The searching behaviour of common shrews, Sorex araneus, was observed as they foraged for prey hidden in a grid of wells in a small arena. The search paths are described and their efficiency compared to that of a simulation model. The shrews searched more efficiently than a random (neutral) model, and were approximately as efficient as the ‘optimal’ model. The efficiency of the model, however, declined more rapidly than that of the shrews with increasing length of search path. It is probable that the shrews were able to increase their searching efficiency by remembering where they had searched: the shrews are shown to have used a simple memory rule, namely alternating right and left turns.