In this article, we present experiments studying path planning under spatial uncertainties. In the main experiment, the participants' task was to navigate the shortest possible path to find an object hidden in one of four places and to bring it to the final destination. The probability of finding the object (probability matrix) was different for each of the four places and varied between conditions. Givensuch uncertainties about the object's location, planning a single path is not sufficient. Participants had to generate multiple consecutive plans (metaplans)--for example: If the object is found in A, proceed to the destination; if the object is not found, proceed to B; and so on. The optimal solution depends on the specific probability matrix. In each condition, participants learned a different probability matrix and were then asked to report the optimal metaplan. Results demonstrate effective integration of the probabilistic information about the object's location during planning. We present a hierarchical planning scheme that could account for participants' behavior, as well as for systematic errors and differences between conditions.