We assessed the performance of high-frequency three-component stations for estimating bearing of regional phases using P waves from 18 mining explosions at a single construction site in Norway. We studied errors in bearing estimates as a function of the parameters frequency, frequency bandwidth, SNR, and analysis window length using seismic signals recorded on a three-component high-frequency (> 50 Hz) seismic element (HFSE) located at the NORESS array in Norway. We found that the time-bandwidth product (TB) and SNR are the most important factors in reducing the bearing estimation error. Bearing errors were minimized for SNR ≧ a factor of 5, and TB ≧ 100. We also found that three-component signals contain bearing information at frequencies as high as 45 Hz when there is sufficient SNR in the signal. Bearing estimation performance can be significantly improved by the appropriate selection of signal parameters, such as window length and frequency. Since these parameters are usually unknown, we propose a parameter variation technique that extracts the best answer by first determining the bearing distribution as a function of the above parameters. The results of this study indicate that with a single three-component station and for high (≧ 5) SNR, the parameter variation method can determine the bearings of events to within a standard deviation of 6°, which is similar to the uncertainty in NORESS array bearing estimates for the same events.