This experiment investigated the effect of children's successive encounters with a large scale environment on their subsequent reconstructions of that environment. Twenty children (10 boys, 10 girls) at each of three grade levels (kindergarten, two, and five) reconstructed from memory the spatial layout of buildings in a large model town. All children walked through the model three times. Half the children at each grade level constructed the town after each walk; the other half constructed the town only after the third walk. Accuracy of construction improved as a function of motor experience with the town (number of walks) for children at all grade levels. On the first construction, fifth graders placed buildings more accurately on both topological and Euclidean measures than did younger children. On successive constructions, age differences diminished greatly on both measures. No significant sex differences in performance were found in any of the analyses. Results are discussed in relation to previous research and theory concerning children's understanding of spatial relationships. (Author/SB)