In an investigation of children's spatial knowledge of a large-scale environment, forty-eight 3- to 6-year-old children (an approximately equal number of boys and girls of each age) were taken through an environment by a specified route. Once the route and landmarks along the route were learned, children were tested on their ability to (1) travel the route in reverse (route reversal knowledge), (2) name the sequence of landmarks along the reverse route (landmark reversal knowledge), (3) infer the relationship between parts of the environment not directly traveled between (inference), and (4) construct a model of the environment. Results indicated that route-knowledge develops before landmark reversal knowledge, and inference ability develops last; also supported was the suggestion that young children's spatial representations are route-like and poorly integrated in comparison with those of older children. In their models, children often exaggerated the route they travelled. No sex