This paper describes two experiments in which children in grades 1, 3, and 5 were given three kinds of spatial perspective-taking problems to solve as quickly as they could: (1) C problems, solvable only by computation (that is, noting which features of a particular object array were closest to another observer in order to estimate how the array appears from that observer's position); (2) R problems, solvable only by applying a rule (such as the generalization that two observers will have the same view of any object array if in the same viewing position and different views of it if in different viewing positions); and (3) RC problems, solvable either by computation or by rule. The task conditions were such that computation-based solutions ordinarily required several seconds to execute whereas rule-based solutions could be achieved with zero-order latencies. Older children proved likelier than younger ones to solve R problems correctly, solve them with zero-order latencies, and verbalize rule use in inquiry. Latency and inquiry data suggested that the children who used the rule to solve R problems also chose to solve RC problems by rule rather than by computation. It was concluded that a number of subjects possessed the rule, consciously and deliberately used it in solving concrete perspective-taking problems, and believed in its general veracity enough to rely on it when they did not have to (RC problems). (Author/SB)