Geographical slope perception is notoriously susceptible to error. Hills tend to appear steeper than they are. Similar errors occur for perceived optical slant of smaller surfaces (frontal tendency). Like perceived optical slant, perceived geographical slope increases with viewing distance. This suggests that slope overestimation may also be due to a frontal tendency in optical slant perception. Ooi et al. (2001) have argued that a frontal tendency makes even flat surfaces appear sloped. How then, do we ever see large flat surfaces as flat? One important cue that may distinguish hills from fields is that hilltops tend to occlude the objects beyond them, whereas fields do not. Although a large flat surface might end at a ravine from which trees emerge, and a gradual hill might end at the base of a building that it does not really occlude, the occlusion of distant objects is generally a reliable indicator that a surface is sloped. We created four immersive virtual scenes in which we manipulated the slope (from 0 to 7 deg) of a grassy surface (40 or 70 m long). We also manipulated the presence or absence of partial occlusion, by that surface, of familiar objects (elephants, person, bus, truck) that have parts (wheels or feet) that normally contact the ground. The objects were simulated as being on a horizontal surface atop the hill (no occlusion) or lowered by one meter as if into a ravine on the crest of the hill (partial occlusion). When occlusion was present verbally-judged slope was significantly steeper (by about 10 deg) than when occlusion was absent. Judgments in the occlusion condition were similar to those typically reported for real outdoor hills (e.g., 5 deg was judged to be 20 deg), whereas judgments in the non-occlusion condition were lower and thus more accurate.