This paper critically examines the widely accepted idea on children's spatial egocentrism. According to the idea which Piaget and Inhelder (1956) advanced, children adhere to their own view when they are required to infer a view from positions other than their own. For the purpose of the critical examination of the empirical validity of the “adherence idea”, the data used by Piaget and other researchers concerning children's performance in the “three-mountains task” are reviewed. I claim that those data demonstrate that children represent mentally a spatial array in relation to its surroundings rather than to their own viewpoint, which is inconsistent with the “adherence idea.” It is concluded, therefore, that children's developmental change should be understood as their acquisition of the ability to generate a spatial representation without relying on the spatial context.