We evaluated the performance of two GPS collars as a tool for studying Japanese black bears (Ursus thibetanus japonicus) in the Kanto region, Japan. Collars were placed at five stationary points to test the effects of GPS antenna orientation and degree of canopy closure on the fix rate and the positional error, and collars were deployed on 16 free-ranging bears to examine the fix rate and the effects of fix interval on fix rate. In the stationary tests, the proportions of no fix and of 2D fix (vs. 3D fix) increased as canopy closure increased and antenna orientation was further from vertical; at an antenna angle ≥90° from the vertical and canopy closure of 76.2–79.1%, the fix rate was <50%. The positional error for 3D fixes (ranged from 6.84 to 16.43 m) was significantly lower than that for 2D fixes (ranged from 30.74 to 43.45 m), and the positional error for both was affected by canopy closure and GPS antenna angle. An expected multipath effect on positional errors by rain was not statistically significant. Mean fix rates on collared bears ranged from 23.6% to 56.4% and were significantly lower than those in the stationary test. The fix rates at 5-min intervals were significantly higher than those at 4-hr intervals. Although our tests revealed some limitations and biases, GPS collar appears suitable for studying the movement and behavior of the bears.