We compared three methods for estimating the location of a radiotransmitter, based on bearings to the transmitter from three radio receiving towers. One was to estimate transmitter locations from the arithmetic averages of the (Cartesian) coordinates of the three points of intersection of the bearings (Arithmetic Average method). The Priddel method is similar but uses the midpoint of the shortest side of the triangle formed by the three bearings when this side is less than half the length of each of the other two sides. The Lenth method is based on a probabilistic model of the system and uses the maximum likehood estimates of transmitter locations. No significant differences in accuracy were found between the Priddel and Lenth methods, but both were superior to the Arithmetic Average method. Estimates of transmitter locations were more accurate: (1) when three rather than two bearings were obtained; (2) for fixed rather than hand-held transmitters; (3) for hand-held transmitters than for transmitters on the backs of birds. By excluding cases where fewer than three simultaneous bearings were recorded, and those in which the bearing triangle was relatively large, the average distance between estimated and true locations of transmitter-carrying birds was estimated to be 15 m when the birds were on our study grid. This is small relative to the dimensions (330 by 330 m) of the grid. Accuracy was lower for birds that were off the grid.