Five experiments examined the integration and competition between body and context objects in locating an object. Participants briefly viewed a target object in a virtual environment and detected whether the target object was moved or not after a 10 s interval. Experiments 1 and 2 showed that performance when both the observer body and the context objects were consistent across study and test was not better than the optimal sum of performances when either one was the only consistent cue across study and test. In Experiments 3 and 4, in the competition conditions, both the body and the context objects were reference points at learning but only one stayed consistent during test. In the no competition conditions, only the body or the context objects were the primary reference points in learning and it stayed consistent in test. Detection performance did not differ between these conditions. Experiment 5 demonstrated the integration and competition between context objects as a reference point. Detection performance based on all four context objects was better than the optimal sum of the performance based on two close context objects and the performance based on two far context objects; detection performance based on two context objects was better when there were only these two context objects during learning than when there were four context objects during learning. These results suggest that body-object (body-target) and interobject (context-target) vectors are encoded independently and combined at test in an optimal way. Body-object and interobject vectors are not encoded in an integrated way and encoding of them does not compete. By contrast multiple interobject vectors are encoded in an integrated way in addition to the representations of individual interobject vectors and encoding close interobject vectors and encoding far interobject vectors interfere with each other. (Contains 1 table and 14 figures.)