Many cognitive studies have examined how to compute, map, or describe the spatial representations of spatial terms. Many of these studies have examined spatial terms that have meanings appropriate for projection onto a two-dimensional upright (2-D) plane. However, the world in which we live and apply spatial terms to is three-dimensional (3-D). This study focused on computing the spatial meanings of projective spatial terms on a level plane in 3-D space from an egocentric viewpoint. Three Japanese spatial terms, ushiro, saki, and mae (similar to behind, ahead and far, and in front of, respectively, in English), were examined in three experiments based on 3-D computer graphics virtual space. This study suggested that prototypical points of certain kinds of projective spatial terms such as mae and ushiro can be computed on a level plane in 3-D space using angles defined by viewpoint and reference object positions under the same spatial condition as used in this study.