Sheep can use spatial memory to locate preferred food items in a background of less preferred items and forage more efficiently if the preferred patches are aggregated (Edwards et al. 1994). However, if a constant proportion of available patches contain food, foraging efficiency is not affected by the total number of patches (Edwards et al. 1996). The objective of this study was to test the hypothesis that spatial memory will be used to a lesser extent as the proportion of patches containing food increases. Not only is the animal faced with a potentially more complex task as the proportion of food patches increases but the cost:benefit ratio of using spatial memory is less. If 10 of 100 potential foraging sites contain food, an animal with perfect spatial memory could locate 5 of these patches in only 5 visits, whereas random searching requires 50 visits. If 50 patches contain food, 5 visits are required with perfect spatial memory but only 10 visits using random searching. Thus the benefit of using spatial memory over random search is reduced from 5:1 to 2:1.