A 2 X 2 between-subjects design was used to test for the tendency of domestic ferrets to take novel shortcuts. The cross maze with shortcuts adapted by Poucet (1985) was used to train ferrets to search for a goal (an empty food bowl) while having the possibility of seeing the shortcuts or not during training (i.e., a screen, which was either transparent or opaque, blocked off the shortcut). In the test sessions in which the animals were given access to the shortcuts, the goal was visible for half of the subjects in each training condition and not visible for the other half. Ferrets were more likely to take the shortcut if they had seen it during training, regardless of whether they could see the goal or not during the test: Visual familiarity with the shortcut is sufficient to account for shortcut taking. When the goal was not visible and the shortcut had not been seen prior to the test, performance was no different from chance: There was no evidence for the ability to infer a shortcut. Pronounced individual differences were obtained when the shortcut was visually unfamiliar yet the goal was visible.