We addressed concerns regarding performance of various Global Positioning System (GPS) collar configurations for describing habitat use by Rocky Mountain elk (Cervus elaphus) in rugged, forested terrain. We tested 8 GPS collars (Lotek Wireless, Newmarket, ON, Canada) in 4 different model and equipment configurations at 2 reference points (an open hilltop and a forested ravine) to determine habitat-specific differences in performance among collar configurations. We then placed individual collars at 60 additional points that were stratified randomly among 4 canopy-cover classes and 3 classes of available sky. All collars exhibited a locational bias of 4 m horizontally west and of 10 m vertically below a reference standard established by position-averaging with a handheld receiver (Garmin 12MAP) calibrated at National Geodetic Survey benchmarks. The GPS collar models that were programmed for longer satellite-acquisition times provided greater location precision than models that had been programmed for short acquisition times to preserve battery power. Canopy cover and available sky had a greater effect on collar location precision and observation rates than slope, slope position, aspect, conifer basal area, tree height, canopy depth, or elevation. Researchers should test collars at known reference points to confirm that location precision and rates of observation are adequate for their particular study objectives. Manufacturers of GPS collars should inform clients of their programming criteria for acquisition time so that customers can make informed decisions regarding trade-offs between precision of locations, data quantity, and battery life.