Prior studies have indicated that on cognitive load, topological spatial terms are more easily selected than projective spatial terms as a selection mechanism, i.e., “ near ” tends to be preferred over “ left ” or “ right ”. To clarify the effect of geometric factors on spatial term selection, we performed a psychological experiment using 3D generated objects with sentences describing each object in spatial terms in Japanese, e.g., hidari ( left ), chikai ( near ), and tooi ( far ). Results showed that the characteristics for determining the acceptability rating of each spatial term varied by term and also by geometric factors such as the position or size of the object and the location of the distractor. We also performed two other experiments to investigate the effects of changing both point of view and the objects' size, with the goal of gaining a better understanding of some projective spatial terms in Japanese, mae ( front ), ushiro ( back ). These experiments confirmed the importance of both viewpoint and the objects' size as factors within the cognitive model. Finally, we discuss a new cognitive model for spatial term selection based on these findings.