The Voyager 2 cosmic ray experiment observed intense electron fluxes in the middle magnetosphere of Uranus. High counting rates in several of the solid-state detectors precluded the normal multiple coincidence analysis used for cosmic ray observations, and we have therefore performed laboratory measurements of the single-detector response to electrons. These calibrations allow a deconvolution from the counting rate data of the electron energy spectrum between energies of about 0.7 and 2.5 MeV. We present model fits to the differential intensity spectra from observations between L values of 6 and 15. The spectra are well represented by power laws in kinetic energy with spectral indices between 5 and 7. The phase space density at fixed values of the first two adiabatic invariants generally increases with L, indicative of an external source. However, there are also local minima associated with the satellites Ariel and Umbriel, indicating either a local source or an effective source due to nonconservation of the first two adiabatic invariants. For electrons which mirror at the highest magnetic latitudes, the local minimum associated with Ariel is radially displaced from the minimum L of that satellite by ~0.5. The latitude variation of the satellite absorption efficiency predicts that if satellite losses are replenished primarily by raidal diffusion there should be an increasing pitch angle anisotropy with decreasing L. The uniformity in the observed anisotropy outside the absorption regions then suggests that it is maintained by pitch angle diffusion. The effective source due to pitch angle diffusion is insufficient to cause the phase space density minimum associated with Ariel. Model solutions of the simultaneous radial and pitch angle diffusion equation show that the displacement of the high-latitude Ariel signature is also consistent with a larger effective source. This source may be created by inelastic scattering, leading to diffusion in energy as well as pitch angle. ¿1991 American Geophysical Union