We tested 30 prototype global positioning system (GPS) radiocollars on brown bears (Ursus arctos) over a 3-year period on the Kenai Peninsula, Alaska. Collars were of 2 design types: GPS units with an Argos (Argos Data collection and Location System) satellite uplink (n = 19) and GPS units where the data were stored on board (n = 10) for retrieval at a later date. All units also contained a conventional VHF (very high frequency) transmitter and weighed 1.7 kg. GPS-Argos units obtained 10-82% of expected GPS fixes, and fix rate declined significantly (P < 0.05) with time after deployment. Argos uplink success (proportion of successful transmissions of stored data) was linearly related to GPS fix rate (r = 0.91, P < 0.001). Storeon-board units obtained significantly more successful fixes when compared with the GPS-Argos units (t = -4.009, P < 0.001). Fix success rate for deployed store-on-board collars ranged from 13-96%; because of the increased number of attempted fixes per day, these collars obtained fixes on 97% of days deployed. Accuracy of the GPS units was less than predicted by the NAVSTAR GPS technology using the course acquisition code. Reduced accuracy was likely a result of the proportion of 2-dimensional versus 3-dimensional fixes obtained, although we could not determine this statistic from recorded data. Increased overstory closure was the only variable measured that partially explained the reduced likelihood of a successful fix. Stem density, stem diameter, and overstory height measured within 3 m of the collar did not affect fix success. GPS fix success rates for collars attached to bears varied more and were lower than fix rates for stationary collars placed in various vegetation types, suggesting that the bear, terrain, and movement all influence both fix and uplink success rate. Application of this new technology to grizzly and brown bear research and comparisons to studies with moose (Alces alces) are discussed.