A study is made of the extent to which the high-frequency cyclical direction-finder devised by Earp and Godfrey is subject to three common types of error. First, the result is presented of theoretical computations on the errors given by a practical fixed-aerial system in wave-interference conditions. The variance of these errors as the phase difference between the incident rays is varied is shown to be the same as that given by the ideal rotating-aerial version of this direction-finder; it is appreciably smaller than that of an Adcock for a system aperture of 4λ, and, in general, much smaller for an aperture of 10λ. Secondly, errors due to aerial interaction are considered for a system of 24 unipoles, 9m high and 1.1 m in diameter, placed on a circle of 100m diameter. It is concluded that the errors in indicated bearing are likely to be negligible, provided that the aerials not in use at any instant are terminated by a resistor equal to their nominal characteristic impedance. Finally, the polarization error is examined for such a system in which surface feeders are used, without an earth mat, on ground of moderate conductivity; it is found to be small, being greatest at the low-frequency end of the band, where the standard-wave error is about 1° at 3 Mc/s. With extended feeders the errors at low frequencies are reduced considerably.