Common marmosets (Callithrix jacchus jacchus), human children, and human adults learned to find a goal that was located in the center of a square array of four identical landmarks. The location of the landmark array and corresponding goal varied across trials, so the task could not be solved without using the landmark array. In Experiment 1, a matrix of discrete goal locations was presented and the landmarks surrounded and were adjacent to the correct location during training. After training, an expansion test was given in which the distance between landmarks was increased. Marmosets, children (ages 5-9), and adults all readily learned to use the landmarks to search accurately during training. On the expansion test, adults uniformly searched in the center of the array. Monkeys and children concentrated their searching near the landmarks rather than in the center. The monkeys, but not the children, searched more often on the directionally appropriate side of the landmarks than on other sides of the landmarks. In Experiment 2, children (ages 3-5) were trained with a continuous search space and with the goal farther from the landmarks so that a beaconing strategy rule could not be used. Several of the children failed to acquire the training task. Of those who learned to find the goal, three searched in the middle on expansion tests but most searched nearer to the landmarks. The "middle rule" strategy that is uniformly used by adult humans does not appear to be a preferred strategy for children or non-human primates.