Laser Doppler anemometers (LDAs) that are arranged to measure nonorthogonal velocity components (from which orthogonal components are computed through transformation equations) are more susceptible to calibration and sampling errors than are systems with uncoupled channels. In this paper uncertainty methods and estimation theory are used to evaluate, respectively, the systematic and statistical errors that are present when such devices are applied to the measurement of mean velocities in turbulent flows. Statistical errors are estimated for two-channel LDA data that are either correlated or uncorrelated. For uncorrelated data the directional uncertainty of the measured velocity vector is considered for applications where mean streamline patterns are desired.