When a visual search target frequently appears in one target-rich region of space, participants learn to search there first, resulting in faster reaction time when the target appears there than when it appears elsewhere . Most research on this location probability learning (LPL) effect uses 2-dimensional (2D) search environments that are distinct from real-world search contexts, and the few studies on LPL in 3-dimensional (3D) contexts include complex visual cues or foraging tasks and therefore may not tap into the same habit-like learning mechanism as 2D LPL . The present study aimed to establish a baseline evaluation of LPL in controlled 3D search environments using virtual reality . The use of a virtual 3D search environment allowed us to compare LPL for information within a participant's initial field of view to LPL for information behind participants, outside of the initial field of view . Participants searched for a letter T on the ground among letter Ls in a large virtual space that was devoid of complex visual cues or landmarks . The T appeared in one target-rich quadrant of the floor space on half of the trials during the training phase . The target-rich quadrant appeared in front of half of the participants and behind the other half . LPL was considerably greater in the former condition than in the latter . This reveals an important constraint on LPL in real-world environments and indicates that consistent search patterns and consistent egocentric spatial coding are essential for this form of visual statistical learning in 3D environments.