Two experiments explored 5-month-old infants' recognition of self-movement in the context of imperfect contingencies between felt and seen movement. Previous work has shown that infants can discriminate a display of another child's movements from an on-line video display of their own movements, even when featural information is removed. These earlier findings were extended by demonstrating self versus other discrimination when the visual information for movement was an unrelated object (a fluorescent mobile) directly attached to the child's leg, thus producing imperfect spatial and temporal contingency information. In contrast, intermodal recognition failed when the mobile was indirectly attached to infants' legs, thus eliminating spatial contingencies altogether and further weakening temporal contingencies. Together, these studies reveal that even imperfect contingency information can drive intermodal perception, given appropriate levels of spatial and temporal contingency information.