a stationary bend and in a rotating straight pipe. The model was then used for the more complex flows in a rotating axial-to-radial bend and in the Ghost and Eckardt's centrifugal impellers. The theoretical results for these impellers showed several of the features observed in the flow measurements.