Three experiments are reported in which the strategy usage for a compass point directions task was investigated. For this task, people may use either an efficient cancellation strategy or an inefficient spatial strategy (see Roberts, Gilmore, & Wood, 1997). Experiment 1, together with Experiment 2, showed that users of cancellation are unlikely to have explicit prior knowledge of this strategy; instead they discover it as a result of experience with the task. Experiment 1 also showed that strategy selection is unlikely to be based on stylistic preferences, and that reducing the cognitive demands of the task does not increase the discovery of cancellation. Experiment 3 showed that evaluation of strategy validity is also of importance: Unsuccessful evaluation may prevent the selection of a newly discovered strategy, even if it is the more efficient method.